Article Text
PDF
Abstract
Objective Chlamydia re-infection data are used to inform and evaluate chlamydia control programmes. We quantitatively investigated the effect of denominator selection on estimating re-infection rates and trends.
Methods Using data on women aged 15–44 years enrolled in Group Health Cooperative (GH), a Pacific Northwest health plan, annual chlamydia re-infection rates from 1998 to 2006 were calculated. Three different denominators were compared using person-years contributed by: (1) all women; (2) women with a prior documented chlamydial infection regardless of whether they were retested; and (3) women with a prior chlamydial infection who were retested within 14 months.
Results From 1998 to 2006, among all women 15–44 years enrolled in GH, re-infection rates increased from 64 to 149 cases per 100 000 person-years. Among women with a prior infection, rates decreased from 10 857 to 8782 cases per 100 000 person-years. Among women with a prior infection who were retested, rates increased from 29 374 to 42 475 cases per 100 000 person-years.
Conclusions Using the same dataset, we demonstrate that it is possible to tell three different stories about the magnitude of rates and trends in chlamydia re-infection among women by using different denominators. All of these strategies have limitations, but restricting the denominator to women with a prior infection who are retested may best represent the population at-risk for re-infection. Still, rates do not account for additional factors influencing the number of re-infections diagnosed, including screening coverage and changes in test technology. Caution is needed in examining and comparing re-infection data.
- Chlamydia Infection
- Surveillance
- Bacterial Infection
Statistics from Altmetric.com
Background
Monitoring chlamydia re-infection rates is a routine part of disease surveillance.1 ,2 Re-infection rates are used to inform chlamydia control programmes, including developing re-screening guidelines.3–5 The calculation of a re-infection rate is mathematically simple: the number of re-infections divided by the population at-risk for re-infection in a specified time period. However, there are a number of challenges in estimating re-infection rates, such as misclassification of re-infections as initial infections and variation in case identification over time, including switching to more sensitive diagnostic tests or increasing screening coverage. The choice of denominator (ie, what population best describes the ‘at-risk’ population for being re-infected) poses an additional challenge.6 Often, denominator choice is limited by available data. For example, surveillance datasets may not contain negative test results to identify who was rescreened, so re-infection rates are calculated among all women with a previous infection.3 Other researchers have calculated re-infection rates using population or census data as a denominator.7 Using data from a Pacific Northwest health plan, we investigated the impact of denominator selection when estimating chlamydia re-infection rates.
Methods
Data were extracted from computerised health plan databases for women aged 15–44 years enrolled in Group Health Cooperative (GH), a mixed model managed care organisation serving Washington State and Western Idaho. Data on chlamydia testing and test results were collected from inpatient, ambulatory care and utilisation databases for women enrolled in the GH integrated group practice (where enrolees receive the majority of their care through GH) between 1 January 1997 and 31 December 2007.8 For each enrolee, annual person-years were calculated based on the time she was enrolled in GH.
Chlamydial infection was defined by a positive chlamydia test. Re-infection was defined as a positive chlamydia test that occurred at least 30 days after a previously documented infection. Annual re-infection rates per 100 000 person-years were calculated using three denominators of annual person-years contributed by GH women enrolees aged 15–44 years: (1) all women enrolled; (2) women with a prior documented chlamydial infection during the study period; and (3) women with a prior documented chlamydial infection during the study period who were retested within the subsequent 14 months. Because current guidelines in the USA recommend rescreening approximately 3 months after treatment or at the first healthcare visit in the 12 months following treatment,9 we used 14 months to allow for time after the initial diagnosis date for treatment and follow-up. For denominator choices #1 and #2, re-infections could occur at any time in the study interval at least 30 days following an initial infection. Year was defined by the year the re-infection was identified, and women could have multiple re-infections. For denominator choice #3, re-infections were restricted to those within 14 months following a previous infection in the study interval. Only the first re-infection was included and year was defined by year of the initial infection.
The re-infection date was defined as the date of the positive result. Because chlamydia is often asymptomatic, this may not be the date when the re-infection actually occurred. Therefore, when evaluating the third denominator choice, we conducted a sensitivity analysis reassigning the date of re-infection to the midpoint between the initial and re-infection test dates (denominator #3a). Analyses were performed in SAS V.9.2 (SAS Institute Inc., Cary, North Carolina, USA). All study protocols received review and approval by the Group Health Institutional Review Board.
Results
During the 9-year study interval, enrolment at GH decreased from 104 300 (1998) to 76 944 (2006) women aged 15–44 years. During the same time period, the number of re-infections identified per year increased, with a large increase the last year (figure 1A). Using denominator #1, all women GH enrolees aged 15–44 years resulted in an increasing re-infection rate between 1998 and 2006, from 64 to 149 cases per 100 000 person-years (figure 1B). Restricting the denominator to women with a previously documented chlamydial infection (denominator #2, women at risk of re-infection) the rate was over 100 times higher than the rate using denominator #1 (figure 1B). In contrast to the increasing re-infection rate using denominator #1, the re-infection rate among women with a previous infection fluctuated over the study interval but decreased between 1998 and 2006, from 10 856 to 8782 cases per 100 000 person-years.
