Prediction of Chlamydia trachomatis infection: application of a scoring rule to other populations

Active case finding is crucial to reduce transmission and consequences of Chlamydia trachomatis (CT) infections. We previously proposed the use of a prediction rule for CT infection for selective screening of high-risk individuals in a population. To support such an application, the prediction rule needs to be validated in other populations.

We studied participants aged 15 to 29 years in a population-based study in Amsterdam (n = 1,788) and an outreach screening project among high-risk youth in Rotterdam (n = 172). Validity was indicated by discriminative ability (area under the receiver operating characteristic curve [AUC]) and by calibration, which was visualized in plots and tested with the Hosmer-Lemeshow (H-L) goodness-of-fit test. Cutoffs of predicted risk were examined for their effect on sensitivity and the fraction of participants that would need to be screened.

Discriminative ability was reasonable both for the Amsterdam study (AUC 0.66; 95% confidence interval [CI], 0.58-0.74) and for the Rotterdam study (AUC 0.68; 95% CI, 0.58-0.79). The observed CT prevalence was lower than predicted in Amsterdam (H-L P = 0.02) and nonsignificantly higher in Rotterdam (H-L P = 0.20). By screening 77% of the Amsterdam population, 93% of the cases would have been detected, whereas in the Rotterdam study, no cases would be missed by screening 75%.

The chlamydia prediction rule showed a reasonable external validity in two studies. These findings support the use of the rule as a tool for selective chlamydia screening, although only a limited fraction of participants can be excluded when a high sensitivity is required.