Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Pharmaceutical sector price and productivity measurement : exploring the role of agency, incentives and… Morgan, Steven George 2000

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2000-486796.pdf [ 7.88MB ]
Metadata
JSON: 831-1.0089638.json
JSON-LD: 831-1.0089638-ld.json
RDF/XML (Pretty): 831-1.0089638-rdf.xml
RDF/JSON: 831-1.0089638-rdf.json
Turtle: 831-1.0089638-turtle.txt
N-Triples: 831-1.0089638-rdf-ntriples.txt
Original Record: 831-1.0089638-source.json
Full Text
831-1.0089638-fulltext.txt
Citation
831-1.0089638.ris

Full Text

P H A R M A C E U T I C A L S E C T O R PRICE AND P R O D U C T I V I T Y M E A S U R E M E N T : E X P L O R I N G ^ T H E R O L E S O F A G E N C Y , INCENTIVES A N D I N F O R M A T I O N 1 by S T E V E N GEORGE M O R G A N B.A.(Hon.), University ofWestern Ontario, London, Ontario, 1993 M . A . , Queen's University, Kingston, Ontario, 1994 A THESIS SUBMITTED IN PARTIAL F U L F I L M E N T OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF G R A D U A T E STUDIES Department of Economics We accept this thesis as conforming to the required standard THE UNIVERSITY" OF BRITISH C O L U M B I A April 2000 © Steven G. Morgan, 2000 In presenting this thesis in partial fulfillment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Economics The University of British Columbia 997- 1873 East Mall Vancouver, BC Canada V6T 1Z1 Abstract This thesis explores how decision-making agency roles played by doctors, pharmacists and government affect the social efficiency of choices in the prescription drug market. The primary objective is to contribute to the quality of expenditure decompositions in this sector and, thereby, to draw attention to the real cost of drug consumption in Canada. Expenditure growth in the pharmaceutical sector may occur because Canadians are purchasing more drugs or more in terms of the health outcomes sought through drug consumption. Prices may also be rising for the drugs that patients consume. Furthermore, with new generations of prescription drugs replacing older and often equally effective ones, expenditure inflation may be due, at least in part, to growing inefficiencies in consumption. Deflating nominal expenditures with traditional economic price indexes is a commonly used approach to decomposing expenditure changes into changes in price, changes in productivity or both. This method may be biased because decision-making agency relationships and non-standard financial incentives give rise to possible inefficiencies in the pharmaceutical sector that would not commonly be found in other sectors. This proposition is explored theoretically and empirically. Potential biases stemming from financial incentives are explored in the context of the measurement problem posed by the entry of generic drugs. Traditional techniques of the economic approach to measurement do not capture the full effect of generic competition because decision-making agents do not always have incentive to consider the full price of drugs consumed. Potential information-related problems in pharmaceutical price and productivity measurement are explored within the context of the hypertension market. Health outcomes based indexes are constructed for this treatment category based on recognized national guidelines for the treatment of hypertension. Economic indexes of price and productivity appear to overstate social productivity in this segment because persistent non-compliance with national guidelines has resulted in higher costs without corresponding health improvements. i i Table of Contents Abstract i i Table of Contents i i i List of Tables v List of Figures vi Acknowledgements vii Introduction 1 1. Regulation and Decision-Making Structures In the Canadian Pharmaceutical Sector.... 4 The "Pre"-History of the Modern Pharmaceutical Sector 4 Prohibited Substances and Mislabeling Laws 5 Safety Regulation and Prescription-Only Drugs 5 Doctors, the Drug Industry and the Marketplace of Ideas 7 Requiring Proof of Efficacy 8 Substitution Laws 9 Evolving Financial Agency 10 Direct-to-Consumer 12 The Current Status and Implications for Measurement 12 2. Productivity Measurement with Economic Indexes of Price and Quantity 14 Introduction 14 Index Numbers and The Economic Approach to Measurement 14 Mathematical Underpinnings of Cost-of-Living Measurement 16 Household theory of bounds 18 Social Measurement 20 3. Pharmaceutical Price Indexes: Black Boxes, Pandora's Boxes 23 Introduction 23 Heath, Health Care and Household Utility 24 Pharmaceutical Decision-making in Practice 25 Insurance and Financial Incentives 26 Agency and Prescription Drug Costs 26 Doctors and Drug Prices 27 Prescribing Appropriateness 28 Health Outcomes as Real Productivity 29 4. Agency, Financial Incentives and Pharmaceutical Price Indexes: The Case of Generic Drugs 31 Introduction 31 Generic Drugs 32 Traditional Approaches to Generic Drugs 33 The Generic Diffusion Problem 35 Generic Drugs and Price Indexes for B C Pharmacare 36 Data 37 Price Indexes 39 Findings 40 ii i Changes in the Incentive to Consider Generics The Impact of Policy Changes Measures of Quantity Conclusion 5. Information (or Lack Thereof) and the Real Cost of Treating Hypertension Information and Medical Practice Guidelines Do Practice Guidelines Guide Practice? Hypertension Treatment Difficulties with Antihypertensive Drug Choice Guidelines for the Treatment of Hypertension Adverse News about Calcium-channel Blockers The Impact of Hypertension Guidelines and Adverse News Analysis of Hypertension Treatment for BC Seniors: 1986 to 1996 Description of Data Elementary Price and Quantity Indexes Elementary Indexes of Real Productivity Findings , Price Indexes Aggregate Drug Exposure Profiles First-Line Treatments Therapeutic Trajectories Advertising and Noncompliance with Guidelines Drug Advertising in Economic Literature Advertising and Information Conclusion Appendix A: The 1997 and 1999 Guidelines Appendix B: Lists of Products Appendix C: New Product Entrants Appendix D: Regression Analysis for Price Indexes Appendix E: Regression Analysis of Patients Receiving Treatments by Classification... Appendix F: Alternative Indexes by Drug Subclass Appendix G: Classification of First-Line Treatment for Non-Persistent Patients Aged 66 Appendix H : Regression Analysis for Trends in First-Line Treatments for Persistent Patients Bibliography iv List of Tables Table 1: Results of Surveys of Physicians' Knowledge of Drug Prices 27 Table 2: DIN and AHFS Numbers for 2-miligram Diazepam Tablets 37 Table 3: Classification of Drugs Purchased 38 Table 4: Rates of Change for Chained Price Indexes 40 Table 5: Major Canadian and American Guidelines for Hypertension Treatment 1977 to 1999 58 Table 6: First-Line Prescribing in 1996 for Elderly People in British Columbia in Relation to Relative Contraindications* 60 Table 7 :Treatment Classification for A l l Seniors Receiving Antihypertensive Drugs 71 Table 8: The Average Cost Per Patient Treated, Grouped by Drug Class 72 Table 9: Classification of First-Line Treatments for Persistent* Patients Aged 66 and Older 74 Table 10: Transitions for 66+ Year Old Patients Who Refill Continuously For Two Years 76 Table 11: Diuretics 84 Table 12: Beta Blockers 84 Table 13: ACE-Inhibitors 84 Table 14: Calcium-Channel Blockers 85 Table 15: Other Antihypertensive Drugs 85 Table 16 :Canadian Launch Dates for Top Selling Antihypertensive Drugs 86 Table 17: Regression Analysis for Trend in Price Indexes 87 Table 18: Regression Analysis for Trends in Patients Receiving Treatments by Classification 88 Table 19: Annualized Price Indexes for Hypertension Treatments, Grouped by Drug Class and Index Type 89 Table 20: Classification of First-Line Treatment for Non-Persistent Patients Aged 66 92 Table 21: Regression Analysis for Trends in First-Line Treatments for Persistent Patients. 93 v List of Figures Figure 1: Decision-Making Structure with Safety Regulations and Prescription-only Drugs 7 Figure 2: Decision-Making Structures with Independent Drug Benefits Providers 10 Figure 3: Decision-Making Structure with Integrated Health and Pharmaceutical Benefits Provision 11 Figure 4: Alternative Paasche Price Indexes for Single and Multi-source Drugs 41 Figure 5: Brands' Average Share of Markets Grouped by Tear of Generic Entry 44 Figure 6: Brands' Share of Total Sales in Multi-Source Diuretic Drug Markets 45 Figure 7: Per-Capita Expenditures and Per-capital Quantity Indexes 47 Figure 8: Alternative Indexes of the Price of Hypertension Treatments 68 Figure 9: Alternative Indexes of the Quantity of Hypertension Treatments 70 Figure 10: Average Cost Per Year of Treatment on Diuretic Drugs as Prescribed 90 vi Acknowledgements I would like to acknowledge a few of the many people who helped me complete my coursework, comprehensives and this thesis. At various stages of the Ph.D. program, I have been assisted by several people in the Department of Economics. Erwin Diewert and Alan White—one of Erwin's students—have been especially supportive of my foray into measurement issues. I have also benefited from working with many people at the Centre for Health Services and Policy Research, my "other home" at U B C . Among them, K i m McGrail has been a constant source of data-related information as well as lighthearted inspiration, and Morris Barer, director of the Centre, has given much encouragement and mentoring. I am extremely fortunate to have worked with Professor Robert Evans. I am truly indebted to Bob for all that he has done for me over the past five years. He has been an inspiring, enthusiastic and supportive supervisor. Finally, I thank my parents, my (growing) family and my friends for all of their support. This research was made possible in part by Health Canada through a National Health Research and Development Program (NHRDP) Research Training Award. Access to the BC Linked Health Database was made possible by the Centre for Health Services and Policy Research, U B C . vii Introduction As of 1996, Canadians spent over $10 billion on pharmaceuticals outside the hospital setting, $7.5 billion of which was spent on prescription drugs. Expenditures on prescription drugs in Canada have grown considerably over the past two decades. Since the mid-1970s, pharmaceuticals have been the fastest growing component of Canadian health care spending. In 1976, drugs accounted for approximately 8 percent of total healthcare expenditures in Canada. By 1996, they accounted for approximately 13 percent of expenditures. There are numerous explanations for the growth in pharmaceutical expenditures. One of the aims of price and productivity measurement in the pharmaceutical sector is to determine whether changes in expenditures are due to changes in aggregate price, changes in aggregate quantities or both. It is clear, for example, that health system restructuring, aging populations and technological progress cause a steady (albeit difficult to quantify) increase in the demand for prescription drugs. Pharmaceutical expenditures could therefore be rising because Canadians are purchasing more drugs or more in terms of the health outcomes sought through drug consumption. On the other hand, price inflation occurs in the pharmaceutical sector as it does in other sectors of the economy. Such price changes will certainly drive a portion of pharmaceutical expenditure inflation. Another source of rising drug costs has received considerable attention in health services research. New generations of prescription drugs tend to replace older, cheaper ones. They often do so despite the absence of substantiated benefits. This implies that increased spending on pharmaceuticals is due, at least in part, to consumption that is "socially inefficient." If costs rise without increases in the social "good" purchased through drug consumption, then implicit social price of drugs rises even if the unit prices of drugs do not. This thesis explores price and productivity measurement in the pharmaceutical sector. Emphasis is placed on how decision-making agency roles played by doctors, pharmacists and government effect the social efficiency of choices in the prescription drug market. This is both a health economics thesis and a measurement thesis because it draws heavily on, and contributes to, literature in both areas. However striking, the empirical significance of distortions caused by financial incentives and imperfect information cannot be generalized to other sectors because the results found in this thesis rely heavily on the contextual setting in which consumption decisions are made. Because this thesis deals with both health economics and measurement theory, it contains two introductory chapters. The first chapter provides a brief history of government regulation and other influences on the decision-making processes that lead to drug consumption. It is most useful for those unfamiliar with the decision-making and decision-influencing roles played by governments, health professions, insurers and manufacturers in pharmaceutical sector. The primary illustration made in this chapter is that interventions in the drug-related decision-making process by government—and, more recently, by insurers— have been motivated by the belief that free markets do not serve social objectives in the pharmaceutical sector. In the case of prescription-only drugs, consumer sovereignty has 1 been abrogated and manufacturers' freedoms restricted because market-style competition and drug selection has failed to protect consumers from health hazards and fraud. A discussion of regulatory policy aimed at manufactures and market structure may appear conspicuously absent from this chapter is. Those issues—many of which are discussed in Morgan (1998a)—are important determinants of market dynamics in the pharmaceutical sector, but do not affect the accuracy of measures of the social costs and benefits of aggregate drug consumption—they will , of course impact on the level of such measurements. The second chapter describes the task of price and productivity measurement and the economic approach to it. The theoretical basis for constructing and interpreting traditional economic price and quantity indexes is reviewed to illustrate that these indexes are not mere statistical estimators of price movements. Readers are reminded that traditional economic price and quantity indexes gauge sector productivity by way of a utility theoretic model of consumer and social welfare. These indexes are efficient estimators of "true" social welfare provided that the assumptions of the traditional approach to measurement are met. The third chapter puts economic theory of measurement within the context of the pharmaceutical industry. This chapter illustrates the fact that important assumptions of the traditional approach to measurement are not met in the pharmaceutical sector. Attention is drawn to the non-standard financial incentives of principal decision-makers: patients and doctors. A majority of patients are insured for their prescription drugs and physicians do not pay for the drugs they prescribe as part of the medical services they sell. Evidence is presented that indicates that doctors generally have poor knowledge of drug prices, regardless of whether they believe that price should be a consideration when choosing prescription drugs. Evidence is also presented to show that, irrespective of drug costs, the expected health benefits of drugs prescribed do not outweigh expected risks of side-effects and adverse reactions on a surprisingly frequent basis. It is argued that price and quantity indexes based on real health outcomes may be compared to traditional economic indexes to gauge the empirical significance of distortions in the drug related decision-making process caused by the non-standard financial incentives and objectives of decision-making agents. The fourth and fifth chapters of this thesis contain empirical studies wherein traditional indexes are compared with alternative approaches to measurement in the pharmaceutical sector. In the fourth chapter, the magnitude and causes of measurement bias related to the use of generic drugs are investigated. American studies show that price indexes designed to account for the price differential between generic and brand-name drugs report lower inflation rates than indexes that do not adjust for product interchangeability. Moreover, adjusting for generic availability as one would account for any "new good" in a traditional price index does not capture generics' long-term impact on average prices. Using data from the B C Pharmacare Plan A, it is shown that traditional techniques of the economic approach to measurement may not capture the full effect of generic competition on the cost of drugs purchased because of the nonstandard financial incentives of decision-makers involved in the brand versus generic choice. This follows from the results of a "natural experiment" that occurred in 1994. Both before and after 1994, the Pharmacare plan would pay the full ingredient cost of generic drugs purchased by its members. After 1994, Pharmacare would not pay for brand-name drugs when lower cost alternatives were available. Beneficiaries were made responsible for the price differences between chemically 2 identical brand-name and generic products i f they chose to remain brand-loyal. This relatively minor change in financial incentive caused an immediate and massive increase in the use of available generics. Standard indexes failed to capture the resulting savings because—at least prior to the policy change—the financial incentives of involved decision-makers were inconsistent with the assumptions of the traditional economic approach to measurement. The fifth chapter also uses data from the B C Pharmacare Plan A . This chapter investigates potential information-related problems in pharmaceutical price and quantity measurement. Attention is focussed on hypertension drugs, which offer a particularly clear illustration of the distinction between scientific information, which might be considered measurable and socially valuable, and the unmeasurable "information" affecting doctors' "preferences" for certain drugs. Indexes of real health outcomes are constructed for this treatment category by using information from national guidelines for the treatment of hypertension to determine an aggregate health-production function for this class of drugs. These "real productivity" indexes are compared with traditional economic indexes of price and quantity for the treatment of hypertension. The validity of assumptions made in constructing the health outcomes based indexes is explored by comparing actual prescribing and utilization patterns with those that guidelines recommend. The implied measurement bias found in this chapter is particularly striking. Economic indexes appear to overstate productivity in this segment because persistent non-compliance with national guidelines has resulted in higher costs without corresponding health improvements. Accumulated over the course of 11 years, the economic index of the price of hypertension treatment grew approximately 15 percent whereas the favoured health outcomes based price index grew 172 percent. The potential hazard of not addressing the measurement bias in traditional price and quantity indexes is that policy may be guided by a perception that the pharmaceutical sector is more "productive" than is in fact the case. Increased price per unit of care delivered stemming from inefficient drug selection is a major determinant of expenditure growth not captured by conventional economic indexes. This occurs because the pharmaceutical sector does not satisfy the conditions that guarantee efficient consumption decisions as per the traditional economic model. Health-outcomes measures of pharmaceutical sector productivity, based whenever possible on recommended prescribing practices, will better indicate whether changes in the true quantity and/or quality of care delivered are consistent with changes in expenditures on drugs. For the hypertension sub-market studied here, such indexes indicate that physicians' persistent non-compliance with recommended prescribing practices cost the British Columbia government far more per unit of care provided than the economic indexes suggest. If the cost per patient treated for hypertension actually had grown according to the rate implied by the traditional economic price index, the BC Pharmacare program would have spent approximately $207 million less on hypertension drugs than it did from 1986 to 1996—$33 million less in 1996 alone! Findings such as those of this thesis may help to focus the attention of researchers and policy-makers on interventions to improve the social efficiency of the prescription decision-making process. 3 1.Regulation and Decision-Making Structures In the Canadian Pharmaceutical Sector Drug related decision-making is a challenging task that can potentially lead to enormous gains in personal and social welfare. It can also lead to wasted resources or even tragic consequences. In response to the potential risks, benefits and costs that have accompanied increasing drug availability, the decision-making process that leads to drug consumption has become increasingly regulated. This process is now like no other. Prescription decision-making, in particular, is characterized by a legally mandated "command" structure, wherein professionals must make decisions on behalf of others. It also involves agents who intervene in the prescription decision-making process through regulation or financial relationships. The roles taken by stakeholders are constantly evolving as governments, professionals, consumers and manufacturers attempt to influence outcomes. This paper offers a brief history of decision-making structures and agency relationships in the pharmaceutical industry. The "Pre"-History of the Modern Pharmaceutical Sector A century ago, the pharmaceutical industry bore only slight resemblance to what it is today. Pharmaceutical companies were relatively small; the arsenal of medicines available was limited; and, apart from professional licensing in medicine and pharmacy, the industry was virtually unregulated in North America. Consumers often chose medicines at their own discretion (a process now described as self-medicating), but could also make choices based on the advice of pharmacists or physicians. They bought medicines through pharmacies, mail-order houses or directly from their doctors. Only a small segment of the pharmaceutical sector—the "ethical" drug market—bore a resemblance to today's prescription-only market segment. The "ethical" drug market was so named because manufacturers would not advertise these products directly to consumers. Manufactures of ethical drugs produced chemical ingredients that were sold in bulk to pharmacists and physicians who compounded them into consumable form according to recipes found in pharmacopoeias. Unlike modern prescription drugs, a prescription was not necessary for the purchase of ethical drugs. Consumers could ask for ethical drugs provided they knew the name of the desired compound or its "recipe"—perhaps found on a doctor's prescription written for a friend or relative. The market segment that dominated the pharmaceutical sector at the turn of the century was that of "proprietary" medicines (sometimes called "patent medicines" though none were patented). Proprietary medicines were generally made with "secret recipes," most of which were little more than coloured water (Temin 1980). Those with any "medicinal" effect typically contained one of a limited number of substances—usually alcohol, though sometimes cocaine or opium (Temin 1980). Proprietary medicines were advertised heavily in popular magazines and newspapers, where manufacturers described 4 their products as cures for all kinds of conditions, including the serious illnesses such as cancer and heart disease. The pharmaceutical industry of the late 19 t h century operated as a free market wherein patients were the primary decision-makers. Though they would often act on doctors' and pharmacists' recommendations, they were free to choose the products (and advice) that they wished. The sector then resembled a market for ordinary commodities more then than it would ever again. Prohibited Substances and Mislabeling Laws Governments in Canada, the US and Britain began to regulate drugs at the turn of the century. The first concern was to ensure the safety of consumers who tried alternative proprietary medicines in search of satisfactory results. Regulation of the pharmaceutical industry in Canada effectively began with the Proprietary or Patent Medicine Act of 1909.1 This Act required that manufacturers register their "secret formulas" with the government. It also banned the use of cocaine and created a list of ingredients that, i f present in the product, must be indicated on the label. The first amendments to this Act came in 1919, limiting the allowable dosage of certain ingredients and prohibiting firms from representing a product as a "cure." Requiring that potent ingredients be labeled increased the quality of information that consumers would have when choosing among products. Since few products on the market could offer more than symptom relief for minor illnesses, the labeling regulations were not intended to help consumers find effective medicines, but rather to protect them from harm caused by dangerous or addictive ones. Intervention through the Proprietary or Patent Medicine Act—which became the Food and Drugs Act in 1920—signified an emerging role for government in this sector and a changing view towards the operations of a free market for drugs. Government was beginning to intervene in the market on behalf of consumers who could not protect themselves. Its decision-making role was limited but primary in the sense that it determined the set of choices available to consumers. The government's role this sector would increase dramatically based on the precedent set by these early restrictions on drug availability and labeling. Safety Regulation and Prescription-Only Drugs In the 1930s, as was consistent the loss of faith in free markets that followed the depression, governments in North America expressed the belief that pharmaceutical market transactions are beyond the understanding and control of consumers (Temin 1980; Bogner 1996). The Canadian government began to limit consumers' decision-making freedom in 1934 by prohibiting consumer directed advertising for medicines to treat certain serious illnesses—including cancer and diabetes (Goyer 1986). This would begin the process of legally delegating product selection to health professionals rather than consumers. The US government also proposed a bill in 1934 that would regulate the drug industry, but there was 1 Drugs were included in the 1875 Inland Revenue Act under sections governing adulteration of foods (Curran 1953). These adulteration laws had little effect because most medicines had "secret formulas." 5 not enough political will to interfere with the free workings of the market. It would take the Sulfanilamide tragedy of 1937—wherein 100 Americans died from the toxic effects of a solvent used in Massengill's Elixir Sulfanilamide—to give the governments in the US and Canada opportunity to regulate drugs further. Economic theory might predict that competition would ensure that only reasonably safe medicines would survive in the market over the long run. The Sulfanilamide tragedy illustrated why extra-normal safeguards are needed in the pharmaceutical sector: the damages done while forces of competition winnowed out dangerous drugs could be substantial and irreversible (Temin 1980, p. 44). Just one year after the tragedy, the US government passed new drug regulations—the 1938 US Food, Drug and Cosmetics Act— that would dramatically alter decision-making structures in the US and elsewhere. In addition to implementing a safety review process for new drugs, the Act required that medicines be clearly labeled with ingredients, indications, cautions and instructions for use—a then unprecedented amount of product information. Not only was a great deal of information required, the laws deemed a drug to be mislabeled i f dangerous when used as recommended. In theory, these regulations would support self-medication by ensuring that consumers had detailed information about the contents of a drug, its uses and its recommended doses. Furthermore, consumers could rest assured that a properly labeled drug would be safe at the recommended doses. In practice, the US laws of 1938 hindered self-medication as never before due to an exemption put in place for drugs labeled with the words "Caution: to be used only by or on the prescription of a doctor, dentist or veterinarian." If a product was labeled (and sold) in this manner, the remaining labeling requirements were dropped—along with the associated liability. The prescription-only classification marked a dramatic change in the industry because it legally entrenched a decision-making hierarchy wherein only doctors could choose certain medicines on behalf of consumers. Interestingly, rather than designate what drugs would become available by prescription only, the US government left the choice to manufactures until 1951. Virtually all products launched in the 1940s were labeled as prescription-only drugs (Temin 1980). The route to prescription-only regulations was slightly more direct in Canada than in the US, though it was based largely on US precedent. Amendments to Canada's Food and Drug act in 1939 gave the Governor in Council authority to define the terms of sale of any drug. By 1941, the government had established the first list of drugs (Appendix IV of the Food and Drugs Act) that would be available only by the prescription of a physician, dentist or veterinarian (Goyer 1985, Curran 1953). Canadian regulators did not have a formal safety approval process until 1951. 6 Figure 1: Decision-Making Structure with Safety Regulations and Prescription-only Drugs Government Regulator Principal Decision-Making Agents Doctor in Private Practice Patient (Consumer) Retail Pharmacist Following the regulatory changes of the late 1930s, there were two legally enforced levels of decision-making agency in the pharmaceutical industry, as depicted in figure 1. Through the regulation of new product safety, the government chose the drugs that would make it to market; then, for prescription-only drugs, physicians decided which products patients would use. Of course, patients had the choice of whether or not to buy medicines prescribed for them, but they could not choose to switch medicines without consent (in the form of another prescription) from their doctors. Doctors, the Drug Industry and the Marketplace of Ideas Under the relatively new prescription decision-making structure, physicians played the most critical role in determining which drugs were demanded. Their responsibility was to scientifically evaluate alternative treatments and determine which would be safest and most effective to treat a patient's illness. The prescription-only category was put into law at the dawn of the therapeutic revolution, when there was a relatively limited number of drugs to choose from (Temin 1980). Shortly after the war, hundreds of new drugs came onto the market for dozens of new indications. In the "marketplace of ideas" that was to "inform" the prescribing process, drug companies managed to capitalize on the tremendous informational "costs" associated with rational prescribing. Since doctors did not pay for the drugs they prescribed, and since—when paid on a fee-for-service basis—they had an incentive to see patients as quickly as possible, entrepreneurial incentives could rationalize prescribing practices based on custom or habit (Temin 1980). Through intense promotional activities, drug companies facilitated the replacement of information-based decision-making with prescribing habits based on emotional assessments of products (Walker 1970). The key to establishing a product as a physician's drug of choice was the personal relationship between the physician and the drug company. Drug company detailers would build relationships that instilled in physicians a sense of brand affinity and trust—emotions that are thought to be critical determinants of customary or habitual decision-making (Temin 1980). To make personal relationships of doctor and detailer critical, larger companies jammed the low-cost channels of marketing by flooding the mailboxes of physicians with fliers and "news-letters." It is estimated that, by 1962, doctors were receiving an average of 4,000 pieces of mail from drug companies every 7 week (Walker 1970). With low-cost information channels jammed, larger drug companies could establish corporate identity through their sales forces—a practice that was too costly for smaller firms (Walker 1970). When "scientific" information was provided by drug companies, it was often dubious. Companies were known to grossly exaggerate benefits and base claims on false inferences. For example, manufacturers of antibiotics in the 1950s relied on evidence of the differential rates of "blood absorption" among competing products—even different brands of the identical drug—to support their claims of superiority. These claims were widely publicized despite condemnation by academics and government regulators who showed that the tests were based on poor methodology and bore no relationship to the therapeutic value of the competing products (Temin 1980 pp.96-102). Requiring Proof of Efficacy The new marketplace had resulted in enormous economic successes for pharmaceutical manufactures in the 1950s. They were so successful, in fact, that they captured the attention of numerous governments and public agencies. Commissions of inquiry into the price of drugs and the conduct of pharmaceutical companies became common in the 1960s. The most important of these because of its timing, scope and nation of origin were the hearings of the US Senate Subcommittee on Antitrust and Monopoly chaired by Senator Estes Kefauver. The Kefauver hearings began in 1959 and concluded in 1961 with a report that was both detailed and damning. Several reports of Canadian investigations followed the Kefauver hearings—almost all were equally scathing in their criticism of industry conduct.2 What was significant about the inquiries conducted in Canada and the US was that regulatory emphasis no longer focussed exclusively on concern for protecting consumers from adulteration and unsafe drugs. New concerns emerged that focussed on doctors' prescribing habits and the practices of drug companies that were supposedly "informing" the prescribing process. Over the course of the Kefauver hearings, many noted physicians testified that haphazard information gathering by individual doctors simply could not result in true knowledge about the effectiveness of a medicine (Temin 1980). In the competitive marketplace, numerous "me-too" medicines, which offered no substantiated advantages over older ones, were being widely prescribed at high costs to consumers. Finding ways to protect consumers from "excessively" high prices, particularly for products of questionable value, became the focus of government policy development. Despite the recommendations of every major inquiry, new regulations were not brought in until a wave of public concern surrounding a highly publicized drug tragedy— just as was the case with earlier regulations. This time, the birth-defects caused by pregnant women's use of Thalidomide precipitated the passing of regulations requiring evidence of safety and efficacy in the US (1962) and Canada (1963). The new regulations added another task for government in the drug related decision-making process—though they actually did little to address the safety issue raised by the Thalidomide tragedy (Temin 1980). 2 Among the Canadian inquiries were the 1963 Combines Investigation Report; the 1963 Restrictive Trade Practices Commission report; the 1965 report of the Royal Commission on Health Services; and the 1967 reports of the Special Committee of the House of Commons on Drug Costs and Prices. 8 The efficacy regulations of the 1960s retained the hierarchical decision structure— from government to physician to patient—while giving government greater authority to restrict which products could compete in the pharmaceutical marketplace. The extension of the F D A authority [in 1962] reflects an additional erosion of the legislative belief that markets protect consumers. Not only can consumers not rely on the general market to weed out inferior products, they cannot even rely on the highly educated medical market—that is, practicing physicians—to weed out worthless drugs. (Temin 1980, p. 128) The scientific definition of clinical effectiveness on which the government would base market restrictions implied a distinction between pharmaceuticals and other goods that could be judged on unmeasurable characteristics. The personal preferences or impressions of individual physicians or patients could no longer justify the existence of a product in the pharmaceutical marketplace. The physician-patient agency relationship remained, but the competency of physicians as independent decision-making agents had suffered a serious challenge. Substitution Laws Public inquires of the 1960s put forth proposals that would give patients and pharmacists some new decision-making authority. At the time, consumers (or pharmacists acting on their behalf) could only make the choice among brand and generic versions of a drug i f the doctor prescribed by its generic name. The near exclusive use of brand-names in marketing and, therefore, prescribing represented a post-patent barrier to competition in the 1960s. To overcome this problem, provincial and state governments —beginning with Alberta in 1962—passed laws that would legally permit substitution among equivalent products even when a particular brand name appeared on the prescription. This type of law created only a minor change in decision-making structures, but signified the belief that doctors were not responsive to price signals in the market. Despite being a small change in decision-making structures, substitution laws were partially responsible for provoking new marketing strategies aimed at patients: trade-dresses. This strategy attempted to ally patients, in part because of their new (though limited) role as decision-makers in the pharmaceutical sector. In 1966, Li l ly was the first company to employ this technique by marking its pills and tablets with what it called "identi-codes"— markings that were unique to Lil ly products (Bogner 1996). They sold the idea to shareholders as a mechanism that would be useful to doctors, pharmacists and police. Truthfully, by marking its tablet, Li l ly was marketing to the patient. Once familiar with the brand-name company's trade-dress (markings), the patient could later identify when a competitor's product had been substituted. Uneducated about the therapeutic equivalence of generic and brand-name drugs, the patient might question what happened to "their" drug i f an interchangeable product had been dispensed in its place. Competitors could not copy trademarks, which, unlike patents, are protected indefinitely. 