UBC Faculty Research and Publications

Assessing the external validity of model-based estimates of the incidence of heart attack in England:… Scarborough, Peter; Smolina, Kate; Mizdrak, Anja; Cobiac, Linda; Briggs, Adam Nov 3, 2016

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

Item Metadata


52383-12889_2016_Article_3782.pdf [ 713.37kB ]
JSON: 52383-1.0362087.json
JSON-LD: 52383-1.0362087-ld.json
RDF/XML (Pretty): 52383-1.0362087-rdf.xml
RDF/JSON: 52383-1.0362087-rdf.json
Turtle: 52383-1.0362087-turtle.txt
N-Triples: 52383-1.0362087-rdf-ntriples.txt
Original Record: 52383-1.0362087-source.json
Full Text

Full Text

RESEARCH ARTICLE Open AccessAssessing the external validity of model-based estimates of the incidence of heartattack in England: a modelling studyPeter Scarborough1*, Kate Smolina2, Anja Mizdrak1, Linda Cobiac3 and Adam Briggs1AbstractBackground: The DisMod II model is designed to estimate epidemiological parameters on diseases wheremeasured data are incomplete and has been used to provide estimates of disease incidence for the Global Burdenof Disease study. We assessed the external validity of the DisMod II model by comparing modelled estimates of theincidence of first acute myocardial infarction (AMI) in England in 2010 with estimates derived from a linked datasetof hospital records and death certificates.Methods: Inputs for DisMod II were prevalence rates of ever having had an AMI taken from a population healthsurvey, total mortality rates and AMI mortality rates taken from death certificates. By definition, remission rates werezero. We estimated first AMI incidence in an external dataset from England in 2010 using a linked dataset includingall hospital admissions and death certificates since 1998. 95 % confidence intervals were derived around estimatesfrom the external dataset and DisMod II estimates based on sampling variance and reported uncertainty inprevalence estimates respectively.Results: Estimates of the incidence rate for the whole population were higher in the DisMod II results than theexternal dataset (+54 % for men and +26 % for women). Age-specific results showed that the DisMod II resultsover-estimated incidence for all but the oldest age groups. Confidence intervals for the DisMod II and externaldataset estimates did not overlap for most age groups.Conclusion: By comparison with AMI incidence rates in England, DisMod II did not achieve external validity forage-specific incidence rates, but did provide global estimates of incidence that are of similar magnitude tomeasured estimates. The model should be used with caution when estimating age-specific incidence rates.Keywords: Myocardial infarction, Incidence, Validity, Modelling, DisModBackgroundFor many diseases, estimates of incidence and preva-lence are often incomplete or based on different datasources making it difficult to compare results [1]. Reli-able and representative population level epidemiologicaldata are needed to inform health care policy and to sup-port decision making processes in health service plan-ning and delivery, and are essential to cost effectivenessanalyses and burden of disease calculations. Due to gapsin directly measured data, models have been establishedthat can estimate incidence and prevalence rates of dis-eases. DisMod II is such a model. It uses available epi-demiological data about a condition to estimate missingdata on incidence, prevalence, remission and case fatalityrates as applicable [2, 3]. Originally developed for theGlobal Burden of Disease studies [4], DisMod II is freelyavailable for use and can be downloaded from http://www.epigear.com/index.htm.The DisMod II model is a multistate life table that fullydescribes the epidemiological progress of a single disease byexploiting the fact that parameters such as incidence, preva-lence, remission, case fatality and mortality rates are not in-dependent variables. By solving a set of differentialequations, Dismod II can estimate age-specific incidence,* Correspondence: peter.scarborough@dph.ox.ac.uk1British Heart Foundation Centre for Population Approaches toNon-Communicable Disease Prevention, Nuffield Department of PopulationHealth, University of Oxford, Oxford, UKFull list of author information is available at the end of the article© The Author(s). 