Matching and Estimation Methods for Capture-recapture Analysis When Individuals Cannot be Identified Uniquely

• Activity Number: 311
• Type: Contributed
• Date/Time: Wednesday, August 14, 2002 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology*
• Author(s): Betsy Cadwell*+ and Phil Smith and Andrew Baughman and Chima Ohuabunwo and Kristine Bisgard
• Affiliation(s): Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention
• Keywords: Capture-recapture ; Population estimation ; bootstrap

Abstract

Classical closed-population capture-recapture methodology is often used to estimate disease incidence. In these studies, two lists are available for matching, and a variant of the Petersen estimate is used to estimate population size. These methods are only applicable if two or more independent, identified lists which allow record matching exist; without unique identifiers, standard capture-recapture methods require modification. As an alternative, we create profiles consisting of attributes common to both lists. Although individuals within each list are not uniquely identified by these profiles, we can estimate the population size for all possible matches between the two lists for a given profile. An aggregate estimate is obtained by weighing each of the estimates by the probability the matches are correct and summing weighted estimates across all profiles. A bootstrap percentile interval is generated for the final population estimate. The validity of this estimation method will be evaluated through simulations. In order to illustrate this method, we applied this approach to estimate the number of pertussis hospitalizations in the U.S. from 1992-1996.