9 Figure 2: Decision-making Structures with Independent Drug Benefits Providers Government Regulator Principal Decision-Making Agents Doctor in Private Practice Patient (Consumer) ^ Retail Pharmacist Drug Insurance Company Evolving Financial Agency The 1980s were an important decade for the pharmaceutical industry. In previous eras, innovation was the result of exhaustive search and serendipitous observation—the roots of antibiotics, anti-psychotics and diuretics for example (Temin 1980; Bogner 1996). Beginning in the 1970s, drug research and development became more deliberate, based on new understandings of the biological events that attended certain illnesses. "Rational drug design" was expensive but profitable because firms could choose to search for therapies of the greatest economic value. Most frequently, drug companies targeted maintenance therapies for chronic conditions, including ulcers, hypertension and depression. In addition to the size of demand that could be expected for these drugs, maintenance therapies also had the advantage of being purchased in the retail setting, outside the demand-side pressures placed on drug prices by hospitals (Bogner 1996, p . l 10). Drug purchasing in the hospital setting had become quite sophisticated in the 1970s. To achieve lower costs and improve the effectiveness of drugs used in hospitals, hospital administrations employed formularies, prescribing protocols and involved pharmacists in the drug-related decision-making process. These tactics kept drug companies' margins from sales to hospitals lower than in the ambulatory setting, where involvement of third-party payers in the drug decision-making process was just beginning to develop. As drug plans in the ambulatory setting evolved from the indemnity plans of the 1970s, providers of drug benefits could employ a number of cost-controlling devices as well. As illustrated in figure 2, the agents that independent drug benefits providers could influence were patients and pharmacists. User-payment schemes were often implemented to curb drug use by patients—though these schemes often had undesirable consequences (financially and in terms of health) when patients made ill-informed decisions regarding which drugs to stop taking (Soumerai et al 1993). Insurance companies also had some influence over pharmacists because insurers often (and increasingly) paid pharmacists directly for drugs dispensed to beneficiaries. The substitution laws passed in the 1960s made it possible for 10 Figure 3: Decision-making Structure with Integrated Health and Pharmaceutical Benefits Provision Government Regulator Principal Decision-Making Agents Doctor in Private Practice Patient (Consumer) Retail Pharmacist If Integrated Insurance Provider payers to encourage generic drug use by offering financial incentives for pharmacists who dispense generics. These pharmacy-directed policies were limited to the brand-versus-generic choice because that was all the pharmacist had agency over. Except in integrated managed care environments, third party payers lacked financial relationships with physicians—the more obvious decision-maker to attempt to influence. Expenditures on drugs grew rapidly in the 1990s, focussing more attention on strategies to curb the use of excessively priced or ineffective medicines. Whenever possible, third party payers began to target the decision-making processes of physicians. Environments where the provider of drug benefits had a financial relationship with physicians provided the greatest scope for intervention. Integrated providers of medical and pharmaceutical benefits, for example, could give doctors incentive to consider drug costs when prescribing through financial rewards or penalties—depicted in figure 3. Leading the way in integrated drug benefits management, pharmaceutical benefits providers in the US, Germany, and the U K implemented various schemes to give physicians incentive to consider the costs (and savings) that result from their prescribing practices. As managed care insurance plans in the US grew, companies saw the potential to reduce the cost of the drug benefits without harming patient outcomes—therefore without increasing costs of other components of these plans. By the 1990s, companies specialized in the management of drug benefits on behalf of major sponsors of health and drug insurance. These Pharmacy Benefits Management companies (PBMs) offered a number of cost-control services, including formularies and prescribing protocols. Using computer networks to track the behaviour of member patients and prescribers, PBMs monitored prescribing patterns and health outcomes. They targeted doctors that computer records showed to have poor prescribing records with "counter-detailing" efforts aimed at educating them about best 11 prescribing practices.3 PBMs were so successful at reducing drug costs and, thereby, lowering drug industry profits that major pharmaceutical companies bought the largest PBMs in 1993 and 1994 (McGahan 1994). In Germany, where doctors are paid out of public funds that must also cover all drug costs, health authorities put a global cap on regional drug budgets for 1993. If doctors prescribed at a rate that exceeded these caps, their own payments would be reduced up to a maximum of 2 percent for that year. In January of 1993, the first month of the global budget system, German doctors prescribed 26 percent less frequently than they did in January of the preceding year. Moreover, they prescribed generic drugs more often and tended to use higher-cost products with more discretion (Munich and Sullivan 1994). In the U K , health authorities gave group practices the option of choosing to be paid on a capitation basis that included funds for drugs per patient. These "fund-holding" doctors would then be responsible for any drug costs in excess of the amounts paid to them on an age/sex-adjusted basis. Numerous practices chose fund-holding. Fund-holding doctors' started prescribing less, used more generics and implemented practice-formularies to reduce drugs of questionable value (Baines et al 1997, Dowell et al 1996). However, some criticize the program for producing only limited short-term savings (Walley et al 1995). Direct-to-Consumer Facing pressures from organized purchasers of medicines, the drug industry in the 1990s has begun to try to "market around" the policies of institutional decision-makers. To do this, drug companies have placed increasing emphasis on marketing directly to consumers. The practice is legal in the US, which influences the Canadian market because Canadians see "cross-border" direct-to-consumer advertising for prescription-only drugs by way of American magazines, television and internet, even though these ads are still illegal in Canada. Billed as "information," direct-to-consumer drug advertising is designed to challenge the traditional as well as evolving decision-making structures in the industry. If a drug company can persuade patients that they need a product, those patients may pressure their doctors to prescribe accordingly. This will likely increase the use of drugs regardless of medical appropriateness because doctors acknowledged that they would rather prescribe an unnecessary drug requested by a patient than loose the patient to another doctor (Schwartz et al 1989, p.579; Hemminki 1975; Sleath et al 1997). Further, i f these patients are covered by drug plans that will not pay for the medicine, perhaps because its proven benefits do not justify its cost, they may pressure the insurer to extend coverage to include the product. The Current Status and Implications for Measurement In 1997, the National Forum on Health recommended that the Canadian government implement a universal public pharmaceutical insurance. New strategies to cost containment in the pharmaceutical sector were thought possible under such a system because, among 3 PBMs would even contact doctors directly if they prescribed medicines that were not on the company formulary, asking them to "switch" the prescription over the telephone while the patient waited to have the prescription filled. 12 other things, administrators would have levers of control over the physicians, the sector's principal decision-makers. Despite this recommendation, the current decision-making structure in the Canadian pharmaceutical sector is fragmented and not dramatically different from that which existed in the 1980s. The only province to adopt a universal4 drug insurance plan, Quebec, has a non-integrated insurance scheme with plans offered by a mix of private and public providers (Castonguay 1993, Morgan 1998b). Doctors remain the principal decision-makers in Canada with limited intervention from providers of benefits in the private and public sectors. No practices are in place that make doctors the residual claimants for the costs or savings that result from their prescribing. Health Canada continues to regulate safety and efficacy. However, Health Canada has come under attack for hastily approving medicines that have not met rigorous standards of safety or efficacy (Lexchin 1994). In part, this may be attributable to the fact that the drug approval process is now done on a cost-recovery basis, which engenders an institutional belief that the industry is Health Canada's client, not the public. Some provincial governments have intervened in this sector through physician education. By way of continuing medical education seminars, information newsletters and face-to-face "academic detailing," governments have attempted to improve the appropriateness and, therefore, cost-effectiveness of physicians' prescribing. Education has been successful in changing prescribing habits in the short-run (e.g., 6 months) but not in the long-run because they have not been able to compete with ongoing pharmaceutical marketing (Anderson and Lexchin 1996; Anderson et al 1996; Haaijer-Ruskamp and Denig 1996; Soumerai et al 1996; Steel et al 1989). As has been discussed here, many stakeholders have been vying for influence over the decision-making process that leads to drug consumption. It is not clear which of these stakeholders can be said to act in a manner most consistently in the public interest. However, decades of public inquiry and criticism regarding the system that influences and informs the current decision-making process indicates that we have not reached a structure that ensures efficient choices. 4 Several provinces have so-called universal plans that cover catastrophic drug expenses. These plans do not affect patients until their annual drug costs have exceeded deductables in excess of $800 per year and more. 13 2.Productivity Measurement with Economic Indexes of Price and Quantity Introduction Aggregate price and quantity indexes—economists' proverbial "black boxes"—can summarize millions of market transactions in a single number. There are many formulae for these indexes, the most common of which are relatively simple to construct and provide among the best approximations of the "true" levels of price and welfare. Traditionally, "truth" is defined within economic models of consumer demand and social welfare. Provided that the models are suitable approximations of reality, economic indexes of price and quantity may facilitate the evaluation of public policy and market performance because they help to gauge the social productivity of a market. Roughly, they do this by telling us how much value for money is being generated by transactions in a given sector. These indexes may be particularly useful in the pharmaceutical sector where government involvement provokes continuous monitoring and the number and complexity of transactions make it difficult to evaluate overall market performance on a transaction by transaction basis. Statisticians and economists often borrow standard economic index formulae for price and quantity measurement in the pharmaceutical sector. Unfortunately, this may result in inaccurate measures of social productivity because the economic model on which common index formulae rest does not describe the pharmaceutical sector well. The measurement of pharmaceutical sector price levels and productivity with common index formulae may be biased due to widespread insurance coverage, public subsidy, agency relationships and externalities in consumption. A l l of these aspects of the pharmaceutical sector make it distinct from the classical market models upon which welfare-theoretic interpretations of common indexes are based. Fortunately, the pharmaceutical sector has a potentially "countable" metric of output—health status—that makes it feasible to devise alternative measures of output and, thereby, price levels and sector productivity. Real output indexes based on health outcomes can be compared against standard measures to indicate the magnitude of pharmaceutical measurement bias with common index formulae. Index Numbers and The Economic Approach to Measurement The objective of index number construction is to combine numerous microeconomic observations of market transactions into scalar indices that in some sense represent the microeconomic data (Diewert 1988b, 1981). At its most micro-level, an observation would be characterized not only by the price paid for a particular type and quantity of product, but also by the identities of the purchaser and seller, the date, time of day, location, and even the state of the world (e.g., the weather). Data at this level are almost impossible to obtain and are probably too fine to be comprehensible. In practice, aggregation begins by grouping transactions considered equivalent from an analytical standpoint. 14 Grouping transaction level data involves the construction of elementary indexes of price and quantity for each "good" (Diewert 1995). Goods are defined by aggregating across similar items, locations and times of sale. Products with similar characteristics (e.g., Granny Smiths and Golden Delicious apples) may be grouped together; purchases within a given "period" (month, quarter, year) may be grouped together; and purchases within geographical areas and across outlet types may be grouped together. Decisions concerning what to aggregate should be based on the economic significance of grouping goods together. If items characterized by similar physical properties, locations and times of sale are close substitutes from consumers' perspectives, they should be aggregated together. For, i f data are "too" finely defined, then quantity observations would frequently be zero for any particular "good," making it difficult to track changes in price and quantity between periods (Diewert 1995 and 1988b). The aggregation of individual transactions into elementary indices results in price and quantity data for N goods, over T periods. The vectors of these elementary indices are p l = (pi1, p2 l ... PN ' ) and x = (x/, x2l ... XN 1 ) . With the elementary indices, the analyst is to construct scalar indexes of aggregate price and quantity: P t and Q T . These aggregate indexes must relate in a meaningful way to the market transactions represented by the elementary indices. Economic indexes are chosen to reflect the objective of the buyers in the market transactions under analysis—such as profits for firms, utility for consumers or social welfare for societies. It is in this manner that the economic approach to measurement differs from a purely statistical approach. Indexes designed in the economic approach are not merely statistical estimators of price and quantity movements; they measure the price and quantity of some metric of production generated by market transactions—profits, utility or social welfare. The metric chosen determines what form of productivity is being gauged with economic price and quantity indexes. The relationship that economic price and quantity indexes bear to a common metric of production is mathematically determined by the 'adding up' condition (Diewert 1981). This is the first restriction on economic indexes. It requires that the product of the aggregate price and quantity indexes equals total expenditures on the underlying goods in each period: P'xQ1 = I n P nV. (1) In light of this restriction, it should be clear that choosing one form of price index over another involves choices about what it is the index "measures the price of." Interpretations of economic price and quantity indexes should therefore be based on the metric behind the index formulae chosen. For example, i f a desired metric of production for computer-related transactions is processing power represented by computer purchases, then the analyst must use a method of converting the number of computers purchased into a measure or total processing power purchased. The adding up condition then uniquely determines the associated price index as total expenditure per unit of processing power. A common metric of market productivity underlies price and quantity indexes any time total expenditures are decomposed into price and quantity indexes in a manner consistent with equation 1 (or its inter-temporal equivalent described below). 15 While elementary indexes of price and quantity measure the cost of their representative products, more general price and quantity indexes involve the aggregation across heterogeneous goods. The resulting indexes are sometimes based on a "countable" unit of output. Nordhaus' (1997) 'true price of light' indexes, for example, measure the cost per lumen (a unit of illumination) generated by purchases of lanterns, oil, light bulbs and electricity. The unit of observation is an objectively measurable metric of output. To estimate the cost of producing the output, all that is needed is information regarding the inputs used in its production—the production function does not have to be "known" because the output is observable thereby making possible the estimation of the production function or points on that function. Once calculated, real-output based indexes may be used as elementary indices in the construction of more general quantity indexes such as corporate production indexes, household cost-of-living indexes or social welfare indexes. While the ultimate objectives of firms (profits and/or revenues) are easily counted, the objectives of consumer and government purchases of goods and services are less easily defined. It is difficult, for example, to quantify the security "purchased" through government spending on the military, police and fire protection or household spending on locks, alarms and fire-detectors. The underlying quantity metric for welfare-theoretic cost-of-living indexes is therefore abstract. Most of these indexes use the utility concept of consumer well-being as the measure of production generated through market transactions. Specifically, self-appraised welfare is the concept that lies behind traditional cost-of-living indexes for households and societies (Diewert 1988a, Pollak 1981). Mathematical Underpinnings of Cost-of-Living Measurement Let U h ( j t h , t ) characterize the utility of household h in period t, which is a function of the household's observable consumption in that period, xh'\ Utility can also be a function of non-market determinants of welfare, including externalities such as one's neighbours' consumption of goods and services, but it is necessary to assume that these are constant between periods.5 Households are assumed to maximize utility subject to the constraint that the sum of each consumer's expenditures, T,npn-x^'\ is less than or equal to his/her income, y h , t . Utility is assumed non-satiating and prices strictly positive. The level of household utility achieved through market transactions can be used as a metric of production. A quantity index based on this metric would be the utility achieved at the observed consumption, xh'1: Q' = U h (?•'). (2) Imposing the adding up restriction determines the corresponding price index: 5 Economists focus on market transactions because these are more readily observable and their price in common units of currency make it possible to base inferences about their relative value on the economic theory of household demand. In sectors where the provision of goods and services are not priced, such as the public provision of roads, parks and national defense, economists use indirect methods to infer the money value that households place on them. Similar methods could be used to estimate the value placed on externalities and other non-market goods. Unless the statistical analyst has data on the resulting set of implicit prices, the quantity of these non-priced goods, services and externalities must be constant for economic measurement to be sensible. 16 P^CInPnWQ' = ( ^ P n ' - ^ / U 1 1 ( X H ) . (3) This is simply the cost per unit of utility. Equations 2 and 3 are unilateral indexes of quantity and price. The unilateral index in equation 2 could be constructed i f the utility function, U h (xh'1), was known by the analyst or the level of utility was directly observable. However, neither utility functions nor utility levels are observable to mortal economists. Because the economic notion of utility is neither observable nor unique, these indexes are of little practical value. Fortunately, bilateral indexes are more readily interpreted and may be inferred from market transactions under the assumptions above. A bilateral cost-of-living index captures changes in the cost of achieving a given level of well-being between two periods. The associated bilateral quantity index gauges the relative change in utility—this ratio is less sensitive to valid transformations of household utility functions. For mathematical purposes, bilateral cost-of-living indexes are typically defined in relation to the household's cost function: C(u,p) = mm x {I n P n -x n :UV)>u}. (4) This function indicates the minimum amount of expenditure necessary to achieve a given level of utility, u, when prices are described by the vector p. A bilateral cost-of-living index is said to be exact for the household with utility function Uh(jch) i f it satisfies the following (Diewert 1981): P ( P y , x ) = C(U h (x),p])/C(Uh (x),p°). ( 5 ) The index on the right hand side of equation 5 is the Koniis true cost-of-living index, which measures the relative cost of achieving the standard-of-living defined by the utility level U h(x) at prices p° and p 1. Although the standard-of-living could be defined according to any utility level, focussing on levels attainable in one of the periods of comparison gives these indexes substantive meaning in welfare analysis. Defined in terms of the utility achieved in the base period, the Koniis price index becomes the Laspeyres-Koniis price index: PL-K(P°,P 1 / ' 0 ) = C(U h (*b V ) / C ( U h (*H'°)>P°) (6) This price index indicates whether a household would require relatively more or less income at current prices than it had in the base period in order to be exactly as well off as it was in the base period. The quantity metric behind the Laspeyres-Koniis index is the level of utility achieved in the base period. The Paasche-Koniis index is defined similarly for the level of utility achieved in the current period. PP.K(P°,P1/>1) = c(uh (^Vyccu 1 1 (* h 'V) (7) The indexes of quantity associated with bilateral price indexes are identified by the inter-temporal equivalent of the adding up condition, the "factor reversal" test. This requires that the product of the price and quantity indexes equals the ratio of expenditures in the two periods: Q(p ,p ,X ) • F( P ,p , X ) - L n P n Xn I L n p n Xn . (8) 17 The bilateral cost-of-living and standard-of-living indexes are intrinsically tied to a common metric by way of this factor reversal condition just as the unilateral indexes were bound by the adding up condition (equation 1). In the case of the Laspeyres-Koniis index, the quantity index is defined as follows: Q L - K C P V ^ V 1 ) • P L - K ( p ° , p 1 ^ h ' V h ' 1 ) = E n p N V ' V l n p N V ' ° , Q L - K ( P V 'V'') = [ E n P n V ' ' / I „ P „ ° X N M ] • [ C ( T J H (xh'°),P°)/C(Uh (/'V)] = [ C ( U h (^•,),p1) / C ( U h (xh'°),P0)] • [ C ( U h (xh'0),p°)/C(Uh (^V)] = C(U 1 X^ h ' , ) ,p 1 )/C(U h (x h ' 0 ) ,p 1 ) . (9) This uses expenditure functions to gauge the household's utility relative to that achieved in the base period. The Laspeyres-Koniis implicit quantity index reports whether a household would require relatively more or less income to attain its current level of utility at current prices than it would require to attain its base period utility at current prices. Noting that cost functions are increasing in utility, this index will be greater than 1 i f utility is higher in the current period and less than 1 i f utility is lower in the current period. The Laspeyres-Koniis price and quantity indexes differ in that the price index measures the relative cost of achieving a constant level of utility at different prices while the quantity index measures the cost of achieving different levels of utility at constant prices. Both indexes, however, are based on a constant metric of productivity: utility achieved in the base period. Household theory of bounds Direct measurement of an exact cost-of-living index, even a bilateral one, is impractical because consumers' utility functions cannot be observed. Implementing the economic approach to cost-of-living measurement, like so many areas of microeconomics, relies on methods of inferring consumer utility from that which is observable: the prices and quantities of market transactions. The route taken to make these inferences in cost-of-living measurement is by way of the cost function because it is more directly related to observable information than is the level of utility. Specifically, the properties of households' cost functions make it possible to approximate the Koniis cost-of-living index with fixed-basket Laspeyres and Paasche indexes. Fixed-basket price indexes are ratios of the cost of a given basket of goods at prices collected from two different periods. Very basic fixed-basket indexes may involve only a few items believed essential to well-being. For example, the earliest known fixed-basket index, which was used to compare the costs-of-living for Oxford students in 1460 and 1707, considered the cost of 5 quarters of wheat, 4 hogsheads of beer and 6 yards of cloth (Diewert 1988b). Laspeyres and Paasche price indexes are simply fixed-basket indexes for which the basket of goods is determined by the observed consumption bundle in the base period or current period, xh'° or xh'{ respectively. The formulae for these indexes are as follows: n /_0 1 h,0 h,K _ v _ 1 h,0 / v _ 0 h,0 _ v h,0 , 1 / 0\ / i n \ P L ( p ,p ,X X ) = I n P n Xn I I n P n X N = E n S„ (p„ / p n ) (10) P P ( p ,p X X ) - L n P n xn I Lnpn xn - Ln sn ( p n / p n ), (11) 18 where s^'0 and snh'] are the shares of total household expenditures on the n t h good in the base period and current period respectively: s^'1 = pnW'1 / Zip^x^'1. The Laspeyres and Paasche index formulae originated at a time when Koniis' concept of cost-of-living had not yet been rigorously defined. Nevertheless, they were commonly accepted in part because they bore an intuitive relationship to the concept of consumer utility. As seen in equations 10 and 11, the Laspeyres and Paasche fixed-basket price indexes weigh the price change of each good by the share of household expenditures on that good—in the base period or current period, respectively. Under the assumptions of the traditional economic model, consumers "reveal" certain components of the cost-of-living to be more important than others by spending more income on these goods relative to others. This notion of revealed preferences often underlies the belief that a cost-of-living index should be most responsive to changes in the price of goods on which consumers spend the greatest share of income. In part because the Laspeyres and Paasche price indexes capture this form of revealed preference, they can be shown to be related to the true utility-based cost-of-living indexes in certain circumstances. This was proven by Koniis when he defined cost-of-living measurement within the duality approach to consumer theory. Koniis proved that the Laspeyres and Paasche fixed-basket indexes are bounds on his utility-theoretic notion of the true costs-of-living. His proof is based on two basic properties of household cost functions. First, when a household maximizes (non-satiating) utility subject to the constraint that its expenditures may not exceed its income, what it spends on consumption in any period is the minimum necessary expenditure needed to achieve the utility obtained in that period. If a household could knowingly achieve its maximum utility level by spending less, it would adjust consumption accordingly and spend the surplus to achieve an even higher utility level. Doing so contradicts the assumption that the household was utility maximizing in the first place. Thus, the utility maximizing household's cost function evaluated at current consumption will always equal current expenditures. It is in this manner that the level of the cost function is directly related to observable information: C(UV'0),p°) = E n P n V'° , (12) C ( U V V ) = SnPnV'1. (13) These equations show that the denominator of the Laspeyres price index is equal to the denominator of the Laspeyres-Koniis formula (comparing equations 10 and 12) and the numerator of the Paasche price index is equal to the numerator of the Paasche-Koniis formula (comparing equations 11 and 13). The second important property of cost functions is that, when prices change, a household will generally demand more of goods that fall in relative price and less of others. This will affect the optimum consumption bundle from the perspective of minimizing the cost of achieving a given level of utility. By responding to price changes, households could hypothetically maintain a constant utility level at lower cost than by holding consumption constant. The numerator of the Laspeyres index holds consumption constant at base period level—thereby holding the hypothetical current utility equal to base period utility—while prices change to current period levels. Because the numerator of the Laspeyres index does not account for rational response to price changes, it must be at least as great as the household's cost function evaluated at the base period utility and current prices. Thus, it must equal or exceed the Koniis numerator, 19 (14) An analogous condition holds for the denominator of the Paasche indexes: S ^ V ' 1 > C(Uh ( x M ) , P ° ) . (15) Equations 12 with 14, and 13 with 15, can be combined to show that the Laspeyres and Paasche indexes bound their Koniis counterparts from above and below, respectively. These bounds are, in fact, efficient bounds of the Koniis indexes. Take the Laspeyres indexes, for example. The base period consumption vector, x ' , serves a critical role in the Laspeyres index: it provides an approximation of what the cost minimizing household would consume in order to maintain a constant level of welfare while facing the new price vector, p 1 . The household will most likely do better by responding to changes in prices, but the optimality of xh'° at p° makes Z„pnXnh'° an efficient estimator of the upper bound on C(U h(x '0),p'). Without knowing more about the household's substitution patterns, it is not possible to find a more efficient upper bound because there are circumstances under which SnPn^n 1 1 '^ ^ U ^ x 1 1 , 0 ) ^ 1 ) . Since the denominator of the Laspeyres index is an exact estimator of the denominator of the Laspeyres-Koniis index, the Laspeyres index is an efficient estimator of the upper bound on the Laspeyres-Koniis true cost-of-living index. An analogous argument holds for the Paasche index that, by way of its inflated denominator, efficiently bounds the Paasche-Koniis index from below. For small changes in price, the Laspeyres and Paasche indexes may lie very close to the Koniis true cost-of-living index, yielding a reasonably close range of estimates for the true cost-of-living. The geometric mean of the Laspeyres and Paasche indexes, the Fisher ideal index, may therefore be an excellent estimator of the true cost-of-living (Diewert 1981). This bounds result is extremely useful because it allows us to approximate the cost-of-living for any utility function with only price and quantity information. It also gives an interpretation to the quantity indexes that are dual to the Laspeyres and Paasche price indexes. The formulae for these indexes are as follows: QL(P°,P V ' V ) = £„P„V / 2nPnV'° = Sn SnK° (xn7xn°) (16) Q P ( P y ,X h-V J) = EnPn1Xnh'1 / ZnPnV'0 = I n Sn^ (XnVxn°). (17) These are expenditure-share weighted averages of quantity changes. Quantity changes for goods on which consumers spend a greater share of income receive greater weight than other goods. Within the traditional economic theory, these indexes gauge the total change in output purchased by market transitions. As with the price indexes, the underlying metric for productivity measurement is related to the households utility level (measured in terms of base and current period welfare for the Laspeyres and Paasche indexes respectively). Mathematically, in accordance with the factor reversal restriction in equation 7, the Paasche quantity index is dual to Laspeyres price index and vise versa. Social Measurement Aggregating market transactions made by numerous households into scalar price and quantity indexes is far more difficult than the single household index problem. In the single household case, a consumption bundle described the household's purchase of the N goods and services. A social allocation is characterized by an N x H matrix, X , that specifies the 20 bundle of consumption for each of the H households in society: X={xl,...jcu}. Households' preferences over the distribution of resources will conflict. This gives rise to the need for a social welfare function, W(X), that ranks all possible allocations of goods and services. This ranking, which may range from egalitarian and paternalistic to plutocratic and consumerist, forms the basis for the underlying metric of output in social cost-of-living and productivity measurement. A social cost-of-living index may be defined with the social cost function, C(W(X),p), which specifies the minimum amount of aggregate income necessary to achieve a specified level of social welfare at given prices, p. A Laspeyres-Koniis social cost-of-living index is exact for the social welfare function, W(X), i f it satisfies the following: PL.K(PV, X 0 ,X ' ) = C(W(XV)/C(W(X°),p°) ( 1 8 ) Just as with household cost-of-living measurement, exact social cost-of-living indexes may depend on information that cannot be "known" by analysts: household utilities. It is more difficult, however, to use the "bounds approach" to implementing social cost-of-living measurement. Unless goods and services are constantly distributed in an optimal manner with respect to the chosen social welfare function, one cannot implement a revealed preference approach to estimating the social cost-of-living. To facilitate the use of the bounds approach to the social measurement problem, attention is typically focussed on Bergson-Samuelson social welfare functions. Appealing to the notion that consumers generally know what is best for them, these make the level of social welfare a function of individual households' appraisals of their own well-being: W(X) = G f U 1 ^ 1 ) , ... ,U H(x H)]. Because Bergson-Samuelson social welfare functions involve no paternalistic judgments regarding household's consumption choices, they guarantee that resource allocations within households are always efficient with respect to the specified social objective. Of course, this relies on the assumption that households behave as described in the preceding section. Efficiency within households does not guarantee the optimality of resource allocations across households, however. To deal with inter-household conflict over income allocations, attention is limited to the set of Pareto-inclusive Bergson-Samuelson social welfare functions. A social welfare function is considered Pareto-inclusive i f for every pair of allocations, A and B such that at least one household prefers A to B and all other households are indifferent between them, the function assigns a higher value to A than B. While it is typically assumed that social welfare functions ought to satisfy the Pareto criterion, this may have as much to do with mathematical tractability than with intuitive appeal. The Pareto criterion limits the set of social welfare functions to functions of positive, linear combinations of household utilities: W(X) = G[ A/xUV) + \2xU2(x2) + ...+ XuxUH(xH)], where Xh > 0 for all h. Yet, even for this limited set of social welfare functions, there are further restrictions to the use of the bounds approach. The practical obstacle to revealed preference social measurement is data. Data are typically limited to observations of aggregate consumption only, which makes it impossible to estimate individual household cost-of-living levels upon which to base social cost-of-living measures. If household cost-of-living indexes could be estimated separately, the analyst could gauge social welfare by summing these according to the weights in the chosen 21 social welfare function. Because this can seldom be done, social measurement must infer the levels of social costs and welfare by other means. "Exact" cost-of-living indexes for Pareto-inclusive social welfare functions can be approximated with aggregated data under two scenarios. Pollak (1981) calls these the "independent society" and the "maximizing society." The "independent society" is not independent so much as its aggregate demands for goods and services are independent of the distribution of incomes. If all households have demand functions with identical, linear income expansion paths, the following Plutocratic index is exact for any Pareto-inclusive social welfare function defined over aggregate consumption patterns (Pollack 1981, Diewert 1988a): PpptX^p 0 ,? ' ) = [ E C V ° , p1) ] / [ Z C V ° , P°) ] (19) This result follows from the fact that, i f aggregate demands are independent of income distributions, all Pareto-inclusive social welfare possibility frontiers will have the same curvature properties in aggregate commodity space. The Plutocratic index for an independent society is bounded by aggregate Laspeyres and Paasche indexes. This bounds result can easily be proven because, for the "independence" assumption to hold over all possible income levels, households must have identical homothetic preferences. Thus, the social measurement problem reduces to a representative household measurement problem. As Pollak notes, ". . . the empirical validity of the independent society assumption is dubious" (1981 p.326.) The second scenario under which social cost-of-living can be inferred from aggregate price and quantity data is less restrictive of household utility functions, but more restrictive of the social welfare function. This scenario is what Pollak described as the "maximizing society." It is based on the assumption that society constantly distributes income among its member households in a way that maximizes its social objectives. If this occurs, and the underlying preferences are the appropriate base for social productivity measurement, then the aggregate Laspeyres index bounds the true cost-of-living index from above. The proof of this is analogous to that of the single household because the optimizing society with a Pareto-inclusive objective function acts as an extended household in its allocation of incomes and, thereby, goods and services. This is certainly a contentious model of social welfare because it implies that the relative importance of a household to social welfare is revealed by the household's wealth. While the optimizing society assumption suffers from many criticisms (including Panglosian optimism6), interpreting aggregate indexes of price and quantity within the context of this scenario may be the best available option for social productivity measurement with limited data. It provides at least some theoretical grounds upon which to estimate social cost-of-living indexes with aggregate data on market transactions. By way of this theory, aggregate Laspeyres and Paasche price and quantity indexes can be interpreted as bounds on "true" indexes that use a plutocratic measure of social welfare as their underlying metric for productivity measurement. 