2016 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Scarborough et al. BMC Public Health  (2016) 16:1135 DOI 10.1186/s12889-016-3782-6prevalence or case fatality rates for a disease, given suffi-cient data on the other (for example, with input data ofage-specific prevalence, case fatality and mortality data fora disease, Dismod II will estimate the age-specific incidencerate for the disease). The model operates by calculating thenumber of people in each of three states: healthy, diseasedand dead at any age. Within the model, there are twocauses of death, either from the disease or from ‘all other’causes, that are assumed to be independent. There are fourtransition hazards which are age specific (assumed to beconstant within a 1-year age interval): incidence, remission,case fatality, and the “all other mortality” hazard. The inputdata for the model are age and sex-specific estimates ofthree out of the four parameters described above for a givenpopulation, and a complete set of parameters (smoothedfrom the original or estimated from the original parame-ters) is the output of the model.Ischaemic heart disease (IHD) is the most commoncause of death in the UK [5] and acute myocardial infarc-tion (AMI) is coded on death certificates as the cause ofapproximately one third of all deaths from IHD [6]. AMImortality and prevalence data for having had an AMI inEngland are routinely collected by the Office of NationalStatistics (ONS) and the Health Survey for England (HSfE)[7] respectively. However, until recently, there have beenno published comprehensive, population-based nationallevel estimates of AMI incidence [8] and these recent esti-mates are unlikely to be routinely updated. Incidence ofAMI is important to researchers and public health policymakers because it serves as an indicator of the effective-ness of preventative measures and management of riskfactors through health promotion and other public healthinitiatives. Without a tool that allows routinely updatedestimates of incidence, measurements of the current bur-den of AMI in England have significant limitations.This study assesses the external validity of DisMod II es-timates of the incidence of first AMI in England by com-parison with estimates generated from the dataset used tosupport a recent series of related papers [8–10]. Establish-ing the external validity of the modelled estimates woulddemonstrate that the DisMod II model could be used as atool for regularly updating estimates of the incidence ofAMI—data that are not routinely collected in England. Itwould also help to establish confidence in studies of non-communicable disease that use DisMod II to estimate in-cidence, such as in modelling studies [11], and studiesestimating disease burden where incidence data are scarceor not regularly updated [12, 13]. Since the 2010 iteration,the Global Burden of Disease study [4] results have beenbased on an updated (but closely related) version of theDisMod model, which is not freely available for use. As-sessments of the external validity of DisMod II offer in-sights into the assumptions used for this important andwidely used global project.MethodsDisease definitionsThe exact definitions that were used in this paper forthe condition under investigation are provided in Table 1.In this table, the model outcome that was comparedwith the external datasets is prefixed with OUTCOME.The remaining model definitions are descriptions of thetheoretical measures that are consistent with the out-come of interest, and the corresponding external datadefinitions describe where the model input data weretaken from, and how similar they are to the theoreticalmeasures.Outcome measuresFor AMI, measured data on prevalence, excess mortalityand remission were used to create modelled data on in-cidence. The modelled incidence data were comparedagainst measured estimates from the dataset used forsimilar results published in the peer reviewed literaturewhich were independent of all input data to the DisModII model [8]. The difference between the modelled andthe measured estimates was calculated and displayed forall ages. Differences were recorded in percentages, withthe measured data as the baseline.Data sourcesModel inputsSingle-year population estimates by sex were taken fromthe Office for National Statistics (ONS), for the year2010. Mortality data where AMI was recorded as theunderlying or contributory cause of death (ICD-10 codesI21-I22) by sex and 5 year age groups were provided byONS. Part of the modelling process for DisMod II is tointerpolate data into single year estimates, in order tohave smooth rates for all the consistent modelled out-puts. For the mortality data, this was achieved using aTable 1 Model definitions, input data sources and external datasourcesModel definition(relevant year inbrackets)External data definition(data