Estimating deaths from industrial injury by capture-recapture: a cautionary tale.

BACKGROUND: Capture-recapture methods are increasingly being used to improve surveillance for a number of diseases. However, concerns persist regarding the validity of estimates obtained. METHOD: Capture-recapture methods were applied to estimate the ability of four separate data sources on occupational fatalities to predict the 237 deaths ('gold standard') we determined from a special in-depth study of medical examiner records. RESULTS AND CONCLUSION: Capture-recapture results based upon the four sources vary according to different models. However, both separately and in aggregate of industry type and cause of death, most models seriously underestimate the gold standard, and give a misleading impression of precision of the estimate of hidden individuals. It is commonly believed that reliable estimates from such methods require lists with high coverage and parsimonious models. Here, to obtain an estimate consistent with the gold standard, the list with almost complete coverage must be discarded and a complex model fitted. It is argued that this conclusion is of widespread application.

Children are not goldfish--mark/recapture techniques and their application to injury data.

OBJECTIVES: Mark/recapture (or capture-recapture) is a simple technique commonly applied to estimate the hypothetical total (including undercount) in a register composed of cases from two or more independent and separately incomplete case lists. This paper seeks to illustrate serious drawbacks in the use of the mark/recapture technique when applied to injuries. SETTING AND SUBJECTS: Northumbrian children under 15 years of age who were seriously injured in motor vehicle accidents (MVAs) over a five year period ascertained from two data sources: police reports and hospital inpatient records. METHODS: Individuals (n) appearing in both police (S) and hospital (H) case lists are identified using various matching criteria. The separate and combined influence of age, sex, and casualty class (cyclist, passengers, pedestrians) on the probability of such matching is estimated using multivariate techniques. The hypothetical total incidence of child MVA victims (N) is calculated from N = (S x H)/n. MAIN OUTCOMES: Estimates of the incidences of "serious" injuries in MVAs under various conditions of stratification and matching. The overall procedure is tested for conformity with accepted criteria for valid use of mark/recapture. RESULTS: About one third of the 1009 police and 836 hospital records could be exactly matched. There were significant variations in matching proportions by class of accident (pedestrian v passenger v cyclist). This selective recapture or "heterogeneity" was not affected by sex, but was independently influenced by the age of the child. Further uncertainty was introduced when matching criteria were slightly relaxed. Estimates of the total population of children with serious injuries vary accordingly from 1729 to 2743. A number of plausible reasons why these two data sources might not be unbiased or mutually independent samples of the total target population are proposed as explanations for this heterogeneity. CONCLUSION: This typical example of two sample mark/recapture estimation in an epidemiological setting can be shown to violate virtually all the requirements for valid use of the technique. Very little can be deduced accurately about the scale or characteristics of an unobserved group by the use of mark/recapture applied to two overlapping health event registers.

Problems with using capture-recapture in epidemiology: an example of a measles epidemic.

Capture-recapture is becoming widely used in epidemiology to estimate disease prevalence or sizes of population at risk. When such estimates are obtained from uncontrolled observation of existing lists, a huge act of faith is required, usually without any scientific justification. Fitting of loglinear models appears to offer some hope but contains major problems of analysis and interpretation. These are illustrated by reanalysis of data on a measles epidemic-see McGilchrist et al. 1996 [J Clin Epidemiol 49, pp. 293-296]--for which the wrong model was selected. It is argued that the measles lists contained so few overlaps that no reliable information is provided by that study about the size of the epidemic.

Estimation of injecting drug users in the City of Edinburgh, Scotland, and number infected with human immunodeficiency virus.

OBJECTIVES: To estimate (1) the number of current and former injecting drug users (IDU) infected with human immunodeficiency virus (HIV) alive in Edinburgh, and (2) the total number of current injectors in the city. METHODS: The number of infected IDU was estimated using a local register of HIV infections with correction for incompleteness of the register. The number of injectors was estimated by two independent methods, one based on the HIV register, the other by log-linear modelling of four lists of IDU interviewed in a city-wide survey, and/or attending drug treatment agencies and family doctors because of drug use. MAIN OUTCOME MEASURES: Estimates for the period 1992-1994 of number of IDU infected with HIV, total number of IDU, and prevalence of injecting. RESULTS: The HIV register indicated that 371 infected drug users who had ever injected were alive and resident in Edinburgh. In all, 95% of infected survey respondents appeared in the register, leading to a corrected estimate of 472 infected ever injectors. From this the number of IDU currently injecting (i.e. in the previous 6 months) was estimated to be 1770 (95% CI: 1340-2240), and the prevalence of injecting as 8.0 (95% CI: 4.8-10.8) per 1000 Edinburgh residents aged 15-59 years. Log-linear modelling gave an estimate of 2070 (95% CI: 1360-2800) current injectors. CONCLUSIONS: The number of HIV-infected IDU in Edinburgh was estimated to be twice that in the larger nearby city of Glasgow, where a higher proportion of young adults currently injected drugs. Knowledge of the high prevalence of HIV in Edinburgh IDU (19.3%), the prescribing of oral substitutes, and counselling by doctors and drug workers are perceived reasons for the reduction in the prevalence of injecting which has occurred in Edinburgh in recent years. Such measures need to be continued to encourage further reduction of injecting.

Interval estimation for mark-recapture studies of closed populations.

Textbooks continue to recommend the use of an asymptotic normal distribution to provide an interval estimate for the unknown size, N, of a closed population studied by a mark-recapture experiment or multiple-record system. A likelihood interval approach is proposed and its implementation demonstrated for a range of models for such studies, including all main effect and interaction models for incomplete contingency tables.