Additional file 1 of Development and validation of prediction models for stroke and myocardial infarction in type 2 diabetes based on health insurance claims: does machine learning outperform traditional regression approaches?

Additional File 1: Additional File 1. Summary of deviations from the model development plan. Additional File 2. Deviations in the set of features during actual implementation compared to the pre-specified model development plan. Additional File 3. TRIPOD Checklist. Additional File 4. Definition of the study population. Additional File 5. Algorithm for the identification of type 2 diabetes patients. Additional File 6. Description of the outcome selection process. Additional File 7. Selection algorithm for identification of outcomes following [2]: Stroke and myocardial infarction. Additional File 8. Identification of targeted macrovascular complications based on ICD-10-GM codes. Additional File 9. Definition of newly recorded diabetes. Additional File 10. Additional criteria (applied to the run-in period Q1-Q4 2014) to identify prevalent diabetes diagnoses in the observation period. Additional File 11. Potential socio-demographic and socio-economic predictors based on previous models. Additional File 12. Potential predictors related health care resource utilization and costs based on previous models. Additional File 13. Identification of other diabetes-related complications and events based on ICD-10-GM, OPS- and EBM-codes. Additional File 14. Potential predictors related to multimorbidity or constituting dimensions. Additional File 15. Operationalisation of the individual dimensions of the aDSCI score. Additional File 16. Potential comorbidity related predictors based on previous models. Additional File 17. Potential medication related predictors based on previous models: Antidiabetic drugs. Additional File 18. Potential medication related predictors based on previous models: Antithrombotic drugs (Anticoagulants). Additional File 19. Potential medication related predictors based on previous models: Anti-hypertensive drugs. Additional File 20. Potential medication related predictors based on previous models: Lipid modifying agents. Additional File 21. Potential medication related predictors based on previous models: Cardiac Medications. Additional File 22. Potential medication related predictors based on previous models: CNS-related drugs. Additional File 23. Potential medication related predictors based on previous models: Other drugs. Additional File 24. Feature engineering– technical details. Additional File 25. Time windows used for prediction modeling. Additional File 26. Software and program codes. Additional File 27. Model development– Technical details. Additional File 28. Details on parameter tuning. Additional File 29. Initial ranges and final values for tuning parameters. Additional File 30. Overview and definition of discrimination metrics. Additional File 31. Overview and definition of calibration metrics. Additional File 32. Overview and definition of classification metrics. Additional File 33. Participant flow chart. Additional File 34. Sample characteristics. Additional File 35. Training and test set characteristics. Additional File 36. Events per variable. Additional File 37. Descriptives on the buffer period. Additional File 38. Results on discrimination performance. Additional File 39. Density plot of predicted MI probability for MI cases and non-cases. Additional File 40. Density plot of predicted Stroke probability for Stroke cases and non-cases. Additional File 41. Plot of classification metrics for MI against threshold. Additional File 42. Plot of classification metrics for Stroke against threshold. Additional File 43. Results on classification performance at the respective optimized classification threshold. Additional File 44. Brier score. Additional File 45. Variable Importance for MI models (10 most important variables for each model). Additional File 46. Variable Importance for stroke models (10 most important variables for each model). Additional File 47. Most important variables for MI prediction models. Additional File 48. Most important variables for stroke prediction models. Additional File 49. Discussion of additional technical aspects.