source in brackets)Incidence OUTCOME: Incidenceof first AMI (2010)Incidence of first AMI since 1998(linked hospital episodes andmortality statistics, 2010).Prevalence Prevalence of everhaving had an AMI(2010)Prevalence of ever having had adoctor-diagnosed AMI (HealthSurvey for England, 2011).Remission Zeroa ZeroExcessmortalityExcess mortality dueto first AMI (2010)Death where AMI is includedanywhere on the death certificate(ONS mortality statistics, 2010).Abbreviations: ONS Office for National Statistics, AMI AcuteMyocardial Infarctiona Remission is zero because the prevalence data measures people who haveever been diagnosed with AMIScarborough et al. BMC Public Health  (2016) 16:1135 Page 2 of 8cubic spline interpolation on a log transformation of theoriginal data.Age and sex-specific estimates of the prevalence ofever having had a doctor’s diagnosis of heart attack weretaken from the Health Survey for England (HSfE) 2011[7]. The HSfE covers all of England and is a nationallyrepresentative sample of those residing at private resi-dential addresses. In the 2011 survey, a sample of 8610individuals between aged 16 and 99 was recruited, witha household response rate of 66 %. Residents of carehomes, prisons and the homeless were excluded whichis thought to contribute to <2 % of the population.Those who were unable to consent and those who wereunable to understand the questions or formulate answers(due to language difficulties, disability or mental illness)were termed non-responders. Figures 1 and 2 show themodel input data (both raw and smoothed) by age formen and women.External validationA recent series of papers examined AMI incidence, casefatality, survival, and trends in event rate, case fatalityand mortality between 2002 and 2010 in a populationbased study using person linked routine hospital andmortality data in England [8–10]. Hospital episode sta-tistics provide information on all patients admitted tohospital whose care is funded by the English NHS. AnAMI event was classified as an emergency hospital ad-mission with primary diagnosis of AMI and a length ofstay of more than 1 day for someone discharged alive, ora death with acute myocardial infarction coded as theunderlying cause of death on the death certificate. Fatalcases were defined as those where AMI was coded asthe underlying cause of death on the death certificate orany death that occurred within 30 days of an admissionfor AMI which was assumed to relate to the same event.Bespoke analyses using the same dataset and methodshave been used to generate measures of the incidence offirst AMI in England in 2010, which are used as the ex-ternal comparator in these analyses. Because the linkeddataset used for these analyses only includes hospital ep-isodes and deaths from 1998 onwards, these measuresare strictly of the incidence of first AMI since 1998 inEngland in 2010 (i.e. a person who had a first heart at-tack before 1998 and then a second heart attack in 2010would be counted as a first heart attack in our dataset).DisMod II settingsPrevalence, remission (set as zero), and disease specificmortality were used as inputs in the DisMod II model.Population numbers and mortality rates for England 2010were also used. The prevalence and mortality data werefitted with a sigmoid mathematical curve in order tosmooth out the data points which have relatively large agegroup widths. Analyses were conducted with and withoutaccounting for trends in AMI incidence rates. Whentrends were applied, an annual change in incidence rate of−4 % and an annual change in case fatality of −1 % for theprevious 28 years was incorporated into the model. Thesetrends are based on a study of trends in CHD since 1978[14]. For a sensitivity analysis, larger trends were also ap-plied that were taken from an analysis of incidence andcase fatality rates over the previous 10 years in England[9]—these were −5 % for incidence and −4 % for case-fatality for both men and women. Application of a trendin DisMod II assumes that each age group is changing atthe same rate as the overall trend. The DisMod II uncer-tainty analysis was conducted to assess uncertainty aroundthe modelled incidence estimates. Here, DisMod II con-ducts a parametric bootstrapping exercise, where theFig. 