Chlamydia Prevalence Trends Among Women and Men Entering the National Job Training Program From 1990 Through 2012

Abstract

Background: Evaluating chlamydia prevalence trends from sentinel surveillance is important for understanding population disease burden over time. However, prevalence trend estimates from surveillance data may be misleading if they do not account for changes in risk profiles of individuals who are screened (case mix) and changing performance of the screening tests used.

Methods: We analyzed chlamydia screening data from a sentinel surveillance population of 389,555 young women (1990-2012) and 303,699 young men (2003-2012) entering the US National Job Training Program. This period follows the introduction of national chlamydia screening programs designed to prevent transmission and reduce population disease burden. After ruling out bias due to case mix, we used an expectation-maximization-based maximum likelihood approach to account for measurement error from changing screening tests, and generated minimally biased long-term chlamydia prevalence trend estimates among youth and young adults in this sentinel surveillance population.

Results: Adjusted chlamydia prevalence among women was high throughout the study period, but fell from 20% in 1990 to 12% in 2003, and remained between 12% and 14% through 2012. Adjusted prevalence among men was steady throughout the study period at approximately 7%. For both women and men, adjusted prevalence was highest among Black and American Indian youth and young adults, and in the Southern and Midwestern regions of the United States throughout the study period.

Conclusions: Our minimally biased trend estimates provide support for an initial decrease in chlamydia prevalence among women soon after the introduction of national chlamydia screening programs. Constant chlamydia prevalence in more recent years suggests that screening may not be sufficient to further reduce chlamydia prevalence among high-risk youth and young adults.

Figure 1

Adjusted and unadjusted chlamydia prevalence among women entering the NJTP (n=389,555) from 1990-2012 (red lines), and men entering the NJTP (n=303,699) from 2003-2012 (blue lines). Adjusted prevalence estimates and 95% CIs account for measurement error associated with use of increasingly sensitive chlamydia screening tests over time (Pathfinder EIA of swabs [1990-1997], Gen-Probe PACE 2 DNA hybridization probe of swabs or urine [1998-2006], and BD ProbeTec ET SDA of urine [2000-2012] or swabs [2007-2012]). Adjusted estimates were modelled using an EM algorithm incorporated into logistic regression. Unadjusted estimates were generated from a logistic regression model, and do not account for changes in the diagnostic accuracy of tests.

Figure 2

Adjusted chlamydia prevalence and 95% CIs among Hispanic (n=61,449), Black (n=224,936), White (n=91,173), and American Indian (n=11,997) women entering the NJTP, 1990-2012 (Panel A) and Hispanic (n=44,822), Black (n=155,045), White (n=95,994), and American Indian (n=7,838) men entering the NJTP, 2003-2012 (Panel B). Adjusted prevalence was modelled using an EM algorithm incorporated into logistic regression to account for measurement error due to changes in the diagnostic accuracy of chlamydia screening tests over the study period.

Figure 3

Adjusted chlamydia prevalence and 95% CIs among women entering the NJTP in four regions of the US, 1990-2012 (Midwest: n=71,291; South: n=174,056; West: n=76,823; Northeast: n=67,385) (Panel A) and men entering the NJTP in four regions of the US, 2003-2012 (Midwest: n=51,361; South: n=149,464; West: n=57,144; Northeast: n=45,730) (Panel B). Adjusted prevalence was modelled using an EM algorithm incorporated into logistic regression to account for measurement error due to changes in the diagnostic accuracy of chlamydia screening tests.