Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The association of young maternal age and low socioeconomic status with poor birth outcomes in urban… Peet, Konnie C. 1993

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

Item Metadata

Download

Media
831-ubc_1993_fall_peet_konnie.pdf [ 19.34MB ]
Metadata
JSON: 831-1.0086285.json
JSON-LD: 831-1.0086285-ld.json
RDF/XML (Pretty): 831-1.0086285-rdf.xml
RDF/JSON: 831-1.0086285-rdf.json
Turtle: 831-1.0086285-turtle.txt
N-Triples: 831-1.0086285-rdf-ntriples.txt
Original Record: 831-1.0086285-source.json
Full Text
831-1.0086285-fulltext.txt
Citation
831-1.0086285.ris

Full Text

THE ASSOCIATIONOF YOUNG MATERNAL AGE AND LOWSOCIOECONOMIC STATUS WITH POOR BIRTH OUTCOMESIN URBAN BRITISH COLUMBIAbyKONNIE C. PEETB.Sc., University of Guelph, 1978B.A., McMaster University, 1985A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCE(Health Services Planning and Administration)inTHE FACULTY OF GRADUATE STUDIESDepartment of Health Care and EpidemiologyWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAOctober 1993© Konnie C. Peet, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature) Department of ^ect^C,otc--. via Ep*,(AlemlcAcz, %The University of British ColumbiaVancouver, CanadaDate Oc..,kc,b es- vcmDE-6 (2/88)Page iiABSTRACTThe reduction of inequalities in health and longevity have been endorsed by variouslevels of government, associations of health professionals and others in Canada.Despite the proportional decline in births to women less than 20 years of age in both theUnited States and Canada, births to young mothers remain of particular interest.Infants born to teenaged mothers have been found to have elevated rates of various poorbirth outcomes, which throughout their lives continue to put them at higher risk forpoor health and reduced life expectancy.There has been considerable research into the apparent association between youngmaternal age (MA) and elevated rates of such outcomes as preterm births, low birthweight and infant mortality. There is an equally large body of research that suggeststhat many of the teenagers having babies live in adverse social and economic circum-stances and that the elevated incidence of various birth outcomes is related to theirsocioeconomic status (SES). Unfortunately only a small portion of the research has beendesigned to allow for either the simultaneous analysis of effect of MA and SES or theanalysis of a potential association between these two risk factors. Additionally, sincethe introduction of Medicare there has only been one study which has addressed anyof these issues in a Canadian context. This present study was undertaken to investigatethe relationships between maternal neighborhood poverty, MA and births outcomes inBritish Columbia (B.C.).Routinely collected data from the Division of Vital Statistics of the B.C. Ministry ofHealth and Ministry Responsible for Seniors for 1985 through 1988 was used to obtainmaternal, birth and birth outcome data for first single births to 2,738 mothers under 20years and 39,540 women living in Vancouver, Victoria, Kamloops, Kelowna and PrincePage iiiGeorge. The mother's postal code was used to link summary SES information, providedby Statistics Canada from the census tract of the mother's residence, to the birth relatedinformation. Birth data were then ranked by this poverty information and divided intoquintiles.There was a significant association between MA and SES with respect to the distributionof births. The percentage of all birth to women under 35 years that occurred to mothersless than 20 years increased from 5.6% in the quintile with the least amount of povertyto 9.7% in that with the highest amount of poverty.Mantel-Haenszel chi-square tests allowed for the assessment of the independentassociation of each of MA and SES with various birth outcomes. Young MA was foundto be a significant risk factor for low birth weight (P<0.001)), infant mortality (P<0.001))and postneonatal mortality (P<0.001). Low SES was found to be significantly associatedwith increased rates of low birth weight (P<0.001), small for gestational age births(P<0.001) and congenital anomalies (P<0.001). Odds ratios and confidence intervals forthese results were also calculated and discussed. The discussion also dealt withresearch and other implications of these results as well as the potential for the methodemployed in this study to be a mechanism for tracking secular changes in inequalitiesin health in British Columbia.Page ivTABLE OF CONTENTSAbstract^ iiTable of Contents^ ivList of Tables viiiList of Figures^ xList of Maps xiAcknowledgements^ xiiChapter 1 INTRODUCTION^ 1REFERENCES^ 5Chapter 2 HISTORY OF BIRTHS TO YOUNG MOTHERS^6TEENAGE BIRTH RATES^ 6TEENAGE BIRTHS RELATIVE TO OTHER BIRTHS^ 13REFERENCES^ 20Chapter 3 THE INDEPENDENT EFFECTS OF YOUNG MATERNAL^22AGE AND POOR SOCIOECONOMIC CIRCUMSTANCES:A REVIEW OF THE LITERATUREINTRODUCTION^ 22THE ASSOCIATION BETWEEN YOUNG MATERNAL AGE AND POOR^25SOCIOECONOMIC CIRCUMSTANCESANALYSIS OF THE IMPACT OF LOW SCIOECONOMIC STATUS AND^27YOUNG MATERNAL AGE ON BIRTH OUTCOMESBirth Weight and Low Birth Weight Rates^ 29I^Impact of Poor Socioeconomic Circumstances^29II^Impact of Young Maternal Age 32III^Interaction Between Socioeconomic Circumstances and^36Maternal AgePage vPrematurity Rates^ 37I^Impact of Poor Socioeconomic Circumstances^37II^Impact of Young Maternal Age^ 38III^Interaction Between Socioeconomic Circumstances and^39Maternal AgeSmall for Gestational Age Rates^ 40I^Impact of Poor Socioeconomic Circumstances^40II^Impact of Young Maternal Age^ 41III^Interaction Between Socioeconomic Circumstances and^42Maternal AgeStillbirth Rates 42I^Impact of Poor Socioeconomic Circumstances^42II^Impact of Young Maternal Age^ 43III^Interaction Between Socioeconomic Circumstances and^44Maternal AgePerinatal Mortality Rates^ 44I^Impact of Poor Socioeconomic Circumstances^44II^Impact of Young Maternal Age^ 46III^Interaction Between Socioeconomic Circumstances and^47Maternal AgeInfant Mortality Rates^ 47I^Impact of Poor Socioeconomic Circumstances^48II^Impact of Young Maternal Age^ 49III^Interaction Between Socioeconomic Circumstances and^50Maternal AgeNeonatal Mortality Rates^ 50I^Impact of Poor Socioeconomic Circumstances^51II^Impact of Young Maternal Age^ 53III^Interaction Between Socioeconomic Circumstances and^55Maternal AgePostneonatal Mortality Rates^ 55I^Impact of Poor Socioeconomic Circumstances^56II^Impact of Young Maternal Age^ 57III^Interaction Between Socioeconomic Circumstances and^59Maternal AgeCongenital Anomaly Rates^ 60I^Impact of Young Maternal Age^ 60II^Interaction Between Socioeconomic Circumstances and^61Maternal AgeSUMMARY 61REFERENCES^ 63Page viChapter 4 RATIONALE AND METHODS^ 69RATIONALE AND PURPOSE^ 69METHOD^ 70Study Population^ 70Data Sources 72I^Maternal and Still/Live Birth Related Data^72a) Physician's Notices of (live and still) Birth 72b) Death Registry^ 73c) Health Status Registry 73II^Socioeconomic Status Date^ 74Procedures 75I^Data Base Development 75II^Variable Definition^ 78Data Analyses 82I^Evaluation of Data Base Development Methods^82II^Evaluation of Variable Definitions^ 83III^Evaluation of Birth and Birth Outcome Data 83REFERENCES 88Chapter 5 STUDY RESULTS^ 91EVALUATION OF DATA BASE DEVELOPMENT METHODS^91Development of Neighborhood Poverty Quintiles 92Analysis of the Effect of Various Groupings of the CMA/CAs on a^97Selection of Birth OutcomesInfant Death Linkage^ 98Congenital Anomalies Linkage^ 98Allocating Stillbirths to NPQs 100Missing Data^ 100EVALUATION OF VARIABLE DEFINITIONS^ 101ANALYSIS OF BIRTH AND BIRTH OUTCOME DATA^ 103Incidence of Single First Live Births^ 103Analysis of Birth Outcomes 104I^Stillbirths^ 104II^Birth Weight 106a) Low Birth Weight^ 108b) Very Low Birth Weight 112III^Small for Gestational Age 115Page viiIV^Infant Mortality^ 119a) Neonatal Mortality 122b) Post Neonatal Mortality^ 124V^Congenital Anomalies 127REFERENCES 131Chapter 6 DISCUSSION OF RESULTS AND IMPLICATIONS^132EVALUATION OF STUDY DESIGN^ 132DISCUSSION OF RESULTS^ 134The Association Between Young Maternal Age and Poor Socioeconomic 134StatusThe Association of Young Maternal Age and Poor Socioeconomic Status 136with Various Birth OutcomesConclusions^ 145LIMITATIONS 146IMPLICATIONS^ 150Research Implications^ 150Other Implications 153Conclusion^ 153REFERENCES 155APPENDIX A^ 156APPENDIX B 157APPENDIX C^ 162Page viiiLIST OF TABLES2.1 Births per 1000 Women by Year of Age, for all Women; United States,1955-198372.2 Births per 1000 Women by Year of Age, for all Women; British Columbia,1959-1988102.3 Population by Age Groups, for all Women; British Columbia, 1959-1988 165.1 Data Base Development 925.2 Development of Neighborhood Poverty Quintiles 935.3 Distribution of Data Between the Five CMAs/CAs 945.4 Income Segregation in the Five CMAs/CAs 965.5 Infant Death Linkage 985.6 Registrations to the B.C. Health Status Registry Made Withinthe First Year of Life995.7 Congenital Anomalies Linkage 995.8 Allocation of Stillbirths into Neighborhood Poverty Quintiles 1005.9 Distribution of the Percentage of Children Living Below the Low 102Income Cutoffs5.10 Rates of Birth Outcomes 1035.11 Incidence of First Single Live Births by Maternal Age Groups and NPQs 1045.12 Stillbirths by Maternal Age Groups Stratified by NPQs 1055.13 Stillbirths by NPQs Stratified by Maternal Age Groups 1065.14 Birth Weights by Maternal Age Groups and NPQs 1075.15 Low Birth Weight Births by Maternal Age Groups Stratified by NPQs 1085.16 Low Birth Weight Births by NPQs Stratified by Maternal Age Groups 110Page ix5.17 Very Low Birth Weight Births by Maternal Age Groups Stratified by^113NPQs5.18 Very Low Birth Weight Births by NPQs Stratified by Maternal Age 114Groups5.19 Gestational Ages by Maternal Age Groups and NPQs 1165.20 Small for Gestational Age Births by Maternal Age Groups Stratifiedby NPQs1165.21 Small for Gestational Age Births by NPQs Stratified by Maternal 118Age Groups5.22 Infant Mortality by Maternal Age Groups Stratified by NPQs 1205.23 Infant Mortality by NPQs Stratified by Maternal Age Groups 1215.24 Neonatal Mortality by Maternal Age Groups Stratified by NPQs 1235.25 Neonatal Mortality by NPQs Stratified by Maternal Age Groups 1245.26 Postneonatal Mortality by Maternal Age Groups Stratified by NPQs 1255.27 Postneonatal Mortality by NPQs Stratified by Maternal Age Groups 1265.28 Congenital Anomalies by Maternal Age Groups Stratified by NPQs 1285.29 Congenital Anomalies by NPQs Stratified by Maternal Age Groups 129B.1 Birth Outcomes to Sub-Groups of Teenaged Mothers 158B.2 Births to Sub-Groups of Teenaged Mothers by Neighborhood Quintiles 159B.3 Birth Outcomes to Sub-Groups of Teenaged Mothers by Neighborhood 160Poverty QuintilesC.1 Chi-Square Test Comparing Births to Teens with Births to Mothers 162Aged 20-34 YearsC. 2 ANOVA Comparing Birth Weight by Maternal AGe Groups and NPQs 162C.3 ANOVA Comparing Gestational Age by Maternal Age Groups and 162NPQsPage xLIST OF FIGURES^2.1^Ratio of the Live Births to Women Aged < 19 Years Relative to Live^15Births to Women Aged > 20 Years (British Columbia, 1950-1988)2.2^Ratio of the Live Births to Women Aged 19 Years Relative to Live^18Births to Women Aged > 20 Years (British Columbia, 1950-1988)2.3^Ratio of the Live Births to Women Aged 17 and 18 Years Relative to^19Live Births to Women Aged > 20 Years (British Columbia, 1950-1988)2.4^Ratio of the Live Births to Women Aged < 16 Years Relative to Live^19Births to Women Aged > 20 Years (British Columbia, 1950-1988)5.1^Income Variation^ 95Page xiLIST OF MAPSA.1 Location of Neighborhood Poverty Quintiles in Vancouver (1)^156aA.2 Location of Neighborhood Poverty Quintiles in Vancouver (2)^156bA.3 Location of Neighborhood Poverty Quintiles in Vancouver (3)^156cA.4 Location of Neighborhood Poverty Quintiles in Victoria (1) 156dA.5 Location of Neighborhood Poverty Quintiles in Victoria (2)^156eA.6 Location of Neighborhood Poverty Quintiles in Kamloops 156fA.7 Location of Neighborhood Poverty Quintiles in Kelowna (1)^156gA.8 Location of Neighborhood Poverty Quintiles in Kelowna (2) 156hA.9 Location of Neighborhood Poverty Quintiles in Prince George^156iPage xiiACKNOWLEDGEMENTSIn undertaking the research reported in this thesis I was supported by a National HealthM.Sc. Fellowship from Health and Welfare Canada. The research was also supportedby grants from the British Columbia Health Research Foundation (Grant # 171(91-1))and the British Columbia Medical Services Foundation (Grant # 91-31).The data for this project were provided by the Division of Vital Statistics of the BritishColumbia Ministry of Health and Ministry Responsible for Seniors, the Planning andStatistics Division of the British Columbia Ministry of Finance and Statistics and theOccupational and Environmental Health Research Section of Statistics Canada.I gratefully acknowledge the suggestions of R. Wilkins and my committee members C.Hertzman and S. Manson-Singer. Thanks are especially extended to S.B. Sheps for hisinterest and patient supervision and to S. Wiggins for her untiring support throughoutall phases of the project.Page 1CHAPTER ONEINTRODUCTIONHealth and longevity have been unequally distributed in human populations through-out history. Even within developed countries such as Canada, morbidity and mortalitydifferentials continue. Health and Welfare Canada, in Achieving Health For All: AFramework For Health Promotion',  has identified the reduction of these inequalities inhealth as a high priority. Various provincial governments, associations of healthprofessionals and others have endorsed similar objectives.Within this context teenage pregnancy is of particular interest. Infants born to teenagedmothers have been found to have elevated rates of poor birth outcomes, which areassociated in later life with poor health and reduced life expectancy. There has beenconsiderable research into the apparent association between young maternal age (MA)and elevated rates of preterm births, low birth weight and infant mortality. It is alsoacknowledged that many teenage women having babies live in disadvantaged socialand economic circumstances which are also known to be associated with poor birthoutcomes. Research has shown the incidence of low birth weight, infant mortality andpreterm births to be inversely related to maternal socioeconomic status (SES).Unfortunately there is much confusion in the literature regarding the independent andrelative impact of both young MA and low SES on poor birth outcomes. The results ofsome studies suggest that young MA has a substantial negative impact on birthoutcomes 2,3,4,5,6. In other instances it is concluded that the combined presence ofteenage and poverty result in elevated risks 7. At the other extreme there are studyresults which lead to the suggestion that when the effect of low SES and related variablesPage 2are controlled the incidence of various negative birth outcomes is lowest among teenmothers and increases with advancing MA 8,9 . Thus it is not at all clear if young MA perse is a risk factor for poor birth outcomes independent of poverty, in association withpoverty, or is not a risk factor at all.There are a number of issues which contribute to this apparent confusion in theunderstanding of the impact of MA on birth outcomes. Not the least of these is the factthat SES can only be measured by proxy variables such as occupation, education,income, or race or by a composite of proxy variables such as a weighted sum ofeducation and occupational measures. There does not appear to be a clear consensusas to which proxy is the best or most relevant measure or if the best measure changeswithin different contexts. Additionally, the choice of variable(s) to represent SES in anystudy is often driven by available information rather than by any theoretical considera-tion of which measure is felt to be most relevant.Another important issue is that of context. When considering the relative impact of lowSES on health outcomes the operative political, social and health care systems must alsobe considered. Any of these may mediate the effect of low SES on birth outcomes. Forexample, the existence of publicly funded medical insurance as well as the eligibilitycriteria for and the magnitude of social assistance payments have an impact on theability of those with the lowest SES to access health care and other services during andimmediately following pregnancy. This in turn affects the outcome of the pregnancy.For this reason it is not surprising that studies carried out in jurisdictions with differentsocio-political environments lead to different conclusions about the impact of youngMA and low SES on birth outcomes.Page 3The utilization of a wide variety of research methods may also have some bearing on theapparent confusion surrounding the independent impact of young MA on birthoutcomes. Designs range from population based retrospective studies utilizing second-ary data collected for some other purpose (usually administrative) to hospital based,small sample studies designed specifically to consider the existence and magnitude ofthe impact of teenage on pregnancy outcomes. Biases resulting from different types ofstudy populations and designs may be substantial enough to significantly contribute tothe apparent contradictions in the literature.Within the Canadian context there do not appear to be any studies carried out since theintroduction of Medicare which have sought to assess the influence of young MA andlow maternal SES on birth outcomes. The study described in this thesis was designedto utilize a relatively inexpensive ecological design involving linkages between variousdata sets created from routinely collected data in order to examine the effect of bothyoung MA and low SES on birth outcomes in urban British Columbia (B.C.). Thepurpose of this study was to study relationships between the proportion of childrenunder 18 years living in poverty within a given urban neighborhood (neighborhoodpoverty level), MA and birth outcomes such as birth weight, gestational age, etc.The remainder of this thesis, which describes the above study, is organized as follows:Chapter 2 chronicles the history of teen birthrates relative to the rates for other womenin order to establish the magnitude of the issue,Chapter 3 reviews studies that consider both MA and SES. First there is a review ofarticles which consider the possibility of an association between young MAand SES. This is followed by a critical summary of the literature that eitherPage 4simultaneously considers the impact of young MA and poor SES in theiranalysis of the factors that are associated with poor birth outcomes to youngmothers or considers one of the two risk factors while controlling for theother. Particular attention is paid to the designs and analyses employed inthe various studies in an attempt to discover if these are a factor when thereappears to be no consensus about the independent impact of young MAand low SES on a particular birth outcome,Chapter 4 describes the rationale and methods used in this study,Chapter 5 presents the results of the study and describes the strengths and weaknessof the study design,Chapter 6 discusses the results and limitations of the study, and related implications.Page 5REFERENCES1.Epp J. Achieving health for all: A framework for health promotion. Ottawa, Ministerof Supply and Services, Canada, 1986.2. Stickle G. Overview of incidence, risks, and consequences of adolescent pregnancyand childbearing. In: McAnarney E, Stickle G ed. Pregnancy and childbearingduring adolescence. New York: Alan R. Liss, Inc., 1981: 5-17.3. Dunn H. Social Aspects of Low Birth Weight. Can Med Assoc J 1984;130:1131-140.4.McCormick M, Shapiro S, Starfield B. High-risk young mothers: Infant mortality andmorbidity in four areas in the United States, 1973-1978. Am J Public Health1984;74(1):18-23.5. Committe to Study the Prevention of Low Birthweight. Preventing low birthweight.Washington, D.C.: National Academy Press, 1985: 284.6. Dollfus C, Patetta M, Siegel E, Cross A. Infant mortality: A practical approach to theanalysis of the leading causes of death and risk factors. Pediatrics1990;86(2):176-83.7. Lieberman E, Kenneth J, Monson R, Scoenbaum S. Risk factors accounting for racialdifferences in the rate of premature birth. N Engl J Med 1987;317:743-8.8. Sukanich A, Rogers K, McDonald H. Physical maturity and outcome of pregnancy inprimiparas younger than 16 years of age. Pediatrics 1986;78(1):31-6.9. McAnarey E. Young maternal age and adverse neonatal outcome. Am J Dis Child1987;141:1053-9.Page 6CHAPTER TWOHISTORY OF BIRTHS TO YOUNG MOTHERSAlthough a great deal has been written about teenage pregnancy it is notable that therehas been so little discussion of the longitudinal trends in teenage birth rates and theirmagnitude relative to birth rates to older mothers. This is unfortunate because it is onlywith this kind of data that the magnitude of the problem of teenage births can beestablished.TEENAGE BIRTH RATESTable 2.1 was abstracted from American birth information published by Hollingsworthand Felice'. From these data it can be seen that for mothers 15 years and under birth ratespeaked in the early 1970s and thereafter declined such that rates in 1983 were similar towhat they had been in 1950. Birth rates for mothers aged 16 through 19 years appear tohave peaked much earlier; in the late 1950s. Since that time they also have declined sothat in 1983 they were 30% to 53% below their highest values.There is support in the literature for the suggestion that birth rates to American teenagewomen 16 years and older began to decline much earlier than the rates for youngerwomen. The age specific birth rate for women aged 15 through 19 years was found tohave increased from 56.9/1000 in the 1940s to 94.2/1000 by 1956 2 . Since that time itdeclined to 81.3/1000 in 1962 2, 68.3/1000 in 19703, and 58.1/1000 in 1974 2 . In the 1980sthe rates for this age group dropped even further. In 1980 they were reported to be 55.5/100034, 51.7/1000 in 19832 and 50.9/1000 in 1984 3 . The only anomaly in the literatureregarding birth rates for this group of women was a rate of 96/1000 recorded in 1981.Page 7Unfortunately there is not enough information provided in McAnarney and Hendee'sarticle to allow for an explanation of this anomaly. Data for 18 and 19 year olds alsofollow this pattern of declining rates after the late 1950s. The 1961 rate of 160/1000 haddropped to 90/1000 by 1973 6 and by 1980 it was reported to be 82.6/1000'. The 1960 rateof 56.9/1000 for mothers age 16 and 17 years had dropped to 42.0/1000 by 1978 7 .Table 2.1Births per 1000 Women by Year of Age, For ALL Women: United States, 1955 - 1983Year <14 15 16 17 18 191955 - 1959 6.0 20.1 45.7 85.8 136.2 184.01960 - 1964 5.4 17.8 40.2 75.8 122.5 169.21965 5.2 16.5 36.0 66.4 105.4 142.41966 5.3 16.4 35.5 64.8 101.8 136.11967 5.3 16.5 35.3 63.2 97.5 129.51968 5.7 16.7 35.2 62.6 95.7 125.21969 6.0 17.4 35.8 63.1 95.7 124.51970 6.6 19.2 38.8 66.6 98.3 126.01971 6.7 19.2 38.3 64.2 92.4 116.11972 7.1 20.1 39.3 63.5 87.1 105.01973 7.4 20.2 38.8 61.5 83.1 98.51974 7.2 19.7 37.7 59.7 80.5 96.21975 7.1 19.4 36.4 57.3 77.5 92.71976 6.8 18.6 34.6 54.2 73.3 88.71977 6.7 18.2 34.5 54.2 73.8 89.51978 6.3 17.2 32.7 52.4 72.2 88.01979 6.4 17.2 32.8 52.5 73.5 90.41980 6.5 17.4 33.1 53.1 74.6 92.51981 6.3 17.0 32.1 51.5 72.2 98.51982 6.4 17.2 32.5 51.9 72.1 98.11983 6.4 17.1 32.1 51.1 70.8 83.6Source: abstracted from Hollingsworth D, Felice M. Teenage pregnancy:A multiracial sociologic problem, Am J Obstet Gynecol 1986;155:741-6.The pattern in U.S. birth rates for women aged 15 and younger also appears to besubstantiated by other sources although the rates fluctuate somewhat (probably be-cause different sources use different denominators in calculating their rates and mostPage 8often do not describe what these are). Stickle presented age specific fertility rates forAmerican women less than 16 years'. In 1950 the rate for these women was 9.1/1000births. It rose to a high of 11.5/1000 in 1973 and then declined to 9.9/1000 in 1978. By1983 and 1984 the rates for women less than 15 years were 1.1/1000 and 1.2/1000 3 . Asimilar pattern can be observed for mothers aged 10-14 years at the time of delivery. Inthe 1940s, 1950s and early 1960s the birth rate for this group was approximately 1.0/10002s. By 1975 the rate had risen to 1.3/10002 but by the early 1980s it had returned toits earlier value of 1.1/1000 2s. Klerman presented rates for mothers age 14 years andunder which were consistent with the values given by Lee and Zuckerman. In 1966there were 0.8 births per 1000 women in this age group while by 1972 the rate had risento 1.2/1000 9 .Some rates presented in the literature do not appear to follow either of these twopatterns. One source suggested that in 1961 the birth rate for women aged 14 through17 years was approximately 32/1000 while in 1974 the rate was 29/1000 6. A possibleexplanation why the two rates were similar may be that the group was comprised ofwomen 15 years and younger, for whom birth rates rose until approximately 1973, aswell as older women for whom birth rates already began to decline in the late fifties. Inanother study2 rates for mothers aged 15 through 17 years did not follow this pattern ofdeclining rates for older teens quite as nicely; probably because of the influence of the15 year old mothers whose rates only peaked in the early 1970s. In 1966 the rate for thisgroup was reported to be 37.7/1000 while in 1977 it was up slightly to 39.2/1000.Although not exhibiting identical patterns in birth rates to teenaged women, Canadiandata present a picture that has many parallels with that of the United States (U.S.). Table2.2 presents birth rates by various years of age for British Columbia from 1959 throughPage 91988. The data were produced as part of a studya at the University of British Columbia.As was the case with the American data, older teenage mothers appear to haveexperienced their highest birth rates in the late fifties. Teens in B.C. aged 17 through 19years experienced their highest birth rates in 1959 or 1960 after which, for all three MAs,these declined till 1988. Younger mothers also exhibited patterns that showed somesimilarity with their American counterparts. Birth rates for mothers aged 14 years andyounger plateaued in the late 1960s and early 1970s after which they experienced asomewhat irregular and gradual decline. The pattern for mothers aged 15 and 16 yearswas somewhat different from that experienced in the U.S., in that in B.C. birth ratesbegan to decline in the 1960s but regained their previous high values in 1969 and 1970.In the early 1970s they began to fall again and continued to do so ti111988. An interestingfeature of the decline in the birth rates for teen mothers from 14 through 19 years of agewas the substantial decline in rates between 1984 and 1985; a drop that was not sharedby older mothers.Besides the similarity in the general birth rate pattern for teenaged mothers between theU.S. and B.C. there are a number of differences that should be noted. In the late 1950sthe birth rates for older B.C. and American teens were very similar. The B.C. rates of31.1/1000 for 16 year olds and 71.9/1000 for 17 year olds (Table 2.2) were somewhathigher than those in the U.S., while the rates of 130.7/1000 for 18 year olds and 180.3/1000 for 19 year olds were a bit lower. By 1983 the B.C. rates of 13.7, 23.3, 34.6 and 51.0/1000 for 16, 17, 18, and 19 year olds were lower than those of their American counter-parts. For B.C. women aged 16 through 18 years the rates were less than half and for 19year olds they were over 30/1000 less then those of American women of the same age.a. study by K. Peet, S. Wiggins and S.B. Sheps, University of British Columbia, 1992.Page 10Table 2.2Births per 1000 Women, by Year of Age, For All Women:British Columbia, 1959 - 1988Year 14 * 15 16 17 18 19 < 19 > 20 Total1959 0.39 9.4 31.1 71.9 130.7 180.3 81.6 135.9 127.31960 0.51 8.6 28.5 66.4 119.9 185.1 78.8 135.8 126.31961 0.31 7.2 28.8 62.2 120.7 171.5 74.1 130.4 120.71962 0.27 5.9 22.1 60.4 111.9 169.6 69.2 128.5 117.81963 0.32 7.7 24.5 55.0 105.9 167.6 67.1 124.1 113.31964 0.32 5.9 25.0 61.1 96.1 145.1 62.7 115.4 104.91965 0.35 6.5 24.6 59.4 110.4 133.3 63.8 102.8 94.81966 0.49 7.0 23.0 55.0 97.4 138.3 62.4 94.5 87.81967 0.31 7.7 21.8 52.1 88.2 129.9 58.6 91.5 84.61968 0.66 8.6 22.5 50.5 81.0 118.4 55.0 70.6 83.21969 0.43 9.6 26.2 48.6 82.6 117.0 56.3 91.8 84.21970 0.44 9.2 26.5 55.6 84.1 112.6 57.2 91.7 84.21971 0.36 7.7 23.5 47.0 78.1 95.8 49.2 84.4 76.81972 0.35 7.8 22.4 43.3 66.8 94.2 46.1 80.3 72.91973 0.40 7.8 21.8 40.8 49.8 84.8 42.8 76.5 69.31974 0.29 7.5 21.9 40.3 58.9 76.6 41.1 75.6 68.21975 0.36 7.7 21.3 38.9 52.6 73.6 38.6 74.2 66.71976 0.42 6.2 16.6 33.2 47.9 69.5 34.4 72.3 64.31977 0.43 6.9 18.0 31.3 45.7 64.4 33.5 72.4 64.51978 0.44 5.7 17.0 28.2 42.8 61.0 31.3 72.6 63.51979 0.28 5.4 15.4 26.7 45.3 58.1 30.6 71.6 63.71980 0.27 6.0 13.9 28.8 41.1 56.3 29.8 71.3 63.61981 0.29 5.3 14.6 27.5 39.1 55.4 29.4 71.1 63.71982 0.33 5.5 12.6 24.0 40.4 51.9 28.5 71.8 64.51983 0.24 5.2 13.7 23.3 34.6 51.0 27.0 71.2 64.21984 0.23 5.0 12.3 21.8 34.1 51.0 25.7 71.9 64.71985 0.08 2.4 7.2 16.1 27.2 37.2 18.1 71.2 63.31986 0.13 2.2 6.5 16.6 26.5 38.1 17.6 69.1 61.41987 0.07 2.4 6.4 15.4 24.5 37.9 12.3 68.1 60.61988 0.10 2.8 8.0 15.0 26.6 40.4 19.0 68.8 61.5* number of births to women^years divided by the number of females aged 10-14 yearsPage 11Younger B.C. women had rates in 1959 that were substantially lower than those in theU.S. at that time. The rate for women 14 years and younger was 0.39/1000 and that forwomen aged 15 years it was 9.4/1000 (Table 2.2) compared with 6.0/1000 and 20.1 forthose ages in the U.S. (Table 2.1). By 1983 the differences in birth rates were even morepronounced because these two rates for B.C.had fallen below their 1959 values to 0.24/1000 and 5.2/1000, while the U.S. rates remained similar to what they had been.Although the differences between B.C. and the U.S. in the birth rate trends for these twoMAs is probably real, some of the difference in the magnitude of the rates for mothersaged 14 years and under, may have been caused by differences in the method ofcalculation of birth rates.B.C.'s teen birth rate is also low compared to that of other Canadian provinces. A recentreport from Manitobalo presented age specific fertility rates among females aged 15through 19 years for Canada and all the provinces for 1986. B.C.'s rate of approximately22/1000 was the third lowest in the country and below the Canadian average of 24/1000. In this same report a the history of fertility rates for women aged 15 through 19years was presented for Manitoba and Canada. In 1959, B.C.'s fertility rate of 81.6/1000was higher than that of either Manitoba or Canada. Although the rates in both thecountry as a whole and in Manitoba have fallen since that time, in neither case has thedecline been as great as in B.C..Of particular interest is the drop in B.C. birth rates for the various teen MAs between1984 and 1985. Between these two years the age specific birth rates dropped for all teenwomen (Table 2.2). The drop was between 19% (for 18 year olds) and 65% (for those 14years and younger) with the larger drops occurring among the younger teenagers. Thisdramatic drop in birth rates was not experienced by older women nor has it beenreported in the literature for other groups of young mothers. Teenaged women inPage 12Manitoba experienced a fertility rate of 36.4/1000 in 1984 and 35.1/1000 in 1985 10 . Therealso does not appear to be a documented explanation for this drop among youngmothers in B.C.Also of note, although somewhat less spectacular, is the slight increase in birth rates toteen mothers between 1987 and 1988 (Table 2.2). The rate for all teens combinedincreased from 13.5/1000 to 19.0/1000. The percentage increases varied from a high of30% for mothers less than 14 years to a low of 3% for mothers aged 17 years. This trendwas not seen among the adult mothers. The birthrate for women 20 years and older onlyincreased 1%.It must also be noted that this review of birth rates does not allow any for anyconclusions to be made concerning pregnancy rates. Complete statistics on teenagepregnancy should include data on live births, stillbirths, therapeutic abortions, ectopicpregnancies and spontaneous abortions. Only one of the papers reviewed included anyinformation concerning trends in therapeutic abortion rates. This paper 10 presented therates (calculated by dividing the number of therapeutic abortions by the number ofwomen aged 15-44 years) for Canada and Manitoba between 1971 and 1986. TheCanadian rates increased from 6.6/1000 in 1971 to 11.6/1000 in 1979 and then declinedslightly to 10.2/1000 in 1983. The rates in Manitoba were less steady then the Canadianrates but increased from 4.1/1000 in 1971 to 10.2 in 1986. Although this data appearsto support a suggestion that increases in the rates of therapeutic abortions during theseyears were responsible for the decreases in the B.C. birth rates for both mothers aged15 through 19 years as well as those aged 20 through 44 years, more research is requiredin this area. It is not certain that the B.C. therapeutic abortion rates followed the samepattern as they did nationally or in Manitoba. Even if they did, it is also not known whatthe patterns of therapeutic abortion rates for the various maternal age sub-groups were.Page 13TEENAGE BIRTHS RELATIVE TO OTHER BIRTHSThe contribution of births to teenaged mothers to the total birth rate is not only afunction of the magnitude of the teen birth rate. The size of the female teenagepopulation relative to the size of the adult female population and the birth rate of thelatter must also be considered. As a consequence the proportion or percentage of allbirths that occur to teenaged mothers may follow a different pattern than does theteenage birth rate.Data gleaned from the American literature suggests that teen births as a percentage ofall births increased till the late 1970s after which they commenced a steady decline. Twosources suggested that in 1960 births to women less than 20 years made up approxi-mately 14% of births to all women11,1 , that by 1966 this had risen to 17% 12 and was 19%of total births in 1975 3,13. A study which considered changes in Baltimore's childbearingpopulation between 1972 and 1977 found that births to white women less than 20 yearsdeclined from 21.5% to 19.7% of all births, while the drop for black women was from39.2% to 35.2%. This latter