1 The data inputs used for the modelling exercise: prevalence of acute myocardial infarction by age and sex, actual data and smootheddata. The blue lines show actual and smoothed data for men and the red lines show actual and smoothed data for womenScarborough et al. BMC Public Health  (2016) 16:1135 Page 3 of 8input data are allowed to vary according to a specified dis-tribution. We allowed the estimates of prevalence of hav-ing had a heart attack to vary with a normal distributionaccording to the uncertainty reported in the HSfE. We didnot specify any uncertainty in either the mortality or theremission estimates. Due to lack of computing power, itwas not possible to conduct an uncertainty analysis forthe DisMod II runs where trends were assumed.The complete set of results and the data used to runthe DisMod II model for these analyses are availablefrom the authors upon request.ResultsFigures 3 and 4 show the age-specific estimates of inci-dence of AMI for men and women separately. The greenlines show the estimates from the external dataset withaccompanying 95 % confidence intervals and the bluelines show the estimates from DisMod II with accom-panying 95 % credible intervals (i.e. the distance betweenwhich 95 % of the iterations of the uncertainty analysisfell). The red line shows the estimates from DisMod IIwhere trends in incidence and case fatality have been in-corporated. The figures show that, for both men andFig. 2 The data inputs used for the modelling exercise: excess mortality rates for acute myocardial infarction, actual data and smoothed data. Theblue lines show actual and smoothed data for men and the red lines show actual and smoothed data for womenFig. 3 The incidence rate per 100,000 of first acute myocardial infarction in males in England in 2010. The green lines are estimates from theexternal dataset with 95 % confidence intervals. The blue lines are estimates from DisMod II with 95 % credible intervals that do not account fortrends in incidence and case fatality. The red line is the estimate from DisMod II that does account for trends in incidence and case fatalityScarborough et al. BMC Public Health  (2016) 16:1135 Page 4 of 8women, the DisMod II estimates tend to over-estimatethe incidence of AMI at younger age groups and under-estimate for the oldest age groups (greater than 85 inmen, and greater than 75 in women). For both men andwomen, for the majority of age groups the confidenceintervals between the external dataset and the DisMod IIestimates did not overlap, suggesting poor external valid-ity of the DisMod II estimates as AMI incidence ratesincrease. Applying trends in incidence and case fatalityrates to the DisMod II estimates resulted in reduced in-cidence rates at all ages than without adjustment and ans-shaped curve for AMI incidence in women where themodel estimated a lower incidence in women in their70s compared to those in their 60s. The sensitivity ana-lysis with larger trends in incidence resulted in furtherreductions in the estimates of MI incidence for bothmen and women, but the s-shaped curve in women didnot persist (data not shown). Table 2 shows how thenon-trended DisMod II estimates of incidence rates forthe total population produced over-estimates of the inci-dence of MI by approximately 54 % in men and 26 % inwomen, but this masks larger differences for age-specificsubgroups, including under-estimates of incidence in theoldest age groups.DiscussionThis study assessed the external validity of DisMod II esti-mates of the age-specific incidence rate of AMI in Englandin 2010. Although the modelled estimates and the externaldataset resulted in incidence rates in the whole populationthat were of similar magnitude, the age-specific rates werenot consistent with the external dataset; they over-estimated rates in younger age groups and under-estimated rates in the oldest age groups. Incorporatingtrends in the incidence and case fatality of AMI in theDisMod II estimates resulted in estimated and observedAMI incidence being more closely matched at youngerage groups than without trends but with more divergentresults at older ages, including for one set of results theimplausible scenario of women in their 70s having a lowerincidence of AMI than women in their 60s. This study im-plies that DisMod II is not an appropriate source of age-specific estimates of the incidence of AMI for Englandand future estimates should be based on measuredoutcomes.Modelled estimates of the burden of disease in differ-ent populations are extremely important for policymakers and health care planners. They allow for an as-sessment of where to direct scarce resources in terms oftreatment and care, and assist in planning for futurehealth care requirements. They can also be used as thebasis for comparing the burden of disease that is attrib-utable to different behavioural risk factors (e.g. the glo-bal comparable risk assessment exercise, Global Burdenof Disease [3, 15], or modelling studies to assess the ef-fectiveness or cost-effectiveness of public health inter-ventions [11, 16, 17] which in turn influences howpublic health resources are directed).The DisMod II model is the most recent freely avail-able version of the model that is used for the GlobalFig. 