Background Chlamydia is the most commonly-diagnosed bacterial STI worldwide. Models have been developed to estimate chlamydia prevalence from surveillance data in Australia and the UK, respectively, by Ali, Cameron et al. (AC) and Lewis & White (LW). To assess robustness, we compared the models’ prevalence estimates when applied to the same data.

Methods The models were applied to Australian 2001–2016 surveillance data to produce annual prevalence estimates in age-sex categories of 15–19, 20–24, 25–29 years for each sex. Two sets of input parameters (the “prior” and “posterior” parameters from the AC modelling study) were used.

Results The LW model produced higher prevalence estimates than the AC model in every age-sex category, with both “prior” and “posterior” parameterisation. Prevalence estimates for Australian women aged 15–29 in 2015 were 2.5%(95%CrI:2.4%–2.7%) and 5.1%(95%CrI:4.0%–6.0%) from the AC model and LW model (using “prior” parameters), respectively; the corresponding empirical estimate from literature was 3.3%(95%CI:2.1%–4.5%). Averaging over all years, the LW model produced prevalence estimates that were 2.5x higher than the AC model for Australian men aged 15–29, and 1.9x higher in women. Neither model agreed perfectly with the empirical prevalence estimates; the LW model tended to be closer in younger age-categories and the AC model closer in older age-categories. The AC model was closer to empirical estimates in men than women.

Conclusion Substantial differences were observed between chlamydia prevalence estimates produced by the two models. These findings have important implications for researchers, policymakers and healthcare professionals, as estimation methods must be robust before they are used to inform public health policy, e.g. assessing the impact of chlamydia-control interventions. Health care systems and associated surveillance systems vary by country, and work to understand the reasons for the models’ differences is planned, including applying the models to English data, in collaboration with the Universities of Bern, New South Wales, and Otago.

Disclosure No significant relationships.