The applications of capture‐recapture models to epidemiological data

Abstract

Capture-recapture methodology, originally developed for estimating demographic parameters of animal populations, has been applied to human populations. This tutorial reviews various closed capturerecapture models which are applicable to ascertainment data for estimating the size of a target population based on several incomplete lists of individuals. Most epidemiological approaches merging di erent lists and eliminating duplicate cases are likely to be biased downwards. That is, the ÿnal merged list misses those who are in the population but were not ascertained in any of the lists. If there are no matching errors, then the duplicate information collected from a capture-recapture experiment can be used to estimate the number of missed under proper assumptions. Three approaches and their associated estimation procedures are introduced: ecological models; log-linear models, and the sample coverage approach. Each approach has its unique way of incorporating two types of source dependencies: local (list) dependence and dependence due to heterogeneity. An interactive program, CARE (for capture-recapture) developed by the authors is demonstrated using four real data sets. One set of data deals with infection by the acute hepatitis A virus in an outbreak in Taiwan; the other three sets are ascertainment data on diabetes, spina biÿda and infants' congenital anomaly discussed in the literature. These data sets provide examples to show the usefulness of the capture-recapture method in correcting for under-ascertainment. The limitations of the methodology and some cautionary remarks are also discussed.A. CHAO ET AL. ascertained by three sources: (i) P-list, records based on a serum test taken by the Institute of Preventive Medicine, Department of Health of Taiwan -there were 135 identiÿed cases; (ii) Q-list, local hospital records reported by the National Quarantine Service -122 cases were found; (iii) E-list, records collected by epidemiologiststhere were 126 cases. Merging the three lists by eliminating duplicate records resulted in 271 ascertained cases. This data set has the advantage of a known true number of infected because a screen serological check for all students was conducted after the three surveys. In Section 5, we use this data set to show the need of correction for undercount. 2. Hook et al. [2] and Regal and Hook [3] presented a data set on spina biÿda collected in New York State from 1969-1974. Three lists were collected from birth certiÿcates, death certiÿcates and medical rehabilitation ÿles. There were 513, 207 and 188, cases, respectively, on the three lists and in total 626 ascertained cases. A method of estimating the number of missed cases was discussed by the above authors to assess the completeness of the survey and to accurately estimate the prevalence rate. 3. Bruno et al. [4] collected a data set on diabetes in a community in Italy based on the following four records: diabetic clinic and=or family physician visits (1754 cases); hospital discharges (452 cases); prescriptions (1135 cases), and pu...

Drug-Related Deaths in France in 2007: Estimates and Implications

“…Its theoretical foundations were formalized later on (Bishop, Fienberg, & Holland, 1975;Wittes, Colton, & Sidel, 1974;Wittes & Sidel, 1968), and Bishop et al (1975) hypothesized its plausible application in estimating the size of substance user populations. The flexibility of the CR technique made it possible to simultaneously incorporate three or more information sources (Chao, Tsay, Lin, Shau, & Chao, 2001;Cormack, 1989;Hay, 1997;Hay & Gannon, 2006;Hook & Regal, 1995). Its advantages and limitations have been discussed elsewhere (Cormack, 1999;Hook & Regal, 1995;LaPorte, 1994).…”

Prevalence of blindness in Western Australia: a population study using capture and recapture techniques

“…The data are illustrated in a Venn diagram (figure 1) and in an incomplete contingency table (table 1). Incomplete contingency tables were used to estimate the number of individuals in the missing cell of the table using log-linear modelling 13 14…”

Utility of Capture-Recapture Methodology to Assess Completeness of Amyotrophic Lateral Sclerosis Case Ascertainment

“…Thus, an estimate of that total (N) may be derived ((S 1, 2 /S 1 = S 2 /N); see fig. 1) [12,23]. However, such analysis requires adherence to strict requirements, including source independence, population homogeneity and that the population be closed (i.e.…”

“…Positive dependencies result in underestimates of the total true population size, and negative dependencies result in overestimation of the population size. As truly independent sources are rare in epidemiological practice, multiple-source logistic techniques are most often used [12,23,24,25,26,27,28,29] to control for these dependencies, as they can account for interdependencies between sources. …”

“…To account for interaction between the individual data sources, a saturated model was fit accounting for all two- and three-way interaction terms, followed by a stepwise reduction (eliminating nonsignificant p values at each step), selecting the final model based on the lowest deviance and Akaike information criteria. The four-way interaction term was assumed nonsignificant [13,23,33]. Population prevalence estimates were calculated based on SEER data from the 2003 census population of people over 18 years of age in the 5 counties of metropolitan Atlanta [34].…”