proportion is somewhat high and suggests that the trend toa declining number of births to women less than 20 years commenced somewhat earlierin Baltimore than in the rest of the U.S. given that other sources suggest that thepercentage of all births occurring to this group had declined to 17% by 197814, and 19%in 197912. By 1980 there was some national evidence that the proportion of total birthsoccurring to young American mothers was beginning to fall. In 1980 the percentage wasapproximately 15.5% 4,11,1 and by 1983 it had dropped to approximately 13.5% 1,3,2 . InEngland, Scotland and Wales, the percentage of singleton births to women less than 20years was 5.8% in 1958 and 9.8% in 1970, suggesting a pattern with some similarities tothe experiences in the U.S..Page 14This pattern of increasing contribution of births to American teenage mothers to allbirths till the late 1970s followed by decreasing percentages was also experienced bysubgroups of teen mothers. Births to mothers aged 18 and 19 years made up 8.2% of allbirths in 1950, 9.5% in 1960 and 11.7% in 1973 7 . In 1978 the percentage had declined to10% 7, suggesting the beginning of a decline in the percentage of total births occurringto young mothers (although it is hard to judge with only one data point). Thepercentages for mothers aged 17 years and under mimicked this pattern as well. In 1950the percentage of total births in this group of women was 3.8%, rising to 8% in 1973 7 Asin the previously described age groups, by 1978 the percentage had declined to 6.4% 7 .Data from Tasmania showed that the percentages of total births for this age group were4.4% in 1976 and 2.2% in 1982 15 .This pattern of an increasing contribution of births to women less than 20 years to allbirths followed by declining contributions till the mid 1980s is similar to the Canadianexperience. Data produced as part of the previously described study at the Universityof British Columbia (Figure 2.1) showed that the ratio of births to women less than 20years relative to births to women greater than 20 years increased to 0.175 in the latter halfof the 1960s and began to decline in the early 1970s. The decline continued to reach aratio of 0.044 in 1985 at which time the ratios began to exhibit a slight increase. With thedata available it is not really possible to discern if this latest increase was a randomfluctuation or a real increase.Page 15Figure 2.1: Ratio of Live Births to Women Aged 5_ 19 Years Relative toLive Births to Women Aged 20 Years (British Columbia, 1959-88)0.175 —0.150 —0.125 —0.100 —0.075 —0.050 - 0.025 —0.000^•1•1•1•1•1•1•1.1.1.1•Ir1•1•1.I58^60^62^64^66^68 70^72^74^76^78^80^82^84^86^88YearsIn order to fully understand the contribution made by the changes in the birth rates toboth young and older mothers in B.C. to the changes in ratios of their births it isnecessary to also be aware of changes in the number of females in each of these twogroups during this time. Table 2.3 presents B.C. population information for sub-groupsof females aged 15 through 19 years and for females aged 20 through 44 years for 1959through 1988. The number of women aged 20 through 44 years increased steadily from263,200 in 1959 to 589,800 in 1988. Women aged 15 through 19 years increased from52,500 to 119,100 between 1959 and 1975 and then declined 102,100 in 1988.The consequence of a steady increases in the number of women aged 20 through 44(Table 2.3) coupled with a fairly uniform decrease in their age specific fertility rates(Table 2.2) was a smaller than 10 percent increase in the number of births to this groupof mothers. The pattern of increasing and then decreasing number of mothers aged 15through 19 years (Table 2.3) combined with their declining age specific fertility rate(Table 2.2) resulted in very small changes in the number of births to this group ofPage 16mothers till approximately 1975, after which the number of births rapidly declined toless than half this number in 1988. These two distinctly different patterns in the numberof births resulted in the pattern of ratios presented in Figure 2.1.Table 2.3Population by Age Groups, For All Women: British Columbia, 1959-1988Year15-16 15-19 20-44 17-18 191959 21,400 50,100 263,200 19,300 9,4001960 22,700 52,500 264,800 20,100 9,7001961 23,900 54,900 265,200 21,000 10,0001962 25,700 58,400 265,800 12,300 10,4001963 27,700 62,800 268,600 14,100 10,9001964 29,700 68,100 274,900 26,400 11,9001965 31,300 72,800 283,000 28,500 13,0001966 32,600 77,700 293,700 30,700 14,4001967 34,200 82,200 308,100 32,500 15,5001968 35,700 86,300 320,000 34,200 16,5001969 37,100 90,100 331,200 35,700 17,4001970 38,900 94,700 344,100 37,400 18,4001971 41,500 98,400 356,800 38,400 18,4001972 42,500 102,700 371,700 40,900 19,3001973 43,800 107,200 388,600 42,900 20,5001974 45,300 111,400 409,600 43,900 22,0001975 47,200 119,100 423,800 45,100 22,4001976 48,500 116,900 441,300 46,200 22,3001977 47,200 117,500 454,700 47,300 23,0001978 47,000 118,300 471,100 47,900 23,3001979 77,100 118,400 488,600 47,200 24,2001980 46,200 118,300 513,600 47,800 24,3001981 44,200 117,000 537,800 48,700 24,1131982 41,100 112,700 553,500 47,000 24,7001983 39,700 107,000 564,000 43,300 24,0001984 40,300 103,300 575,500 40,600 22,4001985 41,700 101,500 581,900 39,400 20,6001986 42,400 102,100 584,200 40,100 19,5131987 40,700 102,100 589,800 41,700 19,7001988 38,700 102,500 599,500 42,700 20,882Source: data provided by the Division of Vital Statistics, B.C. Ministry of Health and Ministry Responsiblefor Seniors and the Planning and Statistics Department, B.C. Ministry of Finance.Page 17The numbers of B.C. women aged 15 and 16,17 and 18, and 19 years all followed patternsthat were similar to the pattern of all teen mothers (Table 2.3). The number of womenin each age group increased from 1959 until the late 70's (younger teen mothers) or early80's (older teen mothers), after which they all declined slightly. The birth rates beganto decline earlier for older teen mothers than for the younger teens (Table 2.2). Thesedifferences in the timing of changes in population sizes and birth rates were responsiblefor patterns in the number of births for each of the sub-groups that were similar, but notidentical, to the pattern for all teen mothers. As a consequence the pattern exhibited bythe ratios of births to teen mothers to births to older women was similar to the patternsof the various subgroups of teen mothers. The ratio of births to 19 year old mothersrelative to mothers 20 years and older increased to 0.072 in 1966 and 1967 after whichtime it declined to 0.018 in 1985 (Figure 2.2). In 1987 and 1988 it increased slightly to0.019 and 0.020. The maximum ratio for births to mothers aged 17 and 18 years alsooccurred in 1966 (Figure 2.3). It remained at approximately 0.084 ti111970, after whichit fell to 0.021 in 1985, rising slightly to 0.022 in 1988. The ratio of births to mothers aged16 years and under relative to births to mothers aged 20 years and older reached amaximum of 0.023 a few years later in 1969 and did not begin to decline till 1976 (Figure2.4). It reached a low of 0.004 in 1986 after which it leveled off and increased somewhatto 0.005 in 1988.Between 1985 and 1984, the ratios of births to each group of teen mothers relative toadult women dropped between 25 and 43 percent with the largest drops occurringamongst the youngest mothers. This drop was consistent with the drop in all agespecific birth rates for teenage women between those two years. It is also interesting tonote that, between 1987 and 1988 the ratios of births to the various groups of teensrelative to older mothers all rose slightly; once again as a result of a similar change inteen birth rates between those two years.Page 18Figure 2.2: Ratio of Live Births to Women Aged 19 Years Relative toLive Births to Women Aged 20 Years (British Columbia, 1959-88)0.175 -0.150 -0.125 -0.100 -0(12' 0.075 -0.050 -0.025 -0.000 i^i^ I^•^I58^60^62^64^66^68^70^72^74^76^78^80^82^84^86^88YearsThese data for B.C., which for the most part are supported by other Canadian andAmerican data, suggest that both the teenage birth rate and the percent of all births thatare occurring to teenaged women are declining. This decline might suggest that theissue of teenage pregnancy is losing its urgency and no longer needs to be a focus ofconcern. Granted, these declines can only been seen as positive but this does not suggestthat energy no longer needs to expended on issues relating to teenage pregnancy. Thisis not the case for a number of reasons. Firstly, teens continue to have babies withelevated risks for poor birth outcomesb and it remains important to ascertain if youngMA is a risk factor for poor birth outcomes independent of poverty, in association withpoverty, or is not a risk factor at all. Secondly, within an environment that includesincreased access to contraceptives and abortion services it becomes important to knowwhether all teens or primarily those who are either economically advantaged ordisadvantaged are having babies. Only by understanding what factors related to youngMA are associated with elevated rates of poor birth outcomes and knowing theb. For example the Division of Vital Statistics of the B.C. Ministry of Health and Ministry Responsible forSeniors found that in 1987 the low birth weight rate for mothers less than 20 years was 6.1% comparedwith 5.0% for older mothers while in 1988 the rates were 5.8% for the younger mothers and 4.9% for theolder mothers0.04 -0.000.01 -Page 19economic circumstances of the teenagers having babies will it be possible to begin toconsider strategies for efficiently and effectively reducing the number of infants bornto teenaged mothers that have elevated rates of poor birth outcomes.Figure 2.3: Ratio of Live Births to Women Aged 17 and 18 Years Relativeto Live Births to Women Aged 20 Years (British Columbia, 1959-88)0.175 -0.150 -0.125 -0.100 -0tr 0.075 -0.050 -0.025 -0.000 tilt58^60^62^64^66^68^70^72^74^76^78^80^82^84^86^88YearsFigure 2.4: Ratio of Live Births to Women Aged 5. 16 Years Relative toLive Births to Women Aged 20 Years (British Columbia, 1959-88)0.05 -58^60^62^64^66^68^70^72^74^76^78^80^82^84^86^88YearsPage 20REFERENCES1. Hollingsworth D, Felice M. Teenage pregnancy: A multiracial sociologic problem.Am J Obstet Gynecol 1986;155:741-6.2. Lee K, Corpuz M. Teenage pregnancy: Trend and impact on rates of low birth weightand fetal, maternal, and neonatal mortality in the United States. ClinPerinatol 1988;15 (4):929-42.3. Wegman M. Annual summary of vital statistics-1985. Pediatrics 1986;78(6):983-94.4. Adams M, Oakley G, Marks J. Maternal age and births in the 1980s. JAMA1982;247(4):493-4.5. McAnarney E, Hendee W. Adolescent pregnancy and its consequences. JAMA1989;262(1):74-7.6. Blum R, Goldhagen J. Teenage pregnancy in perspective. Clin Pediatrics 1981;20(5):335-40.7. Stickle G. Overview of incidence, risks, and consequences of adolescent pregnancyand childbearing. In: McAnarney E, Stickle G ed. Pregnancy andchildbearing during adolescence. New York: Alan R. Liss, Inc., 1981: 5-17.8. Zuckerman B, Walker D, Frank D, et al. Adolescent pregnancy: Biobehavioraldeterminants of outcome. J Pediatrics 1984;105(6):857-63.9. Klerman L. Adolescent pregnancy: A new look at a continuing problem. Am J PublicHealth 1980;70(8):776-778.10. anonymous. Teenage pregnancy in Manitoba: A statistical report. 1989:11. Strobino D, Kim Y, Crawley B, al. e. Declines in nonwhite and white neonatalmortality in Mississippi, 1975-80. Pub Health Rep 1985;100(4):417-27.12. Hutchins F, Kendall N, Rubino J. Experience with teenage pregnancy. ObstetGynecol 1979;54(1):1-5.13.Taffel S. Factors associated with low birth weight: United States, 1976. WashingtonD.C.: U.S. Government Printing Office, 1980: 1-22.14.National Centre for Health Statistics. Vital statistics of the United States, 1978. Vol.I DHHS No. (PHS) 82-1100. Washington, D.C.: U.S. Government PrintingOffice, 1982:Page 2115. Correy J, Kwok P, Newman N, Curran J. Adolescent pregnancy in Tasmania. MedJ Aust 1984;141:150-4.Page 22CHAPTER THREETHE INDEPENDENT EFFECTS OFYOUNG MATERNAL AGE AND POOR SOCIOECONOMIC CIRCUMSTANCES: A REVIEW OF THE LITERATURE INTRODUCTIONThere has long been an interest in births to teenaged mothers. The first publicationdocumenting increased low birth weight rates (called prematurity in the article) amongyoung primiparae appeared in the John Hopkins Hospital Bulletin in 1922 1 . In 1961Aznar and Bennett 2 presented an overview of the results of sixteen studies that hadbeen published between the first article in 1922 and and their own. The number ofwomen in these study populations varied between 23 and 1083 with most including afew hundred cases. In all studies these cases were drawn from a specific hospital orother institution. Unfortunately there was substantial inconsistency in the age distribu-tion of the teen and comparison mothers, the racial compositions and the extent ofprenatal care between the studies. Despite these differences, it was found that amongadolescents prematurity rates were universally increased over the national averages forall women during those years.Fortunately there was some recognition of the hazards of generalizing from theserelatively small, institutionally based studies to all teenagers and the need for studieswith larger and broader based samples. In 1963 two studies 3A were published whichattempted to address some of these issues. The first 3 was only a marginal improvementover previous studies. The obstetric experience of 636 mothers less than 15 years werecompared with those of women aged 15 through 19 years delivered at the hospital, allPage 23women over 15 years delivered at the same Baltimore hospital, as well as other mothersthat gave birth in Baltimore during the same time period. Although the young mothersstill showed unfavorable pregnancy outcomes there was still sufficient potential for biasin this design to make it unwise to generalize from these results. The second study; oneby Israel and Woutersz 4 , collected data through the Obstetrical Statistical Cooperativewhich represented the experiences of ten hospitals in the United States (U.S.). Thesehospitals drew a large number (40,709) of patients from a wide variety of racial groupsand socioeconomic strata. The adolescent mothers in this study did not differ signifi-cantly from older mothers in terms of fetal, neonatal or perinatal deaths. The youngmothers did, however, have significant increases in low birth weight (called prematu-rity in the article) rates. What was also interesting were the racial differences observedin the study. A later reanalysis of these data found that both nonwhite race and agebelow 20 years enhanced the possibility of prematurity and that race was a more potentfactor than age 5 .With this latter research a new door opened with respect to the study of teenagepregnancy. Although it was acknowledged that "the burden of young motherhood"falls heavily on the children of these mothers there was a recognition that anyconsideration of the problem had to go beyond age as the only and/or primary riskfactor. The American Committee to Study the Prevention of Low Birthweight in its 1985report 6 suggested that teenage mothers, particularly the youngest, have many otherrisk factors that could be responsible for an adverse pregnancy outcome. First births aremore likely than later births to be low-weight and teen mothers are more likely to behaving first births. Young American mothers are also more likely to be black, of low SES,have completed less formal education, to report late for prenatal care, and to beunmarried. They also tend to be shorter and lighter than older mothers. Kramer, in areview of determinants of intrauterine growth and gestational duration for the WorldPage 24Health Organization 7, in addition to the factors mentioned above, suggested that lowerweights-for-height and poorer nutritional status might also be factors. Additionally,increased cigarette smoking, alcohol consumption and drug use among teenagersmight also put them at elevated risk for poor birth outcomes.The interrelation of these various factors is not clear. Kramer 7 cautioned againstsuggesting that all the above mentioned factors be considered true confounders ofyoung maternal age (MA) since young age may be an indirect cause of preterm birth orintrauterine growth retardation (or other poor birth outcome) through its effect onstature, weight, gestational nutrition, or cigarette, alcohol, or drug use. Unfortunately,the risk factor literature to date has not been very helpful in unravelling these issues. Itis fraught with methodological and conceptual problems which make its interpretationvery difficult 5,8,6. Birth outcomes and risk factors are often defined differently indifferent studies. Studies of small groups often do not produce information that can begeneralized to other populations. Study designs and statistical analyses are often notwell thought out. For example Kramer7 noted that although he located 144 articlesconcerning the effect of MA on gestational age, preterm birth, birth weight or intrauter-ine growth retardation only 17 satisfactorily met design standards that allowed forfurther consideration in his assessment of risk factors. Although the volume of theliterature is great its utility is somewhat questionable. For these and other reasons, theinterest in teenage pregnancy has not been matched by an expansion of useful informa-tion about the "problem".One of the many issues which still has not been fully explored is one that was raised byIsrael and Woutersz's study 4. It is the issue of the independent and relative impact ofboth young MA and low socioeconomic status (SES) on poor birth outcomes. It is notclear if young MA is a risk factor for poor birth outcomes independent of poverty, inPage 25association with poverty, or is not a risk factor at all. As was the case for teenagepregnancy, there is a formidable body of literature that deals with the impact of youngMA and poor SES on the risk for various poor birth outcomes. A search of the Englishliterature from Western developed countries resulted in well over 300 articles. Whenthey were reviewed it was noted that approximately one third of the articles dealt withthe impact of young MA without considering the effect of SES, another third consideredthe impact of low SES without dealing with MA and the last third were articles thatconsidered the impact of both these risk factors. Given that there is evidence that youngMA and poor SES are associated it was felt that only articles that dealt with both factorscould add to our understanding of the independent effect of both young MA and poorSES on birth outcomes. For this reason the literature review included in this chapter islimited to those articles that considered both MA and SES.The purpose of this chapter is to provide a critical summary of studies published in thescientific literature, since 1963, that have considered both MA and SES in their analysis.The first group of studies that will be reviewed are those which considered thepossibility of an association between young MA and poor SES. After this, studies whichutilized a design or statistical analysis that allowed for the assessment of the independ-ent effect of either young MA or poor SES on one or more birth outcomes will becritiqued.THE ASSOCIATION BETWEEN YOUNG MATERNAL AGE AND POOR SOCIO-ECONOMIC CIRCUMSTANCESThere was very little in the literature that dealt with the possibility of an associationbetween young MA and poor SES. Only one study was retrieved that had theinvestigation of a possible association between young MA and SES as its specificPage 26purpose. In this Scottish study, Smith 9 calculated birth rates to women aged 13 through15 years between 1980 and 1990 in areas of different SES in the region of Tayside,Scotland. Postal codes of the mothers' residences were assigned to postal code sectors(similar to census tracts in Canada), each of which had been assigned a deprivationcategory. These categories were based on the percentage of the people in the sector whowere unemployed, who were in social classes IV and V (presumably two of the socialclasses in the British Registrar General's Scale, although this was not stated in thearticle), who had no car, and who lived in overcrowded housing. Live and stillbirthsfor these young women varied between 1.2/1000 in category 1 (least deprived) to 8.8/1000 in deprivation category 7 (the most deprived): a seven fold increase.There were also a number of studies which, although this was not the focus of the study,analyzed the possibility of an association between young MA and SES. Each of the sixstudies found that as SES decreased the births to adolescents increased. Four of thesestudies were American 10,11,12,13 one was from Canada 14 and another 15 was fromScotland. It is of particular note that there was an association between these twovariables in all the studies given the different political and social climates in these threecountries.The results of the Canadian study are of particular interest given that the present studywas also undertaken with Canadian data from British Columbia. Wilkins's 1991study14 was based on census tract income in metropolitan areas. Income quintiles werecreated by ranking and grouping the census tracts according to the percentage of personbelow 18 years living in the census tract whose economic family income was below theStatistics Canada low income cut-off. Wilkins found that the percentage of all motherswho were less than 20 years increased from the quintile with the lowest percentage ofpoverty to the quintiles with the highest. The rates were 2.2%, 2.9%, 3.9%, 5.6% and 8.8%Page 27in quintiles 1 through 5. This pattern and range was very similar to that observed bySmith 9 in his recent Scottish study. Although these differences were not statisticallyanalyzed in either of these two studies, two of the American studies 12,13 which also usedmedian family income of the census tract of the mother's residence found that for bothblack and white mothers there was a significant increase in the percentage of births toyoung mothers as median income declined.Although not large in number, these studies do suggest the existence of a persistentassociation between birth rates to teenaged mothers and poor SES.ANALYSIS OF THE IMPACT OF POOR SOCIOECONOMIC CIRCUMSTANCESAND YOUNG MATERNAL AGE ON BIRTH OUTCOMESIn the review of the literature on the independent impact of young MA and poor SESon birth outcomes published studies from the developed world (primarily the U.S.)have been grouped into three groups: population based, institution or organizationbased and case-control. The focus of this critical review was on population basedstudies. Because of their large potential for bias, institutionally/organizationally basedstudies (case-control or otherwise) were, for the most part, evaluated relative to theirsupport of any conclusions drawn from the population based studies.The outcomes of these studies were evaluated relative to such things as the indicatorsutilized for MA and SES, the control of parity and multiplicity, the inclusion of otherpotential risk factors in the design and/or analysis, the reasonableness of the analysisand the clarity of the results.Page 28The control of parity and multiplicity is important because both have been acknowl-edged as being related to birth outcomes. There are studies which indicate that thereis an increased riskof low birth weight 16,17 as well as perinatal 18  infant 19 uii1nennatallmortality 20,21 with twin pregnancy. Taffel 17 also noted that for infants in multipledeliveries there was a differential risk for these birth outcomes by different MA groups.There was a much higher risk of low birth weight in a multiple birth when the motherwas a teenager.Birth order also appears to be related to some, but not all birth outcomes. There doesnot appear to be conclusive evidence of any impact of parity on either prematurity orgestational age. In a review of the literature, Kramer 7 found that of the three studies22,23,24 that were identified as satisfactorily designed only one reported a significantassociation between parity and gestational age, and the association was negative andof trivial magnitude. Likewise, only one 25 of three studies carried out in developednations 26,25,27 reported a significant effect of parity on the risk of prematurity. The dataon birth weight are much clearer. Kramer 7 found that 12 of 17 satisfactorily designedstudies he reviewed reported a positive effect of increasing parity on mean birth weight.Hardy and Mellitis 28, Taffe1 17 and the U.S. Committee to Study the Prevention of LowBirth Weight 6 presented U.S. data that showed that the risk of low birth weightdecreased with increasing birth order. Data based on the first British Perinatal MortalitySurvey also showed that there was a decline in low birth weight rates as parityincreased26 . In all these sources it was recognized that there was an interaction betweenMA and birth order. There was a U shaped relation between MA and the risk of lowbirth weight for all birth orders 26,17,6,29 For each MA group, however, there weresubstantial differences in the incidence of low birth weight related to the birth order ofthe child. Of particular note within the context of this project was the observation thatPage 29for teenaged mothers the incidence of low birth weight rose very sharply for third andhigher order births, probably a function of the short interpregnancy interval 17,6,29.Although not as well investigated, there are also studies which have found an associa-tion between parity and mortality. J shaped relationships have been noted betweenparity and perinatal mortality 30,31,32,33, stillbirths 30,34,35, neonatal mortality 30,34,36,and postneonatal mortality 30,37,34,38. The interaction between MA, parity and thesebirth outcomes has also been noted 17,6,7. It is these associations between various birthoutcomes and multiplicity and parity as well as their interactions with MA that madeit important to note differences with regard to controlling for these factors within thedesign or analysis of the studies that were reviewed in the following sections.Birth Weight and Low Birth Weight Rates I Impact of Poor Socioeconomic Circumstances There were eleven population based studies 22,39,40,41,42,43,44,45,29,12,46 which attemptedto evaluate the independent effects of young MA and low SES on either birth weight orlow birth weight rates. All of these studies found poor SES to be associated with eitherdecreased birth weight or increased low birth weight rates. This result was impressivegiven that authors utilized different measures for SES.Two studies 39,29 utilized four or five categories of education based on the number ofyears of formal education as their measure of SES. Gortmaker 41 included poverty status( yes or no according to predefined standards that were adjusted for family size),maternal and paternal education level, and hospital insurance as dichotomous vari-ables representing SES. He found that poverty status and education level of the parentsPage 30were not directly associated with low birth weight but acted indirectly through theireffect on the possession of hospital insurance. Black or white race as well as a numberof educational categories were included in five of the studies 40,42,43,45,46. Gould andLeRoy 12 also included race as a marker for SES but substituted the 1979 median familyincome of the maternal residence for education as a marker for the level of SES of themother. Correy44, in the only non-American study included in this group of populationbased studies, used the occupation of the main wage earner divided into five categories.If the mother was not married her occupation was utilized and the occupation of thehighest wage earner was chosen if she was married.The results regarding the impact of SES on birth weight and low birth weight rates werenot as clear cut for institutionally/organizationally based studies as they were forpopulation based ones. Of the thirteen studies 47,59,48,49,50,51,52,53,54,55,56,57,58 in theinstitution/organization group there were ten 47,59,48,50,51,52,54,55,56,57  in which it wasconcluded that poor SES was associated with significant increases in either low birthweight rates or decreases in mean birth weights. In these studies there was also quitean array of variables used as surrogates for SES. All studies except one included at leasttwo measures. Black or white race was utilized in one study 54 . Race and education wereboth included in two studies 51,57 and race and some measure of income were assessedin another two 50,56. Race, employment status and education were all included in threestudies 59,52,55 while Crosby 47 included family income and education.There was one study 49 in which it was not possible to draw a conclusion about theimpact of SES and there were also two studies 53,58, in which the conclusion was thatpoor SES was not a risk factor for low birth weight. The one inconclusive study 49 wasdesigned such that race (the variable used to measure SES) was controlled. This madeit impossible to draw conclusions regarding the impact of SES. One of the two studiesPage 31that found that SES was not a risk factor for birth weight was a study 53 which used datafrom one hospital in each of the countries participating in an international collaborativematernity care monitoring project. A model was developed based on information fromall the participating countries and then each of the six participants' data were fittedagainst this model. Sweden was the only country whose results were considered givenit was the only developed country in the project. The fact that it was the only developednation may have been responsible for the lack of a significant role for education (thevariable used to measure SES) in Sweden's risk factor model or it may be that Swedenhas effectively eliminated SES gradients (as measured by education or some othervariable) in poor birth outcomes given their aggressive social and other programs. Inthe second study 58 in which SES was found not to be significant, education was also thevariable used to measure SES. The authors of the study acknowledged that the studypopulation was an indigent one and in the description of the demographic character-istics of this population it was noted that there was very little variation in the levels ofeducational attainment of the participants. This may have been responsible for thefailure of education to be a significant variable in the regression analysis.There were eight case-control studies which looked at the impact of young MA whilecontrolling for SES 60,61,62,50,63,64,65. In all cases mothers were drawn from either loweror lower middle class neighborhoods. Since the studies did not include wide SESspectrums they were not able to add to the discussion of the impact of SES on low birthweight and so will not be commented on at this point .This survey of the literature suggests that there is an impact of SES, when consideringboth MA and SES simultaneously or while controlling for MA. When SES decreasedthere was either an increase in low birth weight rates or a decrease in mean birthweights. In only two of the studies reviewed was there the suggestion that this was notPage 32the case and in both of these studies methodological issues may have been a factor in theconclusion arrived at. It is also of note that despite a number of different measures used,SES remained significantly associated with birth weight or low birth weight rates in allof the population based studies. These observations suggest that there is a significantassociation between SES and birth weight or low birth weight rates apart from anyassociation the latter might have with MA.II Impact of Young Maternal AgeThe results regarding the impact of young MA on birth weight and low birth weightrates were not as clear cut as those for SES. In six of the eleven population based studies22,39,41,42,12,45 it was found that there was an association between MA and birth weightand in the other five 40,43,44,29,46 no significant association was noted. There did notappear to be any glaringly obvious differences in the study designs or analyses thatwould explain this difference in outcomes.Although there was some variation in the size of the study populations they were alllarge studies and the ranges were similar in both groups. The smallest of the studiessuggesting an association between young MA and increased low birth weight included6,185 births 41 while 127,000 was the size of the largest. The studies suggesting that therewas no relationship between young MA and low birth weight varied in size from48,00044 to 2 million 29 subjects.There was some difference in the definition of young MA in the two groups of studies.The six that found there was an association tended to restrict their teenage mothersgroup(s) to slightly younger mothers. Three 39,42,12 of the six limited young mothers tothose who were 17 years and under, one 45 restricted the group to those under 17 years,Page 33another 22 set 15 years as the upper cut-off. The one exception was Gortmaker 41 ,whoclassified all women under 20 years of age as young mothers. In the group of studiesthat found no association between MA and birth weight one study 43 included allwomen under 20 years and three 40,44,29 included all women under 18 years. The fifthstudy 46 classified all women under 18 years as young mothers and also looked atmothers aged 15 years and under separately from those aged 16 and 17 years. There wasalso a slight difference in the number of studies in the two groups that either limitedtheir study population to primips and controlled for parity as part of their design oranalysis. In the group that found an association between young MA and increased ratesof low birth weight, parity was not restricted or controlled for in two of the studies 12,45as compared to only one study 46 in the second group.There were also some differences between the two groups of studies in the othervariables included in the study, in whether the dependent variable was birth weight (acontinuous variable) or low birth weight (a categorical variable) and in the analysesundertaken but none of these appeared to be substantial. Of the six studies that foundthat young MA was significantly associated with birth weight, one 39 did not includeany other variables beyond the ones relating to parity, multiplicity and SES. Four22,41,42,12 included one other variable usually relating to medical complications orprenatal care and one study 45 included three other variables; pregnancy complications,gestational age and infant's sex. The five studies that found that there was noassociation with birth weight included marginally more variables. One 44 did notinclude any additional information. Two 43,29 included one extra variable each;prenatal care in the first and marital status in the other. Eisner 40 and Lee 46 includedthree and four additional variables respectively. Both included variables on prenatalcare and marital status. In this last study the population was limited to infants of at least40 weeks gestation.Page 34This different, (but not significant) pattern was also observed when the dependentvariables and analytical techniques were compared. In the first group of studies birthweight was the dependent variable in two 41,42. Both of these employed a log-linearanalysis. Low birth weight was the dependent variable in the other four stud-ies22,39,45,12 . One of these 39 was an observational study , another 22 employed a non-parametric regression, and the last two 45,12 used multiple linear regression and logisticregression. In the second group of studies which did not find young MA to be a riskfactor, the one 44 that had birth weight as its dependant variable used an ANOVA testwithin SES groups. Of the four studies that had low birth weight as the outcome ofinterest, one utilized log-linear analyses 43. One 46 employed logistic regression andanother one 40 used both Mantel-Haenszel chi square tests and logistic regression.Kleinman and Kessel's study 29 included an entire population and no statistical analysiswas undertaken. It should be noted that in this latter group of studies in which youngMA was found not to be associated with either birth weight or low birth weight, in twostudies 40,29 young MA was associated with increases in very low birth weight rates forprimaparous mothers. It was noted that in one other study 46 in this group the odds ratiofor the risk of term low birth weight was 1.00 for mothers less than 15 years andincreased to 2.14 for mothers aged 20-24 years.This review of the population based literature that examined the impact of both youngMA and SES has not unearthed any well defined relationship between young MA andeither birth weight or low birth weight. If any association exists it is not strong andprobably does not apply to all women less than 20 years. It may be that only for the veryyoung teenage mother and those mothers experiencing their second or greater birth isthere a risk of either decreased birth weight or increased probability of a low birthweight birth. It may also be that the design of both the study and the analysis,particularly the inclusion of variables such as marital status, prepregnancy weight andPage 35certain health habits which are closely associated with young MA influence whether ornot young MA is found to be a risk factor.The institutionally/organizationally based studies did not shed any further light on the,situation. Of the thirteen studies included in this group, five 59,49,52 54,56 found thatyoung MA was not a risk factor, seven 47,50,51,53,55,57,58 found the opposite and in onestudy 48 the authors found the evidence to be inconclusive.When the studies were scrutinized more closely, one difference of interest was noticedbetween the two groups. In the majority of these thirteen studies the data were analyzedwith a multiple linear or logistic regression. The five studies that found a significantrelationship included fewer variables in their analysis. In addition to SES, MA andparity variables, one study 49 did not include any additional variables, one 59 includedtwo extra variables, another 54 included three and two studies 52, 56 included fouradditional variables each. This was in contrast to the seven studies which found youngMA not to be a risk factor. One study 47 included four additional variables, anotherincluded five 53, and two studies 50,55 included eight. One study 58 included nineadditional variables, another 51 included eleven, and the last study 57 included 30variables in total. The additional variables included in the analyses were items such asmaternal height, weight and height-weight ratio. Variables relating to prenatal care andmedical complications of pregnancy were frequently included. Smoking habits anddrug and alcohol use were also often included. These are all variables that have beennoted to be associated with teenage pregnancy 6,7 and low birth weight 6,7. It may bethat some of these variables were acting as confounders for young MA and thus resultedin the latter not being significantly related to birth weight or low birth weight. This issueof the number and nature of the other variables included in the analysis may also be afactor in the different outcomes among the population based studies. Although not asPage 36marked a difference as in this group of studies, it was noted that the studies which didnot find young MA to be a significant risk factor included a slightly larger number ofvariables in their analyses. Unfortunately the case-control studies could not shed anyfurther light on this issue since in only one study 62 was a regression analysisundertaken.III Interaction Between Young Maternal Age and Poor Socioeconomic CircumstancesThere were only four studies in which the possibility of an interaction between youngMA and poor SES with respect to birth weight or low birth weight rates was considered.Of the four there were two 43,53 that did not find an interaction between MA andeducation, the latter being the variable included in these two studies as a proxy for SES.In the former study 43 the possibility of an interaction between SES and race was alsotested. This was also not found to be significant, which was different than the resultobtained in a study by Horon, Strobino and McDonald 50. Although there was a raceand MA interaction in this latter study it was not tested for significance. Finally, in astudy by Gould and LeRoy 12 it was noted that there was a significant interactionbetween MA and median income in the census tract of the mother's residence at the timeof the birth.Considering that there were so few studies which addressed the issue of interactionbetween MA and SES and that the results of these analyses were not consistent it is notpossible to draw any conclusion concerning the existence of an interaction betweenthese two variables with respect to low birth weight.Page 37Prematurity RatesThere were no population based studies in the literature reviewed that simultaneouslyconsidered the impact of both SES and young MA on prematurity. The studiesreviewed were all organizationally (i.e. hospital ) based.I Impact of Poor Socioeconomic CircumstancesOnly four studies were found which either considered the impact of young MA and SESsimultaneously or considered one risk factor while controlling for the other. In all ofthese studies 62,66,67,68 prematurity was defined as the spontaneous onset of labour atless than 37 weeks gestation, although Arbuckle and Sherman did not include infantsthat were small for gestational age. There were two case-control studies. In the one 62the 32 adolescents were matched on ethnicity, clinic payment status and a few othervariables with women over 20 years which precluded any consideration of the impactof poor SES. In the other study 68 this was also the case given that 137 mothers under15 years were matched with older women who were also Black, unmarried, wardpatients, who did not have a history of surgical or medical diseases. In both of theremaining studies poor SES was found to be associated with increased rates ofprematurity despite a number of differences in their designs. Both studies onlyincluded single births but one did not control for parity 67 while the other one did 66 .Both studies initially included income as one of the measures of SES but in both casesit was found not to be significantly associated with prematurity. The measures of SESfound to be significantly associated with prematurity were race (black or white) 67 andincome (low or other) and maternal education (<12, 12-13, or >13 years) 66. Logisticregression was employed in both studies with a total of twelve and eleven variables,Page 38including demographic, medical and behavioral risk factors in each analysis. The onlyvariables common to both studies were smoking, height and prepregnancy weight.As in the case of low birth weight, the fact that prematurity was found to be associatedwith SES in both studies despite these differences suggests that the relationshipbetween these two variables is fairly robust.II Impact of Young Maternal Age These four studies differed with respect to their conclusions about the associationbetween young MA and prematurity. Within a multivariate analysis Abrams andNewman67 found that there was a significant association while Arbuckle and Sherman 66did not. This latter result was the same as that obtained in the two case-controlstudies68,62 .Examining the differences between the four studies provides some clues as to possibleexplanations for the differences in the results with respect to young MA. The young MAvariable in the two case control studies 68,62 and that of Abrams and Newman 67 studiesincluded only women less than 15 and 16 years respectively while in the Arbuckle andSherman study66 women under 20 years were included. Additionally, although thenumber of variables included in the logistic regression analysis in both of the organiza-tionally/institutionally based studies was similar, the actual variables included werenot the same. Parity, prepregnancy weight, height and number of cigarettes smokedwere included in both studies. Other variables included by Abrams and Newman67were illicit drug use, low rate of maternal weight gain, short interpregnancy interval,incompetent cervix, trauma etc, and maternal hypertensive disorder. All but the last ofthese was found to be a significant risk factor. Arbuckle and Sherman 66 includedPage 39alcohol consumption, haemoglobin level, sex of the infant, and serum Vitamin C level.Only sex of the infant was found to be significantly associated with prematurity.Arbuckle and Sherman66 themselves comment on the fact that although young MA wasfound to be a significant risk factor in a univariate analysis this was no longer so in thelogistic regression. Another difference between the four studies was the number ofsubjects included in the various studies. The case control studies included only 64subjects from a prenatal care clinic in New Jersey 62, and 3,005 women from a medicalclinic in New York 68, the Abrams and Newman67 included a sample of 2,228 womenwho participated in a University of California prenatal nutrition project between 1978and 1988 and the Arbuckle and Sherman66 study had a sample of 806 rural Canadianwomen from 10 communities that participated in Nutrition Canada Clinics in 1972.The difference in the results of these studies with respect to the association betweenyoung MA and prematurity may well have been a function of one or a combination ofthe differences described. These results suggested that at very best the associationbetween young MA and prematurity if it exists is weak and was only observed byAbrams and Newman67 because of the fairly large sample size and limiting youngmothers to those aged less than 16 years.III Interaction Between Young Maternal Age and Poor Socioeconomic CircumstancesThere were no studies reviewed that considered the possible interaction between MAand SES with respect to prematurity.Page 40Small for Gestational Age RatesFor small for gestational age (SGA) rates there were also no population based studiesretrieved from the literature that simultaneously considered the impact of both poorSES and young MA. The studies reviewed in this section are all organizationally/institutionally based.I Impact of Poor Socioeconomic Circumstances There were five studies49,65,66,67,69 reviewed that considered both the effect of poor SESand young MA simultaneously or considered one of these variables while controllingfor the other. Two of these were case-control studies in which SES was one of thevariables on which the subjects were matched and so could not provide any informationon the impact of SES. Of the three remaining studies, one 67 found poor SES to besignificantly associated with increased SGA rates and the other two 69,66 found noassociation.There were a number of differences between these three studies. For two of these 67,66some of the differences have been highlighted in the previous section dealing withprematurity. The Abrams and Newman67 study was slightly larger with 2228 subjects,compared with just over 800 in the other two studies69,66 . Scott and Ounsted69 did notprovide a definition for SGA. It was defined as below the 10th percentile for weight bygestational age adjusted for infant sex in both of the other studies 67,66, although in eachcase the standards used for comparison were specific to the geographic jurisdiction inwhich the study was carried out. In the Arbuckle and Sherman66 study SGA infantswere compared to all other infants while in the other two studies large for gestationalage infants were excluded. The Scott and Ounsted 69 study included somewhat fewerPage 41variables than did the other two studies (8 compared with 11 and 12) although somedemographic, medical and behavioral risk factors were included in all of them.Smoking, height and prepregnancy weight were variables common to all three studies.SES was measured differently in all three studies. Abrams and Newman 67 used race(black/white) as the measure while the other two studies utilized manual laborer (yes/no) 69 and income (low/other) and maternal education (<12, 12-13, >13 yrs) 66 .Only the sample sizes and the variable used to measure SES appear to separate the threestudies into the two different results. While race as a surrogate measure of SES providedthe same results as other proxy variables when considering low birth weight, this maynot be the case for SGA. It may be that the choice of variable to act as a surrogate for SESmay be more important for this birth outcome because the relationship between SES andSGA is not as robust as that between SES and low birth weight.II Impact of Young Maternal Age In all five studies no association was found between young MA and SGA rates. This lackof a relationship persisted despite differences in the definition of young MA. In two ofthe studies it was defined as less than 20 years 49,66, in one it was defined as less than 17years 65, in another one it was less than 16 years 67 and in the last study69 it was notdefined because it was not found to be a significant risk factor in a univariate analysis.The fact that young MA remained insignificant as a risk factor despite these differencesin its definition as well as differences in the study designs, other variables included inthe studies and types of analyses attest to the strong likelihood that SGA is not affectedby young MA.Page 42III Interaction Between Young Maternal Age and Poor Socioeconomic CircumstancesThere were no studies reviewed that considered the possible interaction between MAand SES with respect to SGA rates.Stillbirth Rates From the literature review five studies were retrieved that either simultaneouslyconsidered the impact of poor SES and young MA on stillbirth rates or considered theimpact of one while controlling for the other. Two population based studies wereretrieved as well as three organizationally/institutionally based studies. Two of thislatter group were case-control studies .I Impact of Poor Socioeconomic CircumstancesThe two population based studies 34,35 and the one (non-case-control) institutionallybased study 70 all found that there was a significant association between SES andstillbirth rates. Poor SES was one of the variables that subjects were matched on in thetwo case-control studies 61,64 so these could not be used to add any information on theimpact of SES.The conclusions of all three studies were similar with regard to the influence of SESdespite some differences between them. The first two studies 34,35 used informationobtained from administrative data bases in Scotland, and England and Wales, while thethird study7° utilized data from the 44,386 pregnancies that were part of the AmericanCollaborative Perinatal Project. The first two studies34,35 used social class based on thefather's occupation, classified according to the Registrar General's Classification ofPage 43Occupations as the measure of SES while the third study" utilized a SES index createdfrom the combined scores of the mother's education, occupation, and family income.The two population based studies controlled for parity and did not deal with multiplic-ity34,35, while Naeye7° only included single births but did not control for parity.Another difference was the method of analysis employed in each of the three studies.Forbes and Pickering34 employed logistic regressions, Murrells, Catford, Smith andMachin35 developed logit models and Naeye7° did chi-square tests within different ageand SES categories.II Impact of Young Maternal AgeYoung MA was not found to be significant in any of the five studies despite somedifferences between them in the definition of young MA and the comparison groups.Forbes and Pickering34 found odds ratios of 0.90, 0.80, and 1.01 in 1960, 1971 and 1981-82 for mothers aged 16 through 19 years relative to mothers aged 25 through 29 years.Murrells, Catford, Smith and Machin35 found that mothers aged 16 through 19 yearshad rates that were lower than mothers aged 20 through 23 years but slightly higherthan those aged 25 through 29. Naeye" noted that 18 and 19 year old mothers regardlessof their SES had stillbirth rates that were lower than those for mothers over 20 years.Mothers under 15 years were compared to mothers aged 19 through 25 in the Duenholtercase-control study 61 . The rates for the mothers under 15 years were found to be lower(although not significantly so). Although the two groups of mothers were aged less than16 years and 20 through 24 years in the second case-control study 64 the results were thesame as in the earlier one.Page 44The consistent lack of an association between young MA and rates of stillbirth in all ofthese studies despite differences in their designs variable definitions and analysessuggests that this lack of association is robust for all young mothers.III Interaction Between Young Maternal Age and Poor Socioeconomic CircumstancesThere was one study 35 reviewed that considered the possible interaction between MAand SES with respect to stillbirth rates. In this study data from 1949/50 and 1975 wereanalyzed. When each year was analyzed separately the interaction between MA andsocial class was not found to be significant. When the data for these two years werecombined the interaction was found to be significant. Unfortunately the nature of theinteraction was not discussed in the article nor was there any information provided onwhich to draw any conclusions.Perinatal Mortality RatesFrom the literature reviewed, five population based studies were retrieved that eithersimultaneously considered the impact of poor SES and young MA on perinatalmortality or considered the impact of one of these variables while controlling for theother. Three case-control studies were also retrieved. These last three studies were allbased on data from the U.S. while the data in the population based studies were fromIsrael, Sweden, New Zealand, Ireland and Scotland.I Impact of Poor Socioeconomic Circumstances The three case-control studies 61,48,49 included subjects with low SES, so these could notadd anything to the discussion of the impact of SES. One of the population basedPage 45studies44 also could not provide any information because the analyses were only carriedout in one social class category. Of the four remaining studies only one 31 found thatthere was not a significant relationship between SES and perinatal mortality and theother three 71,34,72 concluded that perinatal mortality rates increased as SES decreased.There was a design issue which may explain why Smedby and Ericson31 did not finda significant association while the other studies did. In the former study perinataloutcomes to foreign mothers were compared with those to women with Swedishnationality. The assumption was made that foreign mothers would be of lower SES thanthe latter group. This assumption was not tested in any way and may well have beenerroneous. In reviewing the results of their study the authors admit the possibility thatwomen who emigrate may enjoy better health than those from their country of originand so have better prospects for successful pregnancies than the authors initiallyassumed.Among the three remaining studies the definitions of young MA were 16 - 19 years 34 ,14 - 20 years 72 and less than 20 years 71 . Forbes and Pickering34 did not define perinatalmortality while in the other two studies deaths in the first week of life as well asstillbirths of greater than 26 or 28 weeks were included. All three studies controlled forparity and one study was restricted to single births. Logistic regression analyses wereused in two of the studies 34,72 while a binary regression was employed in the third 71 .There was some variation in the way SES was measured in the three studies. In two ofthe studies 71,34 social class was classified by the Registrar General's Classification ofOccupations for the United Kingdom and in the Israeli study ethnic origin was used asa surrogate. There was also substantial variation in the number of variables includedin the analyses in each of the studies. Fifteen were included in the Elwood, Mackenzieand Cran study71 in addition to social class and MA. These included demographic risks,Page 46medical risks predating and in the current pregnancy, and health care risks. A total ofeleven variables were analyzed in the Israeli study 72 while only three; MA, social classand parity were included by Forbes and Pickering34 .This variation in the number of potential risk factors analyzed in the three studies as wellas other design features suggests that the association of SES with perinatal mortality isfairly robust even though one study did not find an association.II Impact of Young Maternal Age Of the eight studies included in this section, five 61,48,71,44,34 found there to be noassociation between MA and perinatal mortality, two 49,72 found an association and inone study 31 the authors did not come to a conclusion.In this last study 31 the perinatal mortality rates to both groups of women (foreign bornand Swedish nationals) were slightly higher for women aged less than 20 years whencompared to those aged 20 - 24 years, but this difference was not analyzed for statisticalsignificance. One of the two studies which found a significant elevation in perinatalmortality rates for young mothers was population based 72 and the other was a case-control study 49. Since all births at Temple University Hospital between July 1970 andDecember 1975 were included in this last study the authors considered it a populationstudy and so did not perform any statistical analysis on the data. They assumed thatperinatal mortality rates of 40.3/1000 birth and 39.8/1000 births were really differentthan the rate of 33.2/1000 for mothers aged 20 - 34 years. In the Israeli, population basedstudy 72 mothers below the age of 20 and over 35 years had a significantly higherfrequency of perinatal death. The odds ratio for mothers aged 14 - 20 years relative tothose aged 21 - 36 years was 1.96 (95% C.L.=1.52, 2.52).Page 47These results stand in contrast to those obtained in the other studies which found noassociation, particularly when it is noted that in a number of these latter studies youngMA was found to protective 61,48,71. For example, in the two case-control studies 61,48the perinatal mortality rates for teen mothers were 3% and 4% compared with 3.8% and5.3% for older mothers. Also, it was noted in the previous section (impact of SES) therewere very few differences between the the three population based studies that could beused to rationalize this result relative to that obtained in the other three populationbased studies as well two of the case-control studies. All this makes it difficult to cometo a definitive conclusion concerning the relationship between young MA and perinatalmortality.III Interaction Between Young Maternal Age and Poor Socioeconomic Circumstances There were no studies retrieved that considered the possible interaction between MAand SES with respect to perinatal mortality.Infant Mortality RatesThere were not many studies that considered infant mortality as opposed to neonataland/or postneonatal mortality. From the literature review three population basedstudies were retrieved that either simultaneously considered the impact of poor SESand young MA on infant mortality rates or considered the impact of one whilecontrolling for the other. One case-control study was also retrieved. Except for one ofthe population based studies, all were based on data from the U.S..Page 48I Impact of Poor Socioeconomic Circumstances All three population based studies 73,74,75 found a significant association between SESand infant mortality. As SES decreased infant mortality rates increased. Only subjectsof lower SES were included in the case-control study 48 and so no insight into theassociation between SES and infant mortality was provided by this study.This association was obtained in all studies despite a number of differences in theirdesigns and analyses. One ecological study 74 used a pooled cross-section time seriessample of eighteen developed nations between 1950 and 1975. Another ecologicalstudy73 used counties in continental U.S with a minimum of twenty deaths between1971 and 1975 as the unit of analysis. The third study 75 utilized linked 1978 live birthand infant death files for the state of California. All three studies used differentmeasures of SES. In the ecological study 73 SES was approximated through a weightingof each county's percentage of the work force in white collar jobs, median level ofeducation, percentage of the families at 125% of the poverty level, median income andthe number of housing units with fewer than one person per room. Race was also avariable in this study. Both the composite index for SES and race were found to besignificantly associated with infant mortality. Years of schooling (0-11,12, and > 13 yrs)and race (black, white) were included in the California study 75 and were found to besignificantly associated with infant mortality . The international study74 included anumber of variables that could be seen as measures of SES. These were each country'sGNP per capita, percentage of the population that was urbanized, social welfareexpenditure per capita, percentage of eligible females enrolled in tertiary education,average unemployment rates and the Gini coefficient (a composite measure of well-being), which is a measure of the income distribution within a country. Of these, GNPPage 49per capita, percentage of eligible women enrolled in tertiary education and averageunemployment rates were found to be significantly associated with infant mortality.Although there were only a few studies which simultaneously considered the effects ofSES and MA on infant mortality, the fact that the there was a significant associationbetween SES and infant mortality in all three despite the above described, as well asother differences, suggests a fairly robust relationship between them.II Impact of Young Maternal AgeOf the one case-control and three population based studies reviewed in this section, theformer 48 and two of the latter 73,75 found a significant association between MA andinfant mortality. Young mothers had higher rates of infant mortality. These threestudies all were based on American data while the one study which did not find thisassociation was an international study including data from 18 developed nations.These four studies also included different numbers and types of variables as well asemploying different analyses. Parity was controlled in one of the studies 75, notcontrolled or included as a variable in two 48,73 and not mentioned in the fourth 74 . Theanalysis employed in the case-control study 48 only allowed for the observation of theeffect of young MA within the context of poor SES. A multiple linear regression analysisincluding SES, MA and race as independent variables was undertaken by Miller andStokes73. Cramer75 and Pampel and Pillai 74 employed logit and least squares tech-niques and also included a number of additional variables. Besides MA, education raceand parity, Cramer75 included marital status, birth weight, and timing of first prenatalvisit as variables in his analysis. The only variables not relating to SES, and MA in thePage 50Pampel and Pillai study74 were the number of physicians, nurses and hospital beds percapita.From the differences described in this and the previous section (impact of SES) theredoes not appear to be any glaringly obvious explanation for the incongruity in the studyresults. One might be inclined to suggest that the international data set and theecological design of the Pampel and Pillai study 74 may have been responsible for thelack of a significant result in this study. Further research will be required before any firmconclusions can be drawn about the independent impact of young MA on infantmortality but there is the suggestion that there is an association between young MA andinfant mortality.III Interaction Between Young Maternal Age and Poor Socioeconomic CircumstancesOf the four studies included in this review only one considered the possibility of aninteraction between MA and SES with respect to infant mortality. Cramer75 found thatthere was no significant interaction between either age and education or age and racewith respect to infant mortality rates. Education and race were the two variablesincluded in the study as proxy measures for SES. Unfortunately the results of one studyare not sufficient to allow for a conclusion about the existence of an interaction betweenthese two variables with respect to infant mortality.Neonatal Mortality RatesNine population based studies were retrieved that either simultaneously consideredthe impact of poor SES and young MA on neonatal mortality or considered the impactof one of these variables while the other was limited. Four case-control studies werePage 51also retrieved. All the case-control studies and six of the nine population based studiesused data from the U.S.. Of the remaining population bases studies two utilized datafrom the United Kingdom and the last was based on data from eighteen developednations.I Impact of Poor Socioeconomic CircumstancesIn all four case-control studies 68,61,49,64 the subjects were limited to those particular SEScategories so these studies could not add to the discussion of the impact of SES. Sevenof the nine population based studies 41,42,34,73,76,74,75 found an association between SESand neonatal mortality; as SES decreased rates of neonatal mortality increased. Theremaining two studies 77,63 did not find a significant association. Although there weredifferences between the two groups of studies, as has been the case with a number ofother birth outcomes, there did not appear to be any obvious explanation for thedifference in their conclusions. There appeared to be as many differences betweenindividual studies within the two groups as there were between the groups.All the studies but one defined neonatal mortality as infant deaths occurring within thefirst 27 days. The the other study 75 the analysis was restricted to early neonatalmortality (deaths within the first week of life). There was somewhat more variation inthe way SES was defined in the various studies. Of the two studies which did not findan association between SES and neonatal mortality one 63 used race as a measure of SESand the other 77 included race as well as a four category ecological variable based on themedian rent of the census tract of the mother's residence at the time of birth. One of theseven studies which found an association was an ecological county based study 73 inwhich race was included as well as another variable which approximated SES througha weighting of each county's percentage of the work force in white collar jobs, medianPage 52level of education and a number of other similar variables. Another two studies whichfound an association included both race and education as measures of SES. Both foundrace to be significantly associated with neonatal mortality. The association witheducation was not significant in one study 42 and was weak in the other 75. The twostudies from Scotland 34, and England and Wales 76 used five social classes based on thefather's occupation classified according to the British Registrar General's Classificationof Occupations. In the last study in the group that found an association, Gortmaker 41classified mothers as poor or not poor based on a predetermined income standardadjusted for family size etc. Six variables, described previously (in the infant mortalitysection) were included as proxy measures of SES by Pampel and Pillai 74 . Of these theGNP per capita, the percent of eligible females enrolled in tertiary education and theaverage unemployment rate were significantly associated with neonatal mortality.Although all the studies employed either log-linear or logit models, least squarestechniques or logistic regressions in their analyses there was also a fair range in thenumber and type of variables included in these analyses. All of the studies, except twoof those which were population based 73,74, either included parity as a variable orrestricted the study to first births. The two studies 63,77 which did not find a significantassociation included birth weight, and prenatal care among their additional variables.One also included gestational age 63 . One of the population based studies 75 alsoincluded birth weight and prenatal care as additional variables. Two included birthweight plus one other variable; either previous fetal death 42 or pregnancy experience 41 .Three studies3473,38 which found an association between SES and neonatal mortalitydid not include any other variables beyond their measure of SES, young MA and parityand the international study 74 included the number of physicians, nurses and hospitalbeds per 1000 populations as measures of the level of health care services.Page 53This lack of a definite pattern of similarities