Table 2.
Model parameters for recalibