Studies relying on passive retrospective cohorts developed from health services data provide biased estimates of incidence of sexually transmitted infections.

OBJECTIVE Passive retrospective cohorts composed of persons who have tested 2 or more times for a sexually transmitted infection (STI) of interest during clinical visits have been used to estimate STI incidence. We hypothesized that the analytic period of a passive cohort might affect the estimate of STI incidence, with shorter periods yielding higher estimates of incidences of infection. STUDY We analyzed data collected from women, 12 to 24 years of age, tested for chlamydia 2 or more times at 6 sites in San Francisco between January 1997 and December 2000. Incidence was calculated for 10 different analytic periods. RESULTS The calculated incidence of chlamydial infection during 1997 was 16.8 (95% confidence interval [CI], 10.9-24.0) per 1000 person-months of follow up. The calculated incidence dropped markedly as the analytic period lengthened, with the incidence estimated to be 9.7 (95% CI, 8.6-10.9) using a study period of 4 years (1997-2000). Estimates of incidence were similar when using the same analytic period, regardless of calendar year, and there was a similar decline in estimated incidence using longer analytic periods. CONCLUSIONS Estimates of STI incidence based on passive cohort data may have limited epidemiologic value because incidence measures may be highly dependent on the analytic period.