or differences makes it difficult toconclusively comment on the existence of a relationship between SES and neonatalmortality. From this review the suggestion is that there is a relationship between thesetwo variables; that rates of neonatal mortality increase as SES decreases. This sugges-tion is made despite the two dissenting studies, in part because in the one study63 racewas the only variable included as a proxy for SES and in the other 77 the ecologicalvariable used may have muted the effect of SES.II Impact of Young Maternal AgeWith the addition of the four case-control studies there were thirteen studies whichconsidered the impact of MA on neonatal mortality. Four of the population basedStUdieS42,73,76,75 and one case-control study49 found a significant association betweenMA and neonatal mortality. The risk of neonatal mortality was increased for teenagedmothers. Five population based studies77,41,34,63,74 and three case-control studies didnot find this association68,61,64 . Once again there was no obvious difference between thetwo groups of studies that would help explain the disparity in the results of thesethirteen studies.Although there were some differences in the definition of young MA in the variousstudies there was no pattern between the two groups of studies. Of those that foundyoung MA was associated with increases in neonatal mortality rates the case-controlstudy 49 had teenaged mothers divided into two groups; those age less than 17 and thosebetween 17 and 19 years. Two of the population based studies 76,75 included womenaged between either 15 or 16 and 19 years. Another 42 included mothers less than 18years and the last 73 included those between 13 and 17 years at the time of birth. In thesecond group the three case-control studies included women that were slightly youngerPage 54than in the other studies. Women less then 15 years were included in two studies 68,61and those less than 16 years were part of the other 64. Three of the population basedstudies in this group 77,41,74 included all women less than 20 years and mothers aged 16through 19 years were considered in another study 34 . Geronimus 63 included four sub-groups of teenage mothers in her study. These were mothers aged 11-13, 14-15, 16-17,and 18-19 years.There was also considerable variation in the sample sizes of the various studies. Thevariation appeared to be random as opposed to separating the studies into two groups.The samples in those studies which did find a significant relationship between youngMA and neonatal mortality varied between 2,000 countries 73 to over 280,000 Americanbirths75 and all the births in England and Wales 76 in two separate years. For studies thatdid not find a difference the number of births included varied between 942 cases andcontrols 61 to all the births in Scotland34 in one year and 305,907 first births in threeAmerican states63 . Although sample size of the study affects its ability to detect adifference, these two groups of studies did not differ radically with respect to the samplesizes within each group. Sample size was not a factor in explaining the difference in theconclusion about the relationship between young MA and neonatal mortality in the twogroups of studies.As was mentioned in the previous section, there was a considerable range in the numberand type of variables included in these analyses. There did not appear to be a patternwithin each of the two groups of studies: those which did and did not find an associationbetween young MA and neonatal mortality. Of the two population based studies whichdid not include parity 73,74 there was one from each of the two groups. Birth weight wasincluded as a variable in two of the studies 42,75 which found an association and threeof the studies 77,41,63 which did not find the association. Similarly, prenatal care wasPage 55included in one 75 of the first group of studies and in two 77,63 of the latter. Of the threecase-control studies only the two which did not find an association 61,64 limited theirsamples to first births and one of these studies 64 also matched cases and controls on sexof the infant and trimester of first prenatal visit.This review shows that there is no consistent or strong association between neonatalmortality and young MA.III Interaction Between Young Maternal Age and Poor Socioeconomic Circumstances Of the thirteen studies included in this review only three considered the possibility ofan interaction between MA and SES with respect to neonatal mortality. Cramer 75investigated possible interactions between MA and race and MA and education withrespect to neonatal mortality and found neither to be significant. Bross and Shapiro 42tested these same two interactions and found the MA and race interaction not to besignificant while the MA and education interaction was significant. Machin, Murrells,Catford, and Smith76 used social class divided into five categories by occupation as ameasure of SES. They found that there was no significant interaction between MA andsocial class with respect to neonatal mortality for either of the years for which they haddata. The fact that only one interaction of the five tested suggests that it is unlikely thatthere is an interaction between MA and SES with respect to neonatal mortality. Morestudy is probably required before this assertion can be made with confidence.Postneonatal Mortality RatesExcept for one case-control study, the studies which were found, that simultaneouslyconsidered the effects of MA and SES on postneonatal mortality or considered thePage 56impact of one of these variables while limiting the other, were a subset of the the studieswhich considered the impact of these risk factors on neonatal mortality. This subsetincluded seven population based studies. Except for the one international study thedata were derived from the U.S. and the United Kingdom.I Impact of Poor Socioeconomic Circumstances SES was restricted in the case-control study 48 so it did not add any information to thediscussion of the impact of SES. In each of the remaining studies 77,41,42,73,38,74,75 SESwas found to be associated with postneonatal mortality. As SES decreased rates ofpostneonatal mortality increased. This result was obtained in each of the studies despitea number of differences in their designs.In all studies but one the postneonatal period was defined as that between 28 days andone year of life. in the one exception, Cramer75 defined as postneonatal all deathsbetween 7 days and one year of life. As was mentioned in the discussion surroundingneonatal mortality, there was somewhat more variation in the way SES was defined inthe various studies. In two studies77,73 race as well as a composite ecological variable(a multi-item factor score, measuring the level of affluence of a predetermined geo-graphical area) were used to measure SES. It was noted in the Shah and Abbey study 77that once adjustments were made for other risk factors race was no longer an importantrisk factor for postneonatal mortality while education continued to be. Two studies 4275included both race and education as measures of SES and found both to be significantlyassociated with neonatal mortality, although race was only significant through itsinteractions with other variables. The study from England and Wales 38 used five socialclasses based on the father's occupation classified according to the British RegistrarGeneral's Classification of Occupations. Gortmaker41 classified mothers as poor or notPage 57poor based on a predetermined income standard adjusted for family size and relatedvariables. Six variables, described previously (in the infant mortality section) wereincluded as measures of SES by Pampel and Pillai 74 . Of these the GNP per capita, thepercent of eligible females enrolled in tertiary education and the average unemploy-ment rate were significantly associated with postneonatal mortality.These and other differences in the study designs and analyses did not affect therelationship between SES and postneonatal mortality. On the basis of these results it canbe stated that the low SES of the mother is a robust risk factor for increased mortalityduring the postneonatal period.II Impact of Young Maternal AgeThe study results regarding the impact of MA on the risk of postneonatal mortality werenot as easy to interpret. Five population based studies 77,42,38,74,75 and the one case-control study 48 found MA to be associated with postneonatal mortality. Youngmaternal age increased the risk of postneonatal mortality. The two remaining studies41,73 did not find this association, although in one of these studies41 the risk ofpostneonatal mortality was elevated for mothers under 20 years. As was the case withyoung MA and neonatal mortality there was no one thing that appeared to explain thelack of agreement between the two groups of studies.There was no pattern in the manner in which young MA was defined in the dissentingstudies as compared with the others. Gortmaker 41 included all mothers less than 20years in the young MA category while Miller and Stokes 73 limited it to mothers between13 and 17 years. In the group of studies which found an association between young MAPage 58and postneonatal mortality young mothers were defined as less than 20 years in twostudies 77,74 , as most mothers less than 19 years3875 and mothers less than 18 years 48A2 .There was also some variation in the sample sizes of the various studies. The samplesin those studies which did not find a significant relationship between young MA andpos tneonatal mortality rates varied between 2,000 U.S. counties 73 and a random sampleof 10,395 white, legitimate, American births 41 . These were somewhat smaller thansome of the samples in studies which did find an association. In this latter group twostudies77,75 had samples greater than 100,000, one included 6% of all births in the U.S.in 1974-75 42 and all the births in Wales and England in 1949-50 and again in 1975 werepart of another 38 . In the Miller and Stokes study 73 the direct effect of the percent ofbirths in the county that occurred to teen mothers was so low (B=0.01) that the fact thatthe study only included 2,000 counties was not the critical factor in the failure to obtaina significant result. In the other non-significant result 41 the chi square value was 7.43,probably close to significance and the coefficient estimate for teen mothers was 0.40.The fact that the study was limited to white, legitimate births was probably moreimportant than the sample size in explaining the non-significance.There was also a range in the number and types of variables included in these analyses.Of the two studies which did not find an association, one 73 included no additionalvariables beyond those measuring SES and young MA. In addition to these variablesthe second study 41 included birth weight, previous pregnancy loss, pregnancy expe-rience and parity. This was also the study that was limited to legitimate births to whitefemales. The group which found a significant association included studies with asimilar range of variables. The case-control 48 and one other study 38 did not include anyvariables beyond those dealing with SES and young MA. The international study 74,included three additional variables but all dealt with measuring the level of health carePage 59services in each of the countries. The last three 77,38,75 had a number of additionalvariables including parity and birth weight.This lack of a definite pattern once again makes it difficult to conclusively comment onthe existence of a relationship between young MA and postneonatal mortality, eitherdirectly or through birth weight. Given that one of the two studies that did not find arelationship between these two variables was limited to legitimate births to whites, andthat in the other study young mothers were compared with all other mothers asopposed to only younger adult mothers, there is the temptation to suggest that theremay be an association between these two variables. The association may be weak andso only be found to be significant relative to mothers with the best birth outcomes(usually those between 20 and 34 years).III Interaction Between Young Maternal Age and Poor Socioeconomic CircumstancesOf the studies included in this review, only two considered the possibility of aninteraction between MA and SES with respect to postneonatal mortality. Machin,Murrells, Catford and Smith38 found a significant interaction between MA and socialclass with respect to postneonatal mortality. Cramer75, on the other hand did not findsignificant interactions with either MA and education or MA and race with respect topostneonatal mortality. Although the measures of SES were different in the two studiesthese contradictory results confirm the need for further study of the possible interactionbetween the two variables.Page 60Congenital Anomaly Rates Unfortunately no studies were found which simultaneously considered the affect ofyoung MA and poor SES on congenital anomaly rates. There were three case-controlstudies68,48,64 which looked at the impact of MA within the context of lower SES. Therewas also one population based study carried out in Tasmania" in which the effect ofMA on congenital anomaly rates was investigated among women in the lowest socialclass. As a consequence it is not possible to discuss past research findings with respectto the effect of SES on congenital anomaly rates.II Impact of Young Maternal AgeAll four studies 68,48,44,64 which looked at the relationship between MA and congenitalanomalies found that there was no association between these two variables. Althoughnot significant, in three studies48,44,64 the rates were lower for young mothers than forthe comparison group. In one study48 the comparison group included all mothers over20 years, in the population based study" this groups was restricted to mothers aged 18through 34 years, in another 64 mothers aged less than 16 years were compared withthose aged 20 to 24 years and in the last study 68 mothers aged less than 15 years werecompared with those 15 years and over. Two of the studies 44,64 were limited toprimiparous births while in the other two 68,48 there was no such restriction. In none ofthe studies was a definition of congenital anomalies provided by the authors. In onestudy 68 congenital anomalies diagnoses prior to discharge from the hospital wereincluded while in the other three 48,44,64 no time frame was provided.Page 61Although the number studies was small, the fact that young MA was found not to beassociated with congenital anomalies despite the lack of consistency of definitions anddesigns suggests that this lack of association is real, at least for mothers with poor SES.III Interaction Between Young Maternal Age and Poor Socioeconomic CircumstancesIt was not possible for any of the studies included in this review to consider thepossibility of an interaction between MA and SES with respect to congenital anomalyrates. This was because each study only considered within the context of poor SES.SUMMARYAlthough the association between the percentage of total births occurring to teenagedmothers and SES has not been widely studied there does appear to be evidence toconclude that a significant association exists. As poverty increases the percentage oftotal births to young mothers increases. One of the studies reviewed 14 did find thisassociation within urban Canada. Unfortunately the association was not tested to seeif it was statistically significant. Further research within a contemporary Canadiancontext would be helpful to confirm this as a significant association within the presentsocial and political contexts within this country.Conclusions are much harder to draw with respect to the independent effect of youngMA and low SES on various birth outcomes. There were instances in which it waspossible to come to a clear conclusion concerning the association between the birthoutcome and the risk factor because of a consensus between the studies reviewed. Thesewere the associations between SES and each of low birth weight, prematurity, still-births, infant mortality, and postneonatal mortality, the lack of associations betweenPage 62young MA and each of small for gestational age births and congenital anomalies. In allother cases the lack of consensus between the studies reviewed precluded any firmconclusions about relationships and suggests the need for further research. This wasalso very much the case with respect to the analysis of possible interaction betweenyoung MA and low SES with respect to the various birth outcomes. Except in the caseswere there was a clear consensus between the articles reviewed with respect to theexistence or nonexistence of a relationship between either young MA or low SES, furtherresearch is required to clarify there relationships. Additionally, only two of the studiesreviewed33,66 regarding the effect of either MA, or SES used Canadian data. It isimportant to examine potential relationships between these two risk factors and poorbirth outcomes within a Canadian context for a number of reasons. As has been notedearlier, different social and political contexts may impact either the existence of a SESgradient and /or the effect of such a gradient on birth outcomes. Social and politicalcontext may also be important if young MA is associated with elevated rates of poorbirth outcomes, not because of the biological immaturity of the mother, but because ofsocial, economic or other circumstances related to adolescent motherhood.It was also noted in a number of studies that relationships between MA and SES andparticular birth outcomes had changed over time. For example Machin, Murrells,Catford and Smith38 found that, between 1949/50 and 1975, age had become a moreimportant risk factor and social class a less important risk factor for postneonatalmortality. In another study76 it was noted that social class had become a moresignificant risk factor for neonatal mortality between 1949/50 and 1975. This suggeststhat as social, political and economic climates change over time or as various interven-tion strategies are implemented it is important to have current analyses with which toevaluate any changes in the relevance of either SES and MA as risk factors for poor birthoutcomes.Page 63REFERENCES1.Harris J. Pregnancy and labour in young primiparae. John Hopkins Hosp Bull1922;33:12-6.2.Aznar R, Bennett A. Pregnancy in the adolescent girl. Am J Obstet Gynecol1961;81(5):934-40.3.Battaglia F, Frazier T, Hellegers A. Obstetric and Pediatric Complications of JuvenilePregnancy. Pediatrics 1963;(Nov.):902-10.4.Israel S, Woutersz T. Teen-age obstetrics. Am J Obstet Gynec 1963;85(4):659-68.5.Grant J, Heald F. Complications of Adolescent Pregnancy. Clin Pediatrics 1972;11(10):567-70.6.Committee to Study the Prevention of Low Birthweight. Preventing low birthweight.Washington, D.C.: National Academy Press, 1985: 284.7.Kramer MS. Determinants of Low BIrth Weight: Methodological Assessment andMeta-Analysis. BUll WHO 1987;65(5):663-737.8.Dott A, Fort A. Medical and social factors affecting early teenage pregnancy. Am JObstet Gynecol 1976;125(4):532-36.9.Smith T. Influence of socioeeconomic factors on attaining targets for reducing teenagepregnancies. Br Med J 1993;306:1232-5.10.Perlman S, Klerman L, Kinard E. The use of socioeconomic data to predict teenagebirth rates. An exploratory study in Massachusetts. Pub Health Rep1981;96 (4):335-41.11.Lieberman E, Kenneth J, Monson R, Scoenbaum S. Risk factors accounting for racialdifferences in the rate of premature birth. N Engl J Med 1987;317:743-8.12.Gould J, LeRoy S. Socioeconomic status and low birth wieight: A racial comparison.Pediatrics 1988;82(6):896-904.13.Collins J, Richard J. The differential effect of traditional risk factors on infantbirthweight among blacks and whites in Chicago. Am J Public Health1990;80(6):679-81.Page 6414.Wilkins R, Sherman GJ, Best P. Birth outcomes and infant mortality by income inurban Canada, 1986. Health Reports 1991;3(1):7-31.15.Williams S, Forbes J, Mcllwaine G, Rosenberg K. Poverty and teenage pregnancy. BrMed J 1987;294:20-1.16.Placak P. Maternal and infant health factors associated with low infant birthweight:Findings from the 1972 national natality survey. In: Reed D, Stanley F ed.Epidemiology of prematurity. New York: Urban and Schwarenberg,1977:17.Taffel S. Factors associated with low birth weight: United States, 1976. WashingtonD.C.: U.S. Government Printing Office, 1980: 1-22.18.Stanley F,Hobbs M. Perinatal outcome in Western Australia, 1968 to 1975. Med J Aust1981;1:414-6.19.U.S. Department of Health and Human Services/ Public Health Service. Nationalbirthweight-specific infant mortality surveillance: preliminary analysis -United States, 1980. MMWR 1986;35 (17):269-73.20.Hendricks C. Twinning in relation to birth weight, mortality and congential anoma-lies. Obstet Gynecol 1966;27(1):47-53.21.Powers W. Twin pregnancy, complications and treatment. Obstet Gynecol1973;42(6):795-808.22.Wiener G, Milton T. Demographic correlates of low birth weight. Am J Epidemiol1970;91:260-72.23.Tennes K, Blackard C. Maternal alcohol consumption, birth weight, and minorphysical anomalies. Am J Obstet Gynecol 1980;138:774-80.24.Polednak A, Janerich D, Glebatis D. Birth weight and birth defects in relation tomaternal spermicide use. Teratology 1982;26:27-38.25.Spiers P, Wacholder S. Association between rates of premature delivery and intra-uterine growth retardation. Dev Med Child Neurol 1982;24:808-16.26.Fedrick J, Adelstein P. Factors associated with low birth weight of infants deliveredat term. Br J Obstet Gynecol 1978;85(1):1-7.27.Newton R, Hunt L. Psychosocial stress in pregnancy and its relation to low birthweight. Br Med J 1984;288:1191-4.Page 6528.Hardy J, Mellitis E. Relationship of low birth weight to maternal characteristics ofage, parity, education and body size. In: Reed D, Stanley F ed. Epidemi-ology of prematurity. New York: Urban and Scharzenberg, 1977.29.Kleinman J, Kessel S. Racial differences in low birth weight: Trends and risk factors.New Engl J Med 1987;317(12):749-53.30.Kessner D, Singer J, Kalk C, Schlesinger E. Infant death: An analysis by maternal riskfactor and health care.. Washington, D.C.: Institute of Medicine, NationalAcademy of Sciences, 1973.31.Smedby B, Ericson A. Perinatal mortality among children of immigrant mothers inSweden. Acta Paediatr Scan 1979;275 (41):41-6.32.Goldstein H. Factors related to birth weight and perinatal mortality. Med Bull1981;37(3):259-64.33.Silins J. Risk factors for perinatal mortality in Canada. Can Med Assoc J1985;133(15):1214-219.34.Forbes J, Pickering R. Influence of maternal age, parity and social class on perinatalmortality in Scotland: 1960-82. J Biosoc Sci 1985;17:339-49.35.Murrells T, Catford J, Smith T, Machin D. The use of logit models to investigate socialand biological factors in infant mortality. II: Stillbirths. Stat Med 1985;4:189-200.36.Murrells T, Catford J, Smith T. The use of logit models to investigate social andbiological factors in infant mortality. III: Neonatal mortality. Stat Med1986;5:139-53.37.Pharoah P, Morris J. Postneonatal mortality. Epid Rev 1979;1:170-83.38.Machin D, Murrells T, Catford J, Smith T. The use of logit models to investigate socialand biological factors in infant mortality. IV: Post-neonatal mortality. StatMed 1986;5:155-69.39.Rosenwaike I. The influence of socio-economic status on incidence of low birthweight. HSMHA Health Reports 1971;86(7):641-9.40.Eisner V, Brazie J, Pratt M, al. e. The risk of low birthweight. Am J Public Health1979;69(9):887-893.41.Gortmaker S. Poverty and infant mortality in the United States. Am Soc Rev1979;44:280-97.Page 6642.Bross D, Shapiro S. Direct and indirect associations of five factors with infantmortality. Am J Epidemiol 1982;115(1):78-91.43.Carlson ED. Social Determinants of Low Birth Weight In a High-Risk Population.Demography 1984;21(2):207-15.44.Correy J, Kwok P, Newman N, Curran J. Adolescent pregnancy in Tasmania. Med JAust 1984;141:150-4.45.Showstack J, Budeatti P, Minkler D. Factors associated with birthweight: An explo-ration of the roles of prenatal care and length of gestation. Am J PublicHealth 1984;74(9):1003-8.46.Lee K, Ferguson R, Corpuz M, Gartner L. Maternal age and incidence of low birthweight at term: A population study. Am J Obstet Gyneco11988;158(1):84-89.47.Crosby W, Metcoff J, Costiloe J, et al. Fetal malnutrition: An appraisal of correlatedfactors. Am J Obstet Gynecol 1977;128(1):22-31.48.Hardy J, Welcher D, Stanley J, Dallas J. Long-range outcome of adolescent preg-nancy. Clin Obstet Gynecol 1978;21(4):1215-32.49.Hutchins F, Kendall N, Rubino J. Experience with teenage pregnancy. ObstetGynecol 1979;54(1):1-5.50.Horon I, Strobino D, McDonald H. Birth weights among infants born to adolescentand young adult women. Am J Obstet Gynecol 1983;146(4):444-9.51.Zuckerman B, al. e. Neonatal outcome: Is adolescent pregnancy a risk factor.Pediatrics 1983;71(4):489-93.52.Anderson G, Blinder I, McClemont S, Sinclair J. Determinants of size at birth in aCanadian population. Am J Obstet Gynecol 1984;150(3):236-44.53.Donaldson P, Billy J. The impact of prenatal care on birth weight Evidence from aninternational data set. Medical Care 1984;22(2):177-88.54.Leppert P, Brickner Namerow P, Barker D. Pregnancy outcomes among adolescentand older women receiving comprehensive prenatal care. J AdolescHealth Care 1986;7:112-7.55.Shiono P, Klebanoff M, Graubard M, Berendes H, al. e. Birth weight among womenof different ethnic groups. JAMA 1986;255 (1):48-52.Page 6756.Giblin P, Poland M, Waller J, Ager J. Correlates of neonatal morbidity: maternalcharacteristics and family resources. J Genetic Psychology 1987;149:527-33.57.Teberg A, Settlage R, Hodgman J, al. e. Maternal factors associated with deliveryofilnfants with birthweight less than 2000 grams in a low socioeconomicpopulation. J Perinato11988;9(3):291-5.58.McLaughlin F, al. e. Randomized Trial of Comprehensive Prenatal Care for Low-Income Women: Effect on Infant Birth Weight. Pediatrics 1992;89 (1):128-32.59.Palti H, Adler B. Body size of Israeli newborn infants in relation to regional origin oftheir mothers. Hum Biol 1977;49:41-50.60.Miller H, Hassanein K. Maternal factors in "fetally malnourished" black newborninfants. Am J Obstet Gynecol 1974;118(1):62-7.61.Duenhoelter J, Facog J, Jimenez J, Baumann G. Pregnancy performance of patientsunder fifteen years of age. Obstet Gynecol 1975;46(1):49-52.62.Scholl T. Early adolescent pregnancy: A comparative study of pregnancy outcomein young adolescents and mature women. J Adolesc Health Care 1984;5:167-71.63.Geronimus A. The effects of race, residence, and prenatal care on the relationship ofmaternal age to neonatal mortality. Am J Public Health 1986;76(12):1416-21.64.Sukanich A, Rogers K, McDonald H. Physical maturity and outcome of pregnancyin primiparas younger than 16 years of age. Pediatrics 1986;78(1):31-6.65.Brown H, Fan Y-D, Gonsoulin W. Obstetric complications in young teenagers. SouthMed J 1991;84 (1):46-8.66.Arbuckle T, Sherman G. Comparison of the risk factors for pre-term delivery andintrauterine growth retardation. Paediatr Perinat Epidemio11989;13:1989.67.Abrams B, Newman V. Small-for-gestational-age birth: Maternal predictors andcomparison with risk factors of spontaneous preterm delivery in the samecohort. Am J Obstet Gynecol 1991;164(3):785-90.68.Coates J. Obstetrics in the very young adolescent. Am J Obstet Gynec 1970;108(1):68-72.Page 6869.Scott A, Moar V, Ounsted M. The relative contributions of different maternal factorsin small-for-gestational-age pregnancies. Europ J Obstet Gynec1981;12:157-65.70.Naeye R. Maternal age, obstetric complications, and the outcome of pregnancy.Obstet Gynecol 1983;61(2):210-6.71.Elwood J, MacKenzie G, Cran G. Observations on single births to women resident inBelfast 1962-1966. Part I. Factors associated with perinatal mortality. JChron Dis 1974;27:517-35.72.Samueloff A, Mor-Yosef S, Seidman DS, Adler I, Persitz E, Schenker JG. Ranking riskfactors for perinatal mortality. Acta Obstet Gynecol Scand 1989;68:677-682.73.Miller M, Stokes S. Teenage fertility, socioeconomic status and infant mortality. JBiosoc Sci 1985;17:147-55.74.Pampel F, Pillai V. Patterns and determinants of infant mortaltiy in developednations, 1950-1975. Demography 1986;23(4):525- 41.75.Cramer J. Social factors and infant mortality: Identifying high risk groups andproximate causes. Demography 1987;24(3):299-322.76.Machin D, Murrells T, Catford J, Smith T. The use of logit models to investigate socialand biological factors in infant mortality . III. Neonatal mortality. Statis-tics in Medicine 1986;5:139-53.77.Shah F, Abbey H. Effects of some factors on neonatal and postneonatal mortality.Analysis by a binary variable multiple regression method. Milbank MemFund Q 1971;49:33-57.Page 69CHAPTER FOURRATIONALE AND METHODSRATIONALE AND PURPOSEFrom the previous chapters it is evident that although the percentage of births toteenaged mothers has diminished they are still occurring in large enough numbers towarrant attention given that the literature suggests that there is a significant associationbetween young maternal age (MA) and poor socioeconomic status (SES) and poor birthoutcomes. Although a number of associations between these two risk factors andvarious birth outcomes have been well established there are instances in which this isnot the case. Further research is required to clarify these relationships. Additionally,for reasons presented in Chapter Three, current Canadian research into these relation-ships would be helpful. Unfortunately a statistical analysis of underlying issues relatedto negative birth outcomes associated with teen pregnancy is difficult because of therelatively small proportion of births which occur to this group and the small percentageof births which have poor outcomes. B.C. has a number of province wide administrativedata bases which routinely collect birth and other health information and which can belinked by way of a unique identifier. The addition of postal code information to thepersonal identifier section in some data bases in 1985 now provides a linkage to federalcensus tracts, for which economic and other information collected by Statistics Canadais available. This method of linking individual and aggregate data has been refined byR. Wilkins of Statistics Canada within the context of an ecological, population based,cross sectional study design and been utilized in a number of Canadian studies 1,2,3 .Wilkins' method was utilized in the project described in this paper. The study designallowed for the examination of the effect of both young MA and low socioeconomicPage 70status (SES) on birth outcomes of infants born to teenaged mothers in urban BritishColumbia (B.C.) from January 1, 1985 through December 31, 1988.The following is a more detailed description of the purposes of the study:1) to determine the association between MA and poverty,2) to determine the independent impact of young MA and poverty on the rates of thevarious birth outcomes.METHODStudy PopulationAll first single live births and stillbirths which occurred in one of five urban areas in theprovince of B.C. from January 1, 1985 through December 31, 1988, to women under 35years made up the study population. This particular time frame was chosen becauseStatistics Canada sociodemographic information from the 1986 Census was utilized inthe study. Since the socioeconomic profile of a census tract may change as peoplemigrate in and out, it was important to have birth outcome data for a time period thatwas as close as possible to the 1986 Census. Postal codes only began to be captured onbirth records in 1985 so this was the first year that could be included in the study.Because of concerns that birth data to teen mothers might be too sparse to allowadequate analysis by SES groups, data were collected for a four year period.Page 71The restriction of the study to five urban areas was necessitated by the way in whichfederal census tracts have been defined. In Canada the term 'census tract' can refer toeither a provincial or federal census tract. The latter are only located in urban areas andwere the ones utilized in this study. These census tracts were established by StatisticsCanada in order to record and report socioeconomic and demographic data collectedin large urban communities through the Canadian census. These tracts are permanentgeostatistical areas usually containing between 2,500 and 8,000 residents, with apreferred average of 4,000 4 . Whenever possible the boundaries have been drawn alongrecognizable divisions between neighborhoods in order to create tracts that are ascompact and socioeconomically homogeneous as possible. This tracting has occurredin Census Metropolitan Areas (CMAs) which include all areas having a population over100,000 based on the previous census and in Census Agglomerations (CAs) whichencompass cities with between 10,000 and 100,000 inhabitants. All CMAs and largerCAs have been tracted. In British Columbia, Vancouver and Victoria were CMAs whilePrince George, Kamloops and Kelowna were tracted CAs in the 1986 census (see mapsin Appendix A). The study population had to be restricted to these five urban areas .From the review of the literature described in Chapter Three there is evidence thatparity and multiplicity affect the rates of various birth outcomes. The affect of thesevariables was controlled for by restricting the study popoulation. Only first single livebirths were included in the study.There is also evidence that MAs greater than 35 years at the birth of the mother's firstchild are associated with increases in negative birth outcomes. There is national datafor the U.S. and Britian as well as smaller studies that shows that first births to mothersgreater than 35 years are at greater risk of LBW then mothers in their twenties 5, 6, 7, 8,9, 10. There is also evidence that perinatal mortality 11, 12 and infant mortality 13, 14Page 72incresases for infants born to these older mothers. Although Kramer 15 and others 16,in a review of the literature found evidence that it may in fact be older MA interactingwith other risk factors that produces these elevated rates of poor birth outcomes it wasfelt that birth outcomes for infants born to teenaged mothers should be compared withthose of infants born to mothers less than 35 years Regardless, these younger motherswould most likely have the highest probability of having positive outcomes. Thiswould allow for the establishment of some measure of the risk associated with youngMA relative to births to women with the lowest levels of risk.Data SourcesI Maternal and Still/Live Birth Related Data Maternal, birth and birth outcome data for January 1, 1985 through December 31, 1988were obtained from the Division of Vital Statistics (DVS), B.C. Ministry of Health andMinistry Responsible for Seniors. The DVS provided information from Physician'sNotices of (still and live) Birth (PNOB), the Death Registry and the Health StatusRegistry (HSR).a) Physician's Notices of (live and still) BirthSince physicians are legally required to complete a PNOB form within 48 hours ofattending a birth, the records capture birth information for more than 99% of the still/live births in the province 17 . The present version of the PNOB is the same for both liveand still births. It was revised in 1985 to include the postal code of the mother's usualresidence at the time of birth of her infanta. Both maternal and infant still/live birtha. UH So Hong, DVS, personal communication, February 1991Page 73dates were obtained from the PNOBs and used in calculating the mother's age at thetime of the still/live birth. Outcome variables extracted from individual PNOB recordsfor live births were birth weight, gender, and gestational age. Birth weight was utilizedto calculate low birth weight (LBW) and very low birth weight (VLBW) rates andtogether with gender and gestational age was used to compute small for gestational age(SGA) rates.b) Death RegistryA physician or coroner is required by law to complete a Medical Certificate of Death,which is passed on to the funeral director. A burial permit is only issued upon thesubmission of the certificate as well as a death registration form to the DVS by thefuneral director 18. Information from these forms is used to update the provincial deathregistry. The registration information includes the date of death and age at death. Thisallowed the identification of infant deaths and the calculation of the infant's age atdeath. Although it has been suggested 17 that this process assures a 99% ascertainmentrate for deaths occurring within the province, this may not be the case for infants. It isnot clear how many infants who die very shortly after birth are not registered. However,as it is a legal requirement to register all deaths it is suggested that the proportion ofinfant deaths not registered is small.c) Health Status RegistryThe DVS has been responsible for the HSR since its establishment in 1952. Thispopulation based data collection system ascertains handicapped children and adultswho are born in or resident in the province 19,20 . To be registered on the HSR individualsmust have a physical, mental, or emotional problem that is likely to be permanentlydisabling, interfere with the individual's ability to obtain an education, or prevent fulland open employment, or they have to have a familial condition or congenital anomalyPage 7421 . Individuals can be added to the registry at any age and, once registered, their recordis updated as new diagnoses become available. These are coded according to the mostrecent edition of the International Classification of Disease (ICD) 22, the latest beingICD-9. Each individual registered may have up to 20 different diagnoses. Only thoseinfants registered with the HSR in the first year of their life were utilized in this study.Cases included in the HSR are registered voluntarily. In the past it has been suggestedthat the resulting threat of under-ascertainment was potentially minimized by the factthat the registry can receive information from over sixty sources, including health units,hospital admissions and separations, and other provincial agencies 23. Unfortunatelyin the last few years there have been organizational as well as other changes which mayhave had some impact on registration rates as well as data quality.II Socioeconomic Status DataSES data were provided by the Occupational and Environmental Health ResearchSection of Statistics Canada from the 1986 Census. Statistics Canada provided summaryinformation by census tracts for Vancouver, Victoria, Kamloops, Kelowna and PrinceGeorge. The values were based on information from the long form of the censusquestionnaire collected from a 20% sample of all the families living in the census tractin 1986. Since the percentages are estimates based on a relatively small sample no valueswere provided for any tract with a non-reserve private household population of fewerthan 250 persons aged less than 18 years. Institutional tracts with few or no privatehouseholds, industrial areas with little or no residential population and most Indianreserves were also excluded 3. The latter were not included because the nativepopulation in Canada is excluded from the Statistics Canada definition of low income.Page 75Procedures I Data Base DevelopmentBirth registration numbers assigned to PNOB (live and stillbirth) records are uniquewithin each calender year. These numbers are also included as identifying informationin the HSR. Consequently, it was possible to use the year of birth and birth registrationnumber together to create a unique identifier to link congenital anomaly informationfrom the HSR to individual PNOB (live birth) records. The Death Registry does notinclude birth registration numbers and so infant mortality data can not routinely linkeddirectly to PNOB records. Fortunately a linked live birth and infant death file preparedby the DVS for an earlier study at the University of British Columbia (U.B.C.) could beused to obtain birth registration numbers for those infants who died in the first year oflife.The required information for each live birth was extracted from the three registries,linked together via the unique identifier created by the year of birth and the birthregistration number. Maternal neighborhood low income information was then addedto the data base using the postal code of the mother's usual residence at the time of thebirth.Although a relatively recent addition to Canadian address identification, postal codeinformation on PNOB forms is very useful and fairly accurate. In a recent Canadianstudy it was found that the postal code error rate in B.C. Vital Statistics data wasapproximately 1 percentb. The usefulness of these codes is a consequence of theirb. Wilkins R., personal communication, February 1991Page 76association with the geographic location of mail pick up, which is usually, but notalways, the ultimate destination of the mail delivery. In urban areas, such as thoseincluded in this study, over 90% of the postal codes are A (private household andsupermail boxes), B (apartment building), E (commercial office building), or G (largevolume receiver). These delivery mode type codes (DMT) codes are linked to the actuallocation of the ultimate destination of the mail 4. Therefore the postal codes associatedwith these DMT codes can be directly converted to the census tract in which theresidence is located. Urban postal codes with a DMT code of H (rural route delivery),J (general delivery), K (group of post office boxes) or M (one post office box) are linkedto the physical location of the post office which receives this mail. Since it is not certainthat the post office which receives the mail is located in the same census tract as theresidence of the individual to whom the mail is addressed, these codes cannot bedirectly converted to the census tract identifiers. In rural areas, postal codes areidentified by a "0" in the second position in the code and they have a blank DMT code.These rural codes are linked to the census tract in which the rural post office is located.Again, there is no assurance that individuals served by this post office live in the samecensus tract as the post office. Even though the study population was urban these lattertwo groups of codes accounted for 10% of the postal codes that had to be convertedbecause a rural fringe is often included within the boundaries of CMA/CAs.The postal codes from PNOB (live and still) records for 1985 - 1988 that had a DMT codeof A, B, E, or G were converted to census tract identifiers by the Population Section,Planning and Statistics Division of the B.C. Ministry of Finance and Corporate Relations(PSD), using their Translation Master Filec (TMF) and by Statistics Canada. Thec. This in-house computer software program does not provide DMT codes when it converts postalcodes to census tract identities and so there is no way of knowing how many were E (commercial)or G (large volume receiver) DMT codes. It is suggested that not many were E or G codes because mostmothers would provide home address information for the birth registration.Page 77"problematic" urban and rural postal codes with a H, K, M, or blank DMT code, as wellas addresses for which the postal code was missing, were manually assigned to censustracts, by the PSD using address information provided by the DVS. The SES informationprovided by Statistics Canada also included a census tract identifier. The linkage of thesummary income information to the PNOB records was by way of this identifier foundin both files.The SES information provided by Statistics Canada was the percentage of the popula-tion less than 18 years in the census tract that was living below the low income cut-offin the year before the Census. For each of the five CMAs and CAs separately, the linkedlive birth data were ranked, in ascending order, according to these percentages. In eachof the CMA/CAs, census tracts were then assigned to one of five neighborhood povertyquintilesd (NPQs) such that each NPQ included approximately one fifth of the firstsingle live births between Jan. 1, 1985 and Dec. 31, 1988 to mothers aged less than 35years (see Appendix A for the location of the census tracts in each of the NPQs). For eachof the five CMA/CAs, NPQ 1 consisted of approximately 20 percent of the births, withthe births coming from those tracts with the lowest percentages of children living belowthe poverty line. NPQ 5 included about the same number of births from tracts with thehighest percentages of children living in poverty. After NPQs were established in eachof the five CMAs/CAs all the similar numbered NPQs were amalgamated. Thus, NPQ1 was referred to as the least poor because it included tracts with the lowest percentagesof children living below the low income cutoff from all 5 urban areas and NPQ 5 wascalled the most poor. Constructing NPQs within each of the CMAs and CAs separatelyand afterwards combining those with the same designation minimized the effect ofd. Following the usual practice at Statistics Canada, the phrase, "a quintile group" is replaced with"quintile".Page 78interurban differences in income, housing and other costs of living. As was noted byWilkins, in his recent study 3, if all census tracts in that study had been ranked beforeconstructing NPQs, over one third of all the births in metropolitan Toronto would havefallen in NPQ 1, while the CMAs south-east of the Saint Lawrence River would not havehad any births in this NPQ. This allowed the focus to be on income heterogeneity withinurban centres, as opposed to income distribution between centres, which was moreconsistent with the purposes of the study.Once the NPQ boundaries had been established the stillbirths were located withinthem. The decision to proceed in this way was based on the fact that the study focus wason birth outcomes and not pregnancies. Postal codes on PNOB (stillbirth) records wereconverted to census tract identifiers using the procedure previously described. Withineach of the CMA/CAs, the NPQ designation of the live birth records with the samecensus tract identifier as any stillbirth record was then added to the latter. Finally,stillbirth records, with the same NPQ designation, from each of the five cities wereamalgamated into the live birth file.II Variable DefinitionMA and SES were the two independent variables in this study. The MA variable wascreated by dividing mothers aged less than 35 years of age into two groups. Womenunder 20 years of age are considered teens in many of the studies that have investigatedthe association of young MA and birth outcomes. Mothers under 20 years of age at thetime of the birth of their first child was the classification that was utilized in this studygiven that a preliminary analysis of data suggested that birth outcomes to subgroupsof young mothers followed similar patterns and that the numbers in various subgroupswould be quite small.(see Appendix B for a discussion of the rationale for subdividingPage 79the teens and the results of an investigation into the merits of doing so). The comparisongroup was restricted to mothers between 20 and 34 years at the time of their first birth.SES is commonly measured by variables such as occupation, education, income, race ora composite generated from the weighted sum of a number of variables 24 . In this studythe choice of a variable based on family income in the census tract of the mother's usualresidence to represent SES was a function of information available by census tractthrough the 1986 Census and of recent studies by Wilkins and Thomson 1,2,3 . The choiceof family income is also supported by studies that suggest that income, rather thaneducation or occupation correlates best with SES differences in mortality 25, 26, 27, 28,29,30, 31,32 .Various census tract-based measures of family income are available from the Canadiancensus. Differences between them relate to their definition of family. According to the1986 Census Dictionary 33, a household refers to a person or group of persons whooccupy a private dwelling and do not have a usual place of residence elsewhere inCanada. A census family includes a husband and wife with children who have nevermarried, (regardless of age), or a lone parent of any marital status with one or morechildren who have never married, living in the same dwelling. Groups of two or morepersons who live in the same dwelling and are related to each other by blood, marriage(including common-law relationships) or adoption are referred to as an economicfamily. The first definition was least consistent with the focus of the study because it isbased on the occupancy of a dwelling rather than any family tie. The choice of eithercensus family or economic family would have been appropriate given that both aremore likely than household to be based on the notion of economic dependancy.Economic family was the definition selected because it was utilized in the previousstudies 1, 2, 3.Page 80In these studies the percentage of persons under 18 years of age, excluding inmates ofinstitutions and residents of Indian reserves, in the census tract of the mother'sresidence whose economic family income in 1985 was below the Statistics Canada lowincome cut-off was calculated. Low income cut-offs were adjusted for family size andpopulation of the CMA or CA 33 . The latter is intended to adjust for differences in thecost of living that are a consequence of the size of the city. These adjusted percentageswere referred to as maternal neighborhood low income values and were utilized asproxy measures of the mother's SES at the time of her child's birth. A poverty variablefor the study was created by ranking the percentages in ascending order and dividingthem into five NPQs.The various birth outcomes were the dependent variables in this study. These werestillbirths, LBW, VLBW, SGA, IM , NM, PNM, and CA. Because these variables wereeither extracted from or created from data taken from DVS records their definitionswere, for the most part, consistent with those of the DVS.During the study period the Vital Statistics Act definition of stillbirth was altered. OnJanuary 2, 1986 the definition changed from, "The complete expulsion or extractionfrom its mother, after at least twenty weeks' pregnancy..." to, "The complete expulsionor extraction from its mother, after at least twenty weeks' pregnancy, or after attaininga weight of at least 500 grams... 34. The latter definition is more inclusive and covers alarger portion of the study period so it was the definition of stillbirth applied in thisanalysis.The definitions of LBW and VLBW were a birth weight of less than 2,500 grams and abirth weight of less than 1,500 grams respectively 34. The SGA standards utilized in thisstudy were those described by Arbuckle and Sherman 35, utilizing information relatingPage 81to all live single births in Canada in 1986. Standards were presented for males andfemales separately for births occurring between 25 and 42 weeks gestation. There were,however, over 400 births at 43 weeks gestation included in the study population whichwould have been lost to SGA calculations based on Arbuckle and Sherman's standards.Consequently a SGA standard based on B.C. data was calculated for this group usingthese births. These cutoff values for both males and females born at 43 weeks werefound to be similar but somewhat larger then the values given by Arbuckle andSherman for 42 weeks. Because they were consistent with the pattern of the presentedrates they were added to them for this study.The definitions for IM, NM and PNM were those utilized by the DVS. Infant mortalityis the death of a child under one year, while NM is death under 28 days and PNM isdeath between 28 days and under one year 34 .The definition of congenital anomaly that was utilized was the registration criteria usedby the HSR. This is that individuals must have a physical, mental, or emotional problemthat is likely to be permanently disabling, interfere with the individual's ability to obtainan education, or prevent full and open employment, or they have to have a familialcondition or congenital anomaly 21 . This registration criteria is obviously a broader thanthe definition of congenital anomalies (physical defects in a person that existed or datafrom birth (ICD-9 categories 740-759) 34). However, it was felt that for infants less thanone year of age the majority of the registrations would be for congenital anomalies.Page 82Data AnalysesI Evaluation of Data Base Development Methods One of the purposes of the study was to evaluate the strengths, weaknesses andlimitations of the design employed in this study. One of the reasons for including anevaluation of the design is because it was the first time this method had been used in astudy employing data from three B.C. data bases. Additionally, using postal codes tolink health data and socioeconomic information is relatively new as is the technique ofapproximating a mother's SES through the use of the percentage of children living inpoverty in the census tract of her usual residence. Given that one of the CAs had asubstantially higher percentage of birth records which could not be located in a NPQ anumber of the birth outcome analyses were redone with various sub-groupings of theCMAs and CAs. The decision to retain this CA in the study was based on the results ofthese analyses.It was also important to establish whether the methods used in this study weresuccessful in establishing a poverty gradient across the quintiles for all five urban areasincluded in the study because previous Canadian studies 1, 2, 3 which have utilizedsimilar methods limited their study populations to individuals resident in CMAs. Therational given for this decision was that neighborhoods are more clearly defined andresidential segregation by income more pronounced in large cities than in small centresand rural areas. While this statement is not to be disputed, it is possible that the threeCAs in B.C. might be large enough to exhibit patterns of income segregation that werenot substantially different than those of Vancouver and Victoria in 1986.Page 83Using PSearch, a commercial software package that allows for computer analyses of1986 Canadian Census data 36, it was possible to do a limited analysis within NPQs.Unfortunately the package did not include the percentage of individuals under 18 yearsof age in each census tract whose economic family income was below the StatisticsCanada poverty level. Instead the average census family income had to be utilized asa surrogate measure of poverty. Consequently the ranking of census tracts and theirallocation within each CMA/CA was not identical, but very similar to that in the study.To examine income segregation between and income homogeneity within NPQs themean, standard deviation and coefficient of variation were calculated for the censusfamily income in each NPQ in each of the five geographical areas. An analysis of theimpact of retaining a CA with a weaker income gradient in the study was made byredoing a number of the outcomes analyses without the data from this CA.II Evaluation of Variable DefinitionsA descriptive analysis of the various independent and dependent variables wasundertaken prior to the more indepth analysis of the various study questions. The ratesof the various independent variables were calculated and compared to B.C. populationrates for a similar time period.III Analysis of Birth and Birth Outcome Dataa) Incidence of Single First Live BirthsThe ratios of first single live births to teenage and older mothers in the various NPQswere analyzed using chi-square tests.Page 84b) Analysis of Birth Outcomes by Maternal Age Groups, NPQs and Both Maternal AgeGroups and NPQsMean birth weights and gestational ages, as well as rates of stillbirth, LBW, VLBW andSGA births, IM, NM, PNM and CAs were calculated by both MA groups and NPQs.Total rates by MA groups and NPQs as well as an overall rate were calculated for eachbirth outcome.Mean birth weights and gestational ages were analyzed with a two way analysis ofvariance (ANOVA). A two way analysis was employed in order to also be able todetermine if there was an interaction between the effects of MA groups and NPQs apartfrom their independent effects.Differences in the rates of the various birth outcomes between the MA groups wereevaluated with chi-square tests for each NPQ and in total. Odds ratios and 95%confidence intervals were also calculated for differences in overall rates between theMA groups. Differences in the rates of the various birth outcomes between the NPQswere also evaluated with chi-square tests for each MA group separately and in total.Differences between NPQ 5 and 1 in the overall rates of the various birth outcomes werealso compared with chi-square tests. Because these were 2X2 tables, odds ratios and95% confidence ratios were also calculated for these last comparisons.The relative impact of MA groups and NPQs for each of the outcomes previouslyanalyzed with chi-square tests was assessed through the application of Cochran-Mantel-Haenszel (CMH) chi-square tests (more commonly referred to as a Mantel-Haenszel chi-square test) 37, 38, 39. This method allows for the assessment of theassociation between one outcome variable and one risk factor while controlling for theeffect of one or more other potential risk factors. The other factors are utilized asPage 85stratifying variables. When attempting to evaluate the association between an exposureand an outcome in the presence of one or more other risk factors it is important toaddress the following:1. Interaction.^Is the association between the exposure and outcome the same forall strata?2. Confounding.^Is the association effected by the presence of another variable thatis associated with both the exposure and outcome variables?3. Association.^^If the association is the same and not mediated by a confoundingvariable, is it significant?4. Estimation.^What is the magnitude of the exposure-outcome association?The first issue was addressed through a visual inspection of stratum specific and overallodds ratios 40, and a chi-square test of homogeneity (Breslow-Day chi square). The nullhypothesis for this test is that the true odds ratio relating outcome to exposure is thesame for each stratum, which if not true suggests there may be interaction between theexposure and stratifying variables. In this case the summary odds ratio becomesmeaningless.Confounding was partially addressed through the design. Some potential confounders,namely parity and multiplicity, were controlled by excluding them from the study. Adata-based method for establishing the presence or absence of confounding requires thecomparison of a crude effect measure with an adjusted measure that controls for thedistortions due to the confounder 40. In this study crude and adjusted odds ratios werecompared for the possible presence of confounding of an independent variable that wassignificantly associated with the dependent variable and was being controlled for in theCMH analysis . Similar odds ratios suggested that there was no confounding of thePage 86controlled variable on the independent variable whose association with the dependentvariable is being tested.The independent presence of association between the exposure and outcomes variableswas measured by the CMH chi-square test. The estimation of magnitude was addressedthrough a summary ratio of the odds of a birth outcome given an exposure relative tono exposure. The effect of MA groups on the various birth outcomes while controllingfor the five NPQs, was estimated with a summary odds ratio produced by the SASroutine that calculated CMH chi-square values.A CMH chi-square test allows for the testing of three alternate hypotheses when theoutcome variable has more than two levels. These are that the outcome variable islinearly related to the exposure, that the outcome is different for different exposures andthat the distribution of the outcome is different for different exposure levels. The firsthypothesis was felt to be the most relevant within the context of this study and so wasthe one utilized in evaluating the effect of NPQs on birth outcomes while controlling forMA.Unfortunately with greater than two levels of the outcome variable it is not possible forSAS to produce a summary odds ratio or to perform a Breslow-Day chi-square test ofhomogeneity. In order to estimate a summary odds ratio for the effect of NPQs on birthoutcomes while controlling for MA groups, the overall rates of the outcomes in NPQs5 and 1 were analyzed. Because these analyses produced 2X2 tables it was thus possibleto calculate the summary odds ratio and Breslow-Day tests of homogeneity.Because of the large number of cases included in the study, the data were manipu-lated in a mainframe computer environment. The SPSSX statistical package 41 wasPage 87utilized to generate descriptive statistics and ANOVA and chi-square tests. SAS 42was employed for the various CMH and Breslow-Day chi-square tests as well as theestimation of odds ratios.Page 88REFERENCES1.Wilkins R, Adams 0, Brancker A. Changes in mortality by income in urban Canadafrom 1971 to 1986. Health Reports 1989;1(2):137-74.2.Thomson M. Association between mortality and poverty. BC Med J 1990;32 (8):337-38.3.Wilkins R, Sherman GJ, Best P. Birth outcomes and infant mortality by income inurban Canada, 1986. Health Reports 1991;3(1):7-31.4.Statistics Canada. Postal code conversion file user guide. Ottawa Canada: Minister ofSupply and Services Canada, 1991:5.Hardy J, Mellitis E. Relationship of low birth weight to maternal characteristics of age,parity, education and body size. In: Reed D, Stanley F. ed. Epidemiologyof prematurity. New York: Urban and Scharzenberg, 1977.6.Fedrick J, Adelstein P. Factors associated with low birth weight of infants deliveredat term. Br J Obstet Gynecol 1978;85(1):1-7.7.Keeping J, Chang A, Morrisson J, Esler E. Birth weight: Analysis of variance and thelinear additive model. Br J Obstet Gynaecol 1979;86:437-442.8.Taffel S. Factors associated with low birth weight: United States, 1976. WashingtonD.C.: U.S. Government Printing Office, 1980: 1-22.9.Stanley F, Hobbs M. Perinatal outcome in Western Australia, 1968 to 1975. Med J Aust1981;1:414-6.10.Kleinman J, Kessel S. Racial differences in low birth weight: Trends and risk factors.New Engl J Med 1987;317 (12):749-53.11.Cox L, McIntosh J, Seglenieks A, Seglenieks Z, Weldaon A. Perinatal mortality inSouth Australia. Med J Aust 1977;2 (14):461-7.12.Correy J, Kwok P, Newman N, Curran J. Adolescent pregnancy in Tasmania. Med JAust 1984;141:150-4.13.Kessner D, Singer J, Kalk C, Schlesinger E. Infant death: An analysis by maternal riskfactor and health care.. Washington, D.C.: Institute of Medicine, NationalAcademy of Sciences, 1973:Page 8914.U.S. Department of Health and Human Services/ Public Health Service. Nationalbirthweight-specific infant mortality surveillance: preliminary analysis -United States, 1980. MMWR 1986;35 (17):269-73.15.Kramer MS. Determinants of low birth weight: Methodological assessment andmeta-analysis. BUR WHO 1987;65(5):663-737.16.Barken S, Bracken M. Delayed childbearing: No evidence for increased risk of lowbirth weight and preterm delivery. Am J Epidemiol 1987;125:101-9.17.Gibson D. Retinopathy of prematurity in British Columbia [Thesis]. University ofBritish Columbia, 1987.18.Vital Statistics BC Ministry of Health and Ministry Responsible for Seniors. SelectedVital Statistics And Health Status Indicators, One Hundred NineteenthAnnual Report, 1990. . Victoria: Crown Publications Inc., 1991:19.Lowry R, Miller J, Scott A, Renwick D. The British Columbia registry for handi-capped children and adults: Evolutionary changes over twenty years.Can J Public Health 1975;66:322-6.20.Colls M, Baird P, Gibson D. Measuring morbidity in a population: The British Co-lumbia health surveillance registry. Can J Public Health 1982;73:313-8.21.Province of British Columbia Ministry of Health Division of Vital Statistics HealthSurveillance Registry. Annual report. H.S.R. 1981;622.International Classification of Diseases, 9th rev. Geneva: World Health Organiza-tion, 1978.23.Baird P. British Columbia health surveillance system. Chronic Diseases in Canada1988;9:66-7.24.Dougherty G. Socioeconomic differences in pediatric mortality in urban Canada:1981 [Thesis]. McGill University, 1986.25.Lerner M, Stutz R, Lapan S. Differential mortaltiy in Maryland by county and inBaltimore by socioeconomic status. Md State Med J 1978;27:35-42.26.Wigle D, Mao Y. Mortality by income level in urban Canada. Minister of NationalHealth and Welfare, Ottawa 1980;27.Egbuonu L, Starfield B. Child health and social status. Pediatrics 1982;69:550-7.Page 9028.Kovar M. Health status of U.S. children and use of medical care. Public Health Rep1982;97:3-15.29.Mare R. Socioeconomic effects on child mortality in the united states. Am J PublicHealth 1982;72(6):539-47.30.Starfield B. Family income, ill health and medical care in the U.S.. J Public HealthPolicy 1982;3:244-259.31.Starfield B. Child health and socioeconomic status. Am J Public Healh 1982;32.Collins J, Richard J. The differential effect of traditional risk factors on infantbirthweight among blacks and whites in Chicago. Am J Public Health1990;80(6):679-81.33.Statistics Canada. Census dictionary. Catalogue no. 99-101E. Ottawa, Ontario:Minister of Supply and Services Canada, 1987:34.BC Ministry of Health and Ministry Responsible for Seniors DoVS. Selected vitalstatistics One hundred twentieth Annual report 1991. . Provinve of BritishColumbia, 1993: 154.35.Arbuckle T, Sherman G. An analysis of birth weight by gestational age in Canada.Can Med Assoc J 1989;140:157-165.36.PSearch. Vancouver: Tetrad Computer Applications Limited, 1990.37.Schlesselman JJ. Case-control studies. . New York: Oxford University Press, 1982:354.38.Kahn H, Sempos C. Statistical methods in epidemiology. New York: Oxford Univer-sity Press, 1989: 292.39.Selvin S. Statistical analysis of epidemiologocal data. New York: Oxford UniversityPress, 1991: 375.40.Kleinbaum D, AKupper L, Morgenstein H. Epidemiologic research Principles andquantitative methods. New York: Van Nostrand and Reinhold Co., 1982:41.SPSSx. Chicago, Ill.: SPSS Inc., 1983:42.Statistical analysis system. . Cary, N.C.: SAS Institute, 1985:Page 91CHAPTER FIVESTUDY RESULTSEVALUATION OF DATA BASE DEVELOPMENT METHODSThere were 170,602 births in the province to residents of B.C. from January 1, 1985through December 31, 1988. Only 4 (<0.001%) of the first single live birth or stillbirthrecords were missing maternal birth date information. As shown in Table 5.1, the 68,795stillbirths and first single live births to women under 35 years were 40.32% of the birthsto B.C.residents. Only 2,122 of these required manual review. They were either missingthe mother's postal code, had an incorrect postal code or had a postal code associatedwith a H, J, K, M,or a blank delivery mode type (DMT) code which precluded theautomated matching of the postal code to the census tract of the mother's residence. TheDivision of Vital Statistics of the B.C. Ministry of Health (DVS) provided maternaladdress information from the PNOB records for these cases and the Population Section,Planning and Statistics Division (PSD) of the B.C. Ministry of Finance and CorporateRelations located the corresponding census tracts and assigned identifiers. Only 189cases were either not in one of the Census Metropolitan Areas (CMAs) or CensusAgglomerations (CAs) or were impossible to locate and 133 were known to be in thestudy area but could not be manually located in a census tract. The remaining 1,800 weresuccessfully assigned.Page 92Table 5.1Data Base DevelopmentNumber PercentTotal B.C. births (1985 - 1988) 170,602 100.00Stillbirths & first single live births to women 534 years 68,795 40.32— with valid postal codes –66,673 –96.9— requiring manual postal code look-up –2,122 –3.1Stillbirths & first single live births requiring manual look-up 2,122 100.00Impossible to locate / not in CMA/CAs 189 8.91In study area 1,933 91.09— manually assigned –1800 –93.1— not assigned –133 –6.9Development of Neighborhood Poverty QuintilesOf the 170,602 births, 67,551 were first, single, live births. Of these, 23,785 births wereexcluded from the study because they were impossible to locate (1.3%) or they felloutside the five CMAs/CAs (98.7%). As shown in Table 5.2, of the 133 records withoutcensus tract information 126 were live birth records. Of the remaining 43,640 cases,1,362 (3.1%) did not have poverty information because they were in census tracts witha non reserve private household population of fewer than 250 persons aged less than 18years. The distribution of the final sample of 42,278 cases and the 126 cases withoutcensus tract information among the various CMAs and CAs is presented in Table 5.3.Of particular note in Table 5.3 is the large number of cases in Prince George which couldnot be located in a census tract, possibly because of a very large urban rural interface.As a consequence 6.35% of the births in this city were excluded from the study.Although this was substantially higher than the rate in the other cities it was decidednot to automatically exclude Prince George from the remainder of the study but toPage 93measure the impact of such a large rate of missing information by analyzing the datawith and without this CA.Table 5.2Development of Neighbourhood Poverty QuintilesNumber PercentFirst single live births to women S34 years 67,551 100.00Births outside the CMA/CAs 23,785 35.21Births without census tract information 126 0.19First single live births to women ..C34 yearsLocated in the CMA/CAs 43,640 100.00Cases not in NPQs 1,362 3.12Cases in NPQs 42,278 96.88The distribution of the cases between the five neighborhood poverty quintiles (NPQs)is also presented in Table 5.3. Twenty percent of cases were not always allocated to eachNPQ in each of the CMAs and CAs. This occurred because the group of births in a censustract could not be split in the process of allocating the tracts to NPQs and there were upto 300 births in a census tract during the years of the study. However when all the samenumbered NPQs from the various CMA/CAs were amalgamated between 19.7% and20.4% of all cases fell into each NPQ.To examine income segregation between and income homogeneity within NPQs themean, standard deviation and coefficient of variation were calculated for the censusfamily income in each NPQ separately for each of the five areas. These are presentedin Table 5.4. In all CMAs and CAs the average census family income increased steadilyfrom NPQ 5 to NPQ 1. In the CMAs, Vancouver (933) and Victoria (935), the meanincome in the first NPQ was 1.9 to 2.1 times as great as in NPQ 5, while in Kamloops(925), Kelowna (915), and Prince George (970) the differences were all approximatelyTable 5.3Distribution of Data Between the Five CMA/CAsCMA/CAValueCMA/CAName# Cases MissingCensus Tracts% of CasesMissing SensusTracts# Cases inQuintile 1# Cases inQuintile 2# Cases inQuintile 3# Cases inQuintile 4# Cases inQuintile 5# Cases inAll Quintiles915 Kelowna 1 0.06 298 385 294 329 324 1,630925 Kamloops 16 1.01 300 359 338 283 297 1,577933 Vancouver 20 0.06 6,516 6,533 6,518 6,450 6,574 32,591935 Victoria 9 0.17 1,016 1,046 1,083 990 1,085 5,220970 Prince George 80 6.35 217 282 226 302 233 1,260Total 126 0.30 8,347 8,605 8,459 8,354 8,513 42,278Percent 19.74 20.35 20.00 19.76 20.14 100.00Page 951.7. The standard deviation of the mean incomes increased as the mean incomes in eachCMA and CA increased. The increase in the coefficients of variation between NPQs 1and 5 in all five areas demonstrated that mean incomes increased more rapidly than didthe standard deviation of these incomes. Figure 5.1 shows this gradient to be strongestin Kamloops. In Vancouver, Victoria and Prince George the increases in the coefficientof variation were somewhat less pronounced, while in Kelowna the increase was evenless marked and the coefficient for NPQ 1 was higher than that for the second NPQ.These data suggested a relatively weak economic gradient in Kelowna and conse-quently a muting of any differential in birth outcomes that are influenced by SES. Aswith Prince George, it was decided to analyze some of the data with and without thedata from Kelowna.80 —70 -0>60 -ti)13o 50 -40 ^0Figure 5.1: Income Variation WithinQuintiles for the Five CMA/CAsJ1^2^3^4Quintile5^6KelownaKamloopsVancouverVictoriaPrince GeorgePage 96CMA/CA QuintileTable 5.4Income Separation in the Five CMAs/CAsMean Income^Standard Deviationof IncomeCoefficient ofVariation*915 1 35,725 22,300 62915 2 34,725 21,160 61915 3 28,233 18,480 65915 4 25,158 17,763 71915 5 21,381 16,006 75925 1 40,914 19,128 47925 2 35,853 19,172 53925 3 35,243 22,251 63925 4 32,846 22,162 67925 5 23,846 19,211 81933 1 48,172 25,028 52933 2 41,856 24,630 59933 3 36,604 23,411 64933 4 31,327 21,922 70933 5 23,329 18,596 80935 1 41,215 22,751 55935 2 39,259 23,306 59935 3 31,390 20,041 64935 4 28,425 19,917 70935 5 21,918 17,037 78970 1 42,653 20,571 48970 2 38,946 22,035 57970 3 35,804 21,437 60970 4 33,089 22,348 68970 5 25,322 17,658 70• (Standard deviation / mean) * 100Page 97Analysis of the Effect of Various Groupings of the CMA/CAs on a Selection of BirthOutcomes In order to consider the effect excluding Kelowna and Prince George from the study anumber of analyses were undertaken first including the data from all five centres andthen with two sub-groupings. The data from Kamloops, Vancouver and Victoria wereanalyzed together as well as that from Kelowna and Prince George. The analysesundertaken were chi-square tests comparing births to teens with those to older mothersin the five NPQs as well as ANOVAs comparing mean birth weights and gestationalages by MA groups and NPQs. The results from each of the various analyses arepresented in Appendix C. In all analyses there were statistically significant differencesbetween the distribution of births to teens and older mothers when all the CMAs andCAs were considered together. Excluding the data from Kelowna and Prince Georgedampened these differences in two analyses and enhanced it in three. However thesechanges were not large and did not alter the conclusions drawn regarding differencesbetween teens and older mothers in any of the five analyses. When the data fromKelowna and Prince George were analyzed separately in four of the five analyses thedifferences between the two groups of mothers were not significant. In interpretingthese results it should be noted that only 2,890 (6.8%) of the cases included in the studywere located in Kelowna and Prince George. Insignificant results could have been asmuch a result of this substantially smaller sample size as the lack of an economicgradient or any other reason inherent in these two cities or any cities of this size. Theinclusion or exclusion of these two cities from the study could not be justified on thebasis of these analyses. It was decided to retain all five cities in the study in order toincrease the generalizability of the study results.Page 98Infant Death LinkageThere were 1,355 infant deaths in B.C. during the study period and all but 15 of them hadbirth registration numbers that could be used to link them with their PNOB records. Ofthe remaining 1,340 deaths, 274 linked to the study population of first single live birthsto mothers aged less than 35 years living in one of the five CMA/CAs (Table 5.5). The274 deaths that linked were 20% of all infant deaths in the province between 1985 and1988 while the study population was 25% of all the births during that time period. Thissuggests that either there was some out-migration from the five urban areas that had theeffect of reducing the number of infant deaths or that the infant death rate in the CMA/CAs was lower than that in the rest of the province. The latter possibility is supportedby the observed intraprovincial variations in infant mortality in B.C. 1 .Table 5.5Infant Death LinkageNumber PercentB.C. Infant Deaths (1985-1988) 1355 100.00no birth registration number 15 1.11birth registration information 1340 98.8Infant Deaths with linking information 1340 100.00infant deaths linked to study population 274 20.4Congenital Anomalies Linkage The HSR file provided to the study contained 11,687 registrations made within the firstyear of life for infants born between Jan. 1, 1985 and Dec. 31, 1988. The number ofregistrations per year and as a percentage of live births in that year is presented in TablePage 995.6. The decrease in registrations as a percent of live births from 1985 through 1988 raisesquestions about the completeness of the registrations in the latter years of the study.The file contained 932 cases without birth registration numbers which precluded theirlinkage to the PNOB file. Of these 465 were lost to the study for a variety of reasons(Table 5.7). The final data set of 11,222 registrations included 3,126 cases which werelinked to the study population of first single live births to women less than 35 years.Thus 27% of the registrations were linked to the study population which itself was 25%of all live births in the province during this time period.Table 5.6Registrations to the B.C. Health Status Registry Made Within the First Year of LifeYear Number of Births HSR Registrations Registrations as a % ofLive Births1985 43,281 3,373 7.751986 42,167 3,113 7.361987 41,949 2,906 6.911988 43,205 2,295 5.30Total 11,687Table 5.7Congenital Anomalies LinkageNumber PercentHSR Registrations received (1985 - 1988) 11,687 100.0Non-B.C. births 100 0.9Inappropriate registrations 15 0.1Birth registration numbers not located 122 1.0Cases merged 183 1.6Duplicate cases 45 0.4HSR Data Set 11,222 96.0Page 100Allocating Stillbirths to Neighborhood Poverty QuintilesThere were 1,244 still births registered in B.C. between January 1, 1985 and December31, 1988. The most recent definition of stillbirth, "The complete expulsion or extractionfrom its mother, after at least twenty weeks' pregnancy, or after attaining a weight ofat least 500 grams..." was applied in the study. Only 926 (79.15%) of the registered caseshad a mass of 500 grams or greater and were at least 20 weeks gestation. The distributionof the 792 stillbirths to women less than 35 years is presented in Table 5.8. There wereonly 432 stillbirths included in the final data set.Table 5.8Allocation of Stillbirths into NPQsNumber PercentBC Stillbirths (1985-1988) 1244 100.00missing gestational age and/or weight information 76 6.10<20 weeks gestation or <500 g 242 19.4520 weeks gestation and .500 g 926 74.44Stillbirths (20 weeks and ?_500 g) 926 100.00Stillbirths to women 534 years 792 85.53Stillbirths to women 534 years 792 100.00No census tract information 7 .88Outside 5 CMA/CAs 338 42.68Within 5 CMA/CAs 447 56.44- cases not in NPQs* -15 -3.36- cases in NPQs -432 -96.6Missing Data In the final study population of 42,278 live births and 432 stillbirths there were very fewmissing data for most birth outcome variables. Only for 14 live birth records was birthPage 101weight information missing. Twenty-three cases were missing gender, gestational ageor birthweight; the variables required for small for gestational age calculations. Noneof the infant death records had missing or obviously incorrect age of death informationsuch that the timing of the death could not be calculated, although 15 had no birthregistration number and so could not be linked to their PNOB record. Only 4 stillbirthswere lost to the study because there had been no live birth in the census tract associatedwith the stillbirth.EVALUATION OF VARIABLE DEFINITIONSOf the 42,278 live births in the study population 2,738 (6.6%) were born to mothers lessthan 20 years at the time of the birth of their infant and 39,540 (93.5%) were to mothersbetween 20 and 34 years at the time of the birth. The mean age for the teen mothers was17.9 years with a standard deviation of 1.1 years. For the older mothers the mean andstandard deviation were 26.8 and 3.7 years.The distribution of the births between the five NPQs, which was presented in Table 5.3,was very close to 20% in all quintiles. The range and mean percentage of children lessthan 18 years of age in a census tract whose economic family income was below theStatistics Canada low income cutoff is presented by NPQs in Table 5.9. The means andranges followed the same pattern as the mean incomes by NPQ and reflected theincrease in the amount of neighborhood poverty from NPQ 1 through 5.Page 102Table 5.9Distribution of the Percentage of Children Living Below the Low Income Cut-offsNPQ1 NPQ2 NPQ3 NPQ4 NPQ5Mean 9.9 15.1 20.0 28.0 41.4Range - lower limit 0.0 9.8 14.0 18.7 31.6- upper limit 15.0 22.7 27.0 35.3 78.1The rates of the various birth outcomes for the entire study population are presented inTable 5.10. Also presented in the table are rates for a number of these outcomes for allof B.C. for each year from 1985 through 1988a. Except for stillbirths and infant mortality,the rates of the various birth outcomes in the study population were reasonably similarto those for the entire B.C. population during this time. A large part of the explanationfor the elevated rates in the study population is that the denominator in the study wasrestricted to stillbirths and first single live births while all live and stillbirths wereincluded in the provincial calculations. The latter denominators included at least twiceas many births as did the study population (for example see Table 5.1) which wouldresult in rates that were half that found in the study. There may also be differences inthe rates of diagnoses and reporting of stillbirths between the urban and rural areas ofthe province. Infant mortality rates were probably affected by differences in access tospecialized medical services between urban and rural B.C. Additionally, the fact thatthe study population was limited to first single birth reduced infant mortality rates inthe study population.a. K. Peet, S. Wiggins and S.B. Sheps, unpublished data, 1992Page 103Table 5.10Rates of Birth Outcomes (per 1000 live births)Study Population1985 - 1988^1985British Columbia1986^1987 1988Stillbirths 10.2 5.8 5.2 5.0 4.4Low birth weight 48.3 48.9 47.2 51.1 50.3Very low birth weight 8.5 9.1 8.8 9.2 8.5Small for gestational age 96.7Infant mortality 6.5 7.8 8.0 8.2 8.2Neonatal mortality 4.2 4.5 4.7 4.5 5.0Postneonatal mortality 2.3 3.2 3.2 3.5 3.1Congenital anomalies 73.9ANALYSIS OF BIRTH AND BIRTH OUTCOME DATAIncidence of First Single Live BirthsThere were 2,738 first, single, live births to teen mothers and 39,540 similar births tomothers aged 20-43 years. The distribution of these births by maternal age groups andthe five NPQs is presented in Table 5.11. Births to the older mothers were fairly evenlyspread between the NPQs, with 20 to 21% found in each NPQ. Births to mothers lessthan 20 years were not as evenly distributed. The percent of all teen births in NPQs 1through 4 increased from 16% to 19% and then jumped to 27% in NPQ 5.Page 104Table 5.11Incidence of First Single Live Birthsby Maternal Age Groups and NPQsMaternal Age NPQ1 NPQ2 NPQ3 NPQ4 NPQ5 Total# (row%) # (row%) # (row%) # (row%) # (row%)<20 years 445 (16) 493 (18) 523 (19) 525 (19) 752 (27) 273820-34 years 7902 (20) 8112 (21) 7936 (20) 7829 (20) 7761 (19) 39540A chi-square value of 105.81 with 4 degrees of freedom (p<0.001) showed that there wasa significant association between births to the maternal age groups and NPQs. Thepercentage of all births to women under 35 years of age that occurred to mothers lessthan 20 years increased as the percentage of children living in poverty in the mother'sneighborhood at the time of the birth increased.b) Analyses of Birth Outcomes I Stillbirths The overall stillbirth ratio for infants born to mothers aged less than 20 years was 0.73per 100 live births while the ratio for mothers 20-34 years was 1.03. In Table 5.12 it isshown that the stillbirth ratio for the offspring of younger mothers was also lower thanthat of the comparison group in all NPQs except the first. This difference in the overallratios for the two maternal age groups was not statistically significant (x 2=2.41, d.f.=1,P=0.120) nor were any of the differences in ratios for the two groups of infants in anyof the five NPQs.Page 105NPQTable 5.12Stillbirths by Maternal Age Groups Stratified by NPQsMaternal Age^Stillbirths (row %)^Live Births (row %) Total1 < 20 5 (1.12) 445 (98.89) 45020 - 34 74 (0.94) 7902 (99.06) 8347X' = 0.154 d.f.=1 p=0.6952 < 20 4 (0.81) 493 (99.19) 49720 - 34 89 (1.09) 8812 (98.81) 8201X2 = 0.38 d.f.=1 p=0.5563 < 20 5 (0.95) 523 (99.05) 52820 - 34 96 (1.21) 7936 (98.79) 8032X2 = 0.262 d.f.=1 p=0.6094 < 20 2 (0.38) 525 (99.62) 52720 - 34 83 (1.05) 7829 (98.95) 7912X2 = 2.22 d.f.=1 p=0.1365 < 20 4 (0.55) 725 (99.45) 72920 - 34 70 (0.89) 7761 (99.11) 7831X' = 0.927 d.f.=1 p=0.3361 - 5 < 20 20 (0.73) 2738 (99.27) 275820 - 34 412 (1.01) 39540 (99.89) 39952X2 = 2.41 d.f.=1 p=0.120When considered by NPQs the stillbirth ratios for all births in the study ranged from 0.86per 100 live births in NPQ 5 to 1.18 in NPQ 3 (Table 5.13) and did not exhibit any patternthrough the NPQs. An insignificant chi-square of 5.10 with 4 degrees of freedom(P=0.277) confirmed the lack of an association between overall stillbirth ratios andNPQs. When stillbirth ratios for adolescent mothers and older mothers were analyzedseparately, for neither group was there a significant association with NPQs (Table 5.13).There appeared to be no association between maternal NPQs and stillbirth rates for thestudy population.These analyses suggested that stillbirth rates were not associated with either maternalage groups or NPQs.Page 106Table 5.13Stillbirths by NPQs Stratified by Maternal Age GroupsMaternal Age NPQ12^< 20 years^345x'=2.611220 - 34 years 345X2 = 4.661234 years^345x2 =5.10Stillbirths (row %)5 (1.12)4 (0.81)5 (0.95)2 (0.38)4 (0.55)d.f.=4 p=0.62474 (0.93)89 (1.09)96 (1.21)83 (1.05)70 (0.89)d.f.=4 p=0.32479 (0.94)93 (1.07)101 (1.18)85 (1.01)74 (0.86)d.f.=4 p=0.277Live Births (row %)445 (98.89)493 (99.19)523 (99.05)525 (99.62)752 (99.45)7902 (99.07)8112 (98.01)7936 (98.19)7829 (98.95)7761 (99.11)8347 (99.06)8605 (98.93)8459 (98.82)8354 (98.99)8315 (99.14)Total4504975285277567976820180327912783184268698856084398587II Birth WeightInfants born to teen mothers had a mean birth weight of 3,340 grams while the mean was3,382 grams for the offspring of mothers aged 20-34 years. From Table 5.14 it can benoted that, except in NPQ 3 mean birth weights for infants born to the younger motherswere slightly lower while their standard deviations were lower in the first two NPQsand higher in the last two. A two way analysis of variance (ANOVA) produced F-valuesof 10.48 (d.f.=1, P<0.001) for birth weight by maternal age group and 1.95 (d.f.=4,P=0.099) for interaction between maternal age groups and NPQs. This suggested thatinfants born to teenage mothers had significantly lower birth weights and that this wasPage 107not affected by an interaction between NPQs and maternal age groups. Althoughstatistically significant, the difference in mean birth weights between the two maternalgroups was only 42 grams.NPQTable 5.14Birth Weights by Maternal Age Groups and NPQsMaternal Age (Years)<20^20 - 34^<_ 341 Mean 3406 3424 3423St.D. 542 549 5492 Mean 3323 3405 3400St.D. 545 550 5493 Mean 3399 3388 3388St.D. 5566 5566 5554 Mean 3300 3354 3351St.D. 604 5543 5475 Mean 3299 3377 3333St.D. 588 576 577Infants included in the study had a mean birth weight of 3,379 grams. Overall meanbirth weights in the NPQs decreased from 3,423 in the first to 3,333 grams in the fifth(Table 5.14). A two way ANOVA produced an F-value of 11.67 (d.f.=4, P<0.001) for birthweight by NPQ group. This significant result was reproduced when both maternal agegroups were analyzed separately. The F-value for the mean birth weights of infantsborn to teenage mothers by NPQ was 4.69 (d.f.=4, P<0.001) and for offspring born to theolder mothers it was 32.70 (d.f.=4, P<0.001). As was previously mentioned, there wasno significant interaction between NPQs and maternal age groups. This suggested thatfor infants from both maternal age groups, as the percentage of maternal neighborhoodpoverty increased the mean birth weights decreased. Again it should be noted thatalthough statistically significant, the difference in mean birth weights between NPQs 1and 5 was only 90 grams.Page 108a) Low Birth WeightThere were differences in the low birth weight (LBW) rates for infants born to the twogroup of mothers. For mothers aged less than 20 years the overall rate was 6.21 per 100live births compared with 4.74 for mothers aged 20-34 years (Table 5.15). This differencewas statistically significant with a chi-square value of 12.09 (d.f.=1, P<0.001). Theassociated odds ratio was 1.33 with 95% confidence limits at 1.13 and 1.57. From Table5.15 it is noted that LBW rates for infants born to the younger mothers were higher inall NPQs except the first. These differences in the rates were only significant in NPQs4 and 5 with chi-squares of 11.63 (d.f.=1, P<0.001) and 4.06 (d.f.=1, P=0.044).NPQTable 5.15Low Birth Weight Births by Maternal Age Groups Stratified by NPQsMaternal Age^< 2500g (row %)^2500g + (row %)^Total1 < 20 17 (3.82) 428 (96.18)^44520 - 34 326 (4.13) 7573 (95.87) 7899X2 = 0.101 d.f.=1^p=0.751 O.R.=0.922 < 20 26 (5.27) 467 (94.73)^49320 - 34 359 (4.43) 7751 (95.51) 8110X2 = 0.780 d.f.=1^p=0.377 O.R.=1.203 < 20 28 (5.34) 495 (94.66)^52320 - 34 369 (4.65) 7566 (95.35) 7935X2 = 0.543 d.f.=1^p=0.461 O.R.=1.60< 20 44 (8.38) 481 (91.62)^52520 - 34 389 (4.97) 7436 (95.13) 7825X2 = 11.633 d.f.=1^p<0.001 O.R.=1.75< 20 55 (7.31) 697 (92.69)^75220 - 34 429 (5.53) 7328 (94.47) 7757X2 = 4.064 d.f.=1^p=0.044 O.R.=1.351- 5 < 20 170 (6.21) 2568 (93.79)^273820 - 34 1872 (4.74) 37654 (95.26) 39526X2 = 12.08 d.f.=1^p<0.001 O.R.=1.33^95% C.I.=1.13, 1.57X CMH= 10.32 d.f.=1^p=0.001 O.R.cmH=1.30^95% C.I.=1.12, 1.53X2 BD= 5.57 d.f.=4^p=0.234Page 109Differences persisted when LBW rates were calculated and compared by NPQs. For allbirths combined the rates increased steadily from 4.11 per 100 live births in NPQ 1 to 5.69per 100 in NPQ 5 (Table 5.16). These differences resulted in a significant chi-square of28.01 with 4 degrees of freedom (P<0.001). When only the values in NPQs 5 and 1 werecompared a chi-square of 22.46 with 1 degree of freedom was produced (P<0.001) andthe associated odds ratio was 1.41 with 95% confidence limits at 1.22 and 1.62. Similarresults were obtained when births to the two maternal age groups were evaluatedseparately. Adolescent mothers experienced an increase in rates from 3.82 per 100 livebirths in NPQ 1 to 7.31 in the last NPQ. Rates for older mothers went from 4.13 to 5.53per 100 births. Both of these increases were significant (Table 5.16) although with thelarge number of chi-square tests performed the P-value of 0.021 for teen mothers couldsuggest that this result was not really significant.Page 110Table 5.16Low Weight Births by NPQs Stratified by Maternal Age GroupsTotal445493523525752Maternal Age^NPQ12< 20 years^345X2 = 11.591220 - 34 years^345X2 = 20.15< 2500 g (row %)17 (3.82)26 (5.27)28 (5.34)44 (8.38)55 (7.31)d.f.=4 p=0.021326 (4.13)359 (4.13)369 (4.65)389 (4.97)429 (5.53)d.f.=4 p<0.0012500 g + (row %)428 (96.18)467 (94.83)495 (94.66)481 (91.62)697 (92.69)O.R.=1.897573 (95.87)7751 (95.87)7566 (95.55)7436 (95.03)7328 (94.47)O.R.=1.30789981107935782577578001 (96.89)8218 (95.52)8061 (95.31)7917 (94.81)8025 (94.31)834486038458835085091^343 (4.11)2 385 (4.48)34 years^3^397 (4.67)4 433 (5.19)5^484 (5.69)X2 = 28.01 d.f.=4 p<0.001X2cmH = 25.95 d.f.=1 p<0.00134 years^5^484 (5.69)1 343 (4.11)X2 = 22.46 d.f.=1 p<0.001X2cmH = 20.15 d.f.=4 p<0.001X2BD = 1.68 d.f.=1 p=0.195^8025 (94.31)^85098001 (96.89) 8344O.R.=1.41 95% C.I.=1.22, 1.62O.R.=1.40 95% C.I.=1.21, 1.61Because both maternal age groups and NPQs were significantly associated with LBWthere was potential for confounding between these two independent variables. Thedata were reanalyzed with CMH chi-square tests. When maternal age group data werereanalyzed with a CMH chi-square test controlling for the five NPQ groups theassociation between maternal age groups and LBW remained significant (x 2cmH=10.32,d.f.=1, P<0.001) although not strongly so given that the associated odds ratio was 1.322with 95% confidence limits at 1.12 and 1.53.Page 111The odds ratios for the individual NPQs were fairly similar to each other and the overallcrude odds ratio, although they ranged from 0.4 below and above this odds ratio (Table5.15). This suggested that there might be some interaction between maternal age groupsand NPQs. This was not supported by the Breslow-Day (BD) chi-square value, whichwas such that the null hypothesis (that the true odds ratio is the same for all strata) couldnot be rejected (X2BD=5.57, d.f.=4, P=0.234) (Table 5.15). The pattern of the strata specificodds ratios and the similarity between the overall crude and CMH odds ratios alsoprecluded the existence of confounding.The result remained significant when the association between LBW rates and NPQmembership was re-examined while controlling for maternal age group membership(Table 5.16). The CMH chi-square value for NPQs 1 through 5 was 25.99 with 1 degreeof freedom (P<0.001). When only the rates in the last and first NPQs were examinedusing the CMH chi-square test the value was 21.41 (d.f.=1, P<0.001). The associatedodds ratio was 1.40 with 95% confidence limits at 1.21 and 1.61. There did not appearto be substantial confounding. When comparing LBW rates in NPQs 5 and 1, the oddsratios for the two maternal age groups were 1.99 and 1.36 compared to the overall rateof 1.40. As well the crude and CMH odds ratios were very similar. Although thevariation in the strata specific odds ratios suggested some interaction the non-signifi-cant BD chi-square (x2=1.68 d.f.=4, P=0.195) did not allow it to be seen as statisticallysignificant.Even when subjected to CMH chi-square analyses controlling for the stratifyingvariable, LBW rates remained significantly associated with both NPQs and maternalage groups. LBW rates increased as either maternal age decreased or the percentage ofmaternal neighborhood poverty increased.Page 112b) Very Low Birth WeightThe rates of very low birth weight (VLBW) were higher for infants born to the youngermothers in all NPQs except for the first (Table 5.17). The rates for all NPQs combinedwere 1.10 per 100 births to teens and 0.83 for births to the older mothers. Neither theoverall rate for teen mothers nor any of the rates by NPQs were significantly differentfrom those of the comparison group (Table 5.17). The chi-square value for the differencein the overall rate was 2.11 (d.f.=1, P=0.147) and the related odds ratio was 1.32 (95%C.L.=0.907, 1.922).Rates of VLBW increased from NPQs 1 through 5. The ranges were from 0.74 to 1.10per 100 live births for all mothers, from 0.75 to 1.08 for mothers aged 20-34 years andfrom 0.67 to 1.33 per 100 births for teenagers (Table 5.18). These differences onlyproduced a significant chi-square value for all births combined (x2=10.06, d.f.=4.P=0.039) (Table 5.18). When only the values in NPQs 5 and 1 for all births werecompared the result was a chi-square of 6.01 with 1 degree of freedom (P=0.014) and theassociated odds ratio was 1.49 with 95% confidence limits at 1.08 and 2.06. The P-valueand confidence limits suggested that the significance was very marginal.Page 113Table 5.17Very Low Birth Weight Births by Maternal Age Groups Stratified by NPQsNPQ Maternal Age < 1500 g (row %) 1500 g + (row %) Total1 < 20 3 (0.67) 442 (99.33) 44520 - 34 59 (0.75) 7840 (99.25) 7899X2 = 0.030 d.f.=1^p=0.862 O.R.=0.902 < 20 4 (0.81) 489 (99.19) 49320 - 34 57 (0.70) 8053 (99.29) 8110X2 = 0.078 d.f.=1^p=0.780 O.R.=1.163 < 20 5 (0.96) 518 (99.04) 52320 - 34 63 (0.79) 7872 (99.21) 7935X2 = 0.162 d.f.=1^p=0.688 O.R.=1.214 < 20 8 (1.52) 517 (98.48) 52520 - 34 66 (0.84) 7759 (99.16) 7825X2 = 2.593 d.f.=1^p=0.107 O.R.=1.825 < 20 10 (1.33) 742 (98.67) 75220 - 34 84 (1.08) 7676 (98.92) 7757X2 = 0.383 d.f.=1^p=0.536 O.R.=1.231 - 5 < 20 30 (1.10) 2708 (98.90) 273820 - 34 329 (0.83) 39197 (99.17) 39526X' = 2.11 d.f.=1^p=0.147 O.R.=1.32^95% C.I.=0.91, 1.92X2 cmH= 1.65 d.f.=1^p=0.153 O.R.=1.28^95% C.I.=0.88, 1.86X2 BD= 1.30 d.f.=4^p=0.862Page 114Table 5.18Very Low Weight Births by NPQs Stratified by Maternal Age GroupsMaternal Age^NPQ^< 1500 g (row %)^1500 g + (row %)^Total1^3 (0.67)^442 (99.33)^4452 4 (0.81) 489 (99.29) 493< 20 years^3^5 (0.96) 518 (99.04)^5234 8 (152)^517 (98.48) 5255^10 (1.32) 742 (98.68) 752X2 = 2.46^d.f.=4 p=0.6521^59 (0.75)^7840 (99.25)^78992 57 (0.70) 8053 (99.30) 811020 - 34 years^3^63 (0.79) 7872 (99.21)^79354 66 (0.84)^7759 (99.16) 78255^84 (1.08) 7673 (98.68)^7757X2 = 8.40^d.f.=4 p=0.0781^62 (0.74)^8282 (99.26)^83442 61 (0.71) 8542 (99.29) 8603_< 34 years^3^68 (0.80) 8390 (99.20)^84584 74 (0.89)^8276 (99.11) 83505^94 (1.10) 8415 (98.90)^8509x2 = 10.06 d.f.=4 p=0.0395_ 34 years^5^94 (1.10)^8415 (98.90)^85091 62 (0.74) 8282 (99.26) 8344X2 6.01^d.f.=1 p=0.014^O.R.=1.49 95% C.I.=1.08, 2.06It is possible that the NPQs were acting as a confounder in the analysis of the effect ofmaternal age groups on VLBW. For this reason the association between maternal agegroups and VLBW was reanalyzed while controlling NPQ membership. The resultingCMH chi-square was insignificant with a value of 1.65 (d.f.=1, P=0.153) and anassociated odds ratio of 1.28 (95% C.L.=0.878, 1.862). The crude odds ratio and the CMHodds ratio were fairly similar and the odds ratios for the individual NPQs only variedbetween 0.4 below and 0.5 above the crude odds ratio. This pattern precluded theexistance of confounding. Although the variation in the strata specific odds ratiossuggested that there might be some interaction this was not supported by the Breslow-Day (BD) chi-square value , which was such that the null hypothesis (that the true oddsPage 115ratio is the same for all strata) could not be rejected (x2BD=1.30, d.f.=4, P=0.862).These results suggested that, although not always significant, there was a trend towardsincreasing VLBW rates with increasing maternal neighborhood poverty levels. Therewas, however, no significant association between VLBW and maternal age groups.III Small for Gestational Age BirthsInfants born to teen mothers had a mean gestational age of 39.3 weeks compared with39.5 weeks for those born to the older mothers. In Table 5.19 it is noted that, except inthe first NPQ, mean gestational ages for infants born to the younger mothers wereslightly lower than those for older mothers and that in all NPQs the standard deviationswere the same as or slightly greater than those of the older group. A two way ANOVAproduced F-values of 5.22 (d.f.=1, P=0.010) for gestational age by maternal age groupand 1.40 (d.f.=4, P=0.233) for interaction between maternal age groups and NPQs. Thissuggested that there was a significant decrease in gestational age for infants born toteens but there was no significant interaction between NPQs and maternal age groups.But as was the case with the analysis of birth weight, it is noted that, althoughstatistically significant, the difference in mean gestational ages between the two groupsof infants was only 0.2 weeks.There was also only a small difference in the small for gestational age (SGA) rates forboth groups of infants either overall or within the various NPQs (Table 5.20). Theoverall rates were 10.08 per 100 births for all mothers aged less than 20 years and 9.59for the older mothers. This difference was not statistically significant (x 2=0.70, d.f.=1,P=0.403), nor were any of the differences in SGA rates between maternal age groupswithin any of the NPQs (Table 5.20).Page 116NPQTable 5.19Gestational Ages by Maternal Age Groups and NPQsMaternal Age (Years)<20^20 - 34 341 Mean 39.6 39.5 39.5St.D. 2.0 2.0 2.02 Mean 39.4 39.5 39.5St.D. 2.1 2.0 2.03 Mean 39.4 39.5 39.5St.D. 2.0 2.0 2.04 Mean 39.2 39.4 39.4St.D. 2.5 2.0 2.05 Mean 39.2 39.4 39.3St.D. 2.3 2.1 2.2Table 5.20Small for gestational Age Births by Maternal Age Groups Stratified by NPQsNPQ^Maternal Age^Yes (row %)^No (row %)^Total1 < 20^35 (7.87) 410 (92.13) 44520 - 34 640 (8.10) 7260 (91.90)^7900X' = 0.03^d.f.=1 p=0.856^O.R.=0.972^< 20 55 (11.16)^438 (88.84)^49320 - 34^704 (8.68) 7406 (91.32) 8110X' = 3.54 d.f.=1 p=0.060^O.R.=1.303^< 20^45 (8.60)^438 (91.40)^52320 - 34 770 (9.70) 7406 (90.30) 7935X' = 0.68^d.f.=1 p=0.409^O.R.=0.88< 20 53 (10.10)^472 (89.90)^52520 - 34^815 (10.41) 7011 (849.59) 7826)e = 0.23 d.f.=1 p=0.631^O.R.=0.97< 20^88 (11.70)^664 (88.30)^75220 - 34 863 (11.12) 6895 (88.88) 7758X' = 0.38^d.f.=1 p=0.536^O.R.=1.231 - 5^< 20 276 (10.8)^2462 (89.92)^273820 - 34^3792 (9.59) 35737 (90.41) 39529X' = 0.70 d.f.=1 p=0.403^O.R.=1.06 95% C.I.=0.93, 1.20x2 cmH= 0.25^d.f.=1 p=0.617^O.R.=1.03 95% C.I.=0.91, 1.18x2 BD= 4.19 d.f.=4 p=0.381Page 117Infants included in the study had a mean gestational age of 39.48 weeks. When all birthsincluded in the study were combined the mean gestational ages in the NPQs decreasedfrom 39.5 weeks in the first to 39.3 weeks in the fifth (Table 5.19). A two way ANOVAproduced an F-value of 6.61 (d.f.=4, P<0.001) for gestational age by NPQ group. Thissignificant result was reproduced when both maternal age groups were analyzedseparately. The F-value for the gestational age of infants born to teenage mothers byNPQ was 3.13 (d.f.=4, P=0.014) and for offspring born to the older mothers it was 8.11(d.f.=4, P<0.001). As was previously mentioned, there was no significant interactionbetween NPQs and maternal age groups. These results suggested that for infants fromboth maternal age groups, as the percentage of maternal neighborhood povertyincreased mean gestational ages decreased. Once again, although this was a statisticallysignificant result, for all births there was only a 0.2 week difference in the meangestational age between NPQs 1 and 5.The SGA rates for all births rose from 8.09 per 100 live births in NPQ 1 to 11.18 per 100live births in the last NPQ (Table 5.21). The associated chi-square value was 58.20 with4 degrees of freedom (P<0.001). When the rate inNPQ 5 was compared with that in NPQ1 the resulting chi-square was 46.04 (d.f.=4, P<0.001) with an odds ratio of 1.430 (95%C.L. = 1.29, 1.59). There were also increases in SGA rates when infants in the twomaternal age groups were considered separately (Table 5.21). Although there was nota significant association for infants born to teenage mothers, for the older group, as thepercent of maternal neighborhood poverty increased there was a significant increase inthe small for gestational age rates.Page 118Table 5.21Small for Gestational Age Births by NPQs Stratified by Maternal Age GroupsMaternal Age^NPQ^Yes (row %)^No (row %)^Total1 35 (7.87) 410 (92.13) 4452^55 (11.16) 438 (88.84)^493< 20 years^3 45 (8.60)^478 (91.40) 5234^53 (10.10) 472 (89.90)^5255 88 (11.70) 664 (88.30) 752X2 = 6.48^d.f.=4 p=0.1661^640 (8.10)^7260 (91.90)^79002 704 (8.68) 7406 (91.32) 811020 - 34 years^3 770 (9.70) 7165 (90.30)^79354^815 (10.41)^7011 (89.59) 78265 863 (11.12) 6895 (88.88)^7758X2 = 55.22^d.f.=4 p<0.0011^675 (8.09)^7670 (91.91)^83452 759 (8.82) 7844 (91.18) 86035_ 34 years^3^815 (9.64)^7643 (90.36)^84584 868 (10.39) 7483 (89.61) 83515^951 (11.18) 7559 (88.82)^8510X2 = 58.20^d.f.=4 p<0.0015_ 34 years^5^951 (11.18)^7559 (88.82)^85101 675 (8.09) 7670 (91.91) 8345X2 = 46.04^d.f.=1 p<0.001^O.R.=1.43 95% C.I.