BackgroundIn British Columbia (BC), we are developing Get Checked Online (GCO), an Internet-based testing program that provides Web-based access to sexually transmitted infections (STI) testing. Much is still unknown about how to implement risk assessment and recommend tests in Web-based settings. Prediction tools have been shown to successfully increase efficiency and cost-effectiveness of STI case finding in the following settings.ObjectiveThis project was designed with three main objectives: (1) to derive a risk prediction rule for screening chlamydia and gonorrhea among clients attending two public sexual health clinics between 2000 and 2006 in Vancouver, BC, (2) to assess the temporal generalizability of the prediction rule among more recent visits in the Vancouver clinics (2007-2012), and (3) to assess the geographical generalizability of the rule in seven additional clinics in BC.MethodsThis study is a population-based, cross-sectional analysis of electronic records of visits collected at nine publicly funded STI clinics in BC between 2000 and 2012. We will derive a risk score from the multivariate logistic regression of clinic visit data between 2000 and 2006 at two clinics in Vancouver using newly diagnosed chlamydia and gonorrhea infections as the outcome. The area under the receiver operating characteristic curve (AUC) and the Hosmer-Lemeshow statistic will examine the model’s discrimination and calibration, respectively. We will also examine the sensitivity and proportion of patients that would need to be screened at different cutoffs of the risk score. Temporal and geographical validation will be assessed using patient visit data from more recent visits (2007-2012) at the Vancouver clinics and at clinics in the rest of BC, respectively. Statistical analyses will be performed using SAS, version 9.3.ResultsThis is an ongoing research project with initial results expected in 2014.ConclusionsThe results from this research will have important implications for scaling up of Internet-based testing in BC. If a prediction rule with good calibration, discrimination, and high sensitivity to detect infection is found during this project, the prediction rule could be programmed into GCO so that the program offers individualized testing recommendations to clients. Further, the prediction rule could be adapted into educational materials to inform other Web-based content by creating awareness about STI risk factors, which may stimulate health care seeking behavior among individuals accessing the website.