Table 1

Summary of description and results of the sensitivity analyses with respect to variations in model structure

Sensitivity analysis	Description	Result
1. Variation in the distribution of risk behaviour across risk groups.*	Explored the impact of variation in the distribution of risk behaviour across risk groups by varying (in univariate analysis) the parameter σ of the distribution of risk behaviour (online supplementary materials), but fixing α at its model-predicted baseline value.	The predicted age-specific Chlamydia trachomatis prevalence distribution was largely invariable despite the variation in the distribution of risk behaviour across risk groups (online supplementary figure S2A).
2. Variation in the sexual mixing by age.*	Explored the impact of variation in sexual mixing by age (in univariate analysis) across the full spectrum starting from proportionate mixing-up to fully assortative mixing. This was done by varying , the parameter describing the degree of assortativity in mixing by age (online supplementary materials).	The predicted age-specific C. trachomatis prevalence distribution was largely invariable despite the variation in the sexual mixing by age (online supplementary figure S2B).
3. Variation in the sexual mixing by risk.*	Explored the impact of variation in sexual mixing by risk (in univariate analysis) across the full spectrum starting from proportionate mixing-up to fully assortative mixing. This was done by varying , the parameter describing the degree of assortativity in mixing by risk (online supplementary materials).	The predicted age-specific C. trachomatis prevalence distribution was largely invariable despite the variation in the sexual mixing by risk (online supplementary figure S2C).
4. Temporal variation in risk behaviour.	Explored the impact of temporal variation in risk behaviour on our estimated partial immunity strength by assuming that 10% of individuals change their risk group every year.	α for the UK data was estimated at 93% (95% UI: 89%–95%) with an uncertainty analysis median of 93%—similar to the original estimate.
5. Removal of latent period in C. trachomatis natural history.	Explored the impact of removing the latent period in C. trachomatis natural history.	α for the UK data was estimated at 93% (95% UI: 88%–97%) with an uncertainty analysis median of 93%— similar to the original estimate.
6. Inclusion of partial immunity for the symptomatically infected individuals.	Explored the impact of inclusion of partial immunity for the symptomatically infected individuals.	α for the UK data was estimated at 93% (95% UI: 89%–96%) with an uncertainty analysis median of 93%— similar to the original estimate.
7. Variation in the duration of the short-term temporary but full immunity.	Explored the impact of varying the duration of the short-term temporary but full immunity over a range of 0–100 days.	Variation in the short-term temporary immunity had limited impact on the estimated effect size of partial immunity (online supplementary figure S3).

All sensitivity analyses were applied to the model fit of the UK data.
*Conducted in view of the fundamental ambiguity in defining ’sexual risk’,26 44–46 and done on the prediction for the age-specific C. trachomatis prevalence distribution, since this distribution is the most prototypical pattern in C. trachomatis epidemiology.
UI, uncertainty interval.