4 The incidence rate per 100,000 of first acute myocardial infarction in females in England in 2010. The green lines are estimates from theexternal dataset with 95 % confidence intervals. The blue lines are estimates from DisMod II with 95 % credible intervals that do not account fortrends in incidence and case fatality. The red line is the estimate from DisMod II that does account for trends in incidence and case fatalityScarborough et al. BMC Public Health  (2016) 16:1135 Page 5 of 8Burden of Disease project and it is currently being usedby non-communicable disease scenario models [18]which have been built in order to estimate the popula-tion health impact of public health interventions. For ex-ample, Cecchini et al. (2010) [11] used DisMod II toestimate the incidence of various cardiovascular diseasesand risk factors when simulating the possible effects onhealth of different diet and physical activity interven-tions. Previous studies have also used DisMod II to esti-mate disease incidence from prevalence, remission, anddisease specific mortality rates where incidence data arescarce. For example, Johnston et al. (2009) [13] esti-mated the global stroke burden and used DisMod II toestimate stroke incidence from mortality and case fatal-ity data for countries where only partial data exist. Rehmet al. (2009) [12] used prevalence, relative risk of mortal-ity, and remission rates to estimate the global incidenceof alcohol-use disorders.Wherever possible, the validity of the DisMod II mod-elled estimates should be assessed by comparison withactual measured data from the population of interest.This study utilised recent results estimating the inci-dence of AMI in England using linked hospital episodestatistics and death certificates, which captures the vastmajority of incident AMIs that occur in England [9] andas such represents a robust dataset for assessment of ex-ternal validity. The present study complies with all ofthe ‘best practices’ identified in the ISPOR modellingguidelines for external validation studies [19].The DisMod II software provides users with a variety ofchoices about how the input data should be manipulatedbefore the outputs are calculated. These choices include:the method used to interpolate input data to single yearestimates (cubic spline or polynomial methods); the shapeof the curve to fit the smoothed age-related input data(linear, quadratic, sigmoid, simple exponent or polyno-mial); and whether or not to allow for time trends in theinput data. These three choices alone would generatetwenty different sets of results, which we did not chooseto display—rather we chose the settings that best suitedthe epidemiology of AMI. The alternative of using crudeinput data was not preferable due to inherent problemswith the crude data. For example, the prevalence dataused in the analyses were taken from the Health Surveyfor England [7], where rates are reported for 10 year agegroups. Using the crude data would lead to large stepchanges in prevalence as age increases, which resulted inimplausible shapes to the modelled incidence data. Inpractice, changing the specific selections for manipulatingthe input data did not improve the comparison betweenthe DisMod II estimates and the external dataset. We de-cided to report the results both for when trends in inci-dence and case fatality were applied and when they werenot as cardiovascular disease rates are decreasing rapidlyin the UK and have been for some time [20]. This is im-portant for the DisMod II model, as the method used forsolving the differential equations assumes a ‘steady state’for the disease being modelled (i.e. that age-specific preva-lence, incidence and mortality rates for the disease arestatic). This allows DisMod II to assume that the preva-lence of AMI at age t equals the prevalence of AMI at aget-1 plus incident cases minus dying cases. But input dataare all taken from the same year (y), whereas data for aget-1 should be taken from y-1. This is not a problem if thedata in year y-1 are equal to those in year y, but if thereare trends in the data (is the case for falling AMI inci-dence rates), this is not the case. For AMI, the incidenceand case fatality rate trend