Enteric disease episodes and the risk of acquiring a future sexually transmitted infection: a prediction model in Montreal residents

Abstract

Objective: The sexual transmission of enteric diseases poses an important public health challenge. We aimed to build a prediction model capable of identifying individuals with a reported enteric disease who could be at risk of acquiring future sexually transmitted infections (STIs).

Materials and methods: Passive surveillance data on Montreal residents with at least 1 enteric disease report was used to construct the prediction model. Cases were defined as all subjects with at least 1 STI report following their initial enteric disease episode. A final logistic regression prediction model was chosen using forward stepwise selection.

Results: The prediction model with the greatest validity included age, sex, residential location, number of STI episodes experienced prior to the first enteric disease episode, type of enteric disease acquired, and an interaction term between age and male sex. This model had an area under the curve of 0.77 and had acceptable calibration.

Discussion: A coordinated public health response to the sexual transmission of enteric diseases requires that a distinction be made between cases of enteric diseases transmitted through sexual activity from those transmitted through contaminated food or water. A prediction model can aid public health officials in identifying individuals who may have a higher risk of sexually acquiring a reportable disease. Once identified, these individuals could receive specialized intervention to prevent future infection.

Conclusion: The information produced from a prediction model capable of identifying higher risk individuals can be used to guide efforts in investigating and controlling reported cases of enteric diseases and STIs.

Keywords: decision support techniques; food-borne diseases; public health surveillance; sexually transmitted diseases.

Figure 1:

Receiver operating characteristic curve of the final logistic regression prediction model. The final logistic regression prediction model included age, sex, residential location, and number of STI episodes experienced prior to the first enteric disease episode, as well as the type of enteric disease acquired, and an interaction term between age and male sex. The area under the model’s receiver operating characteristic curve was 0.77.

Figure 2:

Calibration of the final logistic regression prediction model. The proportion of true cases among each decile of predicted probability is shown. A “rug” is visible on the bottom of the graph depicting all the probability values that were assigned by the prediction model.