Predictive Analytics
AI-driven utilization forecasts, hospitalization risk predictions, and cost trend modeling
0.847
C-Statistic (AUC)
Model discriminative ability. Values >0.8 indicate strong predictive performance.
79.2%
Sensitivity
True positive rate — proportion of actual admissions correctly predicted.
82.4%
Specificity
True negative rate — correctly identifying members who won't be admitted.
71.8%
PPV (Precision)
Positive predictive value — reliability of high-risk flags for intervention targeting.
Utilization Forecast — Next 12 Months
Projected inpatient admissions, ED visits, and per-member costs vs. current trend baseline
Cost Prediction Model
Projected total cost trend with 90% confidence interval
Predicted
90% CI Band
Risk Trajectory
Patient count trends by trajectory direction (last 6 months)
1,842
Improving
↓ Risk tier vs 6mo ago
5,104
Stable
Same tier, ±0.2 RAF
1,486
Worsening
↑ Risk tier or RAF
Month-over-Month Trajectory Shift
Improving
1,842
Stable
5,104
Worsening
1,486
Top Hospitalization Risk — Next 90 Days
Members with highest predicted probability of acute admission
| # | Patient | Risk Score | Predicted Event | Key Drivers | Intervention |
|---|---|---|---|---|---|
| 1 | Maria Gonzalez | 94% | CHF Admission | 3 ER visits, worsening edema, diuretic gap | |
| 2 | Frank Mason | 91% | COPD Exacerbation | FEV1 decline, 2 ER visits, non-adherent | |
| 3 | Dorothy Chen | 88% | Stroke Recurrence | A-fib, BP uncontrolled, anticoagulant gap | |
| 4 | William Sanders | 82% | ESRD Complication | Missed dialysis sessions, hyperkalemia labs | |
| 5 | Robert Ashby | 79% | DM Complication | A1C 11.2%, no foot exam, polypharmacy | |
| 6 | Helen Torres | 76% | CKD Progression | eGFR decline 18%, hypertension, anemia | |
| 7 | James Whitfield | 73% | CAD Event | LDL elevated, statin non-adherent, smoker | |
| 8 | Carol Okafor | 68% | HTN Crisis | BP 168/104 last 3 readings, medication gap |
Showing top 8 of 20 predicted high-risk members