data that we applied were takenfrom the British Regional Heart Study, a cohort study car-ried out in British men aged 40–59 at entry between 1978and 2000 [14]. This study only examines trends in coron-ary heart disease in men, however other studies report de-clines of similar magnitude in both men and women [21,22]. A sensitivity analysis which applied more recent datathat were specific to both men and women did not im-prove the validity of the modelled results. It is not possibleto apply age-stratified trends in DisMod II and thereforethe same trends were applied across all age groups andthe analyses reported here. However, whilst overall therehas been a decline in incidence rates, this decline varies byage with the lowest rate of decline occurring in those agedTable 2 Number of first AMI events in England, 2010Number Incidence rate per 100,000(and 95 % confidenceintervals)ExternalDatasetDisMod IIa External Dataset DisMod IIa %differenceMen0–29 105 1588 1 (1, 1) 16 141230–54 7035 12,558 79 (77, 81) 141 7955–64 8446 13,919 282 (276, 288) 465 6565–74 9674 16,703 451 (442, 460) 779 7375–84 10,451 14,147 828 (812, 844) 1121 3585+ 5978 5487 1521 (1483, 1559) 1396 −8Total 41,689 64,401 162 (160, 163) 250 54Women0–29 38 88 0 (0, 1) 1 13230–54 1752 3091 19 (19, 20) 34 7655–64 2573 6664 83 (79, 86) 214 15965–74 4817 9534 205 (200, 211) 407 9875–84 8640 8490 521 (510, 532) 512 −285+ 8911 5903 1107 (1084, 1130) 733 −34Total 26,731 33,769 101 (100, 102) 128 26aThe DisMod II results are those estimated without the application oftrend dataScarborough et al. BMC Public Health  (2016) 16:1135 Page 6 of 885 and over [9]. Both the mis-match between the externaland modelled datasets and reported declines in AMI inci-dence are age-specific, making this a likely candidate forthe failure of the model to produce externally valid esti-mates of the incidence of AMI in England. However, with-out further investigation using a model that canincorporate age-specific trends in incidence and mortalityit is not possible to prove this assertion.Another important limitation of our validity assess-ment is our input data for mortality. The ideal data forthe DisMod II model would be estimates of the in-creased all-cause mortality rate for people who have pre-viously had a heart attack. We were unable to find directmeasures, so we used data on all deaths where AMI wasindicated as either the primary cause or a contributingfactor. This accounts for the fact that mortality ratesfrom AMI are higher in those that have previously had aheart attack but may not include all increased mortalityrisk for other conditions (e.g. increased risk of pneumo-nia) [10, 23].We found three other studies that compared outputsfrom the DisMod II software with measured epidemio-logical data. Manuel et al. (2007) [24] used AMI incidencedata from linked hospital records and death certificates datato estimate prevalence of having had an AMI in Ontario,Canada and compared these modelled prevalence rateswith estimates derived from a population health survey.The DisMod estimates for both men and women were verysimilar to those derived from the population health survey,and were within the 95 % confidence intervals. However,the authors did not report on age-specific estimates ofprevalence, so it is unclear whether the estimates from thetwo sources showed similar age trajectories. Saha et al.(2008) [25] compared estimates of prevalence and inci-dence of schizophrenia derived from DisMod II with pairedincidence and prevalence estimates from 15 identified stud-ies. They found that the DisMod II estimates of prevalencewere generally higher than those identified in the studiesand the estimates of incidence were generally lower, but noage-specific modelled prevalence or incidence rates werereported. One third of the modelled estimates were within50 % of the estimates from the identified studies. Kruijshaaret al. (2002) [26] compared DisMod estimates of prevalencefor breast, prostate, colorectal and stomach cancer withcancer registry data from the Netherlands. Age-specificprevalence estimates were similar to observed rates forcolorectal and stomach cancer, but considerably higher forprostate and breast cancer (for some ages modelled esti-mates were two and three times higher than measuredrates, respectively). In all three studies the authors sug-gested that inadequate description of trends in the studieddisease limits the accuracy of the modelled estimates. Giventhe findings from these studies and the results presentedhere and the ongoing use of DisMod II in