Accurate risk assessment of an individuals' propensity to develop cardiovascular diseases (CVDs) is crucial for the prevention of these conditions. Numerous published risk prediction models used for CVD risk assessment are based on conventional risk factors and include only a limited number of biomarkers. The addition of novel biomarkers can boost the discriminative ability of risk prediction models for CVDs with different pathogenesis. The present study reports the development of risk prediction models for a range of heterogeneous CVDs, including coronary artery disease (CAD), stroke, deep vein thrombosis (DVT), and abdominal aortic aneurysm (AAA), as well as for Type 2 diabetes mellitus (DM2), a major CVD risk factor. In addition to conventional risk factors, the models incorporate various blood biomarkers and comorbidities to improve both individual and population stratification. An automatic variable selection approach was developed to generate the best set of explanatory variables for each model from the initial panel of risk factors. In total, up to 254,220 UK Biobank participants (ranging from 215,269 to 254,220 for different CVDs and DM2) were included in the analyses. The derived prediction models utilizing Cox proportional hazards regression achieved consistent discrimination performance (C-index) for all diseases: CAD, 0.794 (95% CI, 0.787-0.801); DM2, 0.909 (95% CI, 0.903-0.916); stroke, 0.778 (95% CI, 0.756-0.801); DVT, 0.743 (95% CI, 0.737-0.749); and AAA, 0.893 (95% CI, 0.874-0.912). When validated on various subpopulations, they demonstrated higher discrimination in healthier and middle-age individuals. In general, calibration of a five-year risk of developing the CVDs and DM2 demonstrated incremental overestimation of disease-related conditions amongst the highest decile of risk probabilities. In summary, the risk prediction models described were validated with high discrimination and good calibration for several CVDs and DM2. These models incorporate multiple shared predictor variables and may be integrated into a single platform to enhance clinical stratification to impact health outcomes.

24Cardiometabolic diseases (CMD) impose greater impact on every aspect of health care 25 than any other disease group. Accurate and in-time risk assessment of individuals for their 26 propensity to develop CMD events is one of the most critical paths in preventing these 27 conditions. The principal objective of the present study is to report the development, and 28 validation of a next generation risk engine to predict CMD. UK Biobank population data was 29 used to derive predictive models for six CMD. Missing data were imputed using imputation 30 algorithms. Cox proportional hazard models were used to estimate annual absolute risk and 31 relative risk of different risk factors for these conditions. In addition to conventional risk 32 factors, the applied model included socioeconomic data, lifestyle factors and comorbidities as 33 predictors of outcomes. In total, 416,936 individuals were included in the analysis. The 34 derived prediction models achieved consistent and moderate-to-high discrimination 35performance (C-index) for all diseases: coronary artery disease (0.79), hypertension (0.82), 36 type 2 diabetes mellitus (0.87), stroke (0.79), deep vein thrombosis (0.75), and abdominal 37 aortic aneurysm (0.90). These results were consistent across age groups (37-73 years) and 38 showed similar predictive abilities amongst those with pre-existing diabetes or hypertension. 39Calibration of risk scores showed that there was moderate overestimation of CMD-related 40 conditions only in the highest decile of risk scores for all models. In summary, the newly 41 developed algorithms, based on Cox proportional models, resulted in high disclination and 42 good calibration for several CMD. The integrations of these algorithms on a single platform 43 may have direct clinical impact. 44 45

BackgroundStratification of individuals for their risk to develop cardiovascular diseases can be used for effective prevention and intervention. A significant amount of information for risk assessment can be obtained through repurposing electronic health records (EHR). The objective of this study is to derive and assess the performance of prediction models for cardiovascular outcomes by using EHR-derived data.MethodsWe used the Stanford Medicine Research Data Repository (STARR) data from 2000-2017, containing over 2.1 million patients. A subset of 762,372 individuals with complete International Classification of Diseases (ICD) data was used to fit Cox proportional hazard models for prediction of six cardiovascular-related diseases and type 2 diabetes.ResultsThe derived prediction models indicated consistent high discrimination performance (C-index) for all diseases examined: coronary artery disease (0.85), hypertension (0.82), type 2 diabetes (0.77), stroke (0.76), atrial fibrillation (0.82) and abdominal aortic aneurysm (0.77). Lower prediction abilities were observed for deep vein thrombosis (0.67). These results were consistent across age groups and maintained good prediction abilities among individuals with pre-existing diabetes or hypertension. Assessment of model calibration is ongoing.ConclusionsWe proposed new prediction models for the seven diseases using ICD codes derived from EHR data. EHR data can be used for health risk assessment, but challenges related to data quality and model generalizability and calibration remain to be solved.