Diabetic retinopathy (DR) is a leading cause for blindness among working-aged adults. The growing prevalence of diabetes urges for cost-effective tools to improve the compliance of eye examinations for early detection of DR. The objective of this research is to identify essential predictors and develop predictive technologies for DR using electronic health records. We conducted a retrospective analysis on a derivation cohort with 3749 DR and $94,127$ non-DR diabetic patients. In the analysis, an ensemble predictor selection method was employed to find essential predictors among 26 variables in demographics, duration of diabetes, complications and laboratory results. A predictive model and a risk index were built based on the selected, essential predictors, and then validated using another independent validation cohort with 869 DR and 6448 non-DR diabetic patients. Out of the 26 variables, 10 were identified to be essential for predicting DR. The predictive model achieved a $0.85$ AUC on the derivation cohort and a $0.77$ AUC on the validation cohort. For the risk index, the AUCs were $0.81$ and $0.73$ on the derivation and validation cohorts, respectively. The predictive technologies can provide an early warning sign that motivates patients to comply with eye examinations for early screening and potential treatments.

With the increasing availability of electronic health records (EHR), significant progress has been made on developing predictive inference and algorithms by health-data analysts and researchers. However, the EHR data are notoriously noisy, due to missing and inaccurate inputs, despite abundant information. One serious problem is that only a small portion of patients in the database has confirmatory diagnoses, while many other patients remain undiagnosed because they did not comply with the recommended examinations. The phenomenon leads to a so-called positive-unlabelled situation, and the labels are extremely imbalanced. In this paper we propose a model-based approach to classify the unlabelled patients by using a Bayesian finite mixture model. We also discuss the label switching issue for the imbalanced data and propose a consensus Monte Carlo approach to address the imbalance issue and improve computational efficiency simultaneously. Simulation studies show that our proposed model-based approach outperforms existing positive-unlabelled learning algorithms. The proposed method is applied on the Cerner EHR for detecting diabetic retinopathy (DR) patients using laboratory measurements. With only 3% confirmatory diagnoses in the EHR database, we estimate the actual DR prevalence to be 25% which coincides with reported findings in the medical literature.