=1.29, 1.59Because of the significant association between NPQs and SGA there was a possibilitythat NPQs acted as a confounder in the analysis of maternal age groups and SGA. TheSGA data for the maternal age groups were therefore reanalyzed with a CMH chi-square test while controlling for NPQS. The small difference remained non-significant.The CMH chi-square value was 0.25 with 1 degree of freedom (P=0.617) and theassociated odds ratio was 1.03 (95% C.L.= 0.91,1.18). The crude odds ratio and the CMHodds ratio were fairly similar and the odds ratios for the individual NPQs only variedbetween 0.18 below and 0.25 above the crude odds ratio. This pattern precluded theexistance of confounding. The small variation in the strata specific odds ratios did notsuggest the presence of interaction and this was supported by the insignificant Breslow-Day (BD) chi-square value (x2BD=4.19, d.f.=4, P=0.381).Page 119Although there was a very small but significant association between gestational age andmaternal age groups, SGA rates were not significantly associated with maternal agegroups even when controlling for NPQs. Gestational age and SGA were both associatedwith the five NPQs. As the percent of neighborhood poverty increased the meangestational age decreased and the SGA rate increased. Even while controlling for thetwo maternal age groups, SGA rates increased as the percentage of maternalneighborhood poverty increased.IV Infant MortalityThe overall mortality rate for infants born to teen mothers was 1.24 per 100 live births,which compared unfavorably with a rate of 0.61 for offspring of older mothers. Thisdifference resulted in a significant chi-square value of 16.03 with 1 degree of freedom(P<0.001) and an odds ratio of 2.06 (95% C.L.=1.45 and 2.93). Teenage mothersexperienced higher rates in all NPQs except the third and these were significantlyelevated rates in the second and fifth NPQs (Table 5.22). In these two NPQs the chi-square values were 7.71 (d.f.=1, P=0.006) and 10.93 (d.f.=1, P<0.001).Page 120Table 5.22Infant Mortality by Maternal Age Groups Stratified by NPQsNPQ Maternal Age Yes (row %) No (row %) Total1 < 20 4 (0.90) 441 (99.10) 44520 - 34 46 (0.58) 7856 (99.42) 7902X2 = 0.719 d.f.=1^p=0.4002 < 20 7 (1.42) 486 (98.58) 49320 - 34 39 (0.48) 8073 (99.52) 8112X2 = 7.71 d.f.=1^p=0.0063 < 20 3 (0.57) 520 (99.43) 52320 - 34 50 (0.63) 7886 (99.37) 7936X2 = 0.03 d.f.=1^p=0.8744 < 20 6 (1.14) 519 (98.86) 52520 - 34 49 (0.63) 7780 (99.28) 7829X2 = 2.01 d.f.=1^p=0.1565 < 20 14 (1.86) 738 (94.14) 75220 - 34 56 (0.72) 7705 (99.39) 7761X2 = 10.93 d.f.=1^p<0.0011 - 5 < 20 34 (1.24) 2704 (94.76) 273820 - 34 240 (0.61) 39300 (99.39) 39540X2 = 16.03 d.f.=1^p<0.001 O.R.=2.06^95% C.I.=1.45, 2.93The infant mortality (IM) rate for births included in the study varied from 0.53 in NPQ2 to 0.82 in NPQ 5 . From Table 5.23 it can be seen that this was not a steady increase andresulted in a non-significant chi-square (x2=6.12, d.f.=4, P=0.190). IM rates for infantsborn to mothers aged 20-34 years ranged from 0.48 per 100 live births in NPQ 2 to 0.72per 100 births in NPQ 5. For infants born to the adolescent mothers the IM ratesappeared to fluctuate randomly between 0.57 and 1.86 per 100 live births. When thesewere analyzed separately the differences in IM rates by NPQs for both maternal agegroups were also insignificant (Table 5.23). When the overall IM rate in NPQ 5 wascompared with that in NPQ 1 the chi-square was still insignificant (x 2=2.97, d.f.=1,P=0.085) (OR=1.38, 95% C.L.=0.96, 1.98) (Table 23.5). It is interesting to note that whenthe IM rate for NPQs 4 and 5 combined was compared with the rate calculated for NPQsPage 1211 and 2 the chi-square was significant, but barely (x 2=3.98, d.f.=1, P=0.045). Theassociated odds ratio was 1.311 with 95% confidence limits at 1.005 and 1.711.Table 5.23Infant Mortality by NPQs Stratified by Maternal Age GroupsMaternal Age^NPQ^Yes (row %)^No (row %)^Total1^4 (0.90)^441 (99.10)^4452^7 (1.42) 486 (98.58)^493< 20 years^3 3 (0.57) 520 (99.43) 5234^6 (1.14)^519 (98.86)^5255 14 (1.86) 738 (98.14) 752X2 = 4.87^d.f.=4 p=0.302^O.R.=2.091^46 (0.58)^7856 (99.42)^79022 39 (0.48) 8073 (99.52) 811220 - 34 years^3^50 (0.63) 7886 (99.37)^79364 49 (0.63)^7780 (99.37) 78295^56 (0.72) 7705 (99.28)^7761x2 = 4.03^d.f.=4 p=0.402^O.R.=1.2934 years:5_ 34 years1^50 (0.60)^8297 (99.40)^83472 46 (0.53) 8559 (99.47) 86053^53 (0.63) 8406 (99.37)^84594 55 (0.66)^8299 (99.34) 83545^70 (0.82) 8443 (99.18)^8513X2 = 6.12^d.f.=4 p=0.190X2CMH = 5.59^d.f.=1 p=0.2105^70 (0.82)^8443 (99.18)^85131 50 (0.60) 8297 (99.40) 8347X2 = 2.97^d.f.=1 13=0.085^O.R.=1.38 95% C.I.=0.96, 1.98X2cmn = 2.25^d.f.=1 p=0.133 O.R.=1.33 95% C.I.=0.92, 1.91X2BD= 0.76^d.f.=1 p=0.384Because there was a significant association between the maternal age groups and IM itwas necessary to reanalyze the NPQ data while controlling for maternal age groups.There was no change when the non-significant association between IM and and the fiveNPQs was re-examined while controlling for maternal age group membership. TheCMH chi-square value was 5.59 with 1 degree of freedom (P=0.210). When only therates in the last and first NPQs were compared using the CMH chi-square the value wasPage 1222.25 (d.f.=1, P=0.133) and the associated odds ratio was 1.33 (95% C.L.= 0.918, 1.913).There appeared to be some, but not substantial, confounding. When comparing IMrates in NPQs 5 and 1, the odds ratios for the two maternal age groups were 2.09 and1.24 compared to the overall rate of 1.38. The crude and CMH odds ratios however werevery similar. Although the variation in the strata specific odds ratios suggested someinteraction the non-significant BD chi-square (x 2BD=0.76, d.f.=1, P=0.384) did not allowit to be seen as statistically significant.Thus there was no association between IM and maternal neighborhood low income. Onthe other hand the association of the former with maternal age groups remained strongeven when controlling for the five NPQs. IM rates were elevated for teenage mothersapart from the percentage of maternal neighborhood poverty.a) Neonatal MortalityWhen infant mortality was broken into neonatal mortality (NM) and postneonatalmortality (PNM) a much different picture emerged. Table 5.24 presents rates for NMby the two maternal age groups. In NPQs 1, 3, and 4 infants born to teenage mothersexperienced lower rates of NM than their older counterparts. This was reflected in theoverall rates which were 0.40 per 100 live births for teen mothers and 0.43 for mothersaged 20-34 years. This difference was not significant (x 2=0.04, d.f.=1, P=0.842) (OR=0.94,95% C.L.=0.51, 1.73).Page 123Table 5.24Neonatal Mortality by Maternal Age Groups Stratified by NPQsNPQ^Maternal Age^Yes (row %)^No (row %)^Total1^< 20^1 (0.22) 444 (99.78)^44520 - 34 33 (0.42)^7869 (99.58) 7902X2 = 0.39^d.f.=1 p=0.5342^< 20 2 (0.46)^491 (99.59)^49320 - 34^26 (0.32) 8086 (99.68) 8112X2 = 0.10 d.f.=1 p=0.7473^< 20^1 (0.19)^522 (99.81)^52320 - 34 37 (0.47) 7899 (99.53) 7936x2 = 0.83 d.f.=1 p=0.3624^< 20^1 (1.52)^524 (98.48)^52520 - 34 30 (0.84) 7799 (99.16) 7829X2 = 0.49^d.f.=1 p=0.4825^< 20 6 (0.80)^746 (99.20)^75220 - 34^43 (0.55) 7718 (99.45) 7761X2 = 0.71 d.f.=1 p=0.398Total^< 20^11 (0.40)^2727 (99.60)^273820 - 34 169 (0.43) 39371 (99.57) 39540X2 = 0.04^d.f.=1 p=0.842^O.R.=0.94 95% C.I.=0.51, 1.73NM rates for all births combined as well as for the two maternal age groups separatelywere fairly similar for the first four NPQs and rose in fifth (particularly for infants bornto teenage mothers)(Table 5.25). These trends were not found to be statisticallysignificant for all mothers (x2=7.32, d.f.=4, P=0.120) or for either of the two maternalgroups (Table 5.25).These study results suggest that there is no significant association between NM ratesand either levels of maternal neighborhood low income or maternal age groups.Page 124Table 5.25Neonatal Mortality by NPQs Stratified by Maternal Age GroupsMaternal Age^NPQ^Yes (row %)^No (row %)^Total< 20 years1 1 (0.22) 444 (99.78)^4452^2 (0.41)^491 (99.59) 4933 1 (0.19) 522 (99.81)^5234^1 (0.19) 524 (99.81) 5255 6 (0.80)^746 (99.20)^752X2 = 4.46^d.f.=4 p=0.3471^33 (0.42)^7869 (99.58)^79022 26 (0.32) 8086 (99.68) 811220 - 34 years^3^37 (0.47) 7899 (99.53)^79364 30 (0.38)^7799 (99.62) 78295^43 (0.58) 7718 (98.42)^7761X2 = 5.76^d.f.=4 p=0.21834 years1^34 (0.41)^8313 (99.58)^83472 28 (0.33) 8577 (99.67) 86053^38 (0.45) 8421 (99.55)^84594 31 (0.37)^8323 (99.63) 83545^49 (0.58) 8464 (99.42)^8513X2 = 7 .32^d.f.=4 p=0.1205_ 34 years^5^49 (0.58)^8464 (99.42)^85131 34 (0.41) 8313 (99.58) 8347X2 = 2.44^d.f.=1 p=0.119^O.R.=1.42 95% C.I.=0.92, 2.19b) Post Neonatal MortalityThis was not the case for PNM rates. They were higher for infants born to teen mothersin all NPQs and consequently overall. In all NPQs, except the third, the differences weresignificant (Table 5.26). The overall rates of 0.84 per 100 live births for infants born tothe young mothers and 0.18 for the older mothers resulted in a chi-square of 50.35 with1 degree of freedom (P<0.001). The associated odds ratio was 4.71 with 95% confidencelimits at 2.94 and 7.55.Page 125Table 5.26Postneonatal Mortality by Maternal Age Groups Stratified by NPQsNPQ Maternal Age Yes (row %) No (row %) Total1 < 20 3 (0.67) 442 (99.33) 44520 - 34 13 (0.16) 7889 (99.84) 7902X2 = 5.72 d.f.=1 p=0.0172 < 20 5 (1.01) 521 (98.99) 49320 - 34 13 (0.16) 8099 (99.84) 8112X2 = 16.24 d.f.=1 p<0.0013 < 20 2 (0.38) 521 (99.62) 52320 - 34 13 (0.16) 7923 (99.84) 7936X2 = 1 .33 d.f.=1 p=0.2504 < 20 5 (0.95) 520 (99.05) 52520 - 34 19 (0.24) 7810 (99.76) 7829X2 = 8.65 d.f.=1 p=0.0035 < 20 8 (1.06) 744 (98.94) 75220 - 34 13 (0.17) 7748 (99.83) 7761X2 = 22.38 d.f.=1 p<0.0011- 5 < 20 23 (0.84) 2715 (99.16) 273820 - 34 71 (0.18) 39469 (99.82) 39540X2 = 50.35 d.f.=1 p<0.001 O.R.=4.71^95% C.I.=2.94, 7.5There did not appear to be a significant association between PNM and NPQs. Forinfants born to teenage mothers the PNM rate was lowest in the middle NPQ andhighest in the second and last NPQs (Table 5.27). Rates for offspring born to mothersaged 20-34 years were 0.16 per 100 live births for the first three NPQs and then increasedslightly in the remaining two. The associated chi-square values for PNM rates by NPQswere not significant for either of these groups (Table 5.27) . The pattern of PNM ratesfor all infants was similar to that for the infants born to the older mothers and alsoresulted in a non-significant chi-square value (x 2=3.01, d.f.=4, P=0.556). Even compar-ing the NPQs in NPQs 5 and 1 did not result in significant chi-square values (x 2=0.58,d.f.=1, P=0.445) (OR=1.15, 95% C.L.=0.61, 2.21).Page 126Table 5.27Postneonatal Mortality by NPQs Stratified by Maternal Age GroupsMaternal Age< 20 yearsNPQ12345x2 = 2.17Yes (row %)3 (0.67)5 (1.01)2 (0.38)5 (0.95)8 (1.06)d.f.=4 p=0.704No (row %)442 (99.33)488 (98.99)521 (99.67)520 (99.05)744 (98.94)O.R.=1.58Total4454935235257521220 - 34 years^345x2 = 2.18125.. 34 years^345X' = 3.01X2 cmii= 2.9834 years^51= 0.58X2 cmH = 0.16X2 BD= 0.3213 (0.16)13 (0.16)13 (0.16)19 (0.24)13 (0.17)d.f.=4 p=0.70216 (0.74)18 (0.71)15 (0.80)24 (0.89)21 (1.10)d.f.=4 p=0.556d.f.=4 p=0.57221 (1.10)16 (0.74)d.f.=1 p=0.445d.f.=1 p=0.688d.f.=1 p=0.572^7884 (99.84)^79028099 (99.84) 81127923 (99.84)^79367810(99.76) 78297748 (98.83)^7761O.R.=1.028331 (99.26)^83478587 (99.29) 86058444 (99.20)^84598330 (99.11) 83548492 (98.90)^85138492 (98.90)^85138331 (99.26) 8347O.R.=1.15 95% C.I.=0.61, 2.21O.R.=1.15 95% C.I.=0.59, 2.22Because there was a significant association between the maternal age groups and PNMit was necessary to reanalyze the NPQ data while controlling for this variable in orderto asses the possibility of confounding. There was no change when the non-significantassociation between PNM and and the five NPQs was re-examined while controlling formaternal age group membership. The CMH chi-square value was 2.88 with 1 degreeof freedom (P=0.572). When only the rates in the last and first NPQs were comparedusing the CMH chi-square test controlling for the two maternal age groups, the valuewas 0.16 (d.f.=1, P=0.688) and the associated odds ratio was 1.15 (95% C.L.= 0.59, 2.22).When comparing PNM and rates in NPQs 5 and 1, the odds ratios for the two maternalage groups were 1.58 and 1.02 compared to the overall rate of 1.15 which was the samePage 127as the CMH odds ratios. There appeared to be no substantial confounding. This patternof the odds ratios did not suggest the presence of any interaction which was alsoconfirmed by the non-significant BD chi-square (X2BD=0.32, d.f.=1, P=0.572).PNM rates were not associated with the percentage of maternal neighborhood lowincome. There was however a very significant association between PNM and maternalage groups even while controlling for neighborhood low income: infants born to teenmothers were more likely to die during the postneonatal period.V Congenital AnomaliesCongenital anomaly rates for infants born to the two maternal age groups are presentedin Table 5.28. There did not appear to be any clear difference between the two maternalage groups and chi-square analyses within each of the NPQs did not produce anystatistically significant results (Table 5.28). The overall rate of congenital anomaliesdiagnosed and registered in the first year of the life of infants born to teen mothers was7.45 per 100 live births compared with 7.39 for infants born to the older mothers. Thisresulted in a chi-square value of 0.01 with 1 degree of freedom (P=0.907) and an oddsratio of 1.01 (95% C.L.=0.870, 1.169).Page 128Table 5.28Congenital Anomalies by Maternal Age Groups Stratified by NPQsNPQ Maternal Age Yes (row %) No (row %)1 < 20 25 (5.62) 420 (94.38)20 - 34 517 (6.54) 7385 (93.46)X2 = 0.59 d.f.=1^p=0.441 O.R.=1.182 < 20 24 (4.87) 469 (95.13)20 - 34 573 (7.06) 7539 (92.94)X2 = 3.47 d.f.=1^p=0.063 O.R.=1.493 < 20 43 (8.22) 480 (91.78)20 - 34 560 (7.06) 7376 (92.94)x2 = 1.01 d.f.=1^p=0.316 O.R.=0.854 < 20 46 (8.76) 479 (91.24)20 - 34 596 (7.61) 7233 (92.39)X2 = 0.92 d.f.=1^p=0.339 O.R.=0.865 < 20 66 (1.06) 686 (98.94)20 - 34 676 (0.17) 7085 (99.83)X2 = 0.01 d.f.=1^p=0.951 O.R.=0.991 - 5 < 20 204 (7.45) 2534 (92.55)20 - 34 2922 (7.39) 36618 (92.61)X2 = 0.01 d.f.=1^p=0.907 O.R.=1.01^95% C.I.=0.870, 1.169X2cNfH = 0.03 d.f.=1^p=0.865 O.R.=0.99^95% C.I.=0.851, 1.145X2BD= 6.02 d.f.=4^p=0.198Congenital anomaly rates by maternal age groups for the five NPQs are presented inTable 5.29. Rates for infants born to teenage mothers ranged from 4.87 to 8.78 per 100live births while the rates for the older mothers ranged from 6.54 to 8.71 per 100 livebirths. The associated chi-square values for congenital anomalies by NPQs for bothteens and older mothers were significant (Table 5.29). When all births were consideredtogether the congenital anomaly rates rose from 6.49 in NPQ 1 to 8.72 per 100 live birthsin NPQ 5 with a chi-square of 36.14 with 4 degrees of freedom (P<0.001). When theoverall congenital anomaly rate for NPQ 5 was compared with that of NPQ 1 the chi-square was 29.59 (d.f.=4, P<0.001) with an odds ratio of 1.38 and 95% confidence limitsat 1.23 and 1.54.Page 129Table 5.29Congenital Anomalies by NPQs Stratified by Maternal Age GroupsMaternal Age^NPQ^Yes (row %)^No (row %)^Total< 20 years1 25 (5.62) 420 (94.38)^4452^24 (4.87)^469 (95.13) 4933 43 (8.22) 480 (91.78)^5234^46 (8.76) 479 (91.24) 5255 66 (8.78)^686 (91.22)^752X2 = 10.61^d.f.=4 p=0.0311^517 (6.54)^7385 (93.46)^79022 573 (7.06) 7539 (92.94) 811220 - 34 years^3^560 (7.06) 7376 (92.94)^79364 596 (7.61)^7233 (92.39) 78295^676 (8.71) 7085 (91.29)^7761x2 = 31.17^d.f.=4 p<0.0011^542 (6.49)^7805 (93.51)^83472 597 (6.94) 8808 (93.06) 86055. 34 years^3^603 (7.13) 7856 (92.87)^84594 642 (7.68)^7712 (92.32) 83545^742 (8.72) 7771 (91.28)^8513X2 = 36.14^d.f.=4 p<0.0015 34 years^5^742 (8.72)^7771 (91.28)^85131 542 (6.49) 7805 (93.51) 8347X2 = 29.59^d.f.=1 p<0.001^O.R.=1.38 95% C.I.=1.23, 1.54Because of the significant association between NPQs and congenital anomalies therewas a possibility that NPQs acted as a confounder. The data were reanalyzed with CMHchi-square analyses controlling for NPQs. The analysis again resulted in a non-significant value of 0.03 with 1 degree of freedom (P=0.865) (OR=0.99, 95% C.L.=0.85,1.15). The odds ratios for the individual NPQs were very similar to each other and theoverall crude odds ratio, ranging only from 0.16 below and 0.48 above this odds ratio(Table 28). This homogeneity was supported by a BD chi-square value of 6.02 (d.f.=4,P=0.198) which was not significant. The pattern of the strata specific odds ratios and thesimilarity between the overall crude and CMH odds ratios also precluded the existenceof confounding.Page 130These analyses confirmed the lack of an association between the two maternal agegroups and the rate of diagnosis and registration of congenital anomalies in the first yearof infants' lives. There is a suggestion of a significant relationship between NPQ groupsand congenital anomaly rates. The rate of diagnosis and registration of congenitalanomalies increases with increasing levels of maternal neighborhood low income.Page 131REFERENCES1.BC Ministry of Health and Ministry Responsible for Seniors DVS. Selected vitalstatistics One hundred twentieth Annual report 1991. Province of BritishColumbia, 1993: 154.Page 132CHAPTER SIXDISCUSSION OF RESULTS AND IMPLICATIONSEVALUATION OF STUDY DESIGNThe design of this study was based on a method of linking aggregate and individualdata that has been refined within the Canadian context by R. Wilkins of StatisticsCanada. It was intended to allow for the evaluation of the independent effect ofmaternal age (MA) and socioeconomic status (SES) on various birth outcomes usingmaternal and birth data from three B.C. data sets linked with ecological SES informationprovided by Statistics Canada.The three B.C. data bases provided high quality data for the study. Only 1% of the infantdeath records and 1.4% of the HSR registrations were missing information that wouldnot allow them to be linked to PNOB records. Once the data base was created and thefiles linked fewer than 1% of the cases were missing variables used in the study. Ofparticular note is the fact that B.C. has a HSR that routinely collects congenital anomalyinformation. The availability of this registry provided a unique research tool that allowsfor the study of the association of anomalies with the two risk factors.The postal code of the mother's usual address at the time of birth was critical to linkingthe SES and other data. Fewer than 1% of the 68,795 birth records included in the studywere missing postal codes or had obviously incorrect codes. Although the distributionof these codes throughout the data set may not have been random the total number wasso small that it would not have been responsible for any significant skewing of the studyresults.Page 133The addition of postal codes to the PNOB records in 1985 was invaluable because of theirability to be converted to census tract identifiers which were the linking variables toaggregate income data. It was noted earlier that although previous studies employingthis design 1,2,3 were limited to CMAs it was decided to also include CAs in this study.Only one CA (Prince George) had a substantially higher percentage of birth records forwhich it was not possible to locate a census tract identifier or to attach incomeinformation to the census tract identifier. The results of the analysis of the incomesegregation and homogeneity (Figure 5.1) suggested only for one CA (Kelowna) therewas not a significant and steady gradient between NPQs 1 and 5. A limited analysis ofthe effect of including these CAs in the study (Appendix C) suggested that it was notnecessary to remove them from the study. These analyses suggest CAs do notautomatically need to be excluded from studies employing this type of design. Apreliminary analysis into data quality and income range and homogeneity in thesesmaller urban centres is important however.Even with the inclusion of CAs this study design is limited to urban areas that havecensus tracts. Besides the fact that only urban areas are tracted, in reality it is only largerurban areas that have both substantial range of incomes as well as clustering ofindividuals by income. This economic gradient as well as clustering is critical to aneffective analysis of the impact of SES. The more economic clustering/homogeneitywithin census tracts the smaller the potential for substantial bias that can easily plagueresults obtained with ecological designs.Not withstanding this limitation this design allowed for the relatively inexpensivedevelopment of large population based data set that was required to study relativelyrare occurrences in the population as a whole or to investigate issues within fairly smallsubpopulations. This design would also allow for the investigation of the impact ofPage 134various measures of SES given that, theoretically, all information collected through theCensus is available by census tract.There is an increasing appreciation that the importance of various risk factors, includingyoung MA and low SES, may change as a consequence of policy decisions over time orwith changes in social, economic and political environments. Because this methodrelies on routinely collected data it potentially provides a relatively inexpensive way oftracking changes in various measures of health. One significant limitation of thismethod is that when evaluating the effect of SES quintiles on health it is not possible todetermine if changes are a consequence of a changing relationship between the healthoutcome being considered and SES or a consequence of changes in the SES gradient.Evaluating the SES gradient (as was done in this study) is a possible way of circumvent-ing this limitation.DISCUSSION OF RESULTSThe Association Between Young Maternal Age and Poor Socioeconomic Status An increase in the number and percentage of all births to teenaged mothers with adecrease in SES, has been consistently documented by the few studies which haveexamined this relationship. The association was found in studies based on Scottish,American and Canadian data. The present study found a similar association betweenbirths to young mothers and SES in urban B.C. The pattern of births to young mothersrelative to older mothers in this study was very similar to that observed in the Canadian 2and Scottish studies 4. In all three studies, the number of young mothers and thepercentage of all mothers who were less than 20 years old increased as povertyincreased. There was a 10% to 30% increase in the percentage of births to young mothersPage 135between the most well to do and the next SES category while there was approximatelya 50% increase between the poorest two categories.The range in the distribution of births to young mothers in this study was narrower thanthat observed in Smith's 4 and Wilkins' 2 studies which had increases from approxi-mately 2% to 9% between the highest and lowest SES categories. In the present studyratios of the births to teen mothers relative to mothers aged 20 through 34 years wereused as the unit of comparison instead of rates and these varied between 0.056 in NPQ1 to 0.097 in NPQ 5. Although the units of analysis are not identical in these studies, thisdoes not fully explain why the range between NPQs 1 and 5 was so much smaller thanthe range in the percentages in the other two studies. The reason for the difference liesin the fact that in this study the proportion of all births occurring to teenaged mothersin the least poor NPQs was more than double that in the other two studies.This B.C. study differed from the other two studies in that it was limited to first births,included only births in urban B.C. and the comparison group was restricted to mothersaged 20 through 34 years as opposed to all mothers over 20 years. Although it is possiblethat the 30% of mothers that do not live in urban B.C. have substantially different childbearing patterns than those in urban centres, it is unlikely that these could be so differentas to explain the entire difference in the results. Additionally, the births included inWilkins's2 study were limited to urban centres. Restricting the births included in thisstudy to first births and the comparison group to mothers between 20 and 34 years mayhave had more of an impact on the elevated ratios in the lower NPQs. If it is true thatwealthier women tend to have more children and/or begin to have children at olderages the ratio of births to teen mothers relative to births to other mothers would decreasein the first one or two NPQs. The resulting distribution would then be more similar tothat obtained by both Smith4 and Wilkins 2. It will require another study to assess if thesePage 136or other factors are indeed the reason for the differences observed in this study. It mayalso be that the distribution of births to different MA groups is different in B.C. than inthe rest of Canada or elsewhere.It is important to note that the distribution of live births between NPQs was consideredin this present study. Total pregnancies were not considered. Although it might beappealing to use the association between MA and SES with respect to birth rates tospeculate on patterns of total pregnancy rates for young mothers, it is probably notappropriate to do so. Smith 4 in his recent Scottish study was able to look at totalpregnancy, abortion and birth rates by various SES categories. He found that there wasa SES gradient in the abortion rate for young women. Those in more deprived SEScategories had lower abortion rates. As a consequence the SES gradient for the percentof total pregnancies to women less than 20 years was steeper than that observed forbirths. It is not clear if the same patterns would be observed in B.C. given that theutilization of services is a function of access to these services, moral concerns regardingthe use of abortion services, as well as other issues.The Association of Young Maternal Age and Poor Socioeconomic Status with VariousBirth Outcomes In the present study it was found that young MA, defined as mothers aged less than 20years at the birth of their child, was a significant and independent risk factor for elevatedrates of low birth weight, infant mortality, and postneonatal mortality relative to thecomparison group of women aged 20 through 34 years. For all mothers, the NPQ ofresidence at the time of birth was significantly associated with rates of low birth weight,very low birth weight, small for gestational age births, and congenital anomalies. AsPage 137neighborhood poverty increased so did the rates of these various birth outcomes. Onlylow birth weight was associated with both young MA and low SES.It is interesting to compare the results obtained in this present study with the conclu-sions reached regarding the independent association of either young MA or poor SESwith various birth outcomes in the literature review.No association was found between young MA and stillbirth rates, in any of the fivestudies reviewed. This concurred with the fact that no association was found in thisstudy. The same was found for small for gestational age rates. In none of the five studiesreviewed or the present study were small for gestational age rates significantlyassociated with young MA. Congenital anomaly rates were also found not to beassociated with MA in either the present study or in any of the studies reviewed. It isalso interesting to note that in three of the four studies reviewed the congenital anomalyrate for young mothers was lower than for older mothers. In the present study, althoughthis was not the case, teen mothers had lower rates in two NPQs and the overall ratesfor the two groups of mothers were very similar; 7.45% for the young mothers comparedwith 7.39% for those who were older.In the literature review infant mortality rates were found to be associated with youngMA in three out of the four studies. Given the design and international data set of theone study5 that did not find a relationship, it was concluded that there probably is anassociation between young MA and infant mortality. This present study concurredwith this conclusion in that it found a statistically significant relationship betweenyoung MA and infant mortality rates. The odds ratio was 1.3 with 95% C.L.=1.11 and1.57.Page 138When infant mortality was divided into neonatal and postneonatal mortality there wasnot quite as much congruence between the literature and the results of this presentstudy. In only five of the thirteen studies reviewed that dealt with the independenteffect of MA on neonatal mortality rates was a significant association found. It wasconcluded that from the literature there does not appear to be a consistent or strongassociation between neonatal mortality and young MA. In this study no association wasfound between these two variables. Neonatal mortality rates were not even elevated foryoung mothers compared with the rates for older mothers (Table 5.24). Postneonatalmortality rates were significantly higher for young mothers in six of the eight studiesretrieved. After reviewing these studies it was concluded that there may be anassociation between young MA and postneonatal mortality but that it is probably weakand so only evident relative to mothers experiencing the best birth outcomes; thosebetween 20 and 34 years of age. This study did in fact find that postneonatal mortalityrates were significantly elevated for mothers under 20 years. The association could notbe described as weak given that the odds ratio was 4.8 (95% C.L.=2.98, 7.80).There does not appear to be a readily apparent explanation for this very strongassociation between young MA and postneonatal mortality rates. Given that theneonatal mortality rates for young mothers were not elevated it may be that neonatalmortalities among infants born to young mothers were delayed till the postneonatalperiod and thus inflated these rates. Although the postponement of death to thepostneonatal period is possible it is hard to envisage why this would occur only toinfants born to young mothers. Even if this is the case, it would probably not explainthe entire elevation in rates for infants born to young mothers. An alternativeexplanation may be that infants born to young mothers in British Columbia may be atelevated risk, not due to some biological factor related to young MA (which wouldresult in elevated rates in other jurisdictions as well), but possibly because of risksPage 139related to the stresses of teenage parenting. It is noted in Table 5.26 that 8 of the 23postneonatal deaths of infants born to young mothers occurred in NPQ 5. The oddsratio for young mothers in this NPQ was 6.41, almost 50% greater than the overall oddsratio. It suggests that the elevated risk of postneonatal mortality for infants born toyoung mothers may be exacerbated when teens are parenting within a low SESenvironment.Twenty four articles were reviewed which considered the independent effect of MA onlow birth weight rates. In six of the eleven population based and seven of the thirteeninstitutionally based studies there was a significant association between young MA andeither birth weight or low birth weight rates. It was noted that the number and natureof other variables included in the analysis might be a factor in the observed relationshipbetween young MA and low birth weight. Studies which included more variables; suchas those relating to maternal height and weight, prenatal care and medical complica-tions of pregnancy, and life style issues such as smoking and drug use habits tended notto find an significant association and studies which did not include these were morelikely to note an association. The present study considered the independent effect ofMA while controlling for SES and did find an association between young MA and lowbirth weight rates; with rates increasing with decreasing maternal age. If indeed theresult was significant (because of the large number of tests undertaken it may not havebeen significant) the association was not strong (OR=1.30, 95% C.L.=1.11, 1.53). Thisresult would appear to substantiate the conclusion arrived at upon reviewing theliterature and that made when considering the clinical as opposed to statistical signifi-cance of the differences in mean birth weight of infants born to the two groups ofmothers. Although there were no studies which considered the independent effect ofyoung MA on very low birth weight the insignificant result obtained in this studyPage 140(OR=1.28, 95% C.L.=0.88, 1.86) was consistent with the result obtained for low birthweight.Overall, the results of this study regarding to the independent effect of young MA onvarious birth outcomes were fairly consistent with conclusions from the literature inthis area. This was not the case when comparing study results concerning theassociation between poor SES and various birth outcomes while controlling for MAwith a review of the literature dealing with the same issue.The only congruence obtained between the conclusions from the review of the literatureand the results of this study with regard to SES was in the relationship between poor SESand birth weight or low birth weight rates. Although various measures of SES wereused, all eleven population based and ten of the thirteen institutionally based studiesfound a significant association and it was concluded that there was a significantassociation between poor SES and low birth weight rates. In this study the associationwas also found to be significant with the odds ratio describing the difference in the lowbirth weight rates between NPQs 5 and 1 equal to 1.40 (95% C.L.=1.21, 1.61).The association between poor SES and congenital anomaly rates could not be reviewedbecause in all the articles retrieved SES was either controlled or was a matching variable(case-control studies). In this study there was a significant association between low SESand congenital anomaly rates. The odds ratio comparing the rates in NPQs 5 and 1 was1.38 (95% C.L.=1.23, 1.55).There was disagreement between the conclusions drawn from a review of the literatureand the results of the analysis undertaken in this study regarding the associationsbetween poor SES and the remaining birth outcomes. All three of the studies reviewedPage 141that considered the independent effect of poor SES on stillbirth rates found that theassociation was a significant. This was in contrast with the result obtained in the presentstudy which found that there was not a significant association. In fact the rates in thevarious NPQs were essentially the same (Table 5.13). None of the five articles reviewedregarding the independent effect of poor SES on small for gestational rates found asignificant association between these two variables. Again this was in contrast with thesignificant association between these two variables obtained in this study. The oddsratio comparing the small for gestational rates in NPQs 5 and 1 was 1.43 (95% C.L.