Using infection prevalence, seroprevalence and case report data to estimate chlamydial infection incidence

Abstract

Objectives: To measure the effectiveness of chlamydia control strategies, we must estimate infection incidence over time. Available data, including survey-based infection prevalence and case reports, have limitations as proxies for infection incidence. We therefore developed a novel method for estimating chlamydial incidence.

Methods: We linked a susceptible infectious mathematical model to serodynamics data from the National Health and Nutritional Examination Survey, as well as to annual case reports. We created four iterations of this model, varying assumptions about how the method of infection clearance (via treatment seeking, routine screening or natural clearance) relates to long-term seropositivity. Using these models, we estimated annual infection incidence for women aged 18-24 and 25-37 years in 2014. To assess model plausibility, we also estimated natural clearance for the same groups.

Results: Of the four models we analysed, the model that best explained the empirical data was the one in which longer-lasting infections, natural clearance and symptomatic infections all increased the probability of long-term seroconversion. Using this model, we estimated 5910 (quartile (Q)1, 5330; Q3, 6500) incident infections per 100 000 women aged 18-24 years and 2790 (Q1, 2500; Q3, 3090) incident infections per 100 000 women aged 25-37 years in 2014. Furthermore, we estimated that natural clearance rates increased with age.

Conclusions: Our method can be used to estimate the number of chlamydia infections each year, and thus whether infection incidence increases or decreases over time and after policy changes. Furthermore, our results suggest that clearance via medical intervention may lead to short-term or no seroconversion, and the duration of untreated chlamydial infection may vary with age, underlining the complexity of chlamydial infection dynamics.

Keywords: chlamydia infections; epidemiology; models, theoretical.

Figure 1

Schematics of four models used to estimate chlamydia incidence in women under different assumptions about seroconversion: (A) all seroconversion model, (B) long infection seroconversion model, (C) natural clearance seroconversion model, (D) mixed seroconversion model. At any given time, women can be uninfected-seronegative ($U_0$), asymptomatically infected ($A$), symptomatically infected ($S$) or uninfected-seropositive ($U_1$). Note that for (D) values of $\chi$ differ depending on whether it is applied to symptomatic or asymptomatic infections and which mode of clearance it is attached to. See table 1 for a description of the parameters and see the text and online supplemental appendix for model details.

Figure 2

Y-axis shows (A, B) incidence estimates per 100 000 women in 2014, (C, D) estimates of natural clearance rates for women in 2014 and x-axis shows relative susceptibility of seropositive women compared with seronegative women ($\epsilon$). High values of $\epsilon$ indicate that seropositive and seronegative women are equally susceptible to infection, while low values of $\epsilon$ indicate that seropositive women are less susceptible than seronegative women. Left panel shows incidence for women aged 18–24 years, while right panel shows incidence for women aged 25–37 years. Points show mean estimates, with vertical lines showing IQRs. The black dashed line in panels A and B indicate case reports per 100 000 women. When $\epsilon =1$, models differ in incidence estimates because models differ in the proportion of infected individuals that become seropositive. Thus, models must invoke various incidence estimates to generate a proportion of the population that is seropositive that matches the proportion in the National Health and Nutrition Examination Survey.