epidemiologicalmodelling studies, it is important that the external validityof DisMod II be further examined with different diseaseoutcomes in different populations.Another potential source of error in our analysesis the accuracy of the estimates of prevalence of hav-ing had AMI from the HSfE. Although the HSfEseries is broadly representative of the English popu-lation, it does not include residential care home set-tings in its sample structure. In 2011, around260,000 people aged 75 and over lived in residentialcare homes (about 6 % of this age group in England)[27]. Since the incidence rates for AMI are highestin older age groups, this omission may introducebias for this study. Also, the estimates from theHSfE are based on self report, which could under-estimate true rates.In the absence of measured epidemiological data, theDisMod II model can provide estimates of the incidenceof AMI, which may be helpful for health researchers,health care planners and policy makers. Our researchsuggests that estimates for England may be broadlyaccurate when applied to the whole population, but canconceal large inaccuracies when studied by age group.One reason for this inaccuracy is the ‘steady state’ as-sumption of DisMod II and as such the model should beused with caution when estimating the burden of dis-eases that are changing rapidly within the targetpopulation.ConclusionsThe DisMod II model did not replicate age-specific inci-dence rates of myocardial infarction observed in the ex-ternal dataset and therefore did not achieve externalvalidity.AcknowledgementsWe acknowledge the work of Charlotte Boughton and Premila Webster onearlier versions of this paper.FundingPS is supported by a British Heart Foundation Intermediate Basic ScienceResearch Fellowship (FS/15/34/31656). KS is supported by a BantingPostdoctoral Fellowship from the Canadian Institutes for Health Research.AM is supported by a BHF studentship (FS/13/37/30295). LC is supported bya project grant from the European Commission. AB is supported by aWellcome Trust Research Training Fellowship (102730/Z/13/Z).Availability of data and materialsThe datasets analysed during the current study are available from thecorresponding author on reasonable request.Authors’ contributionsPS conceived the study; PS, KS, AM, LC, AB designed the analyses; AM,AB, KS prepared the datsets; KS, AM conducted the analyses; PS, ABwrote the manuscript; all authors reviewed and agreed the finalmanuscript.Scarborough et al. BMC Public Health  (2016) 16:1135 Page 7 of 8Competing interestsAdam Briggs, Peter Scarborough and Kate Smolina are experts on the 2013Global Burden of Disease study. There are no further conflicts of interest toreport.Consent for publicationNot applicable.Ethics approval and consent to participateNot applicable.Author details1British Heart Foundation Centre for Population Approaches toNon-Communicable Disease Prevention, Nuffield Department of PopulationHealth, University of Oxford, Oxford, UK. 2British Columbia Centre for DiseaseControl, Vancouver, Canada. 3Burden of Disease Epidemiology, Equity andCost Effectiveness (BODE3) Programme, University of Otago, Wellington, NewZealand.Received: 22 July 2015 Accepted: 18 October 2016References1. Chan M, Kazatchkine M, Lob-Levyt J, Obaid T, Schweizer J, Sidibe M, et al.Meeting the Demand for Results and Accountability: A Call for Action onHealth Data from Eight Global Health Agencies. PLoS Med. 2010;7(1):e1000223.2. Barendregt JJ, Van Oortmarssen GJ, Vos T, Murray CJ. A generic model forthe assessment of disease epidemiology: the computational basis ofDisMod II. Popul Health Metrics. 2003;1(1):4.3. Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJL (eds.). GlobalBurden of Disease and Risk Factors. World Bank Publications:Washington, 20064. Murray C, Lopez A. Measuring the global burden of disease. NEJM.2013;369:448–57.5. Townsend N, Wickramasinghe K, Bhatnagar P, Smolina K, Nichols M, Leal J, et al.Coronary heart disease statistics 2012. British Heart Foundation: London, 2012.6. Office for National Statistics. Deaths registered in England and Wales (seriesDR), 2012. Office for National Statistics: Newport, 2013. Accessed at http://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/bulletins/deathsregisteredinenglandandwalesseriesdr/2013-10-22August 2014.7. Joint Health Surveys Unit. Health Survey for England 2011. The InformationCentre: Leeds, 2013.8. Smolina K, Wright FL, Rayner M, Goldacre MJ. Incidence and 30-day casefatality for acute myocardial infarction in England in 2010: national-linkeddatabase study. Eur J Public Health. 