6 L i k e household utility maximization, the assumption of a welfare maximiz ing society is not testable because the optimality of any outcome a tautology. 22 3.Pharmaceutical Price Indexes: Black Boxes, Pandora's Boxes When Pandora received her infamous gift—a box containing all of the evils that could trouble humanity—she was warned not to look inside for fear of its contents. As the gods expected, curiosity got the best of Pandora. Fortunately for humanity, among the contents of the box was hope. Introduction The preceding chapter on productivity measurement lay the theoretical grounds upon which traditional economic indexes are interpreted as measures of productivity. Economic indexes of price and quantity are not merely averages of individual price and quantity changes. They measure the change in the cost of some metric of output purchased through market transactions. The underlying productivity metric is typically a function of self-appraised consumer welfare. Since precise measurement of utility-based productivity metrics is impossible, Laspeyres and Paasche indexes are used as estimators. They provide upper and lower bounds on "true" cost-of-living indexes for utility-maximizing households. For multiple household measurement, aggregate Laspeyres and Paasche indexes bound "true" social price and quantity indexes for the special case where social welfare equals an income-weighted average of households' utility. In this chapter, the economic approach to price and quantity measurement is set specifically within the context of the health and pharmaceutical sectors. The theoretical framework takes on special importance in health care because there are divergent views about how to interpret health care market transactions. One view is that health care is not unlike other goods and services purchased by households. Supporters of this view cite consumer sovereignty, downward sloping health care demand curves, supplier competition and widespread acceptance of economic theory as evidence that market prices are useful in measuring the social value of health care transactions.7 Critics of this view cite the adverse circumstances that attend health care consumption, nonstandard financial incentives, uncertainty and imperfections in principal-agent relationships as reasons to consider health care purchases differently than purchases made in other sectors.8 Berndt and others (1998) provide a brief review of several issues relevant to this controversy, placing them within the context of productivity measurement in the health care sector. What follows is a discussion of how pharmaceutical consumption fits into the economic theory of measurement and how agency relationships and financial incentives may impact on measurement accuracy. 7 Examples of these arguments are often found in the literature on the welfare burden of health insurance—see Feldstein 1973; Pauly 1974; Newhouse 1992; Feldman and Dowd 1993; and Peele 1993. 8 These detractors support alternative methods of evaluating the use and cost of health care services, methods akin to Sen's (1982) call for the use of outside information and careful introspection when evaluating social welfare—see Culyer and Evans 1996; Rice 1998, Reinhardt 1998. 23 Heath, Health Care and Household Utility To understand pharmaceutical price and productivity measurement, it is necessary to see the relationship between health, health care and utility. By making a few assumptions about how to model health and health care in a household's utility function, it is possible to specify what utility-based health care price and productivity indexes might look like. The model used here depicts household utility as a function of non-health care related consumption, health care, and health status: U h ( x h , H C h , HS h (HC h )) . 9 Health status takes on a special role in this model. Broad concepts of health status are often defined to include notions of physiological, psychological and social well-being. Health status, however, should not be confused with utility itself. Evans (1984) makes this distinction clear in his appraisal that were it not for the analytical judgement that health is different from welfare, health economics would encompass all branches of economics and all sectors of the economy. To avoid this, a narrow concept of "health status" is used here. This measure of health will have an upper bound representing the absence of pathology and a lower bound representing a necessity level—without which there is not much sense in speaking of utility or well-being. As will become clear, the concept of output under consideration here is related to the production of health. Health care is a health production input that also has a direct impact on household utility. This differs from, say, the purchase of electricity, an intermediate good that is unlikely to have first-order effects on household utility. Risks, time-costs, stigma and even convenience or amenity values relate health care and utility. Yet, though lowering time-costs, inconvenience or discomfort improves the relative value of health care, no amount of convenience or amenity value wil l render bona fide health care into a consumption-good. Only drug addicts and Munchausen's sufferers would knowingly choose to consume health care in absence of some expected health improvement.10 Furthermore, it is preferable for health to be maintained or improved through the consumption of less, not more, health care, ceteris paribus. It is therefore assumed that first order effects of health care, in and of itself, on utility are always negative, U H C < 0 . Within the context of this model, economic efficiency of patients' health care purchases occurs at the point where the expected marginal rate of substitution between health care consumption and non-health consumption is greater than or equal to the ratio of health to non-health prices: (UHC + H S H C U H S ) / U X > PHC/PX- (1) This inequality will not always be binding due to the nature of health care's impact on health. A specific health care input, such as a drug, may generate significant utility through improved health for a patient. It may do so at a price that is far less than the willingness to pay for the utility gain. However, consumers in this sector cannot simply consume more and 9 This formalization for the household utility function follows Evans (1984). Evans does not develop the mathematical model in particular detail, rather he discusses at length the various ways in which observable choices in the medical marketplace may not resemble "efficient" choices from a constrained utility maximization perspective. 1 0 Munchausen's syndrome is a psychological disorder whereby those effected actually derive pleasure from health care and will feign illness to receive it. As Evans (1984) notes, "anyone who seeks care when he is not sick, is sick." (p.5) 24 more of the health care input until its marginal benefit is equal to its marginal cost. More is not necessarily better in health care. For example, extra consumption of many drugs can render them deadly. Under the assumption that health care consumption is always efficient by the economic definition characterized in equation 1, a "true" Laspeyres-Koniis pharmaceutical cost-of-living index exists for a utility-maximizing household. This index would be no different from any cost-of-living index that uses households' own appraisals of their wellbeing as the metric of productivity.11 The health sector cost-of-living index can be written as follows: PL-K (PHC° , PHC ' ) = C(Uh'°, PHC ' ) / C(Uh'°, PHc°), (2) where prices (PHC°> P H C 1 ) correspond to the prices of health care goods and services only. A sector-specific index for pharmaceuticals is defined exactly as in equation 2 by substituting pharmaceutical price vectors (PRX°, PRX 1 ) for the health care price, vectors (PHC°, P H C 1 ) - The condition for economic efficiency of drug consumption is the same as that for health inputs more generally. It is worth noting, however, that efficiency requires that the most cost-effective treatments be used in all cases. Drugs would be employed only when more cost-effective than non-drug treatments and only the most cost-effective drug would be used in any given circumstance. If drug utilization is always economically efficient, and the chosen social welfare function is a plutocratic one, price and productivity measurement in the pharmaceutical sector should be no different from those other sectors of the economy. The "true" pharmaceutical cost-of-living index, as defined within the context of this model, can be approximated with fixed-basket Paasche and Laspeyres indexes. The approximation may be very good (in fact, exact) i f health care inputs are demanded in fixed proportions across the population (Diewert 1981). This would occur i f society's aggregate health production function were a linear function due to specificity of individual medical "needs." Pharmaceutical Decision-making in Practice Among health care related transactions, drug purchases appear to most closely resemble purchases of ordinary goods and services. Perhaps this is why much economic analysis in this sector, including price and productivity measurement, is based on economic theories applied elsewhere. Despite appearances, however, drug purchases are not like those of other goods and services. Nonstandard financial incentives, incomplete information and imperfect agency relationships in all likelihood distort from social efficiency the allocation of resources in this sector. 1 1 This is true if health status does not fall too low. Berndt and others (1998) note that some illnesses make it impossible to apply the Laspeyres notion of cost-of-living because no amount of wealth can compensate for the illnesses effects on well-being. This difficulty cannot be adequately addressed within the economic model underlying price and productivity measurement. One must therefore assume that changes in health status are small or that shocks to health status will balance out when aggregating across households. 25 Insurance and Financial Incentives One of the distinguishing features of the pharmaceutical sector is that a majority of patients is insured for prescription drug purchases. Approximately 75 percent of Canadians are insured to some extent for the cost of prescription drugs. Public drug insurance plans alone offer (near) full coverage for about a quarter of the population—mainly seniors, social assistance recipients and the chronically ill—which account for half of prescription drug expenditures in Canada. Although household cost-of-living indexes typically focus on out-of-pocket expenses, because these affect the budgets of individual households most directly, gauging social costs and sector productivity requires that analysts consider the full price of drugs. The division between total costs and the costs considered by patients may give rise to serious measurement problems because prices will not necessarily relate to the relative value of pharmaceuticals in terms of social productivity. To see this, note that when a patient is fully insured, conventional economic theory predicts that she/he will choose to consume a drugs until the point where the marginal utility benefit is zero: (URX + H S ^ U H S ) = 0 (3) Prices do not enter this equation because they are not a consideration in the fully-insured patient's decision. More generally, out-of-pocket cost-of-living indexes for insured patients will not equal a social cost-of-living index for the same purchases. The one exception is the rare case where patient payments are a constant percentage of the total cost, say x percent. In these circumstances, changes in out-of-pocket costs captured by the bilateral index will approximate the rate of change in social costs, PL-K(X-PRX°, X-PR, 1) = x-C(UH '°, PRX1) / x-C(UH '°, PR*0) = PL-K(PRX°,PRX1). (4) This follows from the linear homogeneity of the cost function with respect to price. In reality, patients are not the principal decision-makers in the pharmaceutical sector, so their insurance coverage may bear only a weak relationship to the type and quantity of drugs consumed. Agency and Prescription Drug Costs Since patients cannot freely make decisions in this sector, there is some hope that the behaviour of the principal decision-making agents is consistent with societal objectives in this sector. Ignoring for the moment the societal paternalism revealed by mandating a decision-making hierarchy, the notion that agents weigh costs and benefits on behalf of patients may salvage the economic approach to measurement (and market theory more generally) in the pharmaceutical sector. This clearly has intuitive appeal because in the face of a totally price-inelastic demand, one would expect firms to set prices indeterminately high. Something, or someone, must constrain prices in this market. Because physicians may make the decision concerning which drugs patients should consume, they play the most important role in determining whether drug use will be cost-26 Table 1: Results of Surveys of Physicians' Knowledge of Drug Prices Study & Location Doctors #of Replies Accuracy Measure Accuracy Rate Low Cost Drugs High Cost Drugs Silock et al. (1997) U K GPs 627 25% 33% + = Ryanet al. (1996) U K GPs 213 25% 32% + -Kolassa(1995) US GPs 100 25% 19% + -Glickman et al. (1994) US GPs & other 132 25% 20% + -Walzaket al. (1994) US Int. & GPs 137 various 33% Bailey et al. (1993) U K Anesthesia 50 50% 47% + -Miller and Blum (1993) US Family 92 S0.50/unit 37% + -Yule et al. (1991) U K GPs 80 25% 26% + -Hershey et al. (1986) US Residents 48 30% 25% Beringer et al. (1984) US Neurologists 24 15% 10% = = Babington etal. (1983) US Residents 51 20% 5% Long et al. (1983) US Residents 59 25% 42% + -Nagurney et al. (1979) US Various 72 20% 12% + -Lowy et al. (1972) US GPs & other 220 20% 33% + Systematically overestimate prices. - Systematically underestimate prices. = No systematic bias detected. effective or not. Achieving cost-effectiveness will depend on doctors' knowledge of and response to the prices and technical merits of competing products. Doctors and Drug Prices Most physicians have financial incentives that are similar to fully insured patients. Because physicians do not pay for the drugs they prescribe, they have little incentive to consider costs, particularly when their patients are insured. Supposing, however, that doctors do believe that prices should be considered,12 their prescribing decisions will be influenced by physicians' perceptions of price, not actual price. Unlike consumers present at the point of transaction, doctors order drugs on behalf of informed only by their estimates of drug prices. Given that they prescribe numerous times per day—even per hour—one might conjecture that physicians would be reasonably well informed about drug prices. Unfortunately, evidence regarding physicians' knowledge of drug prices suggests that their perceptions of these costs are inaccurate. Table 1 lists findings from surveys of doctors' knowledge of drug costs conducted in various locations in the US and U K over a period of two decades.13 Doctors of varying specialties were asked to estimate the cost of a typical supply (e.g., 30 days) of drugs that were most commonly prescribed in their region. Most of these studies defined an accurate estimate as one within 25 percent of the actual price. Less than a third of all price estimates given by doctors in these studies met with this rather loose definition of "accurate." 1 2 Dennig and Haaijer-Ruskamp (1995) report that several studies have consistently found that approximately 25 percent of doctors do not think that price should be considered. 1 3 The literature on this topic is enormous. The 15 studies summarized here were found by searching Medline and HealthStar databases and the references cited within the literature. 27 For purposes of measuring changes in sector-specific price and productivity, systematic biases in price estimates may not affect the accuracy of economic indexes. If doctors' perceptions of drug prices are biased in a uniform and systematic manner, there may be "excessive" prescribing (if they underestimate all drug prices) or "under-prescribing" (if they overestimate all drug prices) but there wil l not necessarily be a measurement problem. A uniform bias in price estimates would have an impact similar to constant-rate coinsurance. Since cost functions are linearly homogeneous in prices, constant rate of price estimate bias will have no effect on bilateral indexes. Unfortunately, however, the biases in physicians' perceptions of drug prices are not uniform. Nine of the ten studies that looked for bias in price estimates found distinct biases in the estimates of high- and low- cost drugs. This is reported in the right-most columns of Table 1. Nine studies found that physicians systematically overestimated the price of low-cost drugs, and eight of the studies found that physicians underestimated the price of high-cost drugs in a systematic manner. This will clearly cause doctors to over-prescribe high cost drugs relative to low cost drugs. This kind of inefficiency, in drug selection would make economic indexes invalid measures of social costs and, therefore, sector productivity. When consumption patterns shifted from low- to high-cost medicines, an economic index (such as a chained Laspeyres index) would overstate improvements in productivity because doctors' assessments of relative value would be based on underestimated relative price differentials. Prescribing Appropriateness Therapeutic properties are the second component of doctors' knowledge relating to the efficiency of drug use. This knowledge is seldom tested directly. Instead, evaluation is done retrospectively by evaluating actual prescribing habits, how they vary between doctors and the extent to which they may be deemed medically "appropriate." Prescribing habits are known to vary dramatically across physicians. In many settings, a minority of doctors (10 percent) accounts for a majority of prescribing (50+ percent). The high-volume prescribes tend to be older doctors, in general or family practice, who see more patients per day, prescribe more drugs per patient and bill for more services per patient (Lexchin 1998; Tamblyn 1997; Davidson et al. 1994; Davis et al. 1994; Haayer 1982; Hemminki 1975). Doctors are more likely to prescribe to patients of high as opposed to low socioeconomic status (Lexchin 1998; Scott, Shiell and King 1996). Furthermore, doctors are ten times more likely to prescribe to patients who they perceive to be expecting a prescription than to patients they don't perceive to have such expectations (Cockbura and Pit 1997, Britten and Ukoumunne 1997). It would appear that doctors' "preferences" for prescribing vary based on many factors unrelated to the medical needs of the patient or the technical merits of the drugs in question. These non-scientific motivations for variation in prescribing habits may explain some of the findings regarding the medical appropriateness of prescribing. While some definitions of "appropriateness" are consistent with the economic notion of efficiency described in equation 18 (e.g., Buetow et al. 1997 p. 264), others define appropriateness strictly in relation to clinical risks and benefits. The latter definitions require only that the expected health impact of a medicine or procedure be non-negative, H S H C ^ O . Therefore, the clinical appropriateness of prescribing decisions is a necessary but not sufficient condition for socially efficiency. 28 Estimates of the rate of inappropriate prescribing range from 5 to 30 percent (Tamblyn 1997; Buetow et al. 1996). Most appropriateness studies have focussed on the prescribing of drugs for elderly people or the prescribing of specific classes of drugs. For the most commonly prescribed drug classes (antibiotics, non-sedating anti-inflammatory drugs and psychotropic drugs), inappropriate prescribing appears to occur frequently, with estimates of the rate of inappropriate prescribing ranging from 5 to 75 percent (Tamblyn and Perrault 1998). Prescribing appears to be most inappropriate for the elderly—who consume a disproportionate share of medicines, yet are most sensitive to adverse outcomes. As many as half of Canada's elderly receive at least one prescription per year for which clinical benefits do not outweigh clinical risks, irrespective of the cost of the drug (Tamblyn 1997). In a study done in 1993, it was found that 29 percent of Quebec seniors had been prescribed at least one inappropriate drug combination, 36 percent had been prescribed psychotropic drugs (benzodiazepines) for a longer period than is medically appropriate and 15 percent received long-acting psychotropic drugs that are contraindicated for seniors (Tamblyn and Perrault 1998p.223). The effect of inappropriate prescribing on health is evidenced by the data on drug related illnesses. As many as 23 percent of hospital admissions and 1 in 1,000 deaths in Canada are attributable to illness related to ethical drug use (Tamblyn 1997). Clearly, i f doctors are prescribing inappropriately and this is harming patients, then the health-related productivity of the associated purchases is negative. Of course, all drugs involve some risks, and studies of the appropriateness of prescribing consider these when determining what the standards are for accepted tradeoffs between risks and benefits. Physicians are responsible for gathering information and assessing the risks on behalf of their patients. Though this is a costly endeavour, there are widely disseminated professional guidelines to help doctors choose the appropriate medication for patients presenting with many of the most common indications. When prescribing decisions are motivated by personal beliefs contrary to scientific evidence, perhaps due to the influence of drug manufacturers, this represents an imperfection in the principal-agent relationship between patients and physicians. Measures of social welfare should account for such failures when they occur. Health Outcomes as Real Productivity It is difficult to model exactly how doctors choose among alternative medicines when prescribing. One of the reasons for this is the fact that doctors appear to prescribe by habituation, and their habits are determined in large part by emotional self-assessments rather than rational scientific judgements (Temin 1980, ch.5). Moreover, even i f doctors' decision-making processes could be modeled, it may not be easy to relate that decision-making process to the household utility framework upon which economic measurement is based. Imperfections in the physician-patient relationship will not necessarily take a form that can be easily "corrected" for the purposes of productivity measurement. Fortunately, an alternative approach to price and productivity measurement may be used to illustrate the net effects of agency and financial incentives without the need of a formalized agency model. As mentioned above, health outcomes are ostensibly the principal objective of health care consumption. Health outcomes form the basis of most program evaluations in health 29 care that are conducted at a project by project (or product by product) level. It should therefore be possible to use approximate measures of health production to base aggregate price and quantity indexes in the pharmaceutical sector. There is a scientific evidence base upon which to base inferences regarding the technological relationship between drugs and health outcomes or surrogate measures of health outcomes. That is, outside knowledge about pharmacology and therapeutics may be brought to bear on pharmaceutical sector measurement just as Nordhaus (1997) brought knowledge about the production of light to bear on his cost of illumination indexes. To construct quantity metrics based on health outcomes for the purpose of sector price and productivity measurement is not an easy task. It would require that one follow all patients who do and do not receive drugs, document their health outcomes and calculate the relative health improvements created by the drug consumption. It would also require some consensus on what the relevant health outcome measure is, which may be difficult when aggregating across diseases with vastly different effects on the quality of life. Ideally, the outcomes measure in this sector would be a standardized metric such as the Quality Adjusted Life Year (QALY). Short of constructing league tables for the number of Q A L Y s generated by the consumption of all pharmaceuticals, simple and immediate steps towards health outcomes based measures of price and productivity can be made. For some classes of drugs used to treat the same condition, it is relatively easy to rank the relative efficacy of competitors. It is often the case that competitors within narrowly defined therapeutic classes have nearly identical profiles of health risks and benefit. Differences in therapeutic benefits and unwanted side-effects are sometimes inconsistent across clinical studies, indicating that they may be statistically insignificant or the result of study methodologies. Yet, such therapeutically similar products often have dramatically different prices—partly a function of the patent lifecycle and vintage effects in the drug industry. Divergent prices for similar therapies can make the cost per unit of treatment—measured in patients, cases or days— vary across competing products. If the therapies are similar, this implies that the cost per unit of health outcome (or "efficiency unit") will differ as well. Indexes based on measures of price per unit of therapy, rather than measures of the price per pill or tablet, may be compared to traditional economic indexes based which infer relative value from market prices. If agency relationships in the pharmaceutical sector operated in a socially efficient manner, standard (utility-based) economic price and quantity indexes ought to lie reasonably close to indexes based on real health outcomes. Minor differences can be expected due to inevitable imperfections in the task of assigning units of therapy represented by each dosage form of competing drugs. If, however, non-standard financial incentives and/or the effects of agency relationships in the pharmaceutical sector significantly distort the choice of medicines away from social efficiency, it would be expected that economic indexes and health outcomes based indexes would differ significantly. The chapters that follow explore such an approach to measurement in this sector. 30 4.Agency, Financial Incentives and Pharmaceutical Price Indexes: The Case of Generic Drugs Introduction Several researchers have investigated the possibility that standard measures of inflation overstate the "true" rate of price change for pharmaceuticals as defined by the traditional economic approach to measurement.14 Sources of measurement bias include under-accounting for technological change, late introduction biases and elementary index biases caused by changing outlet types (e.g., mail-order pharmacy) (Berndt, Griliches and Rosett 1993). Each of these sources of bias has been investigated in the pharmaceutical sector as well as in other sectors of the economy. Other considerations for price and productivity measurement in the pharmaceutical sector are not as likely to arise in sectors for ordinary consumer goods (Berndt et al. 1998). Widespread insurance and mandated decision-making agency relationships are the most notable aspects of the pharmaceutical sector that deserve special attention in measurement theory. That three-quarters of Canadians' prescription drug expenditures are paid for through public and private drug insurance plans will necessarily influence pharmaceutical price and productivity measurement.15 Although household cost-of-living indexes are typically based on out-of-pocket expenses, calculating social costs and productivity measures requires that analysts consider the full price of drugs. The division between total costs and the costs considered by insured patients in the pharmaceutical sector will give rise to a form of "moral hazard" problem i f consumption goes beyond the point where marginal social costs are equal to marginal benefits. However, patients are not autonomous decision-makers in this sector. Agency relationships between patients, doctors and pharmacists are legally mandated in the process that leads to prescription drug purchases. These principal-agent relationships may help or hinder social cost and productivity measurement depending on whether or not the objectives and incentives of physicians and pharmacists are consistent with social welfare maximization and cost minimization. If drug insurance and/or decision-making agency relationships distort the relationship between prices paid for drugs and their social value, traditional economic price and productivity indexes may serve us poorly in this sector. Fortunately, the existence of a potentially measurable metric of output—health status—makes it possible to verify the 1 4 See, for example, Berndt, Griliches and Rosett 1993; Griliches and Cockburn 1994, 1996; Fisher and Griliches 1995; Suslow 1996; Berndt and Greenberg 1996; Berndt, Cockburn and Griliches 1996; Ellison, Cockburn, Griliches and Hausman 1997; and Baye, Manes and Wiggins 1997. 1 5 Figures from the Canadian Institute for Health Information breakdown prescription drug expenditures for 1996 into 43% public expenditures, 33% private insurance costs and 24% out-of-pocket expenses. 31 relative accuracy of standard economic measures of value in health care. Generic drugs offer a particularly straightforward example of such a comparison. Generic drugs are more-or-less identical substitutes for their brand-name competitors. For a vast majority of patients, the expected clinical outcomes from treatment with generic and brand-name drugs are identical. Thus, patients who switch to an available generic drug experience an implicit price reduction. American studies have shown that price indexes designed to account for the price differential between generic and brand-name drugs report lower inflation rates than indexes that do not adjust for product interchangeability (Griliches and Cockburn 1994, 1996; Berndt, Cockbum and Griliches 1996). These studies have also found that adjusting for generic availability as one would account for any "new good" in a price index does not capture generics' long-term impact on average prices. The reason for this is that generic penetration into brand-name drug markets does not always occur immediately. Because generics eventually do achieve a majority of sales, their gradual diffusion into product markets raises questions about whether utilization patterns reflect socially efficient decisions. Perhaps the slow up-take of generic drugs is a socially inefficient phenomenon induced by nonstandard objectives and incentives of decision-makers in this sector. To date, this has not been investigated within the context of price and productivity measurement. This chapter brings the measurement theory concerning generic drugs and new goods to bear on prescription drug prices in Canada using a comprehensive database for all prescription drugs purchased under British Columbia's public drug insurance plan for senior citizens, Pharmacare Plan A. The focus of this analysis is on financial incentives facing patients and pharmacists—the primary decision-makers in the choice between brand-name and generic drugs. Generic Drugs Generic drugs are chemically equivalent competitors of brand-name pharmaceuticals. Approximately 40 percent of Canadians' prescriptions are dispensed with generic drugs; but, in part because the average generic is priced 30 percent below its brand-name competitor, generics account for only 15 percent of Canadian drug sales. While generic drugs are typically sold under the generic name of their active ingredients (for example, "diazepam" is the generic name for the active ingredient in the brand-name drug Valium), they should not be regarded as merely the pharmaceutical equivalent of generic consumer goods such as "no-name" peanut-butter. Like "no-name" consumer products, brand-name drug manufacturers sometimes sell generic versions of their own products, known as "pseudo-generics." (Pseudo-generics now account for approximately 20 percent of Canada's generic drug market.) Unlike "no-name" consumer goods, all generic drugs are certified by the government as chemically equivalent to the original brand-name product. Before being brought to market, they must meet Health Canada's standards for chemical identity, purity and potency, as well as manufacturing standards and labeling restrictions. In clinical terms, there is no scientific evidence of systematic differences in outcomes across countries, drug plans or hospitals with widely varying rates of generic drug use (Walker 1995). Canadian hospitals, for example, use available generic drugs almost exclusively, without jeopardizing the quality of care or the 32 confidence of patients treated therein. Moreover, the federal and provincial governments in Canada have long encouraged generic drug use in the ambulatory setting in ways that include accepting liability when generics are substituted for their brand-name competitors (Gorecki 1992 ch.3, Anis 1994). Differences between drugs that are government-certified as chemically equivalent are limited to packaging, non-active ingredients16 and possibly colour or shape. These minor differences may be important to some individuals. Some patients may react to the non-active ingredients in the brand or the generic version of a particular drug;1 7 some 1 R patients may greet a change in the colour or shape of their medicine with anxiety; and some patients may be "loyal" to certain brands. In these cases, the patient "prefers" a particular version of a product. Measuring the extent and magnitude of such preferences is important for the measurement of price and productivity in the pharmaceutical sector. Despite government certification of product interchangeability, brand-name drugs often remain in wide use after lower-cost generics become available. Several studies have shown that when generic drugs enter the U.S. market at substantial discounts relative to their brand-name competitors, large numbers of patients (in many cases a majority) continue to purchase the brands (Caves, Whinston and Hurwitz 1991; Frank and Salkever 1992, 1997; Grabowski and Vernon 1992). Furthermore, patients continue to purchase brand-name drugs even as the discrepancy between generic and brand-name prices increases over time. In response to entry by generic competition, brand-name drug manufacturers often engage in a form of "cream-skimming," whereby they increase their price and serve a market segment of "brand-loyal" customers (Caves et al. 1991, Frank and Salkever 1992 and 1997). That many consumers continue to purchase brand-name drugs at prices well above generics appears to indicate that consumers perceive the brand-name drug to be of significantly more value than the generic alternatives. To the extent that differences in prices accurately reflect consumers' perceptions of value, common inferences regarding social cost-of-living can be made from price indexes based on the traditional economic approach to measurement (Fisher and Griliches 1995). It is possible, however, that brand-loyalty is exaggerated due to doctors', patients' and/or pharmacists' lack of incentive to consider price differentials. Traditional Approaches to Gener ic Drugs Consumers who switch from a brand-name drug to a newly available generic experience a form of savings equivalent to that generated by any new good. The "new goods effect" of product entry is an implicit reduction in the cost-of-living that is roughly equivalent to the difference between the price paid for initial purchases of the new good and the cost of a utility-equivalent alternative. Consumers "reveal" new products to be of "value" relative to existing products by spending a share of their income on the newly introduced goods. 1 6 Active ingredients are the chemical ingredients with the pharmacological properties for which the product is consumed. Non-active ingredients include binding agents, casings, liquids etc. 1 7 Such therapeutic differences between a brand-name and generic drug are attributable to patient-specific responses, not product-specific deficiencies. Though this point is often overlooked, it is possible that certain patients will better tolerate a generic than its brand-name counterpart (Anonymous 1998). 