A Closer Look at Chlamydia Re-Infection Rates: Denominators Are Important

Background: Increases in chlamydia (CT) re-infection rates have been hypothesized as a reason for recent increases in reported CT infection rates. However, there are numerous methodological challenges that must be considered when analyzing and interpreting CT re-infection rates.

Objectives: To demonstrate how re-infection trends are influenced by choice of denominator.

Methods: Computerized data on women aged 15-44 enrolled in Group Health Cooperative, a Pacific Northwest health plan, were abstracted. CT re-infection was defined as a positive CT test ≥30 days after a previous CT infection. Crude annual CT re-infection rates, using person-years, were calculated with different denominators from the same population.

Results: From 1997-2007, CT infection rates approximately doubled. During this time, there were 676 CT re-infections during 816,543.8 person-years of follow-up. The number of re-infections per year increased four-fold from 20 to 89. Among all women in the plan, re-infection rates increased six-fold from 2.2 to 13.6 per 10,000 person-years. However, when the denominator was restricted to women previously infected, re-infection rates decreased 2-fold from 1,692.1 to 724.2 per 10,000 person-years.

Conclusions: Using all women in the plan as a denominator showed an increasing re-infection rate, but did not account for the changing size of the population at-risk for re-infection. Restricting the denominator to women with a prior infection is likely more appropriate; however, crude rates do not account for additional factors that could influence the number of re-infections diagnosed, including test and re-test coverage, who is getting retested (e.g. age, reason for re-testing), when they are getting retested, and changes in test technology. 

Implications for Programs, Policy, and/or Research: Trends in re-infection rates are influenced by denominator choice and are also likely to be confounded by the characteristics of the population re-tested, which is often unknown in sentinel surveillance databases. Further investigation is required to elucidate contributing variables.