Developing accurate and equitable screening protocols can lead to more targeted, efficient, and effective, teen dating violence (TDV) prevention programming. Current TDV screening protocols perform poorly and are rarely implemented, but recent research and policy emphasizes the importance of leveraging more trauma-focused screening measures for improved prevention outcomes. In response, the present study examined which adversities (i.e., indices of family violence), trauma-focused risk factors (i.e., threat and reward biases) and strengths (i.e., social support and racial/ethnic identity) best classified concurrent and prospective risk for physical and psychological forms of TDV-perpetration. Participants included 584 adolescents aged 12-18 years (M[subscript Age] = 14.43; SD = 1.22), evenly distributed across gender (48.9% female), race (35% African American; 38.5% White) and ethnicity (40% Hispanic). Surveys completed at baseline and 1-year follow-up were analyzed using an evidence-based medicine (EBM) analytic protocol (i.e., logistic regression, area-under-the-curve; (AUC), diagnostic likelihood ratios (DLR), calibration curves) and compared to machine learning models. Results revealed hostility best classified risk for concurrent and prospective physical TDV-perpetration (AUCs > 0.70; DLRs > 2.0). Additionally, domestic violence (DV) exposure best forecasted prospective psychological TDV-perpetration (AUC > 0.70; DLR > 3.0). Both indices were well-calibrated (i.e., non-significant Spiegelhalter's Z statistics) and statistically fair. Machine learning models added minimal incremental validity. Results demonstrate the importance of prioritizing hostility and DV-exposure for accurate, equitable, and feasible screening for physical and psychological forms of TDV-perpetration, respectively. Integrating these findings into existing prevention protocols can lead to a more targeted approach to reducing TDV-perpetration.