Annual number of chlamydia re-infections (A) and estimated re-infection rates (B) among women aged 15–44 years using different denominators, Group Health Cooperative, 1998–2006.
Restricting the denominator further to women who had a previous infection in the study interval and were retested within 14 months of their initial infection (denominator #3, women at risk for detection of recent re-infection) resulted in a still higher estimated rate. Between 1998 and 2006, rates increased from 29 374 to 42 475 cases per 100 000 person-years with a steady increase over the last half of the study interval (figure 1B). Using this denominator also reduced the annual number of re-infections diagnosed (the numerator for the rate) by approximately 35%, as re-infections had to occur within 14 months of the initial infection and only the first re-infection was included (figure 1A). Finally, reassigning the re-infection date to the midpoint between the dates of the two positive tests resulted in a slightly higher estimated rate, with an average per cent increase of 6.6% (range 4.4%–9.3%) (figure 1B).
Discussion
True re-infection rates are difficult to estimate as most re-infections are asymptomatic and never identified. Studies that rescreen participants at regular intervals can overcome this limitation,4 ,10 but most estimates of re-infection rely on opportunistic data from surveillance or administrative data systems. Using data from computerised health plan databases, we demonstrate that it is possible to tell three different stories about the magnitude of and trends in chlamydia re-infection among women by using different denominators.
Using a denominator of person-time contributed by all women aged 15–44 years enrolled in GH accounted for the changing population size of GH over time, but many of the women never had a first chlamydial infection diagnosed and thus were not at risk for having a re-infection. Consequently, this denominator is inflated, and the estimated rate is erroneously low. The number of infections eligible to be called a re-infection grew annually resulting in an increasing numerator regardless of any changes in re-infection incidence or population size. Thus, similar to other estimated re-infection rates using a population denominator,6 it is unclear whether increasing trends in estimated rates are a consequence of changes in re-infection incidence or the choice of denominator. Although this type of denominator is readily available from census data and can be applied to routinely collected surveillance data, it results in misleading re-infection rates and trends.
Restricting the denominator to women with a documented previous infection better represents the population at-risk. Excluding the person-time contributed by women without a previous infection deflated the denominator, resulting in a much higher estimated re-infection rate compared with using all enrolled women (denominator #1). The majority of published studies of chlamydia re-infection use a denominator restricted to women with a previously documented infection.2 However, this denominator does not take into account how many women were retested following their initial infection. As chlamydial infections (and re-infections) are usually asymptomatic, increases in the proportion of women who are retested following their initial infection (eg, providers become more persuasive in encouraging women to return for retesting) can cause estimated rates to increase even if re-infection incidence is stable.
Restricting the denominator to women with a documented previous infection who were rescreened within 14 months of their initial infection may best represent the population at-risk for detection of re-infection among our three denominator choices. In the GH data, rates estimated using this denominator were higher than the rates using the other two denominators, as the population at-risk was reduced substantially by excluding women without a previous documented infection and women with a documented previous infection who were not rescreened within 14 months. Our sensitivity analysis to account for the time between actual re-infection and re-infection detection by reassigning the re-infection date to the midpoint between the dates of the initial and second positive test produced a slightly higher re-infection rate, as it reduced the time at-risk in the denominator. It is likely that the bias resulting from using the re-infection diagnosis date would be more pronounced if we had used a longer follow-up time.
Restricting the denominator to women with a previously documented infection who were retested in a defined time period best represents the population at-risk in our example; however, there are limitations. This denominator does not account for the healthcare seeking behaviour of women. Women who returned for retesting may have more risk behaviours than those who did not return, particularly if they returned due to a symptomatic infection, inflating the re-infection rate. Additionally, we have only investigated the effect of denominator choice. Numerators also matter! Using more sensitive diagnostic tests and increasing screening coverage can impact how many infections and re-infections are diagnosed. Finally, we have not distinguished between re-infection and treatment failures. Although re-infection data are useful to inform and evaluate prevention programmes, most databases do not include all the data needed to evaluate these issues. Consequently, caution is needed in examining and comparing re-infection data.
Footnotes
-
Contributors All authors conceived the research question and reviewed drafts of the manuscript. ET led the writing of the manuscript. OY and DS led the data management and completed the data analysis.
-
Funding This work was supported by funds from the Division of STD Prevention, Centers for Disease Control and Prevention, Atlanta, GA. The findings and conclusions in this presentation are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
-
Competing interests None.
-
Ethics approval Group Health Institutional Review Board.
-
Provenance and peer review Not commissioned; externally peer reviewed.