Objective: To extend beyond the current approach of predicting warfarin benefit and harm independently in new atrial fibrillation (AF) patients by refining methods to identify predictors of the four combined benefit/harm outcome groups - i) no stroke/no bleed; ii) no stroke/bleed; iii) stroke/bleed; iv) no stroke/no bleed. Methods: We analyzed patient-level data from the Atrial Fibrillation Investigators RCT database (n=9155) and an observational database of AF patients managed by Kaiser Permanente Colorado (n=5475). We classified patients based on the four benefit/harm outcome groups and applied decision tree modeling (CART) and polytomous logistic regression (PLR) to identify patient factors predicting each outcome group. Statistical significance was set at alpha=0.05. Results: CART and PLR consistently identified age and warfarin use as predictors for all outcome groups. Both techniques identified predictors of stroke/no bleed and no stroke/bleed not previously included in AF stroke and bleed risk-assessment tools that predict these outcomes independently (e.g., CHADS2 and HEMORR2HAGES). Methodology strengths and limitations were evident. CART provides a visual algorithm approach to risk. However, there is a lack of quantitative measurement (e.g., odds ratios [OR], confidence intervals) for predictors. While PLR results were thorough and predictor parameter estimates could be converted to ORs to indicate strength of association, the result of PLR is number-intensive. To calculate a patient's probability for each of the four outcome groups, the patient's data must be inputted into three separate equations. While both techniques can be used to calculate an individual patient's probability for each outcome group, PLR likely has more scope for application in a clinical setting. Once refined, a clinical prediction rule could be created based on identified predictors and their ORs. Conclusion: While methods under study need further refinement, these individual patient data analyses provide a useful step forward in the movement towards evidence-based individualization of drug therapy.