Challenge
A leading regional healthcare network faced rising 30-day readmission rates, leading to increased costs and strain on resources. Their existing reporting tools lacked predictive capabilities, making it hard to proactively identify at-risk patients.
Solution
We designed a machine learning model that aggregated patient demographics, historical health records, and real-time vitals. Using a random forest algorithm, we predicted discharge risk scores and integrated these into an interactive dashboard for care teams.
Implementation Steps
- Data Audit & Cleaning: Standardized 5 years of EHR data across multiple systems.
- Model Development: Trained and validated a predictive model with 85% accuracy.
- Dashboard Integration: Built a Tableau dashboard with live patient risk scores.
- Training & Support: Conducted workshops for physicians and nurses on using alerts.
Results
- 35% improvement in recovery tracking and early intervention.
- 28% reduction in 30-day readmission rates within 3 months.
- Positive stakeholder feedback and 90% adoption of the dashboard across departments.