2012;22(6):848–53.9. Smolina K, Wright FL, Rayner M, Goldacre MJ. Determinants of the declinein mortality from acute myocardial infarction in England between 2002 and2010: linked national database study. BMJ. 2012;344:d8059.10. Smolina K, Wright F, Rayner M, Goldacre M. Long-term survival andrecurrence after acute myocardial infarction in England, 2004 to 2010. CircCardiovasc Qual Outcomes. 2012;5(4):532–40.11. Cecchini M, Sassi F, Lauer JA, Lee YY, Guajardo-Barron V, Chisholm D.Tackling of unhealthy diets, physical inactivity, and obesity: health effectsand cost-effectiveness. Lancet. 2010;376:1775–84.12. Rehm J, Mathers C, Popova S, Thavorncharoensap M, Teerawattananon Y,Patra J. Global burden of disease and injury and economic cost attributatbleto alcohol use and alcohol-use disorders. Lancet. 2009;373:2223–33.13. Johnston SC, Mendis S, Mathers CD. Global variation in stroke burden andmortality: estimates from monitoring, surveillance, and modelling. LancetNeurol. 2009;8:345–54.14. Lampe FC, Morris RW, Walker M, Shaper AG, Whincup PH. Trends in rates ofdifferent forms of diagnosed coronary heart disease, 1978 to 2000: prospective,population based study of British men. BMJ. 2005;330(7499):1046.15. Lozano R, Naghavi M, Foreman K, Lim S, Shibuya K, Aboyans V, et al. Globaland regional mortality from 235 causes of death for 20 age groups in 1990and 2010: a systematic analysis for the Global Burden of Disease Study2010. Lancet. 2012;380:2095–128.16. Cobiac L, Magnus A, Barendregt J, Crater R, Vos T. Improving the cost-effectiveness of cardiovascular disease prevention in Australia: a modellingstudy. BMC Public Health. 2012;12:398.17. Cobiac L, Ikeda T, Nghiem N, Blakely T, Wilson N. Modelling the implicationsof regular increases in tobacco taxation in the tobacco endgame. TobControl. 2014;24(e2):e154–60.18. Webber L, Mytton O, Briggs A, Woodcock J, Scarborough P, Mcpherson K, et al.The Brighton Declaration: the value of non-communicable disease modellingin population health sciences. Eur J Epidemiol. 2014;29(12):867–70.19. Eddy D, Hollingworth W, Caro J, Tsevat, McDonald K, Wong J. Modeltransparency and validation: a report of the ISPOR-SMDM modelling goodpractices task force working group—part 4. ISPOR: New Jersey, 2010.Available at http://www.ispor.org/workpaper/Modeling_Methods/DRAFT-Modeling-Task-Force_Validation-and-Transparency-Report.pdf (accessedApril 2016).20. Scarborough P, Wickramasinghe K, Bhatnagar P, Rayner M. Trends incoronary heart disease 1961–2011. British Heart Foundation: London, 2011.21. Davies AR, Grundy E, Nitsch D, Smeeth L. Constituent country inequalities inmyocardial infarction incidence and case fatality in men and women in theUnited Kingdom, 1996–2005. J Public Health. 2011;33(1):131–8.22. Tunstall-Pedoe H, Kuulasmaa K, Mähönen M, Tolonen H, Ruokokoski E.Contribution of trends in survival and coronar y-event rates to changes incoronary heart disease mortality: 10-year results from 37 WHO MONICAProject populations. Lancet. 1999;353(9164):1547–57.23. Goldacre M, Mason A, Roberts S. Myocardial infarction: an investigation ofmeasures of mortality, incidence and case-fatality. National Centre forHealth Outcomes Development: London, 2001.24. Manuel D, Lim J, Tanuseputro P, Stukel T. How many people have had amyocardial infarction? Prevalence estimated using historical hospital data.BMC Public Health. 2007;7:174.25. Saha S, Barendregt J, Vos T, Whiteford H, Mcgrath J. Modelling disease frequencymeasures in schizophrenia epidemiology. Schizophr Res. 2008;104:246–54.26. Kruijshaar M, Barendregt J, Hoeymans N. The use of models in theestimation of disease epidemiology. WHO Bull. 2002;80:622–8.27. Office for National Statistics. Changes in the older resident care homepopulation between 2001 and 2011. ONS: Newport, 2014. Accessed at http://www.ons.gov.uk/ons/rel/census/2011-census-analysis/changes-in-the-older-resident-care-home-population-between-2001-and-2011/rpt—changes-in-the-older-resident-care-home.html July 2015.•  We accept pre-submission inquiries •  Our selector tool helps you to find the most relevant journal•  We provide round the clock customer support •  Convenient online submission•  Thorough peer review•  Inclusion in PubMed and all major indexing services •  Maximum visibility for your researchSubmit your manuscript atwww.biomedcentral.com/submitSubmit your next manuscript to BioMed Central and we will help you at every step:Scarborough et al. BMC Public Health  (2016) 16:1135 Page 8 of 8


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items