Machine learning algorithms demonstrated 70-76% accuracy predicting 30-day readmissions across four cardiovascular conditions in 123,272 predominantly Black patients (96.6%) from Virginia's statewide database. XGBoost emerged as the top performer for three of four conditions, with atrial fibrillation/flutter showing highest discrimination (76% accuracy) and heart failure the lowest (71%). The LACE Index, Charlson Comorbidity Index, and insurance type consistently ranked as the strongest predictive factors across all models. This represents a significant advance in cardiovascular risk prediction for Black populations, who have historically been underrepresented in clinical algorithms despite experiencing disproportionate cardiovascular burden and readmission rates. The condition-specific approach marks a departure from traditional models that lump all cardiovascular diseases together, potentially enabling more targeted interventions. However, the moderate accuracy levels suggest these tools would supplement rather than replace clinical judgment. As a preprint awaiting peer review, these findings require validation in independent cohorts before clinical deployment. The work provides a promising framework for developing equity-informed prediction models that could help reduce healthcare disparities in cardiovascular outcomes.