=1.29,1.58).There were also differences with respect to infant mortality. All three articles reviewedregarding the independent association between low SES and infant mortality found asignificant association while this study found there to be none (OR=1.33, 95% C.L.=0.92,1.91)a. There was however an increase in rates of infant mortality from NPQs 1 through5 (Table 5.23). This difference between the results of the literature review and thepresent study persisted when infant mortality was divided into neonatal and pos tneonatalmortality. Seven of the nine studies reviewed found an association between SES andneonatal mortality resulting in the conclusion that these two were significantly associ-ated; that neonatal mortality rates increased as the percentage of children in themother's neighborhood living in poverty decreased. The finding of this study was thatthese two were not significantly associated (OR=1.42, 95% C.L.=0.92, 2.19) althoughrates (particularly for young mothers) increased between NPQs 1 and 5. All six of thestudies reviewed which considered the independent association between low SES andpostneonatal mortality rates found a significant increase in postneonatal mortality ratesa. An earlier B.C. study by Thomson3 using a similar design found that infant mortality rose from 5.8%in NPQ1 to 12.8% in NPQ5. This gradiant was substantially greater than that observed in this study. Theexplanation for this difference may be that the Thomson study included all births to all women asopposed to only first single births to women less than 35 years.Page 142with decreasing SES. Again, this was opposite the result obtained in this study. The ORcomparing NPQ 5 with NPQ 1 was 1.15 (95% C.L.=0.59, 2.22).This lack of a gradient is particularly interesting given that it was noted earlier that thereappeared to be a SES gradient in postneonatal mortality rates among young mothers.It may be that young mothers are more vulnerable to the effects of low SES than are oldermothers. It may also be that there is a SES gradient in postneonatal mortality rates forall mothers (although possibly not as strong as for young mothers alone) but somethingabout the ecological nature of the design prevented its detection. Alternatively, it maybe that social, economic and political circumstances are such in British Columbia thatthere is no SES gradient in postneonatal mortality rates for all mothers.There are a number of reasons which may explain the differences in the conclusionsdrawn on the basis of the literature review and the results of this study regarding therelationship between poor SES and various birth outcomes. One of the major factorsmay be the different social and political circumstances of the research reviewedcompared with the present study. This work was carried out with B.C. data collectedbetween 1985 and 1988. None of the research reviewed in this section used Canadiandata. It was drawn primarily from the U.S. with some studies corning from the UnitedKingdom, Sweden, New Zealand and Tasmania. All these countries have differentsocial welfare systems, health care and other systems which will impact the SESgradient in the country. A country with a smaller SES range may well not havesignificantly different rates of various birth outcomes by SES while a country with abroader range does. This may go a long way in explaining why no significantassociations were observed between poor SES and stillbirths and infant, neonatal andpostneonatal mortality. In the case of all these outcomes this study did not findsignificant associations while the consensus from the literature, much of which wasPage 143from the U.S., was that the associations were significant. There is generally acknowl-edged to be a larger gap between the rich and the poor in the U.S. than in Canada.The second issue is that of the measurement of SES. A number of different proxymeasure of SES were used in the studies reviewed. These included race, education,occupation, income or composites generated from combinations of the above. Addi-tionally, sometimes the measure was based on individual data and on other occasionsan ecological measure was created based on a variable most often generated frominformation collected from the neighborhood or census tract of the mother's usualresidence. A critical question is whether the variable used to measure SES makes adifference in the results obtained regarding the impact of this risk factor on a birthoutcome. A recent Australian study6 designed to address this question found socialclass trends with respect to infant mortality of a similar order irrespective of whetheroccupation, education, or income were used as the measure of social class. Of thesevariables, family income has been found to correlate best with SES differences inmortality 7,8,9,10. These results for low birth weight may not apply to the relationshipbetween various measures of SES and other birth outcomes. An article by Liberatos,Link and Kelsey" reviewed the various ways of measuring social class in epidemiol-ogy. They noted that social class (or SES) incorporates economic, political and culturaldifferences that may have an impact on health. Since different measures of SES weightdifferent aspects of SES differently the results regarding the relationship betweenalternative proxy measures of poor SES on various birth outcomes will not be identical.The degree of difference will depend on the relative influence of the various aspects ofSES on the birth outcome in question.A very closely related issue, and one that is relevant to this present project, is the impactof using an ecological variable as opposed to an individual one to measure the SES ofPage 144subjects included in the study. Antonovsky and Bernstein 12 reviewed a dozen early(pre 1977) studies dealing with the relationship between social class and neonatal andpostneonatal mortality. They noted that studies which employed ecological measuresof social class differences tended to find smaller, if any social class differences in infantmortality rates. They concluded that ecological measures of social class tended tounderestimate effects of this measure of SES on infant, neonatal and postneonatalmortality. If this conclusion can be generalized to the ability of ecological measures ofSES to estimate the true effect of SES on birth outcomes the suggestion is that ecologicalmeasures SES underestimate the strength of significant associations with birth out-comes at the individual level. This conclusion allows for the suggestion that significantassociations between low SES neighborhoods and low birth weight, small for gesta-tional age and congenital anomalies observed in this study should be generalizable toindividuals living in low SES situations.There is at least one other issue which may be a factor in explaining some of thedifferences in the conclusions drawn regarding the relationship between SES andvarious birth outcomes. It is the issue of secular trends. There are two questions relatingto this issue. The first is whether there have been changes in the disparity between thoseof the highest SES and lowest SES over time. The second is whether changes in the ratesof various birth outcomes due to technological advances, changing health care practicesetc. have had the same impact at all levels of SES or have these changes worked tonarrow or widen the gap between them. The literature reviewed did not go intosufficient detail in order to allow conclusions to be drawn as to whether or not seculartrends were responsible for some of the differences between conclusions drawn fromthe literature and these study results. It is noted , however, that most of the largerpopulation based studies were undertaken in recent years because the sophisticatedPage 145computer technology required to create and manipulate their data bases has onlyrecently become available. This would minimize the impact of secular trends.Although not an exhaustive discussion, the major factors that may have contributed tothe lack of consistent conclusions with respect to the independent impact of SES, beyondstudy design and analysis, have been acknowledged. A more thorough review of oneor two birth outcomes and a further study of the independent impact of variousmeasures of SES on birth outcomes in a Canadian setting would contribute to ourunderstandings in this area.The issue of interaction between SES and young MA with respect to the various birthoutcomes also requires further research. In this study it was found that there did notappear to be significant interaction between SES and MA with respect to any of the birthoutcomes analyzed. It was not possible to adequately compare these results with theliterature that was reviewed because there were not sufficient numbers of studies thatincluded the investigation of possible interactions between SES and young MA withrespect to various birth outcomes as part of their design.ConclusionsThe association between young MA and low SES that was observed in this study isconsistent with what has been observed in the few studies which have addresses thisissue. Each of these two variables was also found to be a risk factor for a number of birthoutcomes independent of the other one.The significant associations observed in this study between young MA and rates of lowbirth weight, infant mortality, and postneonatal mortality were consistent with thePage 146literature and as such should be seen as an affirmation that the independent associationbetween young MA and these birth outcomes in B.C. is consistent with that observedelsewhere. The significant independent associations observed in this study betweenlow SES and rates of low birth weight, small for gestational age, and congenitalanomalies were not all consistent with the literature. Small for gestational age rateswere not found to be significant risk factors and there was no literature in which theindependent effect of congenital anomalies was analyzed. However the literaturefound low SES to be a risk factor for infant, neonatal and postneonatal mortality whenthis study did not find this to be the case. For the various reasons previously stated,these differences do not preclude the study results from being accurate within thecontext of B.C.In affirming the conclusions reached as a consequence of this study it is also acknowl-edged that there are a number of study limitations which may affect its validity,reliability and the generalizability of the results.LIMITATIONSBirth and death registrations are required by law in B.C.. This assures a very highascertainment rate for births and infant deaths in the province. Although there has notbeen any formal evaluation of the validity and reliability of the data collected on theseforms, the suggestion is that they are fairly good for most of the variables used in thisstudy. One exception might be gestational age which is determined by recall, ultra-sound, or pediatrician's assessment at birth. It is not clear if multiple methods ofdetermining gestational age negatively impact on data quality. Another exceptionmight be stillbirths. A number of records had to be removed from the data base becausethey were not stillbirths as defined by DVS. Additionally, there is also concern aboutPage 147stillbirths not reported. There is uncertainty about the number not reported as well asany SES bias in the under reporting.Apart from concerns raised earlier about the HSR there was a significant concernsurrounding the registration of congenital anomalies. In this study congenital anoma-lies were restricted to those registered in the first year of the infant's life. In order to beregistered a condition must be brought to the attention of individuals and/or agenciesinvolved in registrations. It was not clear if the SES gradient observed in this study wasgoing to be a consequence of differences in the rate of congenital anomalies betweenindividuals in the various NPQs or whether it would be a function of differences in therates of registration of these anomalies. It was speculated that socially and economicallydisadvantaged mothers might not access the medical system as early as other mothersin order to seek a diagnosis. As a consequence, although poor SES was associated withsignificant increases in congenital anomaly rates this conclusion must be interpretedvery cautiously.Because census tracts are limited to CMAs and CAs the results of this study can at bestonly be generalized to urban areas. It is unfortunate that this method cannot be appliedto rural areas given that these risk factors are relevant in rural areas and it is likely thatthe relationships between birth outcomes and the various risk factors are different inrural areas than they are in urban centres.Another limitation related to the study design was that of the ecological nature of theSES variable. Even though the income homogeneity of the census tracts was analyzedand found to be fairly good, there are concerns about generalizing the relationshipbetween the birth outcome and SES within NPQs to individuals. It is not known ifindividuals from low income NPQs themselves have low incomes. To automaticallyPage 148assume that the results of this study apply to individuals would be an ecological fallacy.There is the work of Antonovsky and Bernstein 12, however, which suggests that forecological studies dealing with the association between SES and birth outcomes it maybe appropriate to use the strength of associations at the ecological level to makestatements about theses associations at the individual levelAnother design issue was that of the choice of women aged 20 through 34 years as thecomparison group. In an earlier discussion of the association between MA and the ratesof various birth outcomes it was acknowledged that birth outcomes vary with MA.Only part of this difference is because of confounds such as parity and multiplicitywhich was controlled by limiting the study to first single births. Therefore it must beacknowledged that the results of the study would be influenced by changes in the agesof the mothers in the comparison group. Older mothers would tend to obscure thedifferences because their infants are at higher risk for various birth outcomes and sub-groups of younger mothers might enhance the differences for certain outcomes.The fact that the study only considered first, single births also limits it's generalizability.As discussed earlier, birth outcomes are affected by multiplicity, parity, an interactionbetween MA and parity and the closely related issue of interpregnancy interval.Population based research does not require statistical analyses in that all individuals areincluded that can be included in the study are and the results represent the reality forthat population. Although this was a population based study which included allmothers under 35 years residing in the study area, it was decided to statistically analyzethe data because the data were seen as possibly a subset of births to all urban Canadianwomen under 35 years. The large size of the data set and the large number of analysesdone increased the chance of obtaining significant results by chance. The "p" valuesPage 149were not adjusted to compensate for this fact. Additionally, some of the results thatwere identified as being statistically significant might not have been clinically signifi-cant. A case in point is low birth weight. Although it was identified as beingsignificantly associated with both young MA and SES in this study and the literature,it was noted in a preliminary analysis that the difference in mean birth weight for infantborn to the two groups of mothers was 42 grams while the difference in mean birthweights between NPQs 1 and 5 was 90 grams. These small differences were probablynot medically significant. The small mean birth weight difference between the two MAgroups may help explain why young MA was not a significant risk factor for small forgestational age births.The CMH chi-squared tests used in this study did allow for the identification ofsignificant independent associations between each of young MA and SES and thevarious birth outcomes. Unfortunately, except in the 2 X 2 case it produces onlyhypothesis tests, no effect estimates. In this study odds were estimated for the NPQsby comparing the information in NPQs 5 and 1. Also, this method only estimates aneffect for one variable at a time. A logistic regression analysis would be required toproduce simultaneous estimates of the effect of young MA and SES.Despite these limitations the study did provide a confirmation of the significantrelationship between MA and SES and the independent effect of these two variables onvarious birth outcomes. Given these relationships there are a number of policy andresearch recommendations that can be made.Page 150IMPLICATIONSResearch ImplicationsThis study confirmed Wilkins's 2 earlier finding that within Canada there is an associa-tion between maternal age and poverty. There was a significant increase in thepercentage of all births to women under 20 years of age as the percentage of childrenliving in poverty in the mother's neighborhood at the time of birth increased. Unfortu-nately there has not been a study in B.C. or Canada, such as that undertaken by Smith 4,which considered the effect of SES on total pregnancies and abortions as well as birthsto young mothers. It is only with this information that we will be able to gain a clearerunderstanding of the impact of poverty. Is there a SES gradient in pregnancy ratesamong young women and if there is, is it because of a gradient in attitudes aroundsexual activity or access to and use of contraceptives? Is a pregnancy gradientresponsible for the live birth gradient or is it that there is a SES gradient in availabilityor use of abortion services? Answers to these questions are critical to being able toidentify those areas that require program and policy support and to develop effectivestrategies to reduce both the rates of pregnancies and births to teen mothers.In addition to its association with the increased proportion of all birth occurring toyoung mothers, low SES was also found to associated with increased rates of low birthweight, small for gestational age and congenital anomalies among first single births tomothers less that 35 years. The odds ratio for these increased risks were 1.4 in all threecases. Although the confidence limits were all above one (refer to Chapter Five forvalues), the number of analyses carried out it undermines somewhat the confidence onecan have in these results. Further research to confirm these findings would be helpfulgiven the importance of these birth outcomes as risk factors for infant mortality andPage 151future health. The relationship between congenital anomalies and SES is of particularinterest because of the concerns about the quality of the HSR data that were includedin the study and the lack of a clear understanding about the possibility of a SES gradientin reporting congenital anomaliesas well as the recent finding that folic acid supple-ments can prevent neural tube defects13 which represent a substantial proportions ofthe congenital anomalies in B.C.. Given the existence of the HSR in B.C.the province isin the position of being able to make this unique data base available to address somecritical questions about the association between overall congenital anomaly rates aswell as specific anomalies and SES. If indeed rates of neural tube defects increase withdecreasing SES there could be consideration of the feasibility of developing supportivepublic health policies such as the provision of folic acid supplements to poor women.There are also research issues related to the relationships observed in this studybetween MA and birth outcomes. Young MA was found to be significantly associatedwith low birth weight, infant mortality, and postneonatal mortality but not withneonatal mortality. It is tempting to speculate that the statistically elevated rates of lowbirth weight for young mothers (odds ratio = 1.3) were not clinically significant,particularly in light of the small mean birth weight difference between infants born toboth groups of mothers. Since it is now universally accepted, as well as confirmed ina Canadian context 14, 15, that low birth weight is a dominant factor determiningneonatal mortality it could then be suggested that this small difference in low birthweight rates between the two groups of mothers resulted in the statistically insignificantdifference in the neonatal mortality rates. It could then also be suggested that thesignificant difference in infant mortality rates between the two groups of mothers wasa function of the increased postneonatal rates for the teen mothers which was driven byevents in the postnatal environment. These speculations need to be considered moreseriously because, if correct, confirm the suggestion that infants born to young mothersPage 152are not at increased risk because of the biological immaturity of their mothers butbecause of social, lifestyle and other risk factors.As mentioned earlier (Chapter Five), the substantially elevated rates of postneonatalmortality among infants born to young mothers require further attention. It isimportant to know if the increased risk to infants born to teen mothers is due to causessuch as sudden infant death syndrome or communicable diseases. It is also importantto have an understanding of who those teens are who become pregnant and choose tokeep their babies. Are these young mothers more likely than other mothers to smoke,abuse drugs and alcohol or have elevated rates of other risk taking behaviors? What isthe nutritional status of these young women? What is their level of self-esteem and whatabout their ability to manage stress? Or are these young mothers more vulnerable thanolder mothers to increases in economic, social and other stresses after the birth of theirchild. The psychological and emotional as well as physiological well-being of themother is as important in delivering a healthy infant as in being prepared and able tolook after the child in the months after birth. Continued support for new young mothersafter the birth of that child appears to be an area that requires considerably more energythan it is presently receiving. Determining what form(s) this support should take willrequire interacting with these mothers and those who presently support them toincrease our understanding of the "problems".A closely related issue is that of the lack of an observed SES gradient in postneonatalmortality rates for all mothers included in the study. Further research is required toestablish whether there is or is not a gradient for all mothers.Page 153Other Implications In this study low SES was found to be associated with increases in the proportion of birthoccurring to young mothers, marginal increases in low birth weight, infant mortalityand congenital anomaly rates and possibly increases in postneonatal mortality ratesthrough an association with young MA. The importance of poverty and other socialfactors as risk factors for poor health continue to need to be acknowledged andaddressed.Although it was possible to consider the impact of SES on birth outcomes with this studydesign, the study had to be limited to urban areas and the SES variable was an ecologicalone. For these and other reasons it is suggested that some consideration be made torevising birth registration forms to allow for the collection of valuable sociodemographicinformation that will allow for the direct measurement of the SES of individuals.Serious consideration should also be given to the possibility of using a design similarto that employed in this study to evaluate the health impact of social, economic andother policies and to routinely track the health of the population of B.C. and various sub-populations within the province.ConclusionAlthough the discussin of this study has by no means been exhaustive it does covermany of the major areas of interest.British Columbia is very fortunate to have high quality data bases of routinely collecteddata that have enormous potential for research and health status evaluation. EveryPage 154effort should be made to maintain them as well as to revise the data collections formsto include some SES variables. Making the information included in these data basesavailable for research and evaluation purposes should also be a priority.The findings of this study with respect to the effect of young MA on birth outcomes inB.C. were very consistent with the literature for other western, developed countries. Itmust be acknowledged that young MA is an independent risk factor for elevated ratesof low birth weight, infant mortality and postneonatal mortality in urban B.C. For allof these associations, but especially for postneonatal mortality, there is a need forfurther research as well as policy and program development.The results of the study confirmed, in a B.C. context, the increasing proportion of birthsoccurring to young women with decreasing SES. There was less consistency betweenthis study and the literature with respect to the independent association of SES on birthoutcomes. This should not be seen as a reason for not acknowledging the association oflow SES with elevated rates of low birth weight, small for gestational age, congenitalanomalies and postneonatal mortality (for births to young mothers only) observed inthis study. It is important, however, that the impact of SES be investigated further; ifpossible in studies that include SES information by individuals as opposed toneighborhoods.Page 155REFERENCES1.Wilkins R, Adams 0, Brancker A. Changes in mortality by income in urban Canadafrom 1971 to 1986. Health Reports 1989;1(2):137-74.2.Wilkins R, Sherman GJ, Best P. Birth outcomes and infant mortality by income inurban Canada, 1986. Health Reports 1991;3(1):7-31.3.Thomson M. Association between mortality and poverty. BC Med J 1990;32(8)(Aug.):337-38.4.Smith T. Influence of socioeconomic factors on attaining targets for reducing teenagepregnancies. Br Med J 1993;306:1232-5.5.Pampel F, Pillai V. Patterns and determinants of infant mortality in developednations, 1950-1975. Demography 1986;23(4):525- 41.6.Quine S. Problems in comparing findings on social class cross-culturally-applied toinfant mortality (Australia and Britain). Soc Sci Med 1990;30:1283-8.7.Wigle D, Mao Y. Mortality by income level in urban Canada. Minister of NationalHealth and Welfare, Ottawa 1980;8.Starfield B. Family income, ill health and medical care in the U.S.. J Public HealthPolicy 1982;3:244-259.9.Starfield B. Child health and socioeconomic status. Am J Public health 1982;10.Collins J, Richard J. The differential effect of traditional risk factors on infantbirthweight among blacks and whites in Chicago. Am J Public Health1990;80(6):679-81.11.Liberatos P, Link B, Kelsey J. The measurement of social class in epidemiology.Epidemiologic Reviews 1988;10:87-121.12.Antonovsky A, Bernstein J. Social class and infant mortality. Soc Sci Med 1977;11:453-70.13.Werler M, Shapiro S, Mitchell A. Periconceptional folic acid exposure and risk ofoccurrent neural tube defects. JAMA 1993;269:1257-61.14.Silins J. Risk factors for perinatal mortality in Canada. Can Med Assoc J1985;133(15):1214-219.15. Dougherty G. Socioeconomic differences in pediatric mortality in urban Canada:1981 [Thesis]. McGill University, 1986.Page 156APPENDIX APage 157APPENDIX BANALYSES OF TEEN MATERNAL AGE SUB-GROUPSThere are many instances in the literature when birth outcomes to sub-groups ofteenaged mothers are considered. This may, at least in part, be prompted by thedifferences in the rates of various birth outcomes to infants born to young women ofdifferent ages. For example, Lee and Corpusl presented very low birth rate, low birthrate and fetal mortality rates for American women aged less than 15 years and 15through 19 years between 1950 and 1988. All three rates were consistently higher for theyounger group of women. Since the cause of these differences in American rates is notclearly understood a preliminary investigation into sub-groups of teen births in thisstudy in British Columbia was undertaken. Both biological and social factors wereconsidered in this analysis into the feasibility of creating sub-groups of teenagedmothers. It was decided to split mothers 15 years and under from other teen mothersbecause students are only permitted to withdraw from school at sixteen years of age inBritish Columbia. Contact with the school system, its counseling services etc mightimpact the outcome of the young woman's pregnancy in some way. Additionally, theB.C. Ministry of Social Services deals differently with social assistance applicationsfrom teenagers of different ages 2. Applicants aged 16 years and under are alwaysreferred to a Family and Child Services social worker who is responsible for developinga holistic plan for the individual's care. This plan deals with more than financialassistance and always includes an attempt to involve the young women's family.Women aged 17 and 18 years, for whom there are no child protection concerns, meetwith a rehabilitation officer as part of the process of obtaining income assistance. Thisofficer is responsible for establishing a schooling/retraining plan with the applicant ifthis is appropriate. At 19 years individuals can apply for Guaranteed Available IncomePage 158for Need (GAIN) benefits as an adult. This application process does not require anyinteraction with either a social worker or rehabilitation officer and leaves the individualvery much on their owna.When births to teenaged mothers in this study were divided into these three groups,there were 89 first single live births to mothers 15 years and younger, 1,509 to thosebetween 16 and 18 years, and 1,140 to 19 year olds. The total number of birth outcomesfor each of these three groups is presented in Table B.1. The low numbers of births tomothers aged 15 years and younger and the large number of outcomes without data ledto the decision to amalgamate this group of mothers with those aged 16 to 18 years.Table B.1Birth Outcomes to Sub-Groups of Teenaged MothersBirth Outcomes^_. 15 Years^16 - 18 Years^19 YearsStillbirths 1^10^9Low birth weight^7 88 75Very low birth weight^2 16^13Small for gestational age 6^136 134Infant mortality^0 20^14Neonatal mortality^0 5 6Postneonatal mortality^0^15^8Table B.2 presents the number of live births to teen mothers in each of the neighborhoodpoverty quintiles (NPQs) once the youngest two groups of mothers were combined.The number of births to women aged 18 years and under was similar to the number towomen aged 19 years. Both groups also had fairly similar distributions of births withinthe various NPQs. The lowest percentage of births was in NPQ 1. NPQs 2 through 4had slightly higher percentages while in NPQ 5 they were substantially larger. Froma. Janet Wright, supervisor, B.C. Ministry of Social Services, Nanaimo, personal communication, August24, 1993Page 159Table B.3 it is noted that the overall rates of stillbirths, low birth weight births and infantmortality for mothers 18 years and under and those aged 19 years varied by less than10%. Although the observed number of birth outcomes by NPQ suffered from randomfluctuations because of the small denominators the general trends in the rates of thevarious birth outcomes between the five NPQs was similar for most outcomes. Thepattern in the rates of infant mortality was an exception primarily because of the twoquintiles with no infant deaths among mothers aged 19 years.Table B.2Births to Sub-Groups of Teenaged Mothers by Neighbourhood Poverty QuintilesQuintile 518 Years 19 Years# (column %) # (column %)1 261 (16) 184 (16)2 293 (18) 200 (18)3 287 (18) 236 (21)4 319 (20) 206 (18)5 438 (27) 314 (28)Total 1598 1140Because of this general similarly in the distribution of births and most birth outcomesas well as the low numbers of very low birth weight births and infant, neonatal andpostneonatal deaths in some NPQs it was decided to combine all of the births toteenaged mothers into one group. The study analyses were carried out comparingmothers 19 years and younger with those between 20 and 34 years at the time of the birthof their infant.Page 160Table B.3Birth Outcomes to Sub-Groups of Teenaged Mothersby Neighborhood Poverty QuintilesQuintilea) Stillbirth12345Total5.. 18 YearsNumber3222211Rate*11.56.87.06.34.66.919 YearsNumber^Rate*2^10.92 10.03^12.70 0.02 4.89^7.3b) Low Birth Weight1^12 (2t) 46.0 5 (1t) 27.22 13 (2t) 44.4 13 (3t) 65.03 18 (4t) 62.7 10 (1t) 42.44 25 (5t) 78.4 19 (3t) 92.25 27 (5t) 61.6 28 (5t) 67.8Total 95 59.5 75 60.5c) Small for Gestational Age1^14 53.6 21 114.12 27 92.2 28 140.03 29 101.1 16 67.84 31 97.2 22 106.85 41 93.6 47 113.8Total 142 88.9 134 108.2d) Infant Mortality1^4 (1) [3] 15.3 0 (0) [0] 0.02 1 (0) [1] 3.4 6 (2) [4] 30.03 3 (1) [2] 10.5 0 (0) [0] 0.04 5 (0) [5] 15.7 1 (1) [0] 4.95 7 (3) [4] 16.0 7 (3) [4] 17.0Total 20 12.5 14 11.3* per 1000 live birthst number of very low birth weightsnumber of neonatal deaths[] number of postnatal deathsPage 161REFERENCES1.Lee K, Corpuz M. Teenage pregnancy: Trend and impact on rates of low birth weightand fetal, maternal, and neonatal mortality in the United States. Clin Perinatol1988;15(4):929-42.2.British Columbia Ministry of Social Services. Programs for Independence Manual,Volume 1. British Columbia: 1993.Page 162APPENDIX CTable C.1Chi-Square Test Comparing the Incidence of Births by Maternal Age Groups and NPQsCMA/CAs Chi-Square d.f. ProbabilityAll five CMA/CAs 105.81 1 0.000Vancouver, Victoria, Kamloops 110.56 1 0.000Kelowna, Prince George 5.17' 1 0.270Table C.2ANOVA Comparing Birth Weights by Maternal Age Groups and NPQsCMA/CAsi) by maternal age groupsF-value d.f. ProbabilityAll five CMA/CAs 10.48 1 0.001Vancouver, Victoria, Kamloops 7.21 1 0.007Kelowna, Prince Georgeii) by NPQs5.98 1 0.015All five CMA/CAs 11.68 4 0.000Vancouver, Victoria, Kamloops 12.23 4 0.000Kelowna, Prince George 0.60 4 0.668Table C.3ANOVA Comparing Gestational Ages by Maternal Age Groups and NPQsCMA/CAsi) by maternal age groupsF-value d.f. ProbabilityAll five CMA/CAs 5.22 1 0.022Vancouver, Victoria, Kamloops 3.97 1 0.046Kelowna, Prince Georgeii) by NPQs3.61 1 0.058All five CMA/CAs 1.61 4 0.000Vancouver, Victoria, Kamloops 8.35 4 0.000Kelowna, Prince George 1.82 4 0.122

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.831.1-0086285/manifest

Comment

Related Items