Using capture–recapture methods to assess varicella incidence in a community under active surveillance

Abstract

The Varicella (chickenpox) Active Surveillance Project (VASP) has been conducting active surveillance since 1 January 1995 in the high desert community known as Antelope Valley, CA (population 300,000) among 300 public and private schools, daycares, and healthcare providers. Capture–recapture methods were applied to estimate reporting completeness for 1995 varicella incidence data and these were compared with the national average incidence rates by age reported by the National Health Interview Survey (NHIS). Varicella cases reported among individuals aged <20 years reflect under-reporting in excess of 50%. Despite limitations on accuracy, capture–recapture estimates are a reasonably accurate, quick, and inexpensive approach in epidemiologic studies.

Introduction

Capture–recapture methods, derived from techniques developed for studies of animal abundance, estimate the true population size by evaluating the degree of overlap among incomplete lists of cases from existing data sources. Capture–recapture involves consideration of three basic assumptions: (1) the population is closed, i.e. no significant changes occur in the population under study during the investigation (e.g. due to migration or death); (2) there is no loss of tags so that individuals can be matched from capture to recapture; and (3) each individual has the same probability of being caught in the second sample so that capture in the first sample does not affect capture in the second sample, i.e. the samples are independent. While assumptions (1) and (2) are often met, failure to meet assumption (3) can lead to inaccurate and sometimes misleading results—since in epidemiological investigations, sources often display dependence and heterogeneity of capture probabilities [1].

This is the first paper to apply capture–recapture to varicella, a common communicable disease that displays seasonality. Capture–recapture methods have been used in the 1980s and 1990s in studies to estimate the completeness and prevalence of many infectious and chronic diseases, including pertussis [2], tetanus [3], sexually transmitted disease [4], AIDS [5], [6], measles [7], [8], meningococcal disease (which is also a highly seasonal infection) [9], [10], HIV-1 [11], and cryptococcosis [12].

The major question individuals have regarding capture–recapture is “Will capture–recapture give you the ‘truth’, i.e. an extremely accurate estimate of the incidence of disease?” Simply answered, no—it will not. The estimates presented in most epidemiologic studies are extremely poor, missing 10–90% of the cases, with a high degree of variation [13], [14], [15], [16]. Therefore, the options are (a) not to use capture–recapture and report varicella cases from which incidence rates are almost uninterruptible, (b) try to count every case of varicella which is horrendously expensive and slow, or (c) utilize capture–recapture which, depending on the degree to which the assumptions are satisfied, can be a compromise, reasonably accurate, quick, and inexpensive approach.