1 8 Patients' concern about a change in the colour or shape of a medicine are based on the perception that the two products are not equally effective in clinical terms. As agents for patients, physicians and pharmacists might be obligated to point out that this is not, in fact, the case. 33 In standard, unadjusted indexes, new products are simply "linked" into an index in a manner that tracks changes in price over time, but not the savings implicit in the new products' launch price. Such an approach is consistent with the traditional economic approach to measurement i f quality differences between the new and existing products are reflected in the price difference between them (Diewert 1998, p.52). In the case of generic drugs, it is unlikely that, at least for those who switch to generics upon entry, the differences in price between the brand-name drug and the generic reflect quality differences. The most widely employed method for capturing the "new goods effect" of generic availability is to estimate reservation prices for generic entrants (Berndt et al. 1993; Griliches and Cockburn 1994, 1996; Fisher and Griliches 1995). Reservation prices are estimates of the implicit pre-market prices for entrants that can be used to infer the price reduction that implicitly occurs when a product first enters a market. Suppose, for example, a generic drug enters the market in period 1. Analysts observe price and quantity information for the generic in that period: x g ' , p g \ Moreover, the pre-market quantity of the generic is implicitly observed because total purchases are zero until a product enters the market: i.e., x g = 0. What is not observed is a pre-market price that would be consistent with the (implicit) observation of zero sales in the pre-market period. A reservation price is an estimate of the price at which the demand for the generic in the period preceding its entry would equal zero. In most circumstances, a reservation price can only be estimated well after the launch of a product. Using post-launch price and quantity observations, the analyst would estimate a demand function for the drug, x g( p g , p_g ), and calculate the price that would be consistent with the implicitly observed zero demand in the pre-launch period, given the prices of other goods in the pre-market period. That is, p g R is such that x g( p g R , p0~g) = 0. That demand increases to a positive quantity of sales in the introductory period reflects a response to the price reduction implicit in its new availability. The price reduction captured by this measurement technique is the fall in generic price from the estimated reservation price, p g R , to its introductory price, p g ' , weighted (in the instance of the Paasche index formulae) by the share of total expenditures spent on the generic in its initial period. Provided that the demand curve for a product is downward sloping and preferences remain unchanged (an assumption of the economic approach to measurement), the real reservation price is necessarily higher than the real introductory price. It is possible, during periods of high inflation, that the introductory price will exceed the reservation price in nominal terms, but not in real terms. Fisher and Griliches (1995) discuss reservation prices for generic drugs in some detail. They show that Paasche and Laspeyres indexes calculated with accurate reservation prices will efficiently bound the "true" social cost-of-living index under the traditional economic approach to measurement. While properly estimating reservation prices can be troublesome, generic drugs represent a "particularly simple version of the new goods problem" (Griliches and Cockburn 1994, p. 1213) because the existence of government-certified equivalent competitors makes it possible to place bounds on reasonable reservation prices. The launch price of a generic drug is the lower bound on its reasonable reservation prices. This is used implicitly when generics are "linked" into the indexes with no adjustment to account for interchangeability between the new generic drug and its brand-name counterpart. At the other extreme, the 34 reservation price of the generic drug could equal the price of the brand-name product in the year preceding generic entry because the government certifies generic and brand-name drugs to be interchangeable. This "a pill is a pi l l" reservation price provides an upper bound on reasonable reservation prices for generic drugs because generics would not likely sell at prices greater than their brand-name counterparts. Thus, without estimating the demand function for the generic drug, it is reasonable to conjecture that its reservation price lies between its introductory price and the price of the brand-name drug in the period preceding generic entry: p g R e [pg' , p b° ], where p g ' < p b°. Where one places the reservation price between the two bounds depends on where one feels the value of the generic lies. Griliches and Cockburn (1994) conjecture an estimate of the appropriate reservation price for generics based on a demand model where individual patients, or physicians, have different preferences over brand-name and generic products. Their "preferred" reservation price for a generic drug lies halfway between the two 'bounds' i f preferences are uniformly distributed between the two types of product: p g R (G&C) _ j- + p b ° ) / 2 ]. For the anti-infective drugs Cephalexin and Cephradine, Griliches and Cockburn's indexes adjusted with their preferred reservation prices fell at respective rates of 8.0 and 3.8 percent compared to standard indexes that fell at rates of 5.0 and 3.1 percent over a three-year period. Others have attained similar results by estimating reservation prices for generic drugs from post-entry price and quantity data (Ellison, Cockburn, Griliches and Hausman 1997; Feenstra 1997). Each study indicates that, within the traditional economic approach to price measurement in this sector, failure to account for the new-goods aspect of generic competition would create modest biases in pharmaceutical price indexes. The Generic Diffusion Problem After appropriately adjusting a price index for the initial "new goods" aspect of generic competition, there still remains a potential measurement problem associated with generic drugs. This potential problem is related to the drawn-out process of generics' penetration into brand-name drug markets despite dramatic differences in the prices of these chemically equivalent products. The diffusion process is not necessarily of concern in and of itself. Rather, what is of concern is whether consumption patterns are indicative of socially efficient decision-making in light of the financial incentives of involved decision-makers. Appropriate construction of price indexes, within the traditional economic approach to measurement, includes adjustment for the new-goods aspect of generic entry, but not for the gradual diffusion process. This is because the implementation of the economic theory of cost-of-living measurement requires that decision-makers maximize welfare in the face of standard financial constraints. Purchases of brand-name drugs by such decision-makers after the introductory period of a generic imply that they value the difference between brands and generics by an amount at least as much as the price differential. Gradual generic diffusion could consequently be explained by the accumulation of information about the safety and efficacy of generic versions of a particular drug product (an unmeasured aspect of product quality). At some point, enough information may be accumulated to render the generic drug just preferred to the brand-name drug at their given prices. The decision-making agent who previously chose the brand-name drug will then switch to the generic. 35 Alternatively, gradual diffusion might reflect the gradual arrival of new patients who may be more inclined to use generics than those who have used the brand-name drug in the past. Those accustomed to the brand-name drug might prefer to "stay with what they know." Under either of these scenarios, consumption patterns would reflect socially efficient choices, so no adjustments are needed for the diffusion process. On the other hand, considering that generics and brand-name drugs are government certified interchangeable, systematically deferring the purchase of a generic until later periods may be socially inefficient. The "preferences" revealed by the choice of brand-name drugs may have less to due with beliefs about therapeutic value than with the financial incentives of principal decision-making agents. Unlike the model of demand behind the traditional economic approach to measurement, consumers in this sector do not autonomously maximizing utility subject to standard financial constraints. Most patients are insured, they seek advice from doctors who do not pay for what they prescribe, and they purchase drugs from pharmacists who may earn more for dispensing certain products. As such, although brand-name drug use might be perfectly rational in light of the various decision-makers' financial incentives, it may not indicate the relative social value of the products in question. Consumption patterns may overvalue brand-name prescription drugs if decision-makers have incentive to under-respond to price differences. If this is the case, evidence suggests that the magnitude of the measurement problem caused by gradual generic diffusion into brand-name markets is very large. For example, Griliches and Cockburn (1994) adjusted for the gradual diffusion of generic versions of Cephalexin and Cephradine by treating all switches to generics, regardless of what period they occurred, with the same reservation price technique typically only used in the generics' introductory period. That is, they assigned their preferred reservation price to new purchases of generic drugs in every period: p g R ' ( t" i : ) = [ (pg1 + Pt> ( t l )) / 2 ]. With this diffusion adjustment, Griliches and Cockburn's price indexes fell at rates of 14.7 and 7.6 percent over a three-year period—compared to standard indexes that fell at rates of 5.0 and 3.1 percent, and new-goods-adjusted indexes that fell at rates of 8.0 and 3.8 percent. According to these results, the bias caused by the diffusion problem may be three or more times as great as the new goods aspect of generics. Moreover, the underlying cause of this diffusion-related bias undermines the theoretical basis for traditional economic methods for estimating reservation prices to address the new goods aspect of generic competition. Considering the effects of financial incentives, the "true" reservation price (or social reservation price) for generic drugs may lie closer to the brand's price than researchers have assumed. Generic Drugs and Price Indexes for BC Pharmacare This analysis focuses on the market segment corresponding to the public drug plan for seniors in British Columbia: Pharmacare Plan A. Pharmacare Plan A covers all community-dwelling BC residents who are 65 and over, representing about twelve percent of the B C population and accounting for one-third of all ambulatory prescription drug costs in B C . 1 9 The advantage of looking at purchases under a particular drug plan is that decision-makers' financial incentives are more easily identified than with databases covering diverse 1 9 Separate public drug insurance plans cover inpatient drug costs for all BC residents, whether in an acute-care or long-term-care facility. 36 Table 2: DIN and A H F S Numbers for 2-miligram Diazepam Tablets DIN# AHFS # Generic Name Brand Name Company Name 00013277 282408041 Diazepam tab 2mg Valium tab 2mg Hoffmann-la Roche ltd. 00013757 282408041 Diazepam tab 2mg Vivol tab 2mg Genpharm Inc. 00272434 282408041 Diazepam tab 2mg Novodipam tab 2mg Novopharm Ltd. 00272647 282408041 Diazepam tab 2mg E pam tab 2mg ICN Canada Ltd. 00405329 282408041 Diazepam tab 2mg APO Diazepam tab 2mg Apotex Inc. 00434396 282408041 Diazepam tab 2mg Diazepam tab 2mg Pro Doc Limitee populations with different forms of insurance coverage. Its primary disadvantage is that caution must be used in generalizing the results due to the possibility of age-related differences in drug use2 0 and the lack of data regarding behaviour of populations without insurance coverage of any kind. A l l beneficiaries of Pharmacare Plan A are insured for the ingredient costs of prescription medicines, net of dispensing fees. Applying a traditional economic approach to price measurement in this sector therefore relies on the assumption that patients, physicians or pharmacists weigh the costs and benefits of drugs even though expenses are paid for by the government. Alternatively, one might assume that Pharmacare administrators evaluate the relative value of products when determining what the plan will pay for. Perhaps as a result of performing such evaluations in the early 1990s, Pharmacare administrators twice altered patients' and pharmacists' incentives to consider switching to generic drugs. The resulting changes in consumption patterns and price indexes illustrate how agency and financial incentives relate to the generic drug problem in price measurement. Data Data used to construct price indexes in this study consist of annual aggregates of quantity and sales figures for each type of drug purchased under Pharmacare Plan A from 1986 to 1995, inclusive.21 A subset of the data (reported on below) is available on a quarterly basis. The data were compiled from transaction-level Pharmacare claims files using a unit-value approach to pricing across outlets—i.e., the unit value (price) for a particular product is equal to the total expenditure on that product across outlets divided by the total quantity of that product sold (Diewert 1995). This price reflects the total payment received by pharmacists, excluding dispensing fees. While limited to a subset of the total market as defined by population, the Pharmacare Plan A database is particularly rich in the sense that it provides information on a product-by-product basis for all prescription drugs used by the elderly. Compared to previous studies based on data for limited numbers of drugs, in some cases at the wholesale 2 0 While little is known about differences in the inflation rates for pharmaceuticals between elderly and non-elderly populations, one U.S. based study found no evidence of age-related differentials in aggregate measures of pharmaceutical price inflation (Berndt et al. 1997). 2 1 These data were extracted as part of a larger study at the Centre for Health Services and Policy Research, UBC. 37 Table 3: Classification of Drugs Purchased Year Number of Competing Products Purchased Number of Drug Types Purchased Percentage of Drugs Multi-sourced 1986 2614 1619 26.7% 1987 2512 1589 27.0% 1988 2498 1574 29.0% 1989 2758 1776 27.8% 1990 2895 1809 29.2% 1991 2934 1811 29.4% 1992 3037 1868 29.1% 1993 3106 1870 30.3% 1994 3077 1782 33.2% 1995 3198 1811 33.5% level only, the data used in this study consist of retail observations for 4,740 unique products identified by drug identification number (DLN). 2 2 These DIN-level data have been aggregated across patients as well as outlets; they have not, however, been aggregated across dosage forms or brands. Products identified by DfN are unique to the level of ingredient, strength, dosage form, manufacturer and brand.23 In order to make comparisons across equivalent brand-name and generic drugs, BC Pharmacare assigns a therapeutic classification code to each DIN. Their classification system is a 9-digit version of the American Hospital Formulary Service (AHFS) that is unique to a specific drug by strength and dosage form, but not by manufacturer or brand-name. An example is provided in Table 2, which lists the classification codes for a specific class of drugs: 2-miligram tablets of the psychotropic drug diazepam. This breakdown illustrates how the classification codes identify chemically equivalent competitors. As can be seen, multiple firms may sell a particular type of drug (identified by AHFS number) under a number of brand and generic names. The 4,740 competing products in the database represented only 2,484 unique types of drug by dosage form (i.e., common AHFS numbers). Many new types of drug appeared during the period of analysis, while some older drug types disappeared. Those exiting the market accounted for less than 1 percent of sales in any given year. Drugs manufactured in the same dosage form by more than one competitor are referred to as multi-source drugs. Table 3 lists the number of drug types purchased in each year and the percentage of these that were multi-source drugs. The data contained 889 AHFS-level drug classes that were 2 2 The quantity data for some products had extreme inconsistencies in 1986 through 1988. All of these observations were for injected or inhaled medicines. Experts at BC Pharmacare confirmed that, before standardizing recording methods, there have been problems with the quantity data for these products. This is because a pharmacist dispensing an inhaler consisting of twenty lOmg doses of a medicine might have recorded it as one inhaler, twenty doses or two hundred milligrams of the active ingredient. Quantities and prices for 505 products identified as injected or inhaled medicines (representing approximately 5 percent of total sales) were set to zero in 1986, 1987 and 1988. 2 3 Even in cases where a manufacture sells identical products under different brands—as would be the case with "pseudo-generics"—each product is identified with a unique DIN. 38 multi-source at some point from 1986 to 1995. These classes of drugs had a lower rate of year to year turn-over because they tend to be higher-volume drug types. Price Indexes To provide the simplest illustration of the measurement problems associated with generics in the Pharmacare database, attention is focused on indexes constructed with chained Paasche price index formulae. The quantity weights for all of the indexes reported are updated annually. Three ways of incorporating generic drugs into price indexes are explored. The first is to construct an "unadjusted" index that does not account for interchangeability between chemically equivalent products. Using data at the DLN level, new generics are linked into the unadjusted index in the same manner as new brand-name drugs: without making comparisons to any competitors' prices. This involves linking generics into the index during the first period of their availability using reservation prices equal to their launch prices—the "lower-bound" on reasonable reservation prices. Next, an adjustment is made to account for the new goods aspect of generic competition. This "new goods adjusted" index is constructed with separate price and quantity observations for all 4,740 DLN-level products, using a Paasche formula with conjectured reservation prices for entrants into established product markets. The reservation price for a new generic is set at the average price of chemically equivalent brands in the year preceding its entry. This approximates the "upper bound" on reasonable reservation prices for generic drugs, where p g R = pb°. When a generic drug enters a market previously held exclusively by a brand-name firm, its reservation price is based only on the price of the brand-name product. When a generic enters a market that was already subject to generic competition, its reservation price is set equal to the average unit price of all of its competitors. This is because late entry generics are likely to steal market share from the pioneering brand and from other generics. Because annual data are used to construct these indexes, linking generics into the index during their first period of availability gives consumption patterns an average of 6 months to adjust before measuring the implicit price effect of initial generic availability. The final treatment of generic drugs is to use a "unit value" approach to aggregating generics and brand-name drugs at the elementary index level. The unit value approach to chemically equivalent generic and brand-name drugs is to treat them as identical goods by summing their total sales and dividing this by their total quantity to arrive at a common unit value (Diewert 1995). Rather than track the prices and quantities of equivalent brand-name and generic drugs separately, for each group of chemically equivalent drugs, this method tracks only the average unit price, pdl, and total quantity, xal: p d l = [ (pg l x g l + pb1 x b l ) / (x g l + x b l ) ], x d J = (Xg1 + x b l ) . This "a pill is a pi l l" methodology for creating unit values across generics and brands is similar to how purchases of the same brand from different outlets are treated. It imposes the assumption that brand-name and generic versions of a drug represent equal amounts of productivity—an assumption that can be rationalized based on their average clinical productivity. Whereas the DIN-based indexes use price and quantity observations for 4,740 competing products, the unit value index uses aggregated observations for each of the 2,484 unique dosage forms of drug identified by the AHFS number. The unit value indexes are 39 Table 4: Rates of Change for Chained Price Indexes 86-87 87-88 88-89 89-90 90-91 91-92 92-93 93-94 94-95 AAGR All Drugs Unadjusted Paasche 6.3% 2.5% 4.3% 3.9% 4.5% 3.8% 0.6% -0.6% -1.7% 2.59% New Goods Paasche 6.0% 2.3% 4.1% 3.8% 4.5% 3.8% 0.5% -1.2% -1.9% 2.39% Unit Value Paasche 6.1% 3.0% 3.5% 3.1% 3.5% 3.2% -0.2% -5.7% -4.7% 1.25% Multi-source Drugs Unadjusted Paasche 5.7% 1.3% 4.2% 4.0% 4.1% 3.3% -0.4% -1.8% -2.7% 1.93% New Goods Paasche 5.4% 1.0% 4.0% 3.9% 4.0% 3.3% -0.7% -3.1% -3.4% 1.56% Unit Value Paasche 5.5% 1.9% 3.2% 3.0% 2.6% 2.4% -2.1% -12.0% -10.0% -0.79% Single-source Drugs Unadjusted Paasche 8.7% 7.5% 4.6% 3.4% 5.3% 4.5% 1.9% 0.4% -1.0% 3.87% CPI for BC All Items CPI 3.0% 3.6% 4.5% 5.5% 5.3% 2.7% 3.5% 2.0% 2.3% 3.59% Med & Pharm. CPI 7.7% 6.0% 3.9% 4.2% 6.6% 1.9% 5.5% 1.2% 1.4% 4.24% A ARG: Average annual growth rate. constructed with these aggregated price and quantity data in otherwise standard index formulae. Findings The inflation rates of alternative Paasche price indexes for Pharmacare Plan A are listed in Table 4 along with Statistics Canada's "all items" and "medical and pharmaceutical products" consumer price indexes for British Columbia. The difference between Pharmacare's new-goods-adjusted indexes and the unadjusted indexes indicates the extent to which generic drugs merely represent a "new goods" problem for measurement—one that is remedied with the new goods adjustment. For all purchases under Pharmacare Plan A from 1986 to 1995, this amounted to a 0.2 percent per annum difference between unadjusted and adjusted average annual inflation rates of 2.59 percent and 2.39 percent. The difference between the new-goods-adjusted indexes and the unit value based indexes gives the magnitude of the potential measurement bias caused by gradual generic diffusion. This amounted to 1.14 percent per annum for the price of purchases under Pharmacare Plan A from 1986 to 1995 because the unit value based price index grew at an average annual rate of 1.25 percent. The price indexes for single-source drugs and multi-source drugs purchased under Pharmacare Plan A are illustrated in Figure 4 so that the effect of generic competition (or lack thereof)24 can be seen more clearly. It is clear that the total change in price indexes over the period of analysis is the net result of two very different stages. The differences between the indexes for multi-source drug classes over the 1986 to 1993 period are consistent with the findings of Griliches and Cockburn (1994). The total potential for bias 2 4 By definition of a single-source drug, generics will not have a direct effect on single-source drug price indexes. 40 Figure 4 Alternative Paasche Price Indexes for Single and Multi-source Drugs BC Pharmacare Plan A -1986 to 1995 150 •* * •* 125 100 75 50 Single-source -o- Multi-source "Unadjusted" -a- Multi-source "New-Goods-Adjusted' 25 -o-Multi-source "Unit Value Based" 0 4 1 1 1 1 1 1 1 ' 1 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 resulting from failure to account for the impact of generic competition is significant but moderate over this period, representing approximately 0.8 percent per annum on a pharmaceutical inflation rate between 2.3 and 3.2 percent. Through to 1993, the new goods adjustment accounts for only one-quarter—0.2 percentage points—of the spread between the unadjusted index and the unit value index for multi-source drugs. The gradual diffusion of generics explains the remaining three-quarters of the 0.8 percent spread over these years. (If Griliches and Cockburn's preferred reservation prices were used, the new goods adjustment would account for one eighth of the spread, three eighths would be accounted for by their "diffusion adjustment" and half of the spread would remain unadjusted for.) In 1994, and again in 1995, the spread between the unit value index and the unadjusted index widened dramatically. The difference in growth rates widened from 0.8 percentage points to 8 percentage points. Furthermore, the new goods adjustment explained very little of this change. Just as the difference between the unit value index and the unadjusted index rose, so did the difference between the unit value index and the new-goods-adjusted index. This indicates that, in 1994 and 1995, there was a major shift in market shares for drugs that had already been subject to generic competition. The apparent movements within markets for multi-source drugs are coincident with a policy change that was aimed at increasing the incentive for consumers and pharmacists to consider costs when choosing between brands and generics. 41 Changes in the Incentive to Consider Generics Pharmacists' incentives may ultimately determine generic utilization patterns because pharmacists often choose between brand-name and generic drugs on behalf of patients. When acting as agents for insured patients, pharmacists might dispense brand-name drugs based on the belief that patients would choose brands if, hypothetically, patients possessed pharmacists' professional knowledge. Pharmacies are, however, for profit corporations; therefore, financial incentives (such as mark-ups, sales bonuses and "kickbacks" or the threat of audit) will influence decisions made by pharmacists. Such incentives will be important in determining the extent of generic drug use in the Pharmacare program because Pharmacare pays pharmacies directly for the total ingredient cost of drugs dispensed to its beneficiaries.25 In 1990 and again in 1994, Pharmacare modified the terms of pharmacists' drug cost reimbursement to provide financial incentive to dispense generic drugs (Grootendorst et al. 1996). In doing so, Pharmacare took a more active role in the market, altering the principal-agent relationships between patient and pharmacists in an attempt to reduce the potential for a form of "moral hazard" problem that stemmed from the financial incentives of patients and pharmacists. These reimbursement policies focussed on the dispensing of brand-name versus generic drugs because the financial relationship that Pharmacare had with patients and pharmacists provided "levers of control" in the this decision-making process. This contrasts with the lack of such financial "levers of control" over physicians' decisions in the prescribing process. Prior to 1990, pharmacists were reimbursed for ingredients dispensed according to Actual Acquisition Costs. Although this form of reimbursement reduces markups on dispensed drugs, it does not give pharmacists an incentive to dispense the lowest cost drug in the first place (Gorecki 1992, Anis 1994). Both patient and pharmacist have incentives that could lead to excessive brand name use under Actual Acquisition Cost reimbursement: neither is financially worse off when a brand-name drug is dispensed. Because of this weakness, generic dispensing for Pharmacare beneficiaries was relatively low in the 1980s (Grootendorst et al. 1996). To encourage more generic drug use, Pharmacare implemented the Product Incentive Program for 216 frequently prescribed multi-source drugs in 1990. Under this program, savings that resulted when a pharmacist dispensed a generic were "shared" with the dispensing pharmacy. The program gave the dispensing pharmacy 20 percent of the difference between a fixed base price and the cost of the drug dispensed. It was hoped that this would make generic dispensing more common, but switching patients to generic drugs in order to earn a bonus payable to the pharmacy apparently conflicted with pharmacists' professional self-image.26 Furthermore, the bonus scheme used a base price that was high enough to provide a "bonus" to continue dispensing brand-name drugs, albeit this bonus was smaller than would be received i f the generic was dispensed. Some argue that the objective of maintaining professional self-image dominated the slight incentive for generic dispensing, rendering the Product Incentive Program unsuccessful as a tool for 2 5 The direct payment system, as used by Pharmacare, differs significantly from indemnity plans wherein patients first pay for the drugs out-of-pocket and then seek reimbursement from the drug plan by submitting receipts. 2 6 Grootendorst and colleagues (1996) discuss pharmacists' perceptions of the Product Incentive Program in detail. 42 increasing generic drug use (Grootendorst et al. 1996). It is also possible that manufactures offered pharmacists incentives—such as discounts, merchandising kickbacks or cooperative advertising—to continue dispensing brand-name drugs. On April 1, 1994, the government adopted a new reimbursement policy, called the Low Cost Alternative Program, that would apply to virtually all multi-source drugs. Pharmacists would no longer be paid a bonus for dispensing generic drugs. Instead, Pharmacare only paid the price of the "low cost alternatives" in multi-source classes, even i f the brand-name drug was dispensed. The low cost alternatives were invariably generics.27 Pharmacare would pay the full price of the brand-name drug only in cases where patients could not tolerate a generic drug for valid medical reasons. In effect, the new policy did two things. First, it ensured that Pharmacare only paid a price that reflected the clinically substantiated relative value of competing products. From the perspective of the public drug benefits provider, brand-name drugs are only of added value when there are valid medical reasons contraindicating the use of a generic. Thus, the Pharmacare plan would only pay the proven "social value" for a drug. The cost increment required to secure any additional private value that consumers placed on brand-name drugs would have to be paid out of pocket. This was the second aspect of the policy change: it changed (essentially eliminated ) pharmacists' incentives by placing greater product discretion in the hands of patients. The Impact of Policy Changes There was little change in the rates of growth of multi-source price indexes in 1990 (see Table 4 and Figure 7), nor was there a change in the difference between the alternative indexes. This is consistent with the view that the Product Incentive Plan had little effect on generic drug use. It certainly had little impact on aggregate price indexes. On the other hand, changes in the unit value indexes presented above indicate that consumption patterns within multi-source categories changed dramatically in 1994 and 1995—the periods following the introduction of the Low Cost Alternative Program. In particular, it appears that the generic diffusion process changed substantially in these years. The market shares held by brand-name drugs that were subject to generic competition during the period of study clearly illustrate changes in generic drug use upon the introduction of the Low Cost Alternative Program. Searching the database revealed that 132 products purchased by Pharmacare Plan A went from being single-source to being multi-source between 1988 and 1994.29 The point of first generic entry was clearly This policy is similar in spirit to reference based pricing used in British Columbia and elsewhere. However, the Low Cost Alternative program promotes substitution only between chemically equivalent products, whereas reference based pricing policies typically deal with therapeutically similar, but chemically different, products. 2 8 Under the new policy, pharmacists' financial incentives still could be modeled because they might choose to take a financial loss when dispensing brand-name drugs if they believed the brand-name drug to be of real value to patients. 2 9 Database search criteria identified drug types that were single-source in 1986 and 1987, and then became multi-source drugs in following years. Drug types that ceased to be purchased under the Pharmacare Plan were excluded from the sample. 43 Figure 5 Brands' Average Share of Markets Grouped by Year of Generic Entry BC Pharmacare Plan A1986 to 1995 identifiable i n 114 o f these markets. These markets were grouped according to the year o f first generic entry so that potential cohort effects cou ld be i l lustrated. Cohorts comprised o f 18 to 22 markets except for the 1991 and 1992 groups. O n l y 9 and 6 generic entries were identif ied i n 1991 and 1992, poss ib ly due to the 1987 amendments to Canada ' s compulsory l icens ing p r o v i s i o n for drug patents. 1 Figure 5 contains a plot o f the sales-weighted average share o f market revenues and quantities he ld b y brands i n each cohort. Gener ic entry is identif ied i n F igure 5 by markers unique to each cohort. Despi te the fact that the average entrant offered a 21 percent price discount, generics had little immediate impact on market shares he ld b y almost a l l o f the cohorts. The 1988 to 1992 cohorts o f leading brands lost less than ten percent o f their markets ( in quantity or sales terms) dur ing their first years o f generic compet i t ion. The 1993 cohort lost a re la t ively large 18 percent o f market quantities because two h igh-volume products lost 30 percent o f their markets that year . 3 2 Y e t , even the loss suffered by these products is smal l b y compar ison to the 1994 cohort ' s average in i t i a l loss o f 43 percent o f Initial generic entry could not be established clearly in nine classes because generics appeared and exited periodically over the course of analysis. Nine incumbents, with combined annual sales of only $50,000, appeared to be subject to competition by other brands. 3 1 Canada's compulsory licensing provision for pharmaceutical products allowed competing firms to sell patented drugs during the life of the patent. Patent holders were paid a fixed royalty but could not block these licenses—hence, they were "compulsory." The 1987 amendments (Bill C-22) gave patent holders 10 years of market exclusivity before compulsory licenses would be granted. Patent amendments in 1993 (Bill C-91) repealed the provision altogether. 3 2 In 1993, two dosage forms of the brand-name drug Pepcid lost 30 percent of their $2-million-a-year of Pharmacare sales when three competitors launched generic alternatives at 20 to 30 percent discounts. 44 Figure 6 Brand's Share of Total Sales in Multi-source Diuretic Drug Markets BC Pharmacare Plan A 1986 to 1995 10% 80% 40% Moduret Tablets Lozide Table Generic Entry 89:IV — • Generic Entry (Generic Price 0.73) ^ ^ S D - C W - (Generic Price ( 95:1 —>\ J.78) \ — ' H All Other Multi-Sourced Diuretics 1 Generic Entry in 1988 and Earlier LCA Program Begins April 1994 • — i — i — i — i — i — r — i — i 1 1—i 1 — i — i — i 1—i 1 — i — i — i 1 — i — i — i — i — i — i — i 1 — i — i r \ X V» rJ'*pP_r>!!u\ 1—n-^K-^^*-*?^^ 86:1 87:1 88:1 89:1 90:1 91:1 92:1 93:1 94:1 95:1 market quantities (37 percent o f market revenues). Th i s increase i n in i t i a l market share loss is probably expla ined by the fact that the 1994 cohort is the o n l y cohort o f leaders to have generic competitors enter after the implementat ion o f the L o w Cos t Al te rna t ive program. Fa r more s t r ik ing than changes i n the in i t i a l market share losses are changes i n the generic di f fus ion process—that w h i c h occurs f o l l o w i n g the per iod o f in i t i a l entry. Gener ic diffusion under the Pharmacare p lan was extremely s low before the implementa t ion o f the L o w Cos t Al te rna t ive program. Gener ics avai lable i n 1993 offered a sales-weighted discount o f nearly 40 percent relative to the brands i n the markets i l lustrated i n Figure 5. Y e t , no cohort o f brands depicted i n F igure 5 lost more than 35 percent o f market revenues to generics b y the end o f 1993. C o m b i n e d , Pharmacare 's expenditure on the 114 drug markets depicted i n F igure 5 was $25.5 m i l l i o n i n 1993—a year before the introduction o f the L o w Cos t Al te rna t ive Program. Brands accounted for 76 percent o f these sales. In 1995, a year after the p o l i c y change, the value o f these markets fe l l to $15.2 m i l l i o n , as the brands lost a l l but 2 percent o f market quantities and sales. Further evidence o f the immediate impact o f the 1994 p o l i c y change comes from specific case studies i l lustrated i n Figure 6. Th i s figure plots the share o f quarterly market sales he ld by brands i n markets for mult i -source diuretic drugs. D u r i n g the per iod o f study, there were 12 mult i -source diuretic treatments, 10 o f w h i c h were subject to generic compet i t ion o n or before 1988. The gray plot i n F igure 6 represents the sales-weighted average market share he ld b y these 10 relat ively o l d brand-name products. T h e y had total market sales o f $879,466 i n 1986, and lost on ly 40 percent o f their markets to generics by the first quarter o f 1994. Pharmacare implemented the L o w Cos t Al te rna t ive program at the beginning o f the second quarter o f 1994, precisely when these products lost a l l but 13 45 percent of their markets to generic competitors. One quarter later, they held only 3 percent of their markets. The two other lines plotted in Figure 6 correspond to the market share held by Moduret (the leading brand of 5mg amiloride + 50mg hydrochlorothiazide tablets) and Lozide (the leading brand of 2.5mg indapamide tablets). A generic competitor entered Moduret's market—worth $900,000-a-year in sales to Pharmacare Plan A — i n the fourth quarter of 1989. A second generic entered within a year. Despite the fact that both entrants were priced 27 percent below the brand, Moderet lost only 10 percent of its market during the year following generic entry. A generic competitor entered Lozide's market—worth $450,000-a-year in sales to Pharmacare Plan A — i n the first quarter of 1995. This was also followed by another generic within a year. Priced 22 percent below the Lozide brand, the first entrant stole over 90 percent of the market within a single year. The notable difference between these two examples is that the Lozide product was first subject to generic competition after the introduction of the Low Cost Alternative program. The Moduret brand did not lose significant market share until the Low Cost Alternative program was introduced—more than four years after generics first entered its market—despite the fact that generics remained 30 percent cheaper throughout this period. Measures of Quantity Changes in expenditure can be accounted for by changes in indexes of price, quantity or both. Per-capita expenditures under Plan A grew 253 percent (14.2 percent per annum) from 1986 to 1993, and then remained largely unchanged through to 1995. A n index of per-capita expenditures is illustrated in Figure 7 along with per-capita quantity indexes constructed by deflating per-capita expenditures with the new-goods-adjusted price index and unit value based price index.3 4 From 1986 to 1993, the new-goods-adjusted and unit value based per-capita quantity indexes grew at annual rates of 10.3 and 10.7 percent. Both of these indexes imply that changes in "real output" (or productivity) purchased under Pharmacare Plan A explained much of the changes in expenditures over this period. From 1993 to 1995, however, these indexes tell different stories. According to the new-goods-adjusted quantity index, expenditure inflation appears to have stopped in 1994 due to a pause in real output growth under Pharmacare Plan A. The new-goods-adjusted per-capita quantity index was unchanged between 1993 and 1994, and it rose only 2 percent in 1995. On the other hand, the unit value based per-capita quantity index grew 5 percent in 1994 and again in 1995. Its rate of growth was comparable to that in the preceding two years. The unit value based price and quantity indexes point to falling prices as the cause of the freeze in per-capita expenditures in 1994 and 1995. 3 3 Census population figures for BC residents over the age of 65 were used to control for the slight increase in elderly population over the period of study. 3 4 The unadjusted index is not illustrated because its movements are nearly identical to the new-goods-adjusted index. 46 Figure 7 Per-capita Expenditures and Per-capita Quantity Indexes BC Pharmacare Plan A 1986 to 1995 260 -i 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 The alternative measures of real activity in the sector differ because they assign different levels of "value" to generics purchased after their initial period of availability. The new-goods-adjusted index, which most closely conforms to the traditional economic approach to measurement, uses transaction price as the basis for assigning relative productivity values—sometimes called efficiency units (Fisher and Griliches 1995)—to chemically equivalent brand-name and generic drugs. As described in the chapter on economic measurement, traditional economic quantity indexes (such as the new-goods-adjusted index reported here) weigh the quantity changes of goods according to the share of total income spent on them. According to this methodology, the relatively low price of generic drugs "reveals" them to be of less value per unit (pill, tablet, liquid, etc.) than the brand-name drug because optimizing agents would not otherwise continue to purchase the higher-cost brand-name product. The 1994 reduction in real quantity growth reported by the economic index cannot be easily explained with this set of assumptions except to point out that, at least prior to 1994, consumption patterns could not be used to determine the relative value of generic and brand-name drugs due to decision-makers' nonstandard financial incentives. Conclusion The use of low-cost generics can significantly reduce the cost of prescription drugs, but this may not be reflected in certain price indexes. For drugs purchased under British Columbia's public drug insurance plan for seniors, a price index based on "unit values" for equivalent brand-name and generic drugs grew at an average annual rate of only 1.25 percent between 1986 to 1995. Comparatively, a price index consistent with the traditional 47 economic approach to measurement grew at a rate of 2.39 percent over the same period. The reason for the difference between these alternative indexes is that pharmaceutical prices and consumption patterns do not necessarily indicate social welfare because of the financial incentives of decision-makers. Decision-making agency relationships and nonstandard financial incentives can cause socially inefficient pharmaceutical utilization patterns, breaking down the relationship between the prices paid for drugs and their social value. The 1994 change in Pharmacare's drug-cost reimbursement policy illustrates how the financial incentives of decision-making agents can generate consumption patterns from which revealed preference theory would draw false inferences regarding the true "value" of directly comparable drug products. On April 1, 1994, the introduction of financial incentives for pharmacists and patients to consider the difference in price between brand-name and generic drugs caused almost all beneficiaries of Pharmacare Plan A to switch (or be switched) to available generics. Price indexes consistent with the traditional economic approach to measurement failed to account for the resulting savings because—at least prior to the policy change—the financial incentives of principal decision-makers in this setting were inconsistent with assumptions upon which traditional theory rests. In contrast, the price indexes that impose a common unit value for chemically equivalent brand-name and generic drugs showed a marked decline at the onset of the new policy. The unit value based indexes also displayed a more plausible trend in real output through the policy change. Because patients and/or pharmacists switched to generics en masse in 1994, the common unit value for brand-name and generic drugs appears to be closer to their "true" relative value than market prices indicated. Thus, the unit value based indexes appear to be closer to the "true" social price and quantity indexes. This analysis avoided contentious issues surrounding the appropriateness of prescription drug use by focussing instead on the less controversial issue of generic substitution. There is, however, a large body of literature indicating that prescription drug use—in Canada and elsewhere—is often not cost effective due to systematic errors in physicians' judgement of relative safety, efficacy and cost of drugs (Lexchin 1998, Tamblyn 1997, Ryan et al. 1996, Dennig and Haaijer-Ruskamp 1995). Alternative approaches to measurement are advocated by Berndt and colleagues (1998) who argue that moral hazard and imperfect principal-agent relationships in health care make it untenable to use revealed consumer purchases as a guide to the value of medical care. The results of the present study clearly indicate that this is so even for the apparently simple choice between brand-name and generic drugs. A health-outcomes approach to social price and productivity measurement may therefore be of significant value in the pharmaceutical sector. Such an approach might resemble Nordhaus' (1997) 'true price of light' indexes, focussing not on transaction prices and quantities for intermediates—oil, lanterns, light-bulbs and electricity—but on the cost of the derived output from their consumption—illumination. Scitovsky (1967) built indexes for selected health-care treatments based on such an approach three decades ago. Health economists and measurement theorists ought to pursue this idea again. r 48 5.Information (or Lack Thereof) and the Real Cost of Treating Hypertension Economists routinely make assumptions about what consumers know and how that knowledge relates to what we describe as "rational" choice. Our information-related assumptions are generally optimistic. We usually assume that consumers make judgements about products based on accurate information regarding characteristics. Where this is not so, we generally assume that they chose to make decisions based on incomplete or imperfect information because of the cost of obtaining and verifying the accuracy of additional information. In the pharmaceutical sector, it has long been acknowledged by governments that patients generally do not and cannot know the truth regarding the appropriate use of medicines. The information that is required to make rational choices about drugs is simply beyond the comprehension of those without adequate medical and pharmacological training. In response to this, drug selection has been delegated to medical authorities. Accordingly, drug-related information gathering is one of the primary services expected to be provided by prescribing physicians. In practice, the prescription decision-making process may be based on a sub-optimal information for a number of reasons. Doctors, the principal decision-makers in this sector, do not pay for drugs they prescribe on behalf of their patients. Moreover, when paid on a fee-for-service basis, doctors have incentive to see as many patients in as short a period as possible—subject, among other things, to the constraint of possible malpractice liability. The incentive to treat patients quickly and the lack of incentive to consider costs run counter to the incentive for physicians to engage in costly information gathering. Furthermore, physicians, like other people, have personal preferences and objectives. Many of their preferences are influenced by interactions with manufacturers of products and a desire to remain on the forefront of medicine. If, because of these many influences, doctors' choices are inconsistent with "rational" choice based on scientific information available to them at reasonable cost, traditional economic price and quantity indexes will give biased measures of sector productivity. The direction of this bias will depend on the circumstances of the market in question. This chapter investigates potential information problems in pharmaceutical price and productivity measurement. Attention is focussed on hypertension drugs, which offer a particularly clear illustration of the distinction between scientific information, which might be considered measurable and socially valuable, and the unmeasurable "information" affecting doctors "preferences" for certain drugs. The recommendations of national organizations of professionals interested in improving hypertension treatment are used as a marker of the "social preferences" for drug treatments. These preferences are used to construct health outcomes based price and quantity indexes. Traditional economic indexes are compared with the health outcomes based indexes to gauge the measurement bias that results from decision-making that is inconsistent with socially efficient choices based on accurate, scientific information. Finally, tests are applied to see i f patients who are 49 prescribed the drugs "reveal" preferences that are consistent with those "revealed" by prescribing doctors. / Information and Medical Practice Guidelines With the accumulation of evidence that variations in medical practice can seldom be explained by clinically relevant factors, medical practice guidelines have become commonplace (Grilli and Lomas 1994). Professional associations, government bodies and public health organizations have published guidelines or consensus recommendations for the treatment of countless conditions. In a majority of cases, the aim is to improve the quality of care provided by medical practitioners. In some cases, it has been to improve the cost-effectiveness of care delivered. The notable exception has been the development of industry-sponsored consensus conferences, which are ultimately held for promotional purposes (Sheldon and Smith 1993). Guidelines published by recognized professional bodies are typically developed by means of consensus among recognized experts using the best available scientific information about the relative advantages and disadvantages of treatment alternatives. Guidelines usually deal with specific treatments or conditions. They are disseminated to the target audiences by various means, including direct mailing, journal publications, conferences, continuing medical education seminars and face to face communications. Most are available to physicians at what might be considered reasonable cost—in terms of time and effort (they are invariably distributed free of charge). Do Practice Guidelines Guide Practice? Given the prevalence and purpose of guidelines, a parallel stream of research has evolved for the purposes of evaluating their impact on doctors' behaviour. These researchers ask the question, as Lomas et al (1989) put it, "Do Practice Guidelines Guide Practice?" Such studies have been conducted to evaluate practice patterns before and after the dissemination of guidelines, as well as physicians' awareness of and agreement with them. Evidence regarding the efficacy of guidelines is discouraging. Average compliance rates with guidelines are in the order of 50 percent (Grilli and Lomas 1994), but differ depending on the subject of the guidelines. Guidelines are least likely to be adhered to for complex tasks that have few immediate results and cannot easily be implemented on a trial basis (Grilli and Lomas 1994). Even when physicians report knowledge of and agreement with guidelines, actual knowledge of and compliance with guidelines appears poor (Lomas et al 1989). These results indicate that doctors believe that it is socially desirable to comply with guidelines even when they are not doing so themselves. This is further evidenced by the fact that, averaged across 10 studies, self-reported measures of compliance overestimate actual adherence35 by approximately 27 percent (Adams at al 1999). Discouragingly, the average self-reporting bias appears greater than the average increase in compliance attributed to the dissemination of guidelines (Adams at al 1999). Guidelines appear to affect what doctors believe should be done more than what they actually do. Based on objective measures such as audits of medical charts and prescribing records. 50 The purpose of the present study is not to evaluate hypertension treatment guidelines or their dissemination processes as tools to alter medical practice. Recognized national guidelines in Canada and the US are used here to determine what might be considered "socially desirable" practice patterns over the period of study—1986 to 1996. If actual practice patterns are inconsistent with recognized guidelines—whether driven by physicians' personal incentives, interests or otherwise—sector productivity measurement with traditional economic indexes of price and quantity will be biased. Hypertension36 Hypertension is the condition wherein a patient's resting blood pressure is above normal levels for a sustained period. Measured in terms of millimeters of mercury (mm Hg) using a mercury manometer, a patient's blood pressure is recorded in two statistics corresponding to the two phases in contractions of the heart: systolic pressure and diastolic pressure. Systolic pressure is the pressure exerted when blood in the chambers of the heart is forced outward. Diastolic pressure is the pressure exerted when the chambers of the heart are being filled. Systolic blood pressure is typically reported first. "Normal" blood pressure is approximately 120 over 80mm Hg. Diastolic hypertension is marked by diastolic pressures above 90 to 100mm Hg. Isolated systolic hypertension occurs when diastolic blood pressure is approximately normal but systolic pressures are high—above 150 to 160mm Hg. Elevated blood pressure may be caused by the presence of another illness. If an identifiable illness is the cause, then a patient's hypertension is classified as secondary hypertension because the elevated blood pressure is "secondary" to the (generally more serious) primary illness, such as renal failure. For most hypertensive patients, no attending illness causes the elevated blood pressures. These cases are known as essential hypertension. Essential hypertension is far more common than secondary hypertension. Essential hypertension was the number one diagnosis for visits to doctors in the US and Canada during 1998. Essential hypertension was the primary diagnosis in approximately 5 percent (13.8 million) of all office visits in Canada. It accounts for about twice the number of visits for the second ranking diagnosis, diabetes. According to IMS Health, in 1998, over 80 percent of visits to Canadian doctors ended with a prescription when essential hypertension was the primary diagnosis—compared to 59 percent of all patient visits. Because hypertension is a chronic illness, about 90 percent of doctor visits for essential hypertension are repeat visits by patients already diagnosed with the condition. The Heart and Stroke Foundation of Canada estimates that approximately 22 percent of Canadian adults—26 percent of men and 18 percent of women—have hypertension (HSF 1999, p.30). Yet, it is estimated that approximately 50 percent of Canadians with hypertension are unaware that they have it (HSF 1999, Feldman et al 1999). Those going undetected are more likely to be younger to middle-aged hypertensives, since routine checks 3 6 This section is based on several references. ICES (1998) is the most closely related concise summary of hypertension. 3 7 Figures from IMS HEALTH Canadian Pharmaceutical Industry Review and the National Disease and Therapeutic Index, http://us.imshealth.com/ (accessed on Tuesday 21 December, 1999.) 51 for high blood pressure increase with age, along with the prevalence of hypertension. About 90 percent of seniors reported having had their blood pressure taken by a doctor within the year preceding a 1996/97 survey conducted for Health Canada and the Heart and Stroke Foundation of Canada (HSF 1999 p. 30). One third of seniors reported that they had had high blood pressure diagnosed by a physician (HSF 1999 p. 30). Treatment Hypertension is a concern due to its correlation with the onset of adverse cardiovascular events such as coronary artery disease, congestive heart failure and stroke. The primary goal of hypertension treatment is to reduce morbidity and mortality associated with elevated blood pressures. To do so, treatments aim to reduce blood pressure and, by inference, reduce the risk of cardiovascular events. Drug treatment typically requires that a patient take an antihypertensive drug at least once a day for an indefinite period. Due to the inconvenience and cost of daily drug maintenance, as well as the side effects and risks inherently associated with any drug therapy, non-drug therapies are a preferred first-step for patients with mild to moderate hypertension (JNC, various years; Reeves at al 1993; ICES 1998; Anonymous 1999). It is estimated that 50 to 70 percent of patients with mild hypertension can be successfully treated with diet and lifestyle modifications alone (ICES 1998). Reducing salt, fat and alcohol consumption, losing weight, quitting smoking, exercising and controlling stress are associated with significant reductions in blood pressure levels (JNC, various years; ICES 1998, Anonymous 1999). When lifestyle changes alone are insufficient for bringing a patient's blood pressure down to acceptable levels, drugs may be used. The four most frequently used classes of hypertension treatment are diuretics, beta-blockers, ACE-inhibitors and calcium-channel blockers. Certain other drugs may also be used in the treatment of hypertension; these include alpha 1 blockers, central peripheral sympatholytics and direct vasodilators (TI 1995b). Diuretics are the oldest group of anti-hypertensives, most of which have been off patent for decades and are widely available in low-cost generic form. ACE-inhibitors and calcium-channel blockers, which are the most commonly used hypertension drugs today, began entering the market in the 1970s and 1980s. Newer, patented versions of A C E -inhibitors and calcium-channel blockers (as well as entirely new classes of hypertension drugs) continue to come onto the market today. Difficulties with Antihypertensive Drug Choice There are two clinical aspects of the pharmacological treatment of hypertension that cloud the "information" available during drug-related decision-making. 8 First, the treatment of hypertension involves a substantial placebo effect that makes it difficult for a physician to evaluate real productivity on a trial and error basis. In virtually all properly conducted clinical trials, average blood pressure has been shown to fall consistently with 3 8 There are also a few troubling issues concerning the diagnosis of hypertension and decisions of when to treat. Among these is white-coat hypertension—wherein patients' blood pressure is higher in physicians' offices than at home (MacDonald et al 1999). These issues are beyond the scope of this paper, which addresses drug choice once treatment is indicated. 52 placebo treatment (TI 1995b, Wright et al 1999). Thus, no matter what a doctor prescribes for a patient with high blood pressure, it is likely to appear effective at reducing blood pressure (TI 1995b). Moreover, individual doctors will be unable to accumulate enough information to determine a statistically significant difference among drugs to treat hypertension. The second informational problem with individual assessments of hypertension treatments is that efficacy in reducing blood pressure, in and of itself, is necessary but not sufficient for real productivity. The evaluation of hypertension treatment is frequently based on the assumption that all mechanisms to lower blood pressure wil l have the same pressure-related benefits to health, irrespective of the mechanisms themselves (Wright et al 1999). Reliance on surrogate health outcomes to measure efficacy may be a major informational problem in the evaluation of the real cost of treating hypertension. Hypertension treatments that offer no long-term health benefits may be approved for sale because clinical trials needed for regulatory purposes are short-run trials. If a product is reasonably safe and effective at reducing blood pressures in these short-run studies, it can be sold as a hypertension treatment. Much longer trials are needed to determine whether a drug is effective in reducing long-term morbidity and mortality associated with hypertension. In fact, the best selling anti-hypertensive drugs in the 1980s were later shown to increase long-term risks of death (more below). Since drugs to treat chronic "risk-factors" are intended to reduce unwanted events in the long-run, the true value of such medicines may not be known ex-ante. Evaluating the long-term effects of drugs requires massive, properly designed and conducted randomized clinical trials. A number of these gold-standard trials have been conducted on hypertension treatments, the majority of which have focussed on the older products (Wright et al 1999). If a drug proves to offer no benefits or, worse, harms patients, its consumption may be considered wasteful—certainly when judged from the informational position of the ex-post. Even ex-ante, the use of newer, unproven medicines may be considered costly in terms of risk—perhaps unnecessarily so—if products exist for which there is an established body of evidence regarding safety and efficacy. Thus, the "stock" of true information available about a medicine ought to be considered when evaluating the rationality of medical decision-making. As discussed below, for over 20 years, guidelines for the treatment for hypertension appear to have incorporated this form of risk-aversion in their recommendations. Combined, the existence of strong placebo effects and the need for scientific collection of evidence regarding long-term efficacy make hypertension treatment an ideal candidate for medical practice based on scientific guidelines. It should not be surprising, then, that consensus statements and guidelines for the treatment of hypertension have been generated and disseminated by professional associations for more than twenty years. Guidelines for the Treatment of Hypertension A study by the Veterans Administration published in 1970 offered the first clinically substantiated evidence about whether antihypertensive drugs reduced morbidity and mortality among hypertensive patients. Using scientific information made available from such studies, Canadian and American committees representing interested professional organizations began to publish guidelines for the treatment of hypertension in the late 1970s. 53 The body that publishes the most widely recognized guidelines is the US Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure (JNC). Canadian guidelines are published by the Canadian Hypertension Society (CHS). In addition to various direct dissemination efforts, summary reports of the JNC and CHS guidelines are published in major medical journals—the Archives of Internal Medicine and Canadian Medical Association Journal, respectively—every two to four years. With each publication, these guidelines spawn numerous additional summaries and commentaries in medical, nursing and pharmacy journals. In 1977, JNC published its first report of the consensus-based recommendations of professional organizations interested in the treatment of hypertension (JNC 1977). Members of the 1977 JNC were the American Academy of Family Physicians, American College of Cardiology, American College of Physicians, American Heart Association, Veterans Administration, American Medical Association, National Kidney Foundation, National Medical Association and the United States Public Health Service (JNC 1977). The first Canadian guidelines published by a recognized national organization were also published in 1977. Those guidelines were published by the Canadian Cardiovascular Society, the Canadian Heart Foundation and the Ontario Council of Health—collectively, the predecessors to the CHS. The Canadian and American guidelines of 1977 recommended treatment for "virtually all" patients with diastolic pressures consistently above 105mm Hg, with the aim to reduce diastolic blood pressure to 90mm Hg. Both outlined a "Step Care" approach to drug treatment (JNC 1977, p.259-61). Step care is relatively simple. It begins treatment with a "first-line" drug alone. If blood pressures are not controlled by the initial drug, step care recommends the addition of drugs from other drug classes to the patient's first-line drug regiment. The initial drug would be discontinued only i f it proved to be intolerable to the patient. In 1977, thiazide diuretics were recommended as the first line of drug therapy by both the JNC and CHS because it was thiazides that had been proven to reduce morbidity and mortality in the Veterans Administration study. Doctors were encouraged to prescribe thiazides in low-doses for newly treated patients. The recommended dosage range for hydrochlorothiazide was 50-100mg per day—high by today's standards. The step care protocol then called for an increase in dose, as necessary, or the addition of a second and, possibly, a third drug until blood pressure levels were reasonably controlled. Most non-diuretic antihypertensive drugs that were commonly used in the 1970s are classified in this paper as "other" drugs—that is, they were not beta-blockers, ACE-inhibitors or calcium-channel blockers. In 1980, the JNC task force published a second set of consensus recommendations— JNC II—based on new evidence from clinical trials. The JNC II continued to endorse the "step care" approach beginning with diuretics (JNC 1980, p. 1282). Beta-blockers and selected other drugs were recommended as first-line drugs i f diuretics were contraindicated and if specific co-morbidities were present. After initial therapy, the choice of second and third line drugs was more eclectic and could include beta-blockers or several "other" drugs, such as vasodilators. The recommended drug treatment for elderly patients was oral thiazide diuretics in "smaller than usual doses, " which the JNC II noted "...are frequently effective 54 as the sole agent in controlling hypertension in [the elderly]" (JNC 1980, p.1284). The JNC II recommended that greater caution be used when choosing to add second step drugs for the elderly. In 1984, separate guidelines for hypertension treatment were published by the CHS and the JNC. The 1984 CHS report was the first official publication of the CHS task force on the management of hypertension. Funded by the Medical Research Council, National Health Research and Development Program, and the Ontario government, thirty-one biomedical scientists met at the Canadian Hypertension Society's Consensus Conference on the Management of Hypertension, Toronto, in November of 1983. They published what was hoped to be the first of an annual series of conference recommendations in 1984 (Logan 1984). The CHS and JNC III reports of 1984 were in near consensus regarding the treatment of hypertension. One of the only substantive differences was that the JNC took a more aggressive stance regarding when to treat hypertension (JNC 1984). Treatment protocols for non-elderly patients retained the step care format, with thiazide diuretics or beta-blockers recommended by both guidelines as the first-line drugs of choice (JNC 1984, Logan 1984). However, to reduce the side effects of diuretic use, the dosages of diuretics recommended in the Canadian and American guidelines were about half that recommended in early publications—e.g., 25 to 50mg of hydrochlorothiazide per day. For seniors, the JNC III recommended thiazides alone as first-line treatment or in combination with beta-blockers in "smaller than usual" doses (JNC 1984). The CHS did not address treatment of the elderly in 1984. The CHS published consensus guidelines for the treatment of hypertension in the elderly in 1986 (Larochelle et al 1986). The recommended protocol for treating elderly hypertensives involved step care with thiazide diuretics as the preferred first-line drug. If thiazides were contraindicated, beta-blockers were recommended as alternate therapy (Larochelle et al 1986). The CHS specifically addressed the use of the newer antihypertensive drugs in the treatment of the elderly by noting that ACE-inhibitors and calcium-channel blockers may be useful, "...but further study is required before they can be recommended for the elderly" (Larochelle et al 1986, p.745). The JNC published its fourth round of guidelines, JNC IV, in 1988. These were significant because they marked a deviation from the step care model and would be inconsistent with the Canadian guidelines for the first time. The JNC IV was also the first publication of the JNC that involved the American Pharmaceutical Association—a drug manufacturers trade association—as a member organization on the National High Blood Pressure Education Program Coordination Committee. This committee had to endorse the final report of the JNC before it was published (JNC 1988, endnote p. 1037). Pharmacologic treatment protocols in the JNC IV significantly deviated from earlier models. The JNC IV no longer recommended a clear step care model because second and third "steps" in the JNC IV suggested the addition or the substitution of drugs from other classes rather than strictly adding them to the existing drug regimen (JNV 1988, p. 1027-1028). Perhaps most importantly, choices at each "step" became more eclectic than ever before. The initial therapy recommended in the JNC IV was no longer limited to diuretics 55 and beta-blockers; ACE-inhibitors and calcium-channel blockers were also listed as possible first-line therapies. The choice of drug was to be tailored to the "special considerations" of the patient (JNC 1988, p. 1028). Among the special considerations were lifestyles, physiologic and biochemical measurements, and economic considerations, making it clear that the JNC IV endorsed trial-and-error prescribing by individual physicians. The discretionary approach to drug choice was applied to the JNC IV protocol for treating the elderly (JNC 1988, p.1034), but under "special considerations" it is noted that elderly respond better to diuretics or calcium antagonists than beta-blockers or ACE-inhibitors (JNC 1988, p. 1029). In order to minimize the side-effects of diuretic treatments the recommended dosage levels for diuretics fell, once again, to half the previous recommendations: e.g., 12.5 to 50mg of hydrochlorothiazide per day. In 1989, the CHS published the report of its consensus conference on the pharmacologic treatment of hypertension (Myers et al 1989). In large part, the purpose of this conference was to evaluate evidence about the use of ACE-inhibitors and calcium-channel blockers. Like the JNC IV, the CHS report of 1989 moved away from the step care approach because it advocated the substitution of mono-therapies as second and third-line treatments based on the idea that treatment regiments should be simple. It also endorsed a more discretionary drug choice protocol—or lack thereof (see Spence 1989). However, the 1989 CHS report seems clearer in its endorsement of low-dose thiazide diuretics or beta-blockers as initial therapy for non-elderly patients without coexisting medical conditions (Myers et al 1989, pp.1143-1144). Moreover, the 1989 CHS protocol for elderly patients continued to endorse the use of thiazide diuretics in small doses, with reference to the recommendations published in 1986 (Myers et al 1989, p.1144). The 1989 CHS report also made specific recommendations about certain classes of drugs. Notably, it stated that "Calcium antagonists are generally recommended as second-line therapy," and that "Nifedipine [a calcium antagonist] should be considered as a second or third-line drug..." (p.1145). At the time, Nifedipine was one of the most heavily promoted and widely prescribed drugs in Canada and the US—it was also the drug at the center of the calcium-channel blocker controversy in the 1990s (more below). In 1993, the JNC produced another somewhat unexpected report—the JNC V. The JNC V was then described as "steps forward and steps backward" (Weber and Laragh 1993). The steps forward were that the JNC took a broader approach to classifying hypertensive patients, it emphasized the need to treat patients with isolated systolic hypertension, and it increased the emphasis on the non-drug treatment of hypertension (JNC 1993). The step "backward" was the reversal of the JNC IV recommendations to add ACE-inhibitors and calcium-channel blockers to the list of potential first-line treatments for uncomplicated hypertension. The JNC V protocol reverted to thiazide diuretics or beta-blockers as the preferred first line treatments, followed by their substitution or combination. A C E -inhibitors, calcium-channel blocker or other drugs appear as third-line drugs (JNC 1993, p.170-171). For the elderly, it is noted in the JNC IV that " A l l classes of antihypertensive drugs have been shown to be effective in lowering blood pressure in older patients. However, only diuretics and beta-blockers have been used in controlled trials that have shown a reduction in cardiovascular morbidity and mortality" (JNC 1993, p. 178). On that 56 basis, diuretics and beta-blockers are endorsed as preferred first-line treatments unless contraindicated. The CHS also published a new set of guidelines in 1993, and was happy to report consensus with the recently published JNC V (Carruthers et al 1993, Ogilvie et al 1993, Reeves et al 1993). Unlike previous JNC consensus conferences, the basis of the 1993 CHS report was meetings of several special committees that would each publish a report on specific aspects of hypertension diagnosis and treatment. The new format for CHS conferences also involved new funding sources. In addition to funds from the Medical Research Council and Health Canada, numerous drug companies funded the 1992 CHS consensus conferences. In return for this, representatives of these companies participated in the two-day discussions but did not have a vote on the final recommendations. However, companies did review the final drafts of recommendations from the Diagnosis and Pharmacotherapy working group and from the Elderly and Diabetes group. Perhaps in response to the obvious conflicting interests in the new process, the CHS evolved a system of grading its recommendations. Recommendations in the 1993 reports were graded (from A to D) according to the quality of scientific evidence that they were based upon (Carruthers et al 1993). The CHS would continue to take a more conservative "when to treat" approach than the JNC. Recommended first-line drugs for uncomplicated essential hypertension were diuretics or beta-blockers (grade A). This was followed by substituting the untried first-line drug (grade A), then by the combined use of diuretics and beta-blockers (grade A). As with the JNC V , ACE-inhibitors, calcium-channel blockers and other drugs became third-line choices (grade B) (Ogilvie et al 1993). With these recommendations, the authors of the CHS report of pharmacologic treatment note that... ...the role of diuretics and beta-blockers in initial therapy for mild or uncomplicated hypertension is well supported. The actions of A C E -inhibitors and calcium entry blockers appear to be comparable, and some practitioners argue for the addition of these other drug groups to diuretics and beta-blockers for initial monotherapy, with the expectation of improved outcome for cardiovascular disease. Unfortunately, not enough long-term clinical trials have been done with the main endpoints of illness or survival rates to conclude that these drugs may be recommended along with diuretic or beta-blocker therapy ... Other purported attributes of the newer compounds, such as favourable effects on the quality of life and neutral effects on serum lipid levels, have not as yet been related to improvements in long-term rates of illness or death. Consequently, decisions to favour one treatment over another remain speculative. (Ogilvie et al 1993, p. 577, emphasis added) 3 9 The companies were Merck Frosst Canada, Rhone Poulec Rorer, Knoll Pharmaceuticals Canada, Sender Canada, Abbot Laboratories, Shering Canada, Searl and Co. of Canada, Sandoz Canada, Nordic Merrell Dow, Bristol Myers Squib, Pfizer Canada, ICI Pharma, Wyeth Ltd., Hoffmann-La Roche, Miles Canada, Astra Pharma and Parke-Davis. 57 Table 5: Major Canadian and American Guidelines for Hypertension Treatment 1977 to 1999 Study, Date Step Care First-line for First-line for Elderly HCTZ Non-Elderly Dose / Day JNC I, 1977 Yes Diuretic* 50- lOOmg CJR, 1997 Yes Diuretic* JNC II, 1980 Yes Diuretic* or Beta-Blocker Diuretic* CHS, 1984 Yes Diuretic* or Beta-Blocker 25 - 50mg JNC III, 1984 Yes Diuretic* or Beta-Blocker Diuretic* or Beta-Blocker 25 - 50mg CHS, 1986 Yes Diuretic* 25 - 50mg JNC IV, 1988 Partial Diur.*, Beta-B., ACEI or CCB Diur.*, Beta-B., ACEI or CCB 12.5-25mg CHS, 1989 Partial Beta-Blocker or Diuretic* Diuretic* 25 - 50mg JNC V, 1993 Partial Diuretic* or Beta-Blocker Diuretic* or Beta-Blocker 12.5-50mg CHS, 1993 Partial Diuretic* or Beta-Blocker Diuretic* 12.5-25 mg ACEI = ACE-inhibitor CCB = calcium-channel blockers. CHS = Canadian Hypertension Society. CJR = Canadian Joint Recommendations (by the Canadian Cardiovascular Society, the Canadian Heart Foundation and the Ontario Council of Health). JNC = Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure. * Refers to thiazide or thiazide-like diuretics. Loop diuretics and potassium sparing diuretics to be used only when specially indicated. The 1993 CHS recommendation for first-line drug treatment for elderly with uncomplicated hypertension was low-dose thiazide diuretics (grade A). When thiazides are contraindicated or not preferred, the beta-blockers were recommended as second line drugs (grade B) (Reeves et al 1993). Table 5 summarizes the drug-treatment recommendations of the Canadian and American guidelines from 1977 to 1993. Guidelines were published again in 1997 (JNC VI) and 1999 (CHS). These guidelines, which did not change treatment protocols dramatically, are described in appendix A since they do not relate to the period of study below. Adverse News about Calcium-channel Blockers In the mid 1990s, three scientific studies indicated that patients on a short-acting calcium-channel blocker had a significant increase in risk of death compared to those on other antihypertensive drugs or placebos (see Maclure et al 1998 and Stelfox et al 1998). The particular drug tested was the most popular selling hypertension drug in North America during the 1980s and its long-acting version had become the most popular drug of the early 1990s. Due to the widespread use of the group of drugs implicated by the studies, news about the negative findings circulated well before the studies were published in 1995. Manufacturers of calcium-channel blockers immediately launched a campaign to intimidate and discredit the authors of the critical articles—threatening lawsuits, attempting to block publication and paying for widespread dissemination of "dear doctor" letters that questioned the authors' credibility (Deyo et al 1998). Amidst the controversy, a special edition of the Canadian television show "the Fifth Estate" aired, criticizing the government for 58 downplaying the risks of calcium-channel blockers. Later, Stelfox et al (1998) surveyed authors who had published letters or articles concerning calcium-channel blockers during the debate. They found that 96 percent of the authors who supported the continued use of calcium-channel blockers had financial ties to manufacturers of calcium-channel blockers. In contrast, 37 percent of authors who raised or echoed concerns about the potential risks posed by calcium-channel blockers had such financial ties. Stelfox et al (1998) did not record the magnitude of these financial ties. Debate concerning the safety of calcium-channel blockers continues today. Manufacturers claim that the long-acting versions of these drugs are safe and effective, though conclusive evidence regarding their long-term impact on health status has yet to become available. Perhaps in an attempt to overshadow the controversy, long-acting calcium-channel blockers are among the most heavily promoted hypertension drugs on the market. The Impact of Hypertension Guidelines and Adverse News There have been three studies that attempt to evaluate directly the impact of JNC and CHS guidelines on physicians' prescribing habits (Siegel and Lopez 1997, McAlister et al 1997, Hi l l et al 1988). A priori, one might predict a relatively acceptable rate of compliance with hypertension treatment guidelines because the task involved is not complex (compared to many other procedures), its surrogate measure of efficacy is quickly observed and recommendations could easily be implemented on a trial basis. Despite these favourable conditions, none of the JNC and CHS guideline studies found a positive impact on the prescribing habits of physicians attributable to the publishing of the guidelines. Maryland-based doctors reported prescribing behaviours that were not different following the JNC III report in 1984 than before, despite reasonable awareness of and access to the guidelines (Hill et all 1988). Siegel and Lopez (1997) found that, contrary to the recommendations of the 1992 JNC V, prescriptions for calcium-channel blockers increased from 33 to 38 percent of all hypertension prescriptions in the US between 1992 and 1995. This occurred during the period in which adverse news about calcium-channel blockers was widespread in the media. Siegel and Lopez (1997) also found that ACE-inhibitors increased—from 25 to 33 percent of all hypertension prescriptions over the same period— contrary to recommendations of the JNC. Findings in Canada were similar. McAlister et al (1997) found that in 1995, doctors in Edmonton-based primary care offices and medical referral clinics prescribed diuretics or beta-blockers to only 23 percent of newly diagnosed hypertensive patients. Moreover, they found that only 43 percent of patients who were prescribed drugs other than diuretics or beta-blockers had documented contraindications to either first-line drug class (McAlister et al 1997). By inference, then, the percentage of patients for whom guidelines indicated diuretics or beta-blockers could have been 67 percent or more rather than the 23 percent who received them. Another Canadian study addressed the impact of the adverse news about calcium-channel blockers (Maclure et al 1998). They found that first-line prescribing of calcium-channel blockers for elderly patients in British Columbia fell gradually from 22 percent in 1994 to 15 percent in 1996. Maclure et al (1998) also documented a decreased use of 59 Table 6: First-Line Prescribing in 1996 for Elderly People in British Columbia in Relation to Relative Contraindications* Coexisting Illness n % RCfor Thiazides? % Prescribed Thiazides RCfor Beta-Blockers? % Prescribed Beta-Blockers % Prescribed ACEI, CCB or "Others" None 15189 41% No 41% No 11% 48% Depression 3993 11% No 42% Yes 15% 43% Asthma 2255 6% No 43% Yes 6% 51% PVD 365 1% No 38% Yes 7% 55% HL 1964 5% Yes 36% Yes 14% 50% Diabetes 1925 5% Yes 26% Yes 7% 67% Arrhythmia 949 3% Yes 33% No 19% 48% Gout 749 2% Yes 29% No 13% 58% Two or more 9863 26% ? 