Brewer, D. D., Potterat, J. J., Garrett, S. B., Muth, S. Q., Roberts, J. M., Jr., & Rothenberg, R. B. Comparison of direct estimate and partner elicitation methods for measuring the number of sexual and drug injection partners. Poster presented at the 126th annual meeting of the American Public Health Association, November 16, 1998, Washington, DC. Full paper available at SSRN: http://dx.doi.org/10.2139/ssrn.2548807.

The number of sex partners and number of injection partners an individual has are risk factors for sexually transmitted and blood-borne infections, respectively. The most common method for measuring the number of partners is the direct estimate, in which respondents are simply asked to estimate the number of partners for a given recall period (e.g., “how many people have you had sex with in the last year?”). We assessed the intermethod and test-retest reliability of the direct estimate, described subjects’ direct estimate response strategies and processes, and indirectly evaluated the validity of the direct estimate in data from two studies of persons at presumed high risk for HIV.

The direct estimate showed moderate to high levels of test-retest reliability, as did the number of partners recalled when a respondent was asked to list partners by name or description. The correspondence between the two methods, though, tended to be somewhat lower. Across measurement methods, the number of sex partners was more reliably measured than the number of injection/needle-sharing partners with respect to most, but not all, reliability criteria. A large majority of respondents in one study reported using enumeration as a response strategy for the direct estimate. Direct estimates also showed a noteworthy degree of heaping on multiples of five that was not present in the number recalled. Based on comparisons with the estimated mean number of new partners for a comparable period (derived from counts of recalled partners first encountered between two separate interviews), the direct estimate seemed to provide a better estimate of the true mean number of partners (for all respondents) than the number recalled. Nevertheless, both the direct estimate and the number recalled still appeared to underestimate substantially the true number of partners.

Note: Reference 9 is not available, but a subsequent article includes more comprehensive material on the topic: Roberts, J. M., Jr., & Brewer, D. D. (2001). Measures and tests of heaping in discrete quantitative distributions. Journal of Applied Statistics, 28, 887-896.

full article - pdf (ISR site)    full article (SSRN site)