Table 3. C Statistics for Different Models Predicting Fatal and Nonfatal Opioid Overdosesa.
| Models | C statistic (95% CI) |
|---|---|
| Predicting fatal opioid overdose in 2016 | |
| 1. Demographics | 0.692 (0.679-0.710) |
| 2. Demographics + PDMP | 0.789 (0.780-0.804) |
| 3. Demographics + hospital | 0.823 (0.811-0.839) |
| 4. Demographics + hospital + PDMP | 0.864 (0.855-0.875) |
| 5. Demographics + PDMP + hospital + behavioral services | 0.889 (0.880-0.900) |
| 6. Demographics + PDMP + hospital + behavioral services + criminal justice | 0.894 (0.883-0.903) |
| Predicting nonfatal opioid overdoses in 2016 | |
| 1. Demographics | 0.576 (0.570-0.582) |
| 2. Demographics + PDMP | 0.732 (0.726-0.738) |
| 3. Demographics + hospital | 0.769 (0.762-0.775) |
| 4. Demographics + PDMP + hospital | 0.820 (0.815-0.825) |
| 5. Demographics + PDMP + hospital + behavioral services | 0.847 (0.841-0.850) |
| 6. Demographics + PDMP + hospital + behavioral services + criminal justice | 0.851 (0.846-0.855) |
Abbreviation: PDMP, the Prescription Drug Monitoring Program.
a
All risk information is from 2015. Results are based on bootstrapping (selecting repeated subsamples).