33% ? 11% 56% Total 37252 38% 11% 51% Source: Maclure et al (1998) table 3. * Maclure and colleagues regard relative contraindications as "...disorders that many physicians regard as reasons to avoid diuretics or beta-blockers, although evidence supporting some of these reasons may be weak" (Maclure et al 1998, p.352). ACEI = ACE-inhibitors. CCB = calcium-channel blockers. HL = hyperlipidaemia. PVD = peripheral vascular disease. RC = Relative Contraindication diuretics or beta-blockers as first line treatments, contrary to published guidelines. This documentation is significant because it was based on an audit of diagnostic codes found on medical billings pertaining to this patient population—the same population looked at in the present study. Table 6 summarizes data in Maclure et al (1998) regarding first-line prescribing in 1996 for elderly British Columbians in relation to relative contraindications. They regard relative contraindications as "...disorders that many physicians regard as reasons to avoid diuretics or beta-blockers, although evidence supporting some of these reasons may be weak" (Maclure et al 1998, p.352). Consistent with the findings of McAlister et al (1997), Maclure et al (1998) found that " . . . in 1996 physicians continued to prescribe CCBs or A C E inhibitors as first-line therapy to 42% of newly treated patients, contrary to guidelines" (p.952). Patients who did not have documented relative contraindications to diuretics received them only 40 percent of the time. Surprisingly, those who did have documented relative contraindications received diuretics approximately 30 percent of the time (Maclure et al 1998, p.352). From the profile of coexisting illnesses documented in Table 6, it appears that diuretics or beta-blockers could have been rationally prescribed as first-line therapy, in accordance to guidelines and popular (but sometimes unsubstantiated) beliefs about relative contraindications, to approximately 65 percent of patients. It would appear from these findings that, in Canada and the US, doctors are not prescribing diuretics and beta-blockers in accordance to guidelines. Moreover, the findings 60 of Maclure et al (1998) indicate that variations in what is prescribed to patients bear only a weak relationship to documented coexisting illnesses. In most studies, increased marketing of the newer drugs or physicians' desire to be perceived as being on the forefront of medicine are offered as potential explanations for the failure of practice to resemble guidelines. Analysis of Hypertension Treatment for BC Seniors: 1986 to 1996 Description of Data This analysis is based on a unique database extracted at the Centre for Health Services and Policy Research, U B C with permission form the British Columbia Ministry of Health and Ministry Responsible for Seniors. The database contains an observation for every antihypertensive prescription dispensed to beneficiaries of the B C Pharmacare Plan A in the period of January 1986 to December 1997. Pharmacare Plan A covers all community-dwelling B C residents who are 65 and over. A l l beneficiaries of Pharmacare Plan A are insured for the ingredient costs of prescription medicines, net of dispensing fees. The data used here contain ingredient costs only. Observations in the database consist of the following fields of information: (1) a case study identification number for each patient, (2) the patient's year of birth, (3) the date the prescription was filled, (4) the drug identification number (DIN) corresponding to the drug dispensed, (5) the quantity of the drug dispensed and (6) the ingredient cost of the prescription. The database consisted of 9.89 million observations representing antihypertensive prescriptions filled by over 390,000 patients. For the period of 1986 to 1996, the cost data reflect the transaction cost of the drugs dispensed. Pharmacare has recently implemented a new system of classifying drug costs that makes the data for 1997 unreliable. In 1997, the Pharmacare plan introduced a modified reimbursement scheme for the Plan A—the reference pricing program—the out-of-pocket costs of which appear to be missing from the Pharmacare database used here. Unfortunately, price information— including all price indexes—is therefore only reported for the period 1986 to 1996. The quantity information for all eleven years is accurate and used in the analysis of prescribing patterns. Observations were grouped into five broad categories, corresponding to the class of hypertension treatment the prescription pertained to. These categories were diuretics, beta-blockers, ACE-inhibitors, calcium-channel blockers and "other" drugs used in the treatment of hypertension. Appendix B lists the members of each drug class. While hypertension is the primary indication for the use of all of these drugs, each has uses in the treatment of other cardiovascular illnesses. These other conditions may be present in conjunction with, due to, or independent of hypertension. This complicates the problem of identifying trends in the pattern and cost of treatment for hypertension. Some prescriptions for these drugs may be for other conditions, conditions that may or may not be related to hypertension. To account for the existence of other illnesses, concomitant or otherwise, requires that one track the diagnostic history of patients before and after their use of hypertension drugs. For analytical purposes, patients taking hypertensive medicines could be divided into three possible groups. First, there are the patients who are being prescribed anti-61 hypertensives to treat hypertension in absence of any other cardiovascular illness. These are the simplest cases for measurement purposes because they are a reasonably homogeneous group. Few patients in this group would have medical reasons for starting therapy on drugs other than diuretics. Second, there are patients without high blood pressure taking anti-hypertensive drugs for the treatment of other cardiovascular illnesses, such as patients with angina pectoris ("angina") and no sign of elevated blood pressure.40 These patients would ideally be dropped from the sample when calculating aggregate prices and quantities for hypertension treatment. (Possible methods for identifying them are discussed below.) Finally, there are those patients with hypertension who also have records of other cardiovascular illness. These patients may be using hypertension drugs to treat hypertension in conjunction with the other illnesses. Although they are a less homogeneous group than the uncomplicated hypertension cases, calculating aggregate prices and quantities for the treatment of hypertension requires that one include these patients in the sample. The index formulae should take into account the fact that patients in this group may "need" specific types of hypertension drugs. National guidelines have identified the drugs of choice for many cases where patients have co-existing cardiovascular illnesses. For these cases, it is sometimes medically "appropriate" to prescribe specific hypertension drugs other than diuretics. To determine which group a patient belongs to is a difficult task. Pharmacare prescription records do not identify the diagnoses that underlie dispensed prescriptions. The only way to infer diagnoses is to link Pharmacare claims with secondary data from the Medical Services Plan of BC (the MSP) (and, where relevant, hospital separation records). By searching a patient's MSP records for diagnostic codes indicating cardiovascular illnesses, including hypertension, one can assign the patient to one of the three groups of hypertension drug users. This method would necessarily involve subjective judgements about what criteria for diagnostic and claims records constitute the cut-off lines between the three patient types. Nevertheless, it would be a reasonable means for eliminating those patients who show no record of having elevated blood pressures diagnosed, while distinguishing hypertensives with other cardiovascular complications from those with no complications. Unfortunately, diagnostic codes for physicians' services have only recently been added to the BC Linked Health Database at U B C . They were not available at the time that this study was being conducted. Diagnostic information in the MSP database had not previously been available because of questions regarding the reliability of the coding process. The MSP database contains only the diagnostic codes that accompany each physician's submitted billings for medical services delivered. There has been speculation that physicians code their billings in a haphazard fashion—e.g., using "catch-all" codes for most services— thereby reducing the coding accuracy. In light of recent evidence suggesting that physicians' coding may be more reliable than previously believed, the Ministry of Health granted U B C permission to add these data to the B C Linked Health 4 0 "Angina pectoris ("angina") is a recurring pain or discomfort in the chest that happens when some part of the heart does not receive enough blood. It is common symptom of coronary heart disease (CHD), which occurs when vessels that carry blood to the heart become narrowed and blocked due to atherosclerosis." From the National Institute of Health information pamphlet on Angina Pectoris, 1999. 62 Database. These data have only been made available to U B C for medical records dating back to the 1991/92 fiscal year. Because diagnostic code data were not available when the present study was conducted, the reported results depend upon assumptions about the use of hypertension drugs by non-hypertensives. Specifically, i f the proportion of patients using anti-hypertensive drugs in the absence hypertension is stable over time, the treatment regimens estimated with the unrestricted data used here would have the same movements as those for the sample that excluded patients who consume these drugs for strictly non-hypertension related illnesses. A follow-up study is scheduled to re-evaluate the cost of treating hypertension over a longer period than was studied here. The aim of that analysis will be to extend the current results and to compare findings based on patients belonging to the three possible groups of hypertension drug users. The comparison of results across sample types is only possible following 1991/92—when the MSP diagnostic codes became available. Some inference regarding the validity of the findings of the present study can be made through a comparison with some of the findings of Maclure et al (1998). For one component of their study, Maclure and colleagues used diagnostic information—accessed directly from the Ministry of Health—to remove all patients with markers of cardiovascular illness other than essential hypertension. The resulting sample of patients would be the most restrictive set of patient types—those treating hypertension in the absence of any other cardiovascular illness. For 1996, Maclure and colleagues report that the proportions of first-line prescriptions for patients in their restricted sample were as follows: 38 percent diuretic, 11 percent beta-blocker, 26 percent A C E inhibitor, 16 percent Calcium-channel blocker and 9 percent other drugs. The corresponding41 figures for 1996 from data used in the present study are 35 percent diuretic, 17 percent beta-blocker, 24 percent A C E inhibitor, 14 percent calcium-channel blocker and 10 percent other drugs. In comparison with the restricted sample of patients, the data used in the present study appears to over-sample beta-blockers as a proportion of first-line drug prescriptions. This is probably because beta-blockers are prescribed frequently for patients who have had a myocardial infarction and for those with angina. The restricted sample used by Maclure and colleagues will exclude all patients with indications of either medical history regardless of whether they have hypertension or not. Maclure and colleagues limited their sample in this manner to illustrate the overuse of calcium-channel blockers and ACE-inhibitors in clear cases of uncomplicated hypertension. From the perspective of aggregate price and quantity measurement, it is not appropriate to exclude cases of complicated hypertension. National guidelines note myocardial infarction as a complication in the treatment of hypertension, and recommend that beta-blockers be used to treat hypertension in these cases (CHS various years, JNC various years). Guidelines also specify that i f a patient has both angina and hypertension then the appropriate drug choice is a beta-blocker (CHS various years, JNC various years). In light of these recommendations, the data used here probably "over-sample" beta-blocker use in all cases of hypertension by an amount smaller than the 5 percentage points indicated in a comparison with sample restricted to uncomplicated cases only. It is, however, impossible at this time to conjecture the proportion of patients with records of angina or myocardial infarction that also suffered from hypertension. It was necessary to allocate multiple drug therapies captured in this study using an assignment hierarchy similar to the methods of Maclure et al (1998). 63 For the diuretic, ACE-inhibitor and calcium-channel blocker categories, the data used in this study appear to under-estimate the rates of first-line drug use in a uniform manner. The proportionate under-sampling of these therapies appears to be a direct function of the "over-sampling" of beta-blocker use. That the estimates of treatment incidence hold the same relative ranking and are reasonably proportionate to the Maclure findings lends some support for the claim that the sampling technique used here will not significantly bias the results. The reader should nevertheless interpret the results of the following analysis as preliminary. Elementary Price and Quantity Indexes The first step in the manipulation of the database involved aggregating data into elementary indexes that would be used in the construction of the aggregate economic indexes of price and quantity. Elementary indexes of price and quantity were constructed from this database by calculating the unit value and total quantity of drugs dispensed per quarter. Quarterly observations were chosen because they provide sufficient detail regarding price movements and facilitated the task of counting patients actively receiving hypertension treatment at any point in time (described below). Rather than aggregate to the DIN level, this analysis uses a unit value approach to aggregating across brand-name and generic versions of chemically identical products. This aggregation is based on a 9-digit American Hospital Formulary Service (AHFS) number assigned to each product. AHFS numbers are unique to a specific drug by strength and dosage form, but not by manufacturer or brand-name. The unit value approach to products with the same AHFS number treats them as identical goods by summing their total sales and dividing this by their total quantity to arrive at a common unit value (Diewert 1995). This "a-pill-is-a-pill" methodology imposes the assumption that brand-name and generic versions of a drug represent equal amounts of productivity—an assumption justified on the basis of the results found in the chapter on generic drugs. Elementary Indexes of Real Productivity National guidelines for the treatment of hypertension are used here as the basis of the health-outcomes based real productivity measures. Notwithstanding the increased involvement of pharmaceutical manufacturers in both the Canadian and American consensus processes, the CHS and JNC guidelines have been generated by recognized national bodies interested in promoting best medical care. Their recommendations were based on the best available scientific evidence regarding proven reduction in morbidity and mortality associated with high blood pressure. These recommendations are widely recognized and easily accessed by prescribers. They are taken here as indication of the most socially desirable treatment protocol. Clearly, the perspective that the treatment guidelines represent involves a certain degree of risk aversion or preference for the use of medicines with proven track-records of safety and efficacy. This was indicated by cautions regarding the use of newer medicines found in the guidelines (quoted at length above). This form of preference does not detract from the notion that these guidelines represent "real" objectives in the sense that it is, indeed, desired that medical practice conform reasonably with recommended treatments. This caution and conservatism is justified, as shown by the case of short-acting calcium-channel blockers. 64 Since 1977, the CHS and JNC national guidelines have consistently listed diuretics among (often alone as) the preferred first-line drugs to treat uncomplicated hypertension— especially for elderly, who respond best to thiazide diuretics. It therefore seems reasonable to assume that, unless contraindicated, diuretics are as "productive" or better, from a health outcomes perspective, than any other antihypertensive drugs. This assumption is in stark contrast with the implicit productivity weight that traditional economic indexes must put on diuretic drugs. The average calcium-channel blocker is 150 times more expensive per day of treatment than Hydrochlorothiazide, which is one of the oldest diuretics and the drug most frequently recommended by hypertension guidelines. Because of its low cost, Hydrochlorothiazide would receive 0.0067 times the weight placed on the average calcium-channel blocker in traditional Laspeyres or Paasche indexes of price and quantity. The health-outcomes based productivity measures used here are built on an assumption best described as "a-treatment-is-a-treatment." Real productivity will be defined according to the number of prescribed "patient-days" of therapy and by the number of "discrete patients" treated for hypertension, regardless of the class of hypertensives being used. This is analogous to the "a-pill-is-a-pill" approach to comparing equivalent brand-name and generic drugs. Whereas that approach assigned equal productivity values to chemically equivalent products, the a-treatment-is-a-treatment approach assigns equal values to days of treatment (or patients treated) with chemically different drugs that treat the same condition. The assumption that all treatment regimens are of equal "value" may appear troubling due to the fact that specific drugs may be contraindicated for patients with certain coexisting illnesses. Thus, to those patients, the contraindicated drugs are of no value—or of negative value! Diversity in medical "needs" makes it difficult to assign average values to any medical procedure or drug, making health-related productivity measurement difficult. Fortunately, perhaps, guidelines for hypertension can be interpreted as a fixed coefficient production function. In this light, the optimal relative demands for inputs from different drug classes are determined by the prevalence of relative risk factors in the treated population. Observable patient characteristics indicate when treatments are or are not appropriate first-line therapies. It is therefore possible to gauge the optimal vector of first-line inputs into the hypertension-related social welfare "production function" based on reasonably objective measures. The findings of McAlister et al (1997) and Maclure et al (1998) indicate that diuretics or beta-blockers would be appropriate first-line treatments for approximately two thirds of patients. Movement towards a treatment profile where diuretics or beta-blockers are prescribed for approximately two thirds of newly treated hypertension patients would accordingly indicate an improvement in sector productivity, from a health production perspective. Movements in the opposite direction could be judged as deviations away from the socially preferred mix of treatment inputs and, therefore, a decline in sector productivity. It should be noted that assuming that all drugs used over the course of study are as effective as diuretics might be considered generous in light of virtually all clinical evidence available during the period under study. Specifically, the a-treatment-is-a-treatment assumption does not discount purchases for the uncertainty associated with taking medicines that did not have a body of scientific evidence to substantiate claims about safety and long-65 term efficacy. Such gambles appear to have resulted in exposing patients to increased risks of death in the case of short-acting calcium-channel blockers. The simplified assumptions implicit in the methodology used here are necessary because the available database does not include information regarding comorbidities. Despite their rather gross nature, these assumptions are probably accurate within reason— especially given that the variations in antihypertensive prescribing for elderly patients in British Columbia reported by Maclure et al (1998) bore only a weak relationship to documented coexisting illnesses. Changes in the profile of first-line prescribing and continuing drug use will be evaluated to determine whether prescribing moved toward or away from the recommended drugs. Calculating Drug Treatment Exposure: Patient-Days Two measures of drug exposure and, thus, elementary indexes of real health production were calculated from the observed market transactions in the database. The first was based on the number of days of maintenance therapy represented by the quantity of physical pills and tablets purchased in each quarter. The most common method of determining drug utilization levels by populations is to use dose standardization methods to calculate exposure by standardized patient-days (Merlo et al 1996). Simply, this involves the division of the units of drugs purchased by a standardized dose. Among the possible candidates for standard dosages are the minimum marked dose, the defined daily dose and the prescribed daily dose. The minimum marked dose is the minimum amount of a drug that will give the desired therapeutic effect. It is determined a priori and typically equals the smallest dose of the drug marketed by a manufacturer (Merlo et al 1996). This measure suffers several disadvantages, including the fact that it ignores drug dose titration that occurs in hypertension care, rendering it inappropriate for this study. The defined daily dose is an international standard developed by the World Health Organization for comparing drug utilization across countries and regions. A defined daily dose is determined a priori as the assumed average daily dose of a drug for use in its main indication by adults (Merlo et al 1996). Despite international appeal, the greatest weakness of defined daily doses is that they are updated only periodically to account for changes in average doses used for the main indication of a drug. In the case of certain hypertensives, diuretics in particular, the recommended daily dose has fallen dramatically over the past ten years. A n additional weakness of defined daily doses is that they are not necessarily "defined" for all products available in the Canadian marketplace, leaving estimation to the analyst. The third candidate for dose standards is the prescribed daily dose. This is the average daily dose of a drug prescribed to a given population. Prescribed daily doses are not assigned a priori; rather, they must be calculated for each study. The advantage of the prescribed daily dose is that it reflects actual prescribing habits, thereby accounting for changes in dosage strengths, whether recommended or not. Its primary disadvantage is that it must be estimated from available data. Despite difficulties posed by estimation, this analysis uses prescribed daily doses because these measures will capture the potential cost-reducing effects of prescribing lower dosages of antihypertensive drugs over the period of analysis. 66 Prescribed daily doses were calculated by tracking patients who had repeat prescriptions for antihypertensive medications from the same broadly defined therapeutic class. More than 90 percent of prescriptions in each category were repeat prescriptions— about 80 percent of these were for patients who refilled their prescriptions at least every 90 days. The estimated prescribed daily dose was the average number of pills per day between repeat prescriptions, calculated for each drug type on a quarterly basis. A certain amount of "noise" could be expected in this methodology—due, for example, to intermittent hospitalization of patients. Such noise will make the calculated prescribed daily doses imperfect representations of the true therapeutic dosages used by the recipient population or intended by the prescribing physicians. Nevertheless, there are no reasons to suspect that such noise would change systematically across treatment types or over time. Thus, there should be no expected aggregate measurement biases using this averaging technique. Calculating Drug Treatment Exposure: Discrete Patients The second measure of drug exposure used in this study is based on the number of discrete patients receiving hypertension therapy during a given quarter. Indexes using measures of cost per patient treated were calculated because there is a potential for patient-day exposure measures to bias the price and productivity indexes over the period of study. This stems from the fact that recommended treatment protocols changed from step care models, wherein patients would be prescribed additional drugs i f therapies were only partially effective, to eclectic choice models where substitution from mono-therapy to mono-therapies was endorsed. Because a single discrete patient can consume more than one patient-day of drugs per day, the patient-day quantity and price indexes may be biased. This bias will depend on changes in the prevalence and cost of multiple drug therapy use. If fewer patients were prescribed multiple-therapies, patient-day price indexes will overstate costs because more discrete patients could be treated with a given number of patient-days. To construct the indexes of discrete patients, every patient in the database was assigned to one of eight categories for every quarter of the study period. The categories corresponded to the type of antihypertensive drug received by the patients during the given quarter. The categories were (1) no antihypertensive drugs, (2) diuretics only, (3) diuretics plus any other drug, (4) beta-blockers only, (5) ACE-inhibitors only, (6) calcium-channel blockers only, (7) other antihypertensive drugs only, and (8) combinations of any two or more non-diuretic drugs. Diuretics in combination with other drugs were isolated because the use of diuretics received the majority of recommendations as first-line treatment in hypertension guidelines. Isolating diuretics has no impact on the aggregate indexes themselves, but facilitates evaluation of prescribing habits. Quarterly indexes of the total number of patients and the average cost per patient in each of these categories were constructed. Findings Price Indexes Indexes for all classes of drugs combined are illustrated in Figure 8. The economic index is a chained Laspeyres index, with quantity weights updated each period. It represents an estimate of the true sector-specific cost-of-living index within the traditional economic approach to measurement. Under the assumptions of that approach to measurement, the 67 3.00 2.50 2.00 1.50 1.00 0.50 0.00 Figure 8 Alternative Indexes of the Price of Hypertension Treatments BC Pharmacare Plan A 1986 to 1996 -Cost per Discrete Patient Index -Cost per Patient-Day Index -Traditional Economic Price Index i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—p—i—i—1~ 86:1 87:1 88:1 89:1 90:1 91:1 92:1 93:1 94:1 95:1 96:1 Laspeyres index is an upper bound on the "exact" sector-specific cost-of-living index for two reasons. First, Laspeyres indexes theoretically bound exact cost-of-living indexes from above. Second, new goods are "linked" into the Laspeyres index used here during their second period of availability without adjustment. Thus, only changes in their price levels are captured by the index. That is, the index ignores the "new-goods" effect of entrants. Since numerous ACE-inhibitors and calcium-channel blockers entered the market over the course of study (see Appendix C), this will result in an index that overstates price changes provided the assumptions of the traditional economic approach to measurement are satisfied.42 From 1986 to 1996, the average growth rate of the economic index was 1.5 percent per annum. The cost per patient-day index illustrated in Figure 8 uses the estimated number of patient-days of maintenance therapy (described above) as the basis for productivity measurement. Similarly, the cost per discrete patient index uses the number of discrete patients receiving hypertension treatment as the basis for productivity measurement. This measure accounts for changes in the use of multiple-therapy over time. Growth rates for the cost per patient-day and cost per discrete patient indexes were 9.5 and 10.5 percent per annum, respectively, from 1986 to 1996. The fact that costs per patient increased faster than the cost per patient-day indicates an increase in the use of multiple drugs per patient over the period of study. This is an unanticipated change because the national prescribing guidelines gradually encouraged more mono-therapy, not multiple-therapy, over the period. 4 2 As discussed in the chapter on generic goods, under the assumptions of the traditional economic approach to measurement, the reservation price technique for capturing the "new goods" effect of an entrant always captures a decline in real prices. Otherwise, the new good would not be purchased when it was introduced. 68 There was a statistically significant (at p=0.01) change in the trend of all indexes beginning at 1994.43 The 1994 date marks the beginning of the Low Cost Alternative plan (which increased the incentive for pharmacists' to dispense generic drugs) and the time when major media events covered the calcium-channel blocker controversy. A l l indexes grew at a faster rate before 1994 than after. The rate of change in the health outcomes indexes was not significantly different from zero after 1994, while the economic index fell at a rate that was slow but significant (at p=0.05). As is clear from Figure 8, the traditional economic index is far lower than the health outcomes based indexes. The difference between health outcomes based price indexes and the traditional economic index increased rapidly from 1986 to 1994, then stabled off. Increases in the health-outcomes based indexes relative to the economic index is indicative of changes in consumption patterns from drugs that were low-cost to drugs that were high-cost on a per-treatment basis. Such changes would not be captured by the economic index because, when the assumptions of the traditional economic model are met, increased use of high-cost products would only occur i f their relative price reflected their relative productivity. Thus, switching to high-cost therapies instead of low cost therapies is captured by the economic index as an increase in the quantity of output purchased. This is revealed in Figure 9, which illustrates the quantity indexes that are dual to the price indexes listed in Figure 8. The quantity indexes illustrated in Figure 9 are obtained by deflating expenditure growth by the respective price indexes. It is noteworthy that Pharmacare's annual spending on hypertension drugs nearly quadrupled from $15.8 million in 1986 to $58.7 million in 1996. Regression analyses of these trends are listed in appendix D. 69 The traditional economic index of aggregate quantities attributes most of the observed change in expenditures on hypertension drugs to changes in the quantity of output generated through the market transactions described by the underlying data. The reason for this is that the tradition economic approach weighs changes in the use of drugs by the share of expenditures on them. Thus, increased use of high-cost hypertension drugs is said to generate more output than increased use of low-cost drugs. Recall, however, that the metric of sector-specific output for the traditional economic indexes is a utility-theoretic concept of social welfare that is "revealed" by consumption patterns. The health outcomes based indexes, on the other hand, do not assign units of productivity based on "preferences" revealed by the level expenditures on particular drugs. Rather, the productivity metrics of patient-days of treatment and discrete patients treated are based on the actual number of patients receiving hypertension drugs and the actual number of days that they receive them. The indexes based on these measures of real output grew by 63 and 80 percent, respectively, over the period of study. These changes are modest by comparison to the 280 percent increase (a near quadrupling) in the traditional economic index of quantity. As with the aggregate price indexes, the difference between the health outcomes based quantity indexes and the traditional economic quantity index reflects an underlying increase in the use of higher priced therapies rather than lower priced ones, as well as the use of more therapies per patient treated. 70 Table 7 :Treatment Classification for All Seniors Receiving Antihypertensive Drugs. 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 N* 91986 92663 96648 101832 108673 14424 19838 126157 129472 132459 141384 % Persistent"1" 86% 93% 93% 92% 91% 91% 90% 90% 89% 89% 88% Diuretics 35% 33% 30% 26% 23% 20% 18% 16% 15% 14% 14% Diuretics Plus 22% 21% 20% 18% 17% 16% 15% 15% 15% 15% 16% Beta-Blockers 20% 19% 19% 18% 17% 16% 15% 14% 13% 13% 13% ACE-inhibitors 2% 3% 6% 9% 12% 14% 17% 18% 19% 20% 21% CCB 8% 10% 13% 16% 18% 20% 21% 22% 23% 22% 20% Other 7% 6% 5% 4% 4% 3% 3% 2% 2% 2% 3% ND Combo 6% 7% 8% 9% 10% 11% 12% 12% 13% 13% 14% Annual average number of patients per quarter receiving drugs to treat hypertension. +Seniors who filled prescriptions every quarter or intermittently for two consecutive years or more. CCB = calcium-channel blockers. ND = Combination of non-diuretic drugs. To determine whether the substitution of high cost-treatments for low-cost treatments should be considered an improvement in real productivity, it is necessary to look at how prescribing patterns evolved over the period of analysis. If the movements towards higher cost medicines were coincident with movements toward treatment protocols recommended in national guidelines, then there may be grounds to favour the traditional economic measures of price and quantity in this sector. If otherwise, the health outcomes based indexes should be considered closer to the true social costs and real outcomes generated by transactions in this market segment. Aggregate Drug Exposure Profiles Table 7 shows the annual average number of discrete patients receiving drugs to treat hypertension during each quarter, along with the distribution of these patients across different treatment regimens. The table also indicates what percentage of the drug recipients were persistent in their treatment for two years following their first prescription during the period of study. Persistence was defined as those who filled a prescription for any kind of hypertension drug every quarter for at least eight consecutive quarters, or those did so during at least two quarters of each of two consecutive years. In interpreting this table, and all subsequent tables, attention must be paid to the difference between the measures for 1986 and 1996 as compared to the years in-between. The data used here track the purchase of medicines by many people who used hypertension treatments before 1986 and many who used them after 1996. This wil l make it difficult to determine precisely the nature of patients "persistence" at the tails of the observation period—though transaction records for 1997 facilitate this for the 1996 observations. In addition, since data for prescriptions filled in 1986 account for many patients who would had long been using hypertensives, there is no way of determining which of them recently "started" hypertension treatment using the available database—treatment "starts" are, however, estimated for all other years (more below). 71 Table 8: The Average Cost Per Patient Treated, Grouped by Drug Class 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Diuretics $34 $37 $39 $50 $42 $45 $47 $46 $40 $34 $30 Diuretics Plus $267 $332 $375 $511 $453 $496 $529 $543 $514 $494 $502 Beta-Blockers $125 $152 $168 $221 $188 $197 $204 $202 $181 $161 $152 ACE-inhibitors $367 $376 $372 $464 $386 $401 $416 $414 $371 $373 $391 CCB $399 $467 $492 $635 $529 $554 $572 $568 $554 $536 $543 Other $105 $121 $130 $171 $153 $164 $181 $196 $187 $207 $249 ND Combo $519 $633 $698 $935 $817 $873 $915 $923 $872 $838 $847 CCB = calcium-channel blockers. ND = Combination of non-diuretic drugs. Since diuretics were the most common hypertensive treatment used in the 1970s and early 1980s, it is not surprising that a large percentage of patients receive diuretics in 1986. Once tritrated on a therapy, patients typically remain on that therapy (a proposition examined below). The exposure to diuretics among the treated population of elderly hypertensives in BC fell dramatically over the course of this study. Approximately 57 percent of those receiving hypertension drugs in 1986 received diuretics. By 1996, this figure fell to approximately 30 percent. Exposure to treatment with beta-blockers also fell over this period, from 20 percent to 13 percent. The declining exposure to these drugs was mirrored by increased exposure to ACE-inhibitors and calcium-channel blockers, which rose from 2 and 8 percent to 21 and 20 percent, respectively. Also notable was the increased exposure to multiple-therapies that did not include diuretics—from 6 percent to 14 percent. Of these multiple non-diuretic therapies, the percentage involving beta-blockers fell from approximately 75 percent in 1986 to 50 percent in 1996 (data not shown). Combined, and including combination drug use, the use of diuretics and/or beta-blockers fell from 80 percent in 1986 to 50 percent in 1995. Appendix E contains regression results for the treatment classifications of patients. There were statistically significant (at p=0.01) changes in the trend of drug use following 1994. Exposure to calcium-channel blockers fell over this period, whereas it had increased before 1994. The use of diuretics and beta-blockers increased, whereas they had decreased before 1994. And, the use of ACE-inhibitors grew at a faster rate after 1994 than before. Exposure to combination therapies increased over the period of study, which explains the difference between the indexes based on discrete patients and patient days. Table 8 lists the average cost of treatment in the various categories as used by the patients listed in Table 7. As shown, the cost of treating patients for a year varied dramatically across categories and over time. (Appendix F describes price measurement within the five major therapeutic subclasses—combination therapies are not discussed there.) The cost of treating patients with diuretics alone was only a fraction of the cost of treating patients on any other drug or combination. Calcium-channel blockers, followed by ACE-inhibitors, were the most expensive forms of mono-therapy prescribed to this patient population. By far the most expensive courses of therapy overall involved combinations of non-diuretic drugs. The cost of treatment in this category also rose dramatically with the 72 increased use of ACE-inhibitors in combination with calcium-channel blockers (with or without additional agents)—from about 1 percent to 30 percent of combination therapies (data not shown). Given the changes in treatment profiles listed in Table 7 and the difference in the cost of these treatments listed in Table 8, it is not surprising that the price indexes measured in terms of patient-days and patients treated rose dramatically over time. The question remains whether these changes reflect improvements in therapy or not. First-Line Treatments The source of the decline in diuretic and beta-blocker exposure among all seniors receiving hypertension drugs was a dramatic change in the profile of first-line drug treatments. Table 9 lists the distribution of "first-line" hypertension treatments for patients aged 66 and older who received drugs to treat hypertension for the first time during each quarter. Only patients over 66 years of age were selected to be sure that, for at least one year before the first recorded prescription, these patients had not received hypertension drugs in BC. (Thus, first-line treatment profiles could not be constructed for 1986.) Furthermore, only patients who continuously or intermittently filled hypertension prescriptions over at least a two-year period are listed in Table 9. This is because approximately 25 percent of all patients who receive their first hypertension treatments do not obtain another again for at least two years (the chosen length of time for following up on patients in this study). Combined, patients who do not persist with therapy account for a small but growing share of total costs. These non-persistent patients accounted for less than 1 percent of costs in 1986 and 3 percent of costs in 1996. While the treatment profiles of non-persistent patients includes all therapy types, the share of non-persistent patients that use diuretics was constantly greater than that of patients who were persistent with therapy (See appendix G). This should not, however, be interpreted as proof that diuretics do not work. Since these patients do not "switch" to other hypertension treatments, it appears that they are receiving these drugs for non-chronic conditions or that they are discontinuing therapy for clinical reasons other than drug tolerability (J. Wright, personal communication, Dec 1999). 73 Table 9: Classification of First-Line Treatments for Persistent* Patients Aged 66 and Older 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996+ N 8,405 7,423 8,156 9,143 7,389 7,372 7,762 8,153 8,491 10,214 Diuretics 36% 31% 25% 24% 19% 18% 18% 17% 20% 23% Diuretics Plus 11% 9% 9% 8% 6% 6% 7% 6% 7% 7% Beta-Blockers 19% 16% 16% 15% 14% 13% 13% 13% 14% 16% ACE-inhibitors 6% 13% 17% 19% 26% 27% 27% 29% 28% 26% CCB 19% 23% 24% 24% 27% 26% 24% 24% 18% 13% Other 5% 3% 3% 3% 2% 2% 2% 3% 5% 8% ND Combo 5% 5% 6% 7% 6% 7% 7% 8% 8% 7% "Those who filled prescriptions every quarter or intermittently for two consecutive years or more. """Includes patients only persistent for one year. CCB = calcium-channel blockers. ND = Combination of non-diuretic drugs. First-line treatment was determined by the drugs used during a patient's first quarter of hypertension treatment. This will account for some early-treatment volatility. When newly treated patients are immediately switched from one treatment to another, they will show up in the combination categories. The data in Table 9 clearly indicate why there has been gradual erosion in the share of hypertensive patients being treated with diuretics and beta-blockers over the period of study. Diuretics or beta-blockers—alone or in combination with other drugs—were used as first-line treatments for 68 percent of newly diagnosed elderly hypertensives in 1986. This figure is consistent with the estimated two thirds of seniors for whom these drugs would have been appropriate (see discussion above). Their use as first-line treatments fell to a low of 38 percent in 1994, and then rose to 48 percent by 1996. Statistical analysis of the trend in first-line treatments (listed in appendix H) indicates that there was a significant decline in the number of patients being treated with diuretics or beta-blockers prior to 1994. There was a significant rise in the number of patients treated with ACE-inhibitors and calcium-channel blockers over this period. The total number of patients being first treated for hypertension did not increase over the period of study (the 1996 figure in Table 9 is "large" because it is based on a looser definition of persistence). First-line treatment with calcium-channel blockers declined following 1994, while first-line use of all other drugs increased from 1994 to 1996. The rate of increase in the use of ACE-inhibitors as first-line treatments after 1994 was not significantly different from that before 1994. The full effect of the changes in first-line prescribing is felt over the long-run. Because hypertension treatment requires continuous drug maintenance, today's flow of new patients will affect current and future stocks of patients on drug therapy. The high exposure to diuretics among the elderly population early in the period of study—depicted in Table 7—was a result of the "stock" of long-treated patients who likely had been on diuretics for many years before they "aged into" the database used here. The dramatic changes in first-74 line drug use over the period of study—depicted in Table 9—will influence the average cost of hypertension treatment in B C for many more years. Therapeutic Trajectories To investigate how patterns of drug use have changed for those taking drug therapy, "therapeutic trajectories" were constructed for all newly treated patients who received continuous treatment for their hypertension. This allows for tests of the proposition that once started on therapy, patients are likely to continue on it. It also offers an indirect test to see i f increased use of newer drugs reflected increased productivity not captured by the cost per patient-day and cost per discrete patient indexes. During the 1980s and 1990s, while guidelines consistently recommended diuretics and beta-blockers as first-line therapies, doctors may have prescribed the newer drugs because their patients had voiced dissatisfaction about the side-effects of older medicines. There was no evidence that newer hypertension drugs were more effective, and they were far more expensive than older drugs, so their selection could only be "rationalized" based on perceptions of superior tolerability. If better tolerability was to justify the higher expense of newer drugs, then the persistence on the newer therapies should have been significantly higher than that on the older drugs. After being prescribed older drugs, one might conjecture that hypertensives would voice concerns, thereby initiating a change in therapy, more often than patients who received newer drugs—if the older medicines were less tolerable than new ones. 75 Table 10: Transitions for 66+ Year Old Patients Who Refill Continuously For Two Years 1987-1989 First Line Year One Year Two D D+ B A C o ND D D+ B A C o ND Diuretics 24% 74% 11% 3% 5% 5% 1% 1% 67% 11% 4% 9% 7% 1% 2% Diuretics Plus 11% 9% 53% 8% 10% 11% 3% 5% 11% 43% 9% 12% 14% 3% 8% Beta-Blockers 18% 2% 9% 70% 5% 6% 1% 8% 3% 9% 61% 6% 10% 1% 10% ACE-inhibitors 13% 2% 8% 3% 74% 5% 1% 7% 3% 9% 2% 69% 7% 1% 8% CCB 24% 1% 5% 2% 3% 83% 0% 6% 1% 6% 3% 4% 78% 0% 8% Other 3% 3% 10% 3% 6% 7% 62% 9% 4% 11% 4% 11% 10% 51% 8% ND Combo 6% 1% 8% 11% 11% 20% 2% 46% 2% 9% 10% 11% 25% 2% 41% N 13,280 20% 13% 16% 14% 26% 3% 8% 19% 13% 14% 15% 27% 3% 9% 1990-1992 First Line Year One Year Two D D+ B A C O ND D D+ B A C O ND Diuretics 15% 69% 9% 2% 9% 7% 1% 2% 62% 11% 3% 12% 8% 1% 3% Diuretics Plus 8% 8% 51% 5% 15% 11% 2% 8% 9% 44% 6% 16% 15% 2% 8% Beta-Blockers 14% 1% 7% 71% 5% 7% 0% 10% 2% 8% 62% 6% 10% 1% 12% ACE-Inhibitors 25% 2% 7% 2% 74% 8% 1% 7% 3% 9% 3% 66% 11% 1% 8% CCB 27% 1% 4% 2% 5% 82% 0% 6% 1% 6% 2% 6% 76% 0% 9% Other 2% 1% 5% 1% 7% 9% 68% 9% 4% 6% 3% 10% 11% 56% 10% ND Combo 8% 0% 6% 10% 12% 20% 2% 50% 1% 7% 10% 13% 23% 1% 45% N 13,924 12% 10% 13% 24% 29% 2% 10% 12% 11%. 12% 23% 29% 2% 11% 1993-1995 First Line Year One Year Two D D+ B A C O ND D D+ B A C O ND Diuretics 16% 67% 12% 3% 11% 4% 0% 2% 60% 14% 3% 13% 6% 1% 3% Diuretics Plus 8% 10% 45% 7% 17% 14% 1% 6% 11% 41% 7% 18% 14% 1% 8% Beta-Blockers 14% 2% 6% 72% 5% 6% 1% 9% 2% 8% 64%, 6% . 8% 1% 11% ACE-Inhibitors 29% 3% 7% 2% 74% 7% 1% 6% 4% 9% 2% 67% 9% 1% 8% CCB 23% 2% 6% 3% 5% 77% 0% 8% 3% 7% 4% 7% 69% 0% 10% Other 3% 2% 5% 1% 4% 3% 80% 5% 3% 4% 1% 6% 5% 72% 8% ND Combo 9% 1% 5% 12% 14% 18% 1% 49% 3% 7% 12% 15% 19% 1% 44% N 14,361 13% 10% 13% 28% 24% 3% 10% 13% 11% 12% 27% 23% 3% 11% ND = Combination of non-diuretic drugs. To investigate this possibility, transition matrices were constructed for newly treated hypertensives over the age of 66. They are listed in Table 10. For purposes of parsimony, patients were grouped into cohorts based on when they started treatment. Three matrices are listed here: corresponding to the cohorts that started treatment in the periods 1987-1989, 1990-1992 and 1993-1995. (Recall that the classification system used prevents first-line estimation in 1986 and persistence estimation for 1996.) The leftmost columns of each matrix list the proportion of patients that started in each treatment category along with the total number of patients who started treatment during the three-year period. This is followed by matrices identifying the proportion of patients from each first-line treatment group that went on to use other therapies after one year and two years. The diagonal elements in bold 76 print identify those patients that continued using drags from the same treatment category after one or two years. i The diagonal figures in Table 10 help to substantiate the claim that once started on a therapy, patients are likely to continue using that therapy. This is true, at least, for patients who start on most mono-therapies. Between two-thirds and three-quarters of patients who are on a mono-therapy of diuretics, beta-blockers, ACE-inhibitors or calcium-channel blockers during the first quarter of their treatment remain on those therapies after one and two years of treatments, conditional on the fact that their hypertension is treated persistently. Patients who begin treatment on a multiple-therapy (with diuretics or otherwise), are far more likely to switch to another course of therapy after one or two years. Using multiple drugs within the first quarter is likely a sign of treatment intolerance or inadequacy, so it is not surprising that these patients switch to other courses of therapy. Another trend to note from the matrices in Table 10 is that treatment regimens appear to have become gradually more volatile over time. This is somewhat consistent with the increasingly eclectic nature of national hypertension guidelines. At first glance, the diagonal figures in Table 10 might indicate that diuretics had persistence rates that were worse than other drugs during the 1990s. However, the data in this table indicate that a step-care approach to diuretic therapy is being used. Between 9 and 14 percent of those who begin treatment with diuretics are advanced to the "diuretic plus other drugs" combination category. This would appear to imply that the diuretic drags were tolerable for these patients, but blood-pressures were not adequately controlled. Similar step-care movements can be inferred from the transition patters of patients on beta-blockers, ACE-inhibitors and calcium-channel blockers. Conspicuous in its absence is a major difference between the treatment trajectories of patients starting on diuretics and beta-blockers and those of patients starting on A C E -inhibitors and calcium-channel blockers. After one or two years, patients switch to mono-therapy of the newer drags more frequently than mono-therapies of the older drags, but the net effects of switching are small, even after two years. High rates of treatment persistence, combined with evidence of step-care are not consistent with the hypothesis that diuretics and beta-blockers are substantially less tolerable than ACE-inhibitors and calcium-channel blockers. This should not be surprising in light of the fact that the best available clinical information suggests that low-dose diuretics are as tolerable as or better tolerated than any other group of antihypertensive drags (Wright et al 1999). Advertising and Noncompliance with Guidelines The findings from the previous section suggest that prescribing patterns—driven in large part by trends in first-line prescribing—moved toward the use of newer drags, contrary to prescribing guidelines. Causes for noncompliance with guidelines may include time constraints, financial incentives and possibly patient demand. As alluded to above, the most frequently cited possible cause of continued non-adherence to hypertension prescribing guidelines is advertising. It is certain that the differences in product ages among the hypertension classes are important determinants of market dynamics and especially advertising levels. 77 The lifecycle of an average drug product can be roughly broken into two phases, corresponding to its patent status. In the initial phase, when a product is patented, it is heavily marketed because the resulting sales are captured exclusively by the patent holder. This characterizes many of ACE-inhibitors and calcium-channel blockers during the 1980s and 1990s. After the patent on a drug expires, generic competitors often enter the market. Generic entry results in lost market share—in units and dollar-volume—if consumers (or pharmacists acting on their behalf) choose lower-cost generics. Subject to this form of competition, brand-name companies' incentive to advertise falls. The total market—in units—for the pioneering product and its generic competitors wil l likely decline as advertising declines. This is particularly likely when competing substitutes are patented and therefore promoted. Advertising efforts by newer competitors can dramatically alter the prescribing habits of physicians.44 This may explain the decline in the use of diuretics and beta-blockers. Detailed marketing information was too costly to obtain for this study. However, it is known that ACE-inhibitors and calcium-channel blockers were the most heavily marketed hypertension drugs in the late 1980s and through the 1990s. Due to the high cost of market research data, Wang et al (1999) conducted their own page-audit of the New England Journal of Medicine (NEJM) to determine how heavily alternative hypertension drugs had been promoted between 1985 and 1996—the same period of analysis for this study. The N E J M was chosen as representative of reputable medical journals. They found that calcium-channel blockers had been the most heavily promoted drugs, and that advertising for these products increased steadily over the decade. In 1985, ads for calcium-channel blockers accounted for 4.6 percent of all drug ads in the N E J M ; by 1996, they accounted for 26.9 percent of all drug ads in the journal. This made calcium-channel blockers the most heavily promoted of all classes of drugs advertised, including the entire category of antibiotics combined. Advertising for ACE-inhibitors increased and then declined again over this period. ACE-inhibitor ads accounted for approximately 4 percent of drug ads in the N E J M during 1985 and again in 1996. In contrast, advertisements for diuretics and beta-blockers declined steadily over the period analyzed. They accounted for 4.2 and 12.4 percent of drug ads in 1985 and 0.8 and 0 percent of drug ads in 1996, respectively (Wang et al 1999). In financial terms, makers of calcium-channel blockers spent $54 million (US) advertising their products in the US during 1997, while only $1.4 million was spent advertising beta-blockers and less on diuretics (IMS statistics cited in Anonymous 1999). In Canada, six brands of ACE-inhibitors (five of which were launched during the 1990s) and two brands of calcium-channel blockers were on IMS Health's list of top 50 products by promotional expenditures for 1998. No diuretics or beta-blockers made that list. Drug Advertising in Economic Literature The economics literature has traditionally focussed on narrowly defined measures of the potentially anti-competitive effects of drug advertising (Leffler 1981, Hurwitz and Caves 1988, Riso 1999, Rubin and Schrag 1999). Most frequently, economists focus on indicators of barriers to entry—particularly the frequency and impact of generic entry—and measures 4 4 The ability to shift markets from one product type to another by advertising is so dramatic that brand-name firms will even cannibalize their own markets by launching newly patented modifications of their older products to continue the profitable life of the original innovation (Morgan 1998, Anderson 1997). 78 of reduced price competitiveness—such as decreased cross-price elasticity of demand. While some economists acknowledge the possibility that advertising can be misleading or even fraudulent, none has investigated how consumption patters may be affected by this. Leffler (1981), for example, defines "persuasion" as that which substitutes emotional decisions for rational, evaluative decisions, and argues that "...positive and normative analysis should therefor be prefaced by the particulars of the products advertised, the message delivered, and the buyers addressed" (Leffler 1981, p.46). Leffler (1981) then goes on to assume that advertising always relays truthful information in his empirical analysis of the impact of drug advertising. Several other authors—supporting both sides of the drug advertising and competitiveness debate—have made similarly optimistic assumptions about the "information" contained in advertising signals (Hurwitz and Caves 1988, Riso 1999, Rubin and Schrag 1999). Advertising and Information The health services literature regarding advertising is fare more extensive than the economic literature. It also clearly calls into question the accuracy and completeness of the informational content of drug advertising. In numerous studies, advertising has been shown to have a significant impact on prescribing patterns, often ranking as the most influential form of "information" (Hemminki 1975, Haayer 1982, Orlowski and Wateska 1992, Caudill et al 1996, Lexchin 1989 & 1993, Spingarn et al 1996, Wivell and O'Fallon 1992). Moreover, studies indicate that increased reliance on company-sponsored information is positively correlated with prescribing inappropriateness (see Lexchin 1997 and references therein). Noted physicians acknowledge that they are as susceptible as anyone to the "obvious and not so obvious" forms of persuasion used by drug companies (Squires, 1993, p. 1391, Avorn 1996), which is sensible because were it not the case, drug companies would not spend more on advertising than they do on research and development—as they do now. Most drug advertising serves the purpose of brand building and name-recall. This is clearly the purpose of reminder ads and gift giving. Drug marketing also takes forms that more closely resemble informational services. Among other such activities, drug companies sponsor continuing medical educational programs, consensus conferences and symposia. Drug company representatives (detailers) are also seen as an educational tool, one that accounts for an immense amount of marketing activity and a majority of promotional expenditures in the pharmaceutical industry. According to IMS Health, Canadian drug companies detailed practicing physicians 3.2 million times in 1998—conservatively, about 50 details per practicing physician.45 Canadian physicians rank interactions with drug representatives as either the first or the second most important source of prescribing information (Lexchin 1993). Ostensibly, the claims made by drug representatives are regulated, but the personal nature of details makes effective regulation impracticable (Lexchin 1989). What is known about the "information" content of details suggests that profit motives outweigh the motive to educate because benefits of drugs are typically exaggerated and risks seldom mentioned (Lexchin 1989 & 1993, Hodges 1995). Based on the assumption that all 63,000 practicing physicians are detailers' targets. 79 The second most important channel of drug promotion is medical and professional journal advertising. IMS Health estimates that Canadian drug companies spent almost $69 million on journal advertisements in 1998, $14 million of which was for cardiovascular medicines. Journal advertisements are left to the auspices of self-regulation by the industry, resulting in the implicit endorsement of questionable marketing practices (Lexchin 1997). For example, journal advertising typically involves full-page glossy advertisements found at the beginning of the journal and amidst the articles. "Information" regarding the appropriate indications, doses, cautions, contraindications, side-effects and risks printed on separate pages, usually at the very back of the journal. This information is invariably printed in compressed fonts—often seven-point, which looks like this, but sometimes six-point, which looks like this. Because it is generally inconvenient to search for and then read the information that "accompanies" journal ads, the "informational content" is skipped by most readers (Lexchin 1994). Given that advertising is intended to sell products and that the informational content of advertising is not always complete or accurate, it is not surprising that prescribing has increased for the most heavily promoted drugs in the hypertension market, contrary to the recommendations of national guidelines. In light of the available evidence, it is difficult to accept the standard economic assumption that changes in prescribing influenced by the "information" provided through drug marketing are necessarily optimal. Conclusion From 1986 to 1996, the cost of hypertension drugs used by recipients of Pharmacare Plan A increased from $15.8 million per year to $58.7 million. With expenditures rapidly reaching such high levels, it is useful to know whether British Columbians received more or less value for their money from expenditures on hypertension drugs over time. To conduct such an assessment, researchers typically decompose changes in aggregate expenditures into measures of aggregate price and quantity that allow one to gauge the real productivity represented by purchases in this market segment. One way of performing such a decomposition is to construct indexes using the traditional economic approach to price and productivity measurement. For the hypertension treatments purchased under Pharmacare Plan A , a traditional price index grew a total of 15 percent from the first quarter of 1986 to the end of 1996. Considering that the Canadian consumer price index grew by 34 percent over this period, this transitional economic price index implies that real productivity actually increased in the hypertension sector. Deflating hypertension expenditures with the traditional economic price index yields a quantity index that grew 283 percent over the period. However, the notion of output that underlies the traditional economic price and quantity indexes does not necessarily relate to the number of tablets and pills consumed, nor to the quantity or quality of hypertension treatment achieved in this sector. Traditional economic indexes relate to a utility-theoretic notion of social welfare generated by the purchases accounted for in the aggregate index. Provided the assumptions of the traditional economic model of consumer demand are met, the common economic indexes used would be reasonably good estimators of true social costs and welfare. 80 Among the assumptions that is clearly not. met in this sector is the assumption that decision-making agents consider the full costs and benefits. Physicians, the principal decision-makers in this sector, certainly lack incentives to consider the financial costs of their prescribing. Furthermore, in part due to their financial incentives, physicians may base their decisions on information that is incomplete or inaccurate. Therefore, price and quantity indexes based on the traditional economic formulae wil l not necessarily relate to real social welfare or productivity in this sector. This is evident in the case of the indexes for hypertension drugs purchased under Pharmacare Plan A . The economic indexes differed substantially from indexes based on measures of health outcomes. From 1986 to 1996, price indexes based on the cost per patient treated or per day of treatment purchased grew 172 and 147 percent respectively—ten times as much as the economic indexes. Over this period, the number of Pharmacare Plan A beneficiaries treated for hypertension and the number of days of therapy consumed by those patients (which form the basis for the health outcomes quantity indexes) grew 63 percent and 80 percent, respectively. These alternative indexes are rationalized by the scientific evidence and social preferences expressed in national guidelines for the treatment of hypertension. The considerable difference between the health outcomes based indexes of price and quantity and the traditional economic indexes of price and quantity is due to increased use of high-cost medicines. The economic indexes account for this shift in treatment patterns as an increase in total output purchased, not a price change. This would be appropriate i f purchases were always efficient with respect to maximizing social welfare at least cost. However, it is clear that this was not the case. The lowest cost drugs on the market were consistently judged by the national prescribing guidelines to be the first choice in hypertension care based on scientific evidence regarding their efficacy and tolerability. Despite the recommendations of national guidelines, a declining proportion of hypertensive patients were prescribed these drugs over the period of study. Since the observed changes in anti-hypertensive prescribing were inconsistent with recommended care, the increased costs that resulted from the use of more expensive drugs could not have come with commensurate increases in social welfare. In fact, i f the cost per patient treated for hypertension actually had grown according to the rate implied by the traditional economic price index, Pharmacare would have spent approximately $207 million less on hypertension drugs than they did from 1986 to 1996—$33 million less in 1996 alone! The implied bias from using traditional economic indexes to measure productivity in this market segment is significant, not just for indexes related to this sub-market, but also for overall indexes of pharmaceutical sector price and quantity. Hypertension drugs accounted for approximately 30 percent of all drug expenditures for the elderly in British Columbia each year between 1986 and 1996. Therefore, the measurement bias from the use of traditional economic indexes for hypertension drugs alone wil l have a powerful impact on the measure of overall price and productivity for purchases made under Pharmacare Plan A. Mismeasuring the real cost of hypertension treatment will bias (downward) price indexes for Pharmacare Plan A by about 3 percent per annum between 1986 and 1996. The average annual growth rate of a traditional price index for all Pharmacare Plan A purchases is approximately 1.3 percent over this period. Therefore, the measurement bias from the hypertension segment alone roughly triples the aggregate measure of the cost of care in this sector, making it approximately 4.3 percent per annum. 81 It is clear from the results of this chapter and those preceding it that more research needs to be conducted on alternative approaches to price and productivity measurement in the pharmaceutical sector. The potential hazard of not addressing the measurement bias from the use of traditional economic indexes is that policy may be guided by a perception that the sector is more "productive" (or efficient) than is actually the case. The traditional indexes reported here suggest that, i f anything, utilization rates are the potentially problematic cause of expenditure inflation. Such a suggestion focuses policy attention on utilization-cubing measures that would possibly create further inefficiencies. It seems clear from the health services literature—and from the evidence presented above—that increased price per unit of care delivered stemming from inefficient drug selection is a major determinant of expenditure growth. This is not captured by conventional economic indexes because inefficiencies are simply not possible within the assumptions of the traditional economic model. Health-outcomes measures of pharmaceutical sector productivity, based whenever possible on recommended prescribing practices, will better indicate whether changes in the true quantity and/or quality of care delivered are consistent with changes in expenditures on drugs. Such indexes may focus the attention of researchers and policy makers on strategies to improve the social efficiency of the prescription decision-making process. 82 Appendix A: The 1997 and 1999 Guidelines In 1997, the JNC published its sixth and most recent report—JNC VI (JNC 1997). Like the CHS, the JNC developed a classification system for the evidence upon which its recommendations were based. However, the system used by the JNC was not hierarchical in its "grading" structure. The JNC classification system merely classified evidence into categories such as meta-analysis, randomized clinical trial, opinion article, etc. As with the JNC V , JNC VI contained an increased emphasis on non-drug treatments and called for population-wide strategies for the prevention of hypertension. There were no changes in "when to treat," but the JNC VI lowered the target blood pressure for those receiving treatment to a diastolic pressure below 140mm Hg over 90mm Hg (JNC 1997, p. 2421 emphasis in original). Unless contraindicated, the JNC VI recommends thiazide diuretics or beta-blockers as preferred first-line drugs for non-elderly hypertensives. If a diuretic is not chosen as the first-step, it is "usually indicated" as the second-step in therapy (JNC 1997, p. 158). The JNC VI lists compelling indications for individualizing antihypertensive therapy beyond the simple diuretic or beta-blocker choice of uncomplicated hypertension. Unless contraindicated, the JNC VI recommends low-dose thiazide diuretics or low-dose beta-blockers in combination with thiazide diuretics for elderly patients, and diuretics for those with isolated systolic hypertension. The most recent CHS guidelines were published in 1999 (Feldman et al 1999). As with the 1993 CHS report, the development of the 1999 guidelines was supported by pharmaceutical company partners46 of the CHS (CHS 1999, p.S17). Again, recommendations were graded (from A to D) according to the quality of scientific evidence that they were based upon. The 1999 guidelines offer the same when to treat recommendations as those of 1993, with greater emphasis on non-pharmacologic treatment. Treatment protocols also resemble previous publications, with a few changes. The 1999 CHS drugs of choice for uncomplicated hypertensives under 60 are thiazide diuretics, beta-blockers or ACE-inhibitors (grade A). If these prove inadequate or intolerable, the guidelines recommend substitution among the first-line drugs or combinations with diuretics (grade A). In the 1999 CHS, the drugs of choice for uncomplicated hypertensives over 60 are low-dose thiazide diuretics (grade A) or long-acting dihypdropyridine calcium-channel blockers (grade A , based on evidence from a study published in 1997). An ACE-inhibitor may be considered i f diuretic and calcium-channel blockers are ineffective, contraindicated or intolerable (grade B). 4 6 Company partners to the CHS include AstraZeneca Canada, Bayer Inc., Bristol-Myers Squibb Canada Inc., Hoescht Marion Roussel Canada Inc., Hoffman-LaRoche Limited, Merck Frosst Canada Inc., Novartis, Parke-Davis, Pfizer Canada Inc., Searle Canada, SmithKline Beecham Pharma and Wyeth-Ayerst Canada Inc. 83 Appendix B: Lists of Products Table 11: Diuretics AHFS Generic Name 402800012 Bendroflumethiazide Tab 5Mg 402800014 Bendroflumethiazide Kcl 5/500Mg 402800031 Chlorothiazide Tab 250Mg 402800032 Chlorothiazide Tab 500Mg 402800041 Chlorthalidone Tab 50Mg 402800042 Chlorthalidone Tab 100Mg 402800044 Chlorthalidone Tab 50Mg/Plus 402800061 Hydrochlorothiazide Tab 25Mg 402800062 Hydrochlorothiazide Tab 50Mg 402800063 Hydrochlorothiazide Tab 100Mg 402800064 Hydrochlorothiazide Tab 25Mg/Plus 402800065 Hydrochlorothiazide Tab 50Mg/Plus 402800172 Indapamide Tab 2.5Mg 402800173 Indapamide Tab 1.25Mg 402810011 Amiloride Tab 5Mg 402810012 Amiloride/Hydrochlor Tab 5/50Mg 402810021 Spironolactone Tab 25Mg 402810022 Spironolactone Tab 10OMg 402810023 Spironolactone Tab 25Mg/Plus 402810024 Spironolactone Tab 50Mg/Plus 402810031 Triamterene Hcl Tab 50Mg 402810032 Triamterene Hcl Tab 10OMg 402810033 Triamterene Hcl Plus Tab 50/25Mg Table 12: Beta Blockers AHFS Generic Name 240400012 240400014 240400101 240400102 240400103 240400 240400 240400 240400 240400 240400 240400 240400 240400 240400138 240400139 24040013A 24040013B 24040013C 240400151 240400152 240400153 240400154 240800016 240800017 240800101 240800102 105 106 131 132 133 134 135 136 137 Atenolol Tab 50Mg Atenolol Tab 100Mg Nadolol Tab 40Mg Nadolol Tab 80Mg Nadolol Tab 160Mg Nadolol/Bendroflu. 40/5Mg Nadolol/Bendroflu. 80/5Mg Propanolol Tab 10Mg Propanolol Tab 20Mg Propanolol Tab 40Mg Propanolol Tab 80Mg Propanolol Tab 120Mg Propanolol La Cap 160Mg Propanolol Inj 1Mg Propanolol La Cap 60Mg Propanolol Tab Plus 40Mg Propanolol La Cap 80Mg Propanolol La Cap 120Mg Propanolol Tab Plus 80Mg Timolol Tab 5Mg Timolol Tab 10Mg Timolol Tab 20Mg Timolol Tab 10Mg/Plus Atenolol/Plus Tab 50/25Mg Atenolol/Plus Tab 100/25Mg Metoprolol Tab 50Mg Metoprolol Tab 100Mg 240800103 Metoprolol Sr Tab 200Mg 240800104 Metoprolol Tab 100Mg/Plus 240800106 Metoprolol Sr Tab 100Mg 240800107 Metoprolol Inj Iv 1Mg/MI 240800131 Pindolol Tab 5Mg 240800132 Pindolol Tab 10Mg 240800133 Pindolol Tab 15Mg 240800134 Pindolol/Plus Tab 10/25Mg 240800135 Pindolol/Plus Tab 10/50Mg 240800201 Oxprenolol Tab 20Mg 240800202 Oxprenolol Tab 40Mg 240800203 Oxprenolol Tab 80Mg 240800204 Oxprenolol Slow Tab 80Mg 240800205 Oxprenolol Slow Tab 160Mg 240800211 Labetalol Tab 100Mg 240800212 Labetalol Tab 200Mg 240800213 Labetalol Inj 5Mg/MI 240800251 Acebutolol Tab 100Mg 240800252 Acebutolol Tab 200Mg 240800253 Acebutolol Tab 400Mg Table 13: ACE-inhibitors AHFS Generic Name 240400021 Captopril Tab 25Mg 240400022 Captopril Tab 50Mg 240400023 Captopril Tab 100Mg 240400024 Captopril Tab 12.5Mg 240400025 Captopril Tab 6.25Mg 240800281 Enalapril Maleate Tab 2.5Mg 240800282 Enalapril Maleate Tab 5Mg 240800283 Enalapril Maleate Tab 10Mg 240800284 Enalapril Maleate Tab 20Mg 240800285 Enalapril Maleate Tab 40Mg 240800286 Enalapril Maleate Plus 10/25 240800291 Lisinopril Tab 5Mg 240800292 Lisinopril Tab 10Mg 240800293 Lisinopril Tab 20Mg 240800295 Lisinopril/Hydrochloro Tab 20/25 240800296 Lisinopril/Hydrochloro Tab 20/12.5 240800297 Lisinopril/Hydrochloro Tab 10/12.5 240800311 Quinapril Tab 10Mg 240800312 Quinapril Tab 5Mg 240800313 Quinapril Tab 20Mg 240800314 Quinapril Tab 40Mg 240800321 Fosinopril Tab 10Mg 240800322 Fosinopril Tab 20Mg 240800341 Cilazapril Tab 0.5Mg 240800342 Cilazapril Tab 1 .OMg 240800343 Cilazapril Tab 2.5Mg 240800344 Cilazapril Tab 5.0Mg 240800351 Benazepril Tab 5Mg 240800352 Benazepril Tab 10Mg 240800353 Benazepril Tab 20Mg 240800361 Ramipril Tab 1.25Mg 240800362 Ramipril Tab 2.5Mg 240800363 Ramipril Tab 5.0Mg 240800364 Ramipril Tab 10.0Mg 84 Table 14: Calcium-Channel Blockers AHFS Generic Name 240400071 Diltiazem Tab 30Mg 240400072 Diltiazem Tab 60Mg 240400073 Diltiazem Sr Cap 90Mg 240400074 Diltiazem Sr Cap 120Mg 240400075 Diltiazem Sr Cap 60Mg 240400076 Diltiazem Cr Cap 180Mg 240400077 Diltiazem Cr Cap 240Mg 240400078 Diltiazem Cr Cap 300Mg 240400079 Diltiazem Cr Cap 120Mg 240400111 Nifedipine Cap 5Mg 240400112 Nifedipine Cap 10Mg 240400113 Nifedipinepa Tab 20Mg 240400114 Nifedipine Tab 30Mg 240400115 Nifedipine Tab 60Mg 240400116 Nifedipine Ft Tab 10Mg 240400117 Nifedipine Pa Tab 10Mg 240400161 Verapamil Sr Cap 120Mg 240400162 Verapamil Inj 2.5Mg 240400163 Verapamil Sr Cap 180Mg 240400164 Verapamil Tab 80Mg 240400165 Verapamil Sr Cap 240Mg 240400166 Verapamil Tab 120Mg 240400167 Verapamil Sr Tab 120Mg 240400168 Verapamil Sr Tab 240Mg 240400169 Verapamil SrTab 180Mg 240400291 Nicardipine Cap 20Mg 240400292 Nicardipine Cap 30Mg 240400301 Felodipine Er Tab 5Mg 240400302 Felodipine Er Tab 10Mg 240400303 Felodipine Er Tab 2.5Mg 240400311 Amlodipine Tab 5Mg 240400312 Amlodipine Tab 10Mg Table 15: Other Antihypertensive Drugs AHFS Generic Name 240800031 Clonidine Tab .025Mg 240800032 Clonidine Tab . 1 Mg 240800033 Clonidine Tab .2Mg 240800034 Clonidine Tab . 1 Mg/Plus 240800071 Hydralazine Tab 10Mg 240800072 Hydralazine Tab 25Mg 240800073 Hydralazine Tab 50Mg 240800074 Hydralazine Inj 20Mg 240800091 Methyldopa Tab 125Mg 240800092 Methyldopa Tab 250Mg 240800093 Methyldopa Tab 500Mg 240800094 Methyldopa/Plus Tab 250/15Mg 240800095 Methyldopa/Plus Tab 250/25Mg 240800096 Methyldopa/Plus Tab 250/150Mg 240800097 Methyldopate/Plus 250/250Mg 240800098 Methyldopate/Plus Inj 50Mg 240800111 Minoxidil Tab 2.5Mg 240800112 Minoxidil Tab 10Mg 240800141 Prazosin Cap .5Mg 240800142 Prazosin Tab 1 Mg 240800143 Prazosin Tab 2Mg 240800144 Prazosin Tab 5Mg 240800145 Prazosin Cap 1 Mg 240800146 Prazosin Cap 2Mg 240800271 Terazosin Hydrochloride Tab 1 Mg 240800272 Terazosin Hydrochloride Tab 2Mg 240800273 Terazosin Hydrochloride Tab 5Mg 240800274 Terazosin Hydrochloride Tab 10Mg 240800301 Doxazosin Mesylate Tab 1 Mg 240800302 Doxazosin Mesylate Tab 2 Mg 240800303 Doxazosin Mesylate Tab 4 Mg 85 Appendix C: New Product Entrants Table 16 :Canadian Launch Dates for Top Selling Antihypertensive Drugs Chemical Class Brand Date Enalapril ACE Vasotec Aug-87 Enalapril ACE Vaseretic Oct-90 Diltiazem CCB Cardizem CD Nov-92 Amlodipine CCB Norvasc Sep-92 Nifedipine CCB Adalat XL Apr-92 Nifedipine CCB Adalat Pa20 Apr-87 Lisinopril ACE Zestril Nov-90 Lisinopril ACE Prinivil Nov-90 Lisinopril ACE Zestoretic Nov-92 Lisinopril ACE Prinizide Dec-92 (Source: IMS Health 1999) 86 Appendix D: Regression Analysis for Price Indexes Table 17: Regression Analysis for Trend in Price Indexes Economic Day Based Patient Based Index Index Index Dependent Variable ln(index) ln(index) ln(index) R 2 0.901 0.965 0.957 Quarterly Growth Rate 86-93 0.009* 0.032* 0.032* (0.000) (0.001) (0.001) Quarterly Growth Rate 94-96 -0.007+ -0.001 0.002 (0.003) (0.005) (0.006) Constant 0.049* 0.080* 0.133* (0.010) (0.020) (0.023) 94-96 Dummy 0.359* 0.863* 0.743* (0.101) (0.212) (0.242) Slope Change F(l,40) 35.271 36.227 22.452 •Significant at p=0.01 + Significant at p=0.05 Standard errors are in parentheses. 87 Appendix E: Regression Analysis of Patients Receiving Treatments by Classification Table 18: Regression Analysis for Trends in Patients Receiving Treatments by Classification Total DIR DIR+ BET ACE CCB OTH NDC R 2 0.98 0.97 0.67 0.70 1.00 1.00 0.98 0.99 Constant 84693* 32843* 19665* 18510* -1779* 4998* 6576* 3880* (1252) (407) (398) (153) (271) (195) (85) (194) QI Dummy -193 96 160 -279 81 -185 -102 36 (1182) (384) (376) (145) (256) (184) (80) (183) Q2 Dummy 882 461 323 -115 322 -107 -94 92 (1187) (386) (378) (145) (257) (184) (81) (184) Q3 Dummy -139 435 -26 -142 14 -170 -38 -212 (1178) (383) (375) (144) (255) (183) (80) (183) Quarterly Growth 1323* -432* -47* -25* 799* 765* -118* 381* 86-93 (51) (16) (16) (6) (11) (8) (3) (8) Quarterly Growth 1784* 157 506* 247* 594* -193* 100* 373* 94-96 (265) (86) (84) (32) (57) (41) (18) (41) 94-96 Dummy -19821 -19882* -18877* -10376* 5774+ 31031* -7129* -361 (10400) (3383) (3310) (1273) (2251) (1616) (708) (1611) Slope Change F(l,37) 2.9 45.3* 41.7* 67.9* 12.4* 523.9* 141.7* 0.0 * Significant at p=0.01 + Significant at p=0.05 Standard errors are in parentheses. 88 Table 19: Annualized Price Indexes for Hypertension Treatments, Grouped by Drug Class and Index Type Class: Index 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 ACE-inhibitors: E 1.00 1.08 1.16 1.19 1.24 1.27 1.31 1.29 1.13 1.10 1.11 ACE-inhibitors: D 1.00 0.96 0.92 0.90 0.92 0.94 0.97 0.96 0.84 0.83 0.87 ACE-inhibitors: P 1.00 1.04 1.02 1.00 1.04 1.08 1.11 1.10 0.98 0.98 1.03 Beta-Blockers: E 1.00 1.09 1.16 1.21 1.24 1.29 1.32 1.31 1.16 1.03 0.96 Beta-Blockers: D 1.00 1.13 1.21 1.28 1.34 1.41 1.46 1.45 1.30 1.16 1.09 Beta-Blockers: P 1.00 1.20 1.31 1.39 1.46 1.53 1.58 1.57 1.41 1.26 1.19 CCB: E 1.00 1.11 1.17 1.19 1.23 1.28 1.32 1.33 1.28 1.21 1.18 CCB: D 1.00 1.09 1.12 1.16 1.19 1.24 1.27 1.27 1.22 1.17 1.17 CCB: P 1.00 1.17 1.21 1.26 1.31 1.37 1.41 1.41 1.38 1.33 1.35 Diuretics: E . 1.00 1.04 1.09 1.13 1.17 1.24 1.25 1.23 1.06 0.91 0.84 Diuretics: D 1.00 1.03 1.06 1.08 1.14 1.23 1.27 1.26 1.10 0.93 0.82 Diuretics: P 1.00 1.10 1.13 1.15 1.20 1.29 1.33 1.31 1.15 0.96 0.84 Other: E 1.00 1.09 1.15 1.20 1.25 1.30 1.35 1.35 1.16 1.07 1.06 Other: D 1.00 1.10 1.16 1.23 1.36 1.48 1.61 1.74 1.64 1.76 2.02 Other: P 1.00 1.17 1.25 1.33 1.45 1.57 1.71 1.85 1.74 1.85 2.13 CCB = calcium-channel blockers. E = Economic index (Laspeyres) D = Cost-per-day based index. P = Cost-per-patient based index. 'includes the cost of multiple-therapy across drug classes. Appendix F: Alternative Indexes by Drug Subclass Alternative price indexes for each of the therapeutic subclasses are listed in Table 19. A l l indexes were constructed on a quarterly basis—average annual accumulations are listed here for parsimony. The economic (E) indexes listed are based on a chained Laspeyres index, with quantity weights updated each period. The patient-day (D) based indexes use the estimated number of patient days of maintenance therapy (described above) as the productivity basis for measurement. The cost per discrete patient (P) indexes use the number of discrete patients receiving drug from each therapeutic subclass as the basis for productivity measurement. One of the first things to notice in Table 19 is the slight discrepancy between patient-day and discrete patient based indexes. For example, the cost per patient-day index for ACE-inhibitors fell thirteen percent while the cost per discrete patient index rose three percent. This variation between indexes is due to difficulties in estimating patient-days for newly available products. When new products first enter the market, they are used by relatively few patients, only a portion of whom will refill their prescriptions. Since the estimates of prescribed doses early in a product life-cycle are based on only a few observations, they are less reliable than estimates of the prescribed dose for older products. This "noise" clearly affected the 1986/87 indexes for ACE-inhibitors as a major A C E -inhibitor—Vasotec (Enalapril)—entered the market. 89 Figure 10 Average Cost Per Year of Treatment on Diuretic Drugs As Prescribed BC Pharmacare Plan A 1986 to 1996 $250 $200 $150 .-$100 $50 86:l 87:l 88:l 89:l 90:l 91:1 92:1 93:1 94:1 -95:1 96:1 • Indapamide • Others • Hydrochlorothiazide Comparing economic indexes with the cost per discrete patient indexes within therapeutic subcategories shows that substitution towards more expensive products occurred within the beta-blocker, calcium-channel blocker and (most strikingly) "other" therapeutic subclasses. Changes in market share were of the opposite direction in the ACE-inhibitor market. Movements in the diuretic category, though not resulting in major price changes, are a result of interesting market dynamics. Figure 10 illustrates the cost per year of treatment by alternative types of diuretic drugs. As can be seen, there are enormous differences in the cost per day of treatment within this therapeutic subclass. The oldest and most widely recognized diuretic, hydrochlorothiazide, had been available in generic form throughout the study. Moreover, the prescribed hydrochlorothiazide dose per day fell from 48mg in 1986 to 29mg in 1996, consistent with national guidelines. The price (per pill) of hydrochlorothiazide fell modestly from 1986 to 1993 and then dramatically after the low-cost alternative program increased generic substitution in 1994. The price per tablet of several of the higher price diuretics rose from 1986 to 1993 (explaining the increase in the Laspeyres price index). Included among these drugs was indapamide, the unit price of which rose over 50 percent from 1986 to 1993. Indapamide, a newer diuretic sold under the brand-name Lozide, was not subject to generic competition in B C until 1995. Lozide's share of the diuretic market increased from 1 to 6 percent of patient-days, resulting in an increase from 2 to 30 percent of dollar-volume from 1986 to 1996. Hydrochlorothiazide also gained lost market share the market in terms of patient-days (from 33 to 46 percent), but lost market share in dollar-volumes (7 to 4 percent). The increased market share and price movements of indapimide placed inflationary pressures on the price 90 indexes for diuretic treatments. On the other hand, the increased market share, falling prices and reduced dosages of hydrochlorothiazide put downward pressure on the indexes. As seen in Table 19 above, the net effect over was increased prices (in economic terms and based on real productivity) until policy changes increased generic usage in 1994 and competition entered the indapamide market in 1995. Other than its heavy promotion, there is little explanation for the use of the high cost Lozide product when lower-cost alternatives were widely available. 91 Appendix G: Classification of First-Line Treatment for Non-Persistent Patients Aged 66 Table 20: Classification of First-Line Treatment for Non-Persistent Patients Aged 66 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 N 9,057 7,474 7,148 6,699 6,255 6,154 6,507 6,364 6,267 9,409 Diuretics 54% 49% 43% 41% 38% 36% 35% 33% 33% 30% Diuretics Plus 5% 4% 4% 4% 3% 3% 3% 3% 3% 4% Beta-Blockers 15% 14% 13% 13% 13% 12% 13% 13% 15% 16% ACE-inhibitors 5% 8% 12% 14% 16% 18% 20% 21% 20% 21% CCB 14% 18% 20% 20% 22% 22% 21% 19% 16% 11% Other 6% 5% 5% 5% 5% 6% 5% 7% 11% 13% ND Combo 2% 2% 3% 3% 3% 3% 3% 4% 3% 4% CCB = calcium-channel blockers. ND = Combination of non-diuretic drugs. 92 Appendix H: Regression Analysis for Trends in First-Line Treatments for Persistent Patients Table 21: Regression Analysis for Trends in First-Line Treatments for Persistent Patients Total DIR DIR+ BET ACE CCB OTH NDC R 2 0.52 0.75 0.54 0.62 0.89 0.53 0.78 0.65 Constant 2092.7* 695.4* 219.2* 374.9* 154.5* 451.8* 91.3* 105.5* (132.0) (42.4) (17.3) (25.8) (29.8) (30.7) (13.2) (10.9) QI Dummy 240.2+ 103.2+ 23.9 49.5+ 39.2 16.7 -2.5 10.2 (122.7) (39.4) (16.1) (24.0) (27.7) (28.5) (12.3) (10.1) Q2 Dummy -8.6 45.8 -3.9 -3.9 11.3 -20.8 -15.8 -21.2+ (123.3) (39.6) (16.1) (24.1) (27.8) (28.6) (12.3) (10.1) Q3 Dummy -292.8+ -20.5 -31.0 -47.9+ -82.3* -72.0* -10.6 -28.6* (122.3) (39.3) (16.0) (23.9) (27.6) (28.4) (12.2) (10.1) Quarterly Growth -6.0 -16.3* -3.8* -4.8* 16.1* 2.8+ -1.8* 1.8* 86-93 (6.1) (2.0) (0.8) (1.2) (1.4) (1.4) (0.6) (0.5) Quarterly Growth 77.9* 35.5* 6.5 21.6* 9.5 -16.3* 19.0* 2.1 94-96 (26.2) (8.4) (3.4) (5.1) (5.9) (6.1) (2.6) (2.2) 94-96 Dummy -2542.1* -1498.9* -291.0+ -797.2* 139.7 530.7* -623.0* -2.4 (926.4) (297.8) (121.3) (180.9) (209.1) (215.0) (92.7) (76.1) Slope Change F(l,37) 9.7* 35.9* 8.5* 25.5* 1.2 9.4* 60.4* 0.0 *Significantatp=0.01 + Significant at p=0.05 Standard errors are in parentheses. 93 Bibliography Adams, A.S., S.B. Soumerai, J. Lomas and D. Ross-Degnan. (1999) 'Evidence df Self-Report Bias In Assessing Adherence to Guidelines.' International Journal for Quality in Health Care, 11(3), 187-192. Akerlof, G. (1970) 'The Market for 'Lemons': Qualitative Uncertainty and the Market Mechanism.' Quarterly Journal of Economics, 89,488-500. Anderson, G.M. , and J. Lexchin. (1996) 'Strategies for Improving Prescribing Practice' Canadian Medical Association Journal 154(7), 1013-1017. Anderson, G.M. , K.J . Kerluke, I.R. Pulkins, C.H. Hertzman and M . L . Barer. (1993) 'Trends and Determinants of Prescription Drug Expenditures in the Elderly: Data from the British Columbia Pharmacare Program.' Inquiry, 30(2), 199-207. Anderson, J.F., K . L . McEwan, and W.P. Hrudey. (1996) 'Effectiveness of Notification and Group Education in Modifying Prescribing of Regulated Analgesics' Canadian Medical Association Journal 154(1), 31-39. Anderson, M.J . (1997) Evergreening in the Canadian Pharmaceutical Industry. Queen's Health Policy Research Unit. Queen's University, Kingston. Andersson, F. (1995) 'Why is the Pharmaceutical Industry Investing Increasing Amounts in Health Economic Evaluations?' International Journal of Technology Assessment in Health Care 11 (4), 75-761. Anis, A . H . (1994) 'Substitution Laws, Insurance Coverage, and Generic Drug Use,' Medical Care 32(3), 240-256. Anonymous. (1998) 'Are Generic Drugs OK, or Should You Stick With Brand Names?' Tufts University Health & Nutrition Letter. 16(6): 3. Anonymous. (1999) 'Hypertension: What Works?' Consumer Reports, 64(5), 60-62. Arnst, C. (1998) 'Is Good Marketing Bad Medicine?' Business Week, 4/13, 62-63. Avorn, J. (1996) 'Drug Use in Academic Medical Centers in an Era of Ferment.' PharmacoEconomics. 10(Sup 2), 124-129. Babington, M.A . , L .A . Robinson, and R.A. Monson. (1983) 'Effect of Written Information on Physicians' Knowledge of Drug Prices' Southern Medical Journal 76(3), 328-331. Bailey, C.R., R. Ruggier, and J.N. Cashman. (1993) 'Anaethesia: cheap at twice the price? Staff awareness, cost comparisons and recommendations for economic savings' Anaethesia 48(10), 906-909. Baines, D.L., D.K. Whynes and K . H . Tolley. (1997) 'General Practitioner Fundholding and Prescribing Expenditure Control' PharmacoEconomics 11(4), 350-358. 94 Baumann, A.O, R.B. Deber, and G.G. Thompson. (1991) 'Overconfidence Among Physicians and Nurses: the "Micro-certainty, Macro-uncertainty" Phenomenon' Social Science and Medicine 32(2), 167-174. Baye, M.R., R. Maness, and S.N. Wiggins (1997) 'Demand Systems and the True Subindex of the Cost of Living for Pharmaceuticals,' Applied Economics 29,1179-1189. Berginer, G.B., M . Bile, and D.K. Ziegler. (1984) 'Neurologists' knowledge of medication costs' Neurology 34(1), 121-2. Berndt, E.R., and P.E. Greenberg (1996) 'An Updated and Extended Study of the Price Growth of Prescription Pharmaceutical Preparations,' In Competitive Strategies in the Pharmaceutical Industry, ed. R.B. Helms (Washington: American Enterprise Institute) Berndt, E.R., D . M . Cutler, R.G. Frank, Z. Griliches, J.P. Newhouse, and J.E. Triplett. (1998) 'Price Indexes for Medical Care Goods and Services: A n Overview of Measurement Issues,' N B E R Working Paper 6817. Berndt, E.R., I .M. Cockburn, and Z. Griliches (1996) 'Pharmaceutical Innovations and Market Dynamics: Tracking Effects on Price Indexes for Antidepressant Drugs,' Brookings Papers on Economics Activity, Microeconomics, 133-188. Berndt, E.R., I .M. Cockburn, D.L. Cocks, A . Epstein, and Z. Griliches (1997) 'Is Price Inflation Different for the Elderly? An Empirical Analysis of Prescription Drugs,' N B E R Working Paper, 6182. Berndt, E.R., Z. Griliches, and J.G. Rosett (1993) 'Auditing the Producer Price Index: Micro Evidence From Prescription Pharmaceutical Preparations,' Journal of Business and Economic Statistic 11(3), 251-264. Blomqvist, A . (1991) 'The Doctor as Double Agent: Information Asymmetry, Health Insurance, and Medical Care' Journal of Health Economics 10(4), 411-432. Bogner, W.C. (1996) Drugs To Market: Creating Value and Advantage in the Pharmaceutical Industry. Elsevier Science: USA. Britten, N . and O. Ukoumunne. (1997) 'The Influence of Patients' Hopes of Receiving a Prescription on Doctors' Perceptions and the Decision to Prescribe: A Questionnaire Survey' British Medical Journal 315(7121), 1506-1511. Buetow, S.A., B. Sibbald, J.A. Cantrill and S. Halliwell. (1996) 'Prevalence of Potentially Inappropriate Long Term Prescribing in General Practice in the United Kingdom, 1980-1995: Systematic Literature Review.' British Medical Journal 313(7069), 1371-1374. Buetow, S.A., B. Sibbald, J.A. Cantrill and S. Halliwell. (1997) 'Appropriateness in Health Care: Application to Prescribing' Social Science and Medicine 45(2), 261-271. Canada, Health Protection Branch, Health Canada. (1991) Heath Protection and Drug Laws. Minister of Supply and Services: Ottawa. 95 Canada, Special Committee of the House of Commons. (1967) Second (Final) Report of the Special Committee of the House of Commons on Drug Costs and Prices. Queen's Printer: 1967. Canada. Department of Labour. (1927) Investigation into the Proprietary articles trade association, an alleged combine of wholesale and retail druggists and manufacturers, established to fix and maintain resale prices ofproprietary medicines and toilet articles. Report of Commissioner. Printer to the King: Ottawa. Carruthers, S.G., P.Larochelle, R.B. Haynes, A . Pertasovits, and E.L. Schiffrin. (1993) 'Report of the Canadian Hypertension Society Consensus Conference: 1. Introduction.' Canadian Medical Association Journal, 149(3), 289-293. Castonguay, C , L. Borgeat, L . Champigny-Robillard, D. Leclerc, and Y . Morin. (1996) Drug Insurance: Possible Approaches. Report of the Committee of Experts on Drug Insurance. Bibliotheque nationale du Quebec. Caudill, T.S., M.S. Johnson, E.C. Rich and W.P. McKinney. (1996) 'Physicians, Pharmaceutical Sales Representatives and the Cost of Prescribing.' Archives of Family Medicine, 5, 201-206. Caves, R.E., M.D. Whinston, and M . A . Hurwitz (1991) 'Patent Expiration, Entry, and Competition in the U.S. Pharmaceutical Industry,' Brookings Papers on Economics, Microeconomics, 1-66. Chappell, N.L . , M . Maclure, H . Brunt, J. Hopkinson, and J. Mullett. (1997) 'Seniors' Views of Medication Reimbursement Policies: Bridging Research and Policy at the Point of Policy Impact' Canadian Journal on Aging/Canadian Public Policy 16/23(sup), 147-161. Cockburn, I., and R. Henderson. (1996) 'Public-private interaction in pharmaceutical research.' Proceedings of the National Academy of Sciences, USA, 93(23): 12725-12760. Cockburn, J. and S. Pit. (1997) 'Prescribing Behaviour in Clinical Practice: Patients' Expectations and Doctors' Perceptions of Patients Expectations—A Questionnaire Survey' British Medical Journal 315(7107), 520-523. Culyer, A.J. , and Evans, R.G. (1996) 'Mark Pauly on Welfare Economics: Normative Rabbits from Positive Hats' Journal of Health Economics 15(2), 243-51. Curran, R.E. (1953) Canada's Food and Drug Laws. Commerce Clearing House: New York. Darbourne, A . (1993) 'OTC Switching—Pricing in a Profitable Market' Scrip Magazine December 1993, 14-16. Davidson, W., W. Molloy, G. Somers, and M . Bedard. (1994) 'Relationship Between Physician Characteristics and Prescribing for Elderly People in New Brunswick' Canadian Medical Association Journal 150(6), 917-921. Davis, P., R.L. Yee and J. Millar. (1994) 'Accounting for Medical Variation: The Case of Prescribing in A New Zealand General Practice Sample' Social Science and Medicine 39(3), 367-374. 96 Denig, P., and F . M . Haaijer-Ruskamp. (1995) 'Do Physicians Take Costs Into Account When Making Prescribing Decisions?' PharmacoEconbmics 8(4), 282-290. Deyo, R.A, B . M . Psaty, G. Simon, E.H. Wagner, and G.S. Omenn. (1998) 'The Messenger Under Attack—Intimidation of Researchers by Special-Interest Groups.' The New England Journal of Medicine 336(16), 1176-1180. Diewert, W.E. (1981) 'The Economic Theory of Index Numbers: A Survey' Reprinted in Diewert, W.E., and A.O. Nakamura eds. (1993) Essays in Index Number Theory, Volume 1. Elsevier Science: Amsterdam Diewert, W.E. (1988a) 'Group Cost of Living Indexes' Methods of Operations Research 48:23-45. Reprinted in Diewert, W.E., and A.O. Nakamura eds. (1993) Essays in Index Number Theory, Volume 1. Elsevier Science: Amsterdam Diewert, W.E. (1988b) 'The Early History of Price Index Research' Discussion Paper No.88-26. Department of Economics, U B C . Reprinted in Diewert, W.E., and A.O. Nakamura eds. (1993) Essays in Index Number Theory, Volume 1. Elsevier Science: Amsterdam. Diewert, W.E. (1998) 'Index Number Issues in the Consumer Price Index.' Journal of Economic Perspectives, 12(1), 47-58. Diewert, W.E. (1995) 'Axiomatic and Economic Approaches to Elementary Price Indexes' Department of Economics Discussion Paper 95-01, University of British Columbia. Dowell, J.S., D. Snadden, and J.A. Dunbar. (1996) 'Rabid Prescribing Change, How do Patients Respond?' Social Science and Medicine. 43(11), 1543-1549. Eastman, H.C., R.D. Fraser, J.I. Laskin, and W.A. Kennedy. (1985) Summary of the Report of the Commission of Inquiry on the Pharmaceutical Industry. Supply and Services Canada: Ottawa. Ellison, S.F., I. Cockburm, Z. Griliches, and J. Hausman (1997) 'Characteristics of demand for pharmaceutical products: and examination of four cephalosporins,' RAND Journal of Economics 28(3), 426-446. Evans, R.G. (1974) 'Supplier Induced Demand: Some Empirical Evidence and Implications' In M . Perlman (ed.) The Economics of Health and Medical Care (London). Evans, R.G. (1983) 'Incomplete Vertical Integration in the Health Care Industry: Pseudomarkets and Pseudopolicies' Annals (AAPSS) 464, 60-87. Evans, R.G. (1984) Strained Mercy. Butterworths: Toronto. Evans, R.G. (1990) 'The Dog in the Night-Time: Medical Practice Variations and Health Policy.' In Anderson, T.F and G. Mooney (eds.) The Challenges of Medical Practice Variations. MacMillin Press: London. Evans, R.G. (1995) 'Manufacturing Consensus, Marketing Truth: Guidelines for Economic Evaluation.' Annals of Internal Medicine, 123(1), 59-60. Feenstra, R.C. (1997) 'Generics and New Goods in Pharmaceutical Price Indexes: Comment,' The American Economic Review 87(4), 760-767'. 97 Feldman, R., and B . Dowd. (1993) 'What Does a Demand Curve for Medical Care Measure?' Journal of Health Economics 12(2), 193-200. Feldman, R.D., N . Campbell, P. Larochelle et al. (1999) '1999 Canadian Recommendations for the Management of Hypertension.' Canadian Medical Association Journal, 161(12), S1-S17. Feldstein, M.S. (1973) 'The Welfare loss of Excess Health Insurance.' Journal of Political Economics, 81(2), 251-280. Fisher, F .M. , and Z. Griliches (1995) 'Aggregate Price Indices, New Goods, and Generics,' Quarterly Journal of Economics 60(1), 229-244. Frank, R.G., and D.S. Salkever (1992) 'Pricing, Patent Loss, and the Market for Pharmaceuticals,' Southern Economic Journal 59(2), 165-179. Frank, R.G., and D.S. Salkever (1997) 'Generic Entry and The Pricing of Pharmaceuticals,' Journal of Economics and Management Strategy 6(1), 75-90. Gambardella, A . (1995) Science and Innovation: The US Pharmaceutical Industry During the 1980s. Cambridge University Press: Cambridge. Glickman, L, E.A. Bruce, F.G. Caro, and J Avorn. (1994) 'Physicians' Knowledge of Drug Costs for the Elderly' Journal of the American Geriatrics Society 42(9), 992-996. Gorecki, P.K. (1992) Controlling Drug Expenditures in Canada: The Ontario Experience. Economic Council of Canada: Ottawa Goyer, R. (1986) Regulatory Aspects and Their Influences on Pharmaceutical Research and on the Introduction of Drugs in Canada. Background Study. Commission of Inquiry on the Pharmaceutical Industry. Minister of Supply and Services: Ottawa. Grabowski, H . (1998) 'The Role of Cost-Effectiveness Analysis in Managed-Care Decisions' PharmacoEconomics 14(sup.l), 15-24. Grabowski, H.G., and J.M. Vernon (1992) 'Brand Loyalty, Entry, and Price Competition in Pharmaceuticals After the 1984 Drug Act,' Journal of Law and Economics 35, 331-350. Griliches, Z., and I.M. Cockburn. (1994) 'Generics and New Goods in Pharmaceutical Price Indexes,' The American Economic Review 84(5), 1213-1232. Griliches, Z., and I.M. Cockburn. (1996) 'Generics and the Producer Price Index for Pharmaceuticals,' in Competitive Strategies in the Pharmaceutical Industry, ed. R.B. Helms. (Washington: American Enterprise Institute) Grilli , R., J.L. Lomas. (1994) 'Evaluating the Message: The Relationship Between Compliance Rate and the Subject of a Practice Guideline.' Medical Care, 32(3), 202-213. Grootendorst, P., L . Goldsmith, J. Hurley, B. O'Brian, and L . Dolovich (1996) Financial Incentives to Dispense Low Cost Drugs: A Case Study of British Columbia Pharmacare. Report Submitted to the Health System and Policy Division, Health Canada. (Hamilton: McMaster University) 98 Haaijer-Ruskamp, F .M. , and P. Denig. (1996) 'New Approaches to Influencing Physicians' Drug Choices: The Practice-Based Strategy' In Davis, Peter (Ed), Contested Ground: Public Purpose and Private Interest in the Regulation of Prescription Drugs. Oxford University Press: New York. Haayer, F. (1982) 'Rational Prescribing and Sources of Information' Social Science and Medicine 16, 2017-2023. Hall, E .M. , A . G . Girard, D . M . Baltzan, O J . Firestone et al. (1963) Royal Commission on Health Services—Provision, Distribution and Cost of Drugs in Canada. Queen's Printer: Ottawa. Haynes, R.B., Y . Lacourciere, S.W. Rabkin et al. (1993) 'Report of the Canadian Hypertension Society Consensus Conference: 2. Diagnosis of Hypertension in Adults.' Canadian Medical Association Journal, 149(5), 409-418. Hemminki, E. (1975) 'Review of Literature on the Factors Affecting Drug Prescribing' Social Science and Medicine 9, 111-116. Hershey, C O . , D.K. Poerter, D. Breslau, and D.I. Cohen. (1986) 'Influence of Simple Computerized Feedback on Prescription Charges in an Ambulatory Clinic' Medical Care 24(6), 472-481. Hi l l , M . N , D . M . Levine, and P.K. Whelton. (1988) 'Awareness, Use, and Impact of the 1984 Joint National Committee Consensus Report on High Blood Pressure.' American Journal of Public Health, 78(9), 1190-193 Hodges, B. (1995) 'Interactions with the Pharmaceutical Industry: Experiences and Attitudes of Psychiatry Residents, Interns and Clerks.' Canadian Medical Association Journal, 153(5), 553-559. HSF (Heart and Stroke Foundation of Canada). (1999) The Changing Face of Heart Disease and Stroke in Canada 2000. Ottawa: Canada. Hurwitz, M.A. , and R.E. Caves. (1988) 'Persuasion or Information? Promotion and the Shares of Brand Name and Generic Pharmaceuticals.' Journal of Law and Economics, 31(2), 299-321. ICES (Institute for Clinical Evaluative Sciences). (1998) 'Management of Hypertension: Diagnosis, Evaluation, When to Treat and Non-Pharmacological Therapy.' Internet Web Site, http://www.ices.on.ca/docs/fbl420.htm. Accessed 9/25/98. JNC (Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure). (1977) 'Report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure.' Journal of the American Medical Association, 237(3), 255-26. JNC (Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure). (1980) 'The 1980 Report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure' Archives of Internal Medicine, 140(10), 1280-1285. JNC (Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure). (1984) 'The 1984 Report of the Joint National Committee on Detection, 99 Evaluation, and Treatment of High Blood Pressure' Archives of Internal Medicine, 144(5), 1045-1057. JNC (Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure). (1988) 'The 1988 Report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure' Archives of Internal Medicine, 148(5), 1023-1038. JNC (Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure). (1993) 'The Fifth Report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure (JNC V ) ' Archives of Internal Medicine, 153(2), 154-183. JNC (Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure). (1997) 'The Sixth Report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure' Archives of Internal Medicine, 157(21), 2413-2445. Johannasson, M , and B. Jonsson. (1991) 'Cost-effectiveness Analysis of Hypertension Treatment—A Review of Methodological Issues.' Health Policy 19(1), 55-78. Kelly, G.G. (1997) 'Improving the PPI Samples for Prescription Pharmaceuticals' Monthly Labor Review 120(10), 10-17. Kolosa, E . M . (1995) 'Physicians' Perceptions of Prescription Drug Prices: Their Accuracy and Effect on the Prescribing Decision' Journal of Research in Pharmaceutical Economics 6(1), 23-37. Lang, R.W. (1974) The Politics of Drugs: A comparative pressure-group study of the Canadian Pharmaceutical Manufacturers Association and the Association of the British Pharmaceutical Industry. Saxon House: England. Larochell, P., M.J. Bass, N.J. Birkett, J. De Champlain, and M . G . Myers. (1986) 'Recommendations from the Consensus Conference on Hypertension in the Elderly.' Canadian Medical Association Journal, 135(7), 741-745. Lee, C. (1995) 'Optimal Medical Treatment Under Asymmetric Information' Journal of Health Economics 14(4), 419-441. Leffler, K . B . (1981) 'Persuasion or Information? The Economics of Drug Advertising' Journal of Law and Economics, 24(1), 45-74. Lexchin, J. (1989) 'Doctors and Detailers: Therapeutic Education or Pharmaceutical Promotion?' International Journal of Health Services, 19(4), 663-679. Lexchin, J. (1993) 'Interactions Between Physicians and the Pharmaceutical Industry: What Does the Literature Say?' Canadian Medical Association Journal 149(10), 1401-1407. Lexchin, J. (1994a) 'Canadian Marketing Codes: How Well Are they Controlling Pharmaceutical Promotion?' International Journal of Health Services, 24(1), 91-104. 100 Lexchin, J. (1994b) 'Who Needs Faster Drug Approval Times In Canada: The Public of the Industry?' InternationalJournal of Health Services, 24(2), 253-264. Lexchin, J. (1997) 'Enforcement of Codes Governing Pharmaceutical Promotion: What Happens When Companies Breach Advertising Guidelines?' Canadian Medical Association Journal, 156(3), 351-356. Lexchin, J. (1998) 'Improving the Appropriateness of Physician Prescribing,' International Journal of Health Services 28(2), 253-267. Logan, A . G . (1984) 'Report of the Canadian Hypertension Society's Consensus Conference on the Management of Mi ld Hypertension.' Canadian Medical Association Journal, 131(9), 1053-1056. Lomas, J., G . M . Anderson, K. Doimnick-Pierre, E. Vayda, M.W. Enkin and W.J. Hannah. (1989) 'Do Practice Guidelines Guide Practice? The Effect of a Consensus Statement on the Practice of Physicians.' New England Journal of Medicine, 321(19), 1306-1311. Long, M.J. , K . M . Cummings, and K . B . Frisof. (1983) 'The Role of Perceived Price in Physicians' Demand for Diagnostic Tests' Medical Care 21(2), 243-250. Lowy, D.R., M.D. Lowy, and R.S. Warner. (1972) ' A Survey of Physicians' Awareness of Drug Costs' Journal of Medical Education 47(5), 349-351. MacDonald, M.B. , G.P. Laing, M.P. Wilson, and T.W. Wilson. (1999) 'Prevalence and Predictors of White-Coat Response In Patients with Treated Hypertension' Canadian Medical Association Journal, 161(3), 265-269. Maclure, M . , C. Dormuth, T. Naumann, J. McCormack, R. Rangno, C. Whiteside and J.M. Wright. (1998) 'Influences of Educational Interventions and Adverse News About Calcium-channel Blockers on First-line Prescribing of Antihypertensive Drugs to Elderly People in British Columbia.' The Lancet 352, 943-78. Mattson, W.R., and E.G. Dick. (1995) 'Global Pharma 20/20' Journal of Pharmaceutical Marketing and Management, 10(2/3):3-18. McAlister, F.A., K . K . Teo, R.Z. Lewanczuk, G. Wells, and T.J. Montague. (1997) 'Contemporary Practice Patterns in the Management of Newly Diagnosed Hypertension.' Canadian Medical Association Journal, 157(1), 23-30. McGahan, A . M . (1994) 'Industry Structure and Competitive Advantage' Harvard Business Review November-December, 115-124. Merlo,, J., A . Wessling, and A. Melander. (1996) 'Comparison of Dose Standard Units for Drug Utilization Studies.' European Journal of Clinical Pharmacology, 50(1,2), 27-30. Miller, L .G . and A. Blum. (1993) 'Physician Awareness of Prescription Drug Costs: A Missing Element of Drug Advertising and Promotion' Journal of Family Practice 36(1), 33-36. 101 Morgan, P.P., and L . Cohen. (1995) 'Off the Prescription Pad and Over the Counter: The Trend Toward Drug Deregulation Grows' Canadian Medical Association Journal 152(3), 387-389. Morgan, S. (1998a) 'Issues for Pharmaceutical Policy.' In, Canada Health Action: Building on the Legacy. Papers Commissioned by the National Forum on Health. Volume 4. The National Forum on Health: Ottawa. Morgan, S. (1998b) Quebec's Drug Insurance Plan: A Prescription for Canada? Discussion paper H P R U 98-2D. Centre for Health Services and Policy Research: U B C . Morgan, S., and M . Barer. (1997) Evaluating Amendments to the Canadian Patent Act Discussion Paper H P R U 97:ID, Centre for Health Services and Policy Research, U B C . Munnich, F.E., and K . Sullivan. (1994) 'The Impact of Recent Legislative Change in Germany.' PharmacoEconomics 6(Sup 1), 22-27'. Myers, M.G. , S.G. Carruthers, F.H. Leenen, and R.B. Haynes. (1989) 'Recommendations from the Canadian Hypertension Society Consensus Conference on the Pharmacologic Treatment of Hypertension.' Canadian Medical Association Journal, 140(10), 1141-1146. Nagurney, J.T., R.L. Brham, and G.G. Reader. (1979) 'Physician Awareness of Economic Factors in Clinical Decision-Making' Medical Care 17(7), 727-736. Newhouse, J.P. (1992) 'Medical Care Costs: How Much Welfare Loss?' Journal of Economic Perspectives 6(1), 3-21. Nordhaus, W.D. (1997) 'Do Real-Output and Real-Wage Mesures Capture Reality? The History of Lighting Suggests Not.' In, Bresnahan, T.F., and R.J. Gordon (eds.) The Economics of New Goods. University of Chicago Press: Chicago. Ogilvie, R.I, E.D. Burgess, J.R. Cusson, R.D. Feldman, L . A . Leiter, and M . G . Myers. (1993) 'Report of the Canadian Hypertension Society Consensus Conference: 3. Pharmacologic Treatment of Essential Hypertension.' Canadian Medical Association Journal, 149(5), 575-584. Oppenhiem, G.L., S.H. Erickson, and C. Ashworth. (1981) 'The Family Physician's Knowledge of the Cost of Prescribed Drugs' Journal of Family Practice 12(6), 1027-1030. Palmer, N . (1997) National Pharmacare Cost Impact Study Palmer D'Angelo Consulting Inc.: Ottawa. Pauly, M . V . (1974) 'Overinsurance and Public Provision of Insurance: The Role of Moral Hazard and Adverse Selection' Quarterly Journal of Economics 88(1), 44-62. Peele, P.B. (1993) 'Evaluating Welfare Losses in the Health Care Market' Journal of Health Economics 12(2), 205-208. Pharmaceutical Manufacturers Association of Canada (PMAC). (1997) A Response of the Pharmaceutical Manufacturers Association of Canada to the National Forum on 102 Health's Synthesis Report Entitled 'Directions for a Pharmaceutical Policy in Canada' P M A C : Ottawa Pollak, R.A. (1981) 'The Social Cost of Living' Journal of Public Economics 15(3), 311-336. Reekie, W.D. (1970) 'Some Problems Associated with the Marketing of Ethical Pharmaceutical Products.' Journal of Industrial Economics, 19(1), 33-39. Reeves, R.A., J.G. Fodor, C.I. Gryfe, C. Patterson and J.D. Spence. (1993) 'Report of the Canadian Hypertension Society Consensus Conference: 4. Hypertension in the Elderly.' Canadian Medical Association Journal, 149(6), 815-820. Reinhardt, U.E. (1985) 'The Theory of Physician-Induced Demand: Reflections After a Decade' Journal of Health Economics 4(2): 187-193. Reinhardt, U.E. (1998) 'Abstracting from Distributional Effects, this Policy is Efficient' in Barer, M . , T. Getzen and G. Stoddart (Eds.) Health, Health Care and Health Economics Wiley & Sons: New York. Rice, T. (1998) The Economics of Health Reconsidered. Health Administration Press: Chicago. Rizzo, J.A. (1999) 'Advertising and Competition in the Ethical Pharmaceutical Industry: The Case of Antihypertensive Drugs.' Journal of Law and Economics, 17(1 .pt. 1), 89-116. RTCP (Restrictive Trade Practices Commission). (1963) Report Concerning the Manufacture, Distribution and Sale of Drugs. Queen's Printer: Ottawa. Rubin, P.H., and J.L. Schrag. (1999) 'Mitigating Agency Problems by Advertising, with Special Reference to Managed Health Care.' Southern Economic Journal, 66(1), 39-60. Ryan, M . , B. Yule, C. Bond, and R.J. Taylor. (1996) 'Do Physicians' Perceptions of Drug Costs Influence Their Prescribing?' PharmacoEconomics 9(4), 321-331. Satterthwaite, M . A . (1979) 'Consumer Information, Equilibrium Industry Price, and the Number of Sellers' The Bell Journal of Economics, 483-502. Schwartz, R.K, S.B. Soumerai, and J. Avorn. (1989) 'Physician Motivations for Nonscientific Drug Prescribing' Social Science and Medicine 28(6), 577-582. Scitovsky, A . A . (1967) 'Changes in the Costs of Treatment of Selected Illnesses, 1951 -1965' American Economic Review 57(7), 1182-1195. Scott, A. , A . Shiell, and M . King. (1996) 'Is General Practitioner Decision-making Associated With Patient Socio-Economic Status?' Social Science and Medicine 42(1), 33-45. Sen, A. (1982) Choice, Welfare and Measurement. Basil Blackwell: Oxford. Sheldon, T.A., and G.D. Smith. (1993) 'Consensus Conferences as Drug Promotion.' The Lancet, 341, 100-102. 103 Siegel, D. and J. Lopez. (1997) 'Trends in Antihypertensive Drug Use in the United States: Do the JNC V Reconomendations Affect Prescribing?' Journal of the American Medical Association 278(21), 1745-1748. Silcock, J , M . Ryan, C. Bond, and R J . Taylor. (1997) 'The Cost of Medicines in the United Kingdom: A Survey of General Practitioners' Opinions and Knowledge' PharmacoEconomics 11(1), 56-63. Sleath,. B. , B. Svarstad and D. Roter. (1997) 'Physician vs. Patient Initiation of Psychotropic Prescribing in Primary Care Settings: A Content Analysis of Audiotapes.' Social Science and Medicine 44(4), 541-548. Soumerai, S.B., D. Ross-Degnan, E.E. Fortess, and J. Abelson. (1993) ' A Critical Analysis of Studies of States Drug Reimbursement Policies: Research in Need of Discipline.' The Milbank Quarterly, 71(2), 217-252. Soumerai, S.B., T J . McLaughlin, and J Avorn. (1989) 'Improving Drug Prescribing in Primary Care: A Critical Analysis of Experimental Literature' Milbank Quarterly 67(2), 268-317. Spence, J.D. (1989) 'Stepped Care for Hypertension is Dead, But What Wil l Happen?' Canadian Medical Association Journal, 140(10), 1133-1136. Spingarn, R.W., J A . Berlin and B.L. Strom. (1996) 'When Pharmaceutical Manufacturers' Employees Present Grand Rounds, What Do Residents Remember?' Academic Medicine, 71(1), 86-88. Squires, B.P. (1993) 'Physicians and the Pharmaceutical Industry.' Canadians Medical Association Journal, 149(10), 1391-1392. Steele, M.A . , D.T. Bess, V . L . Franse, and S.E. Graber. (1989) 'Cost Effectiveness of Two Interventions for Reducing Outpatient Prescribing Costs' The Annals of Pharmacology 23, 497-500. Stelfox, H.T., G. Chua, K . O'Rourke, and A.S. Detsky. (1998) 'Conflict of Interest in the Debate Over Calcium-channel Antagonists.' The New England Journal of Medicine 338(2), 101-106. Suslow, V . Y . (1996) 'Measuring Quality Change in the Market for Anti-Ulcer Drugs,' in Competitive Strategies in the Pharmaceutical Industry, ed. R.B. Helms. (Washington: American Enterprise Institute) Tamblyn, R. (1997) 'Medication Use in the Seniors Population: Optimization of Physician Prescribing As a Means of Preventing Drug-Related Illness' Canadian Journal on Aging/Canadian Public Policy 16/23(sup), 147-161. Tamblyn, R., and R. Perreault. (1998) 'Encouraging the Wise Use of Prescription Medication by Older Adults.' In, Canada Health Action: Building on the Legacy. Papers Commissioned by the National Forum on Health. Volume 2. The National Forum on Health: Ottawa. Temin, P. (1980) Taking Your Medicine: Drug Regulation in the United States. Harvard University Press: Cambridge. 104 TI (Therapeutics Initiative). (1995a) 'Drug Choice in the Treatment of Hypertension: Part 1. ' Therapeutics Letter, Issue 7, June 1995. TI (Therapeutics Initiative). (1995b) 'Drug Choice in the Treatment of Hypertension: Part 2. ' Therapeutics Letter, Issue 8, July/August 1995. Tirole, J. (1988) The Theory of Industrial Organization. MIT Press: Massachusetts. Torrance, G.W., W.H. Thomas, and D.L. Sackett. (1972) ' A Utility Maximization Model for Evaluations of Health Care Programs.' Health Services Research, 7(2), 118-133. Towse, A . (1994) 'Health Economics—A Useful Tool or a Threat?' Scrip Magazine May 1994,30-33. Walker, R. (1995) 'Generic Medicines: Reducing Cost at the Expense of Quality?' PharmacoEconomics 7(5), 375-377. Walley, T., R. Wilson and J. Bligh. (1995) 'Current Prescribing in Primary Care in the U K ' PharmacoEconomics 7(4), 320-331. Walzak, D. S. Swindells, and A. Bhardwaj. (1994) 'Primary Care Physicians and the Cost of Drugs: A Study of Prescribing Practices Based on Recognition and Information Sources' Journal of Clinical Pharmacology 34(12): 1159-1163 Wang, T.J, J.C. Ausiello and R.S. Stafford. (1999) 'Trends in Antihypertensive Drug Advertising, 1985-1996.' Circulation, 99(15), 2055-2057. Weber, M . A . , and J.H. Laragh. (1993) 'Hypertension: Steps Forward and Steps Backward—the Joint National Committee Fifth Report' Archives of Internal Medicine, 153(2),T49-152. Wivell, M . K . , and D.A. O'Fallon. (1992) 'Drug Overpromotion: When Is a Warning Not a Warning?' Trial, 82(11), 21-27. Wright, J .M., C H . Lee, and G.K. Chambers. (1999) 'Systematic Review of Antihypertensive Therapies: Does the Evidence Assist in Choosing a First-line Drug?' Canadian Medical Association Journal, 161(1), 25-32. Yule, B., M . Ryan, C. Bond, and R.J. Taylor. (1991) 'Awareness of drug costs among general practitioner trainees' Medical Education 25(6), 536-538. 105 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0089638/manifest

Comment

Related Items