BACKGROUND The association between chlamydia infection and pelvic inflammatory disease (PID) is a key parameter for models evaluating the impact of chlamydia control programs. We quantified this association using a retrospective population-based cohort. METHODS We used administrative health data sets to construct a retrospective population-based cohort of women and girls aged 12-24 years who were resident in Manitoba, Canada, between 1992 and 1996. We performed survival analysis on a subcohort of individuals who were tested for chlamydia to estimate the risk of PID diagnosed in a primary care, outpatient, or inpatient setting after ≥ 1 positive chlamydia test. RESULTS A total of 73 883 individuals contributed 625 621 person years of follow-up. Those with a diagnosis of chlamydia had an increased risk of PID over their reproductive lifetime compared with those who tested negative (adjusted hazard ratio [AHR], 1.55; 95% confidence interval [CI], 1.43-1.70). This risk increased with each subsequent infection: the AHR was 1.17 for first reinfection (95% CI, 1.06-1.30) and 1.35 for the second (95% CI, 1.04-1.75). The increased risk of PID from reinfection was highest in younger individuals (AHR, 4.55 (95% CI, 3.59-5.78) in individuals aged 12-15 years at the time of their second reinfection, compared with individuals older than 30 years). CONCLUSIONS There is heterogeneity in the risk of PID after a chlamydia infection. Describing the progression to PID in mathematical models as an average rate may be an oversimplification; more accurate estimates of the cost-effectiveness of screening may be obtained by using an individual-based measure of risk. Health inequalities may be reduced by targeting health promotion interventions at sexually active girls younger than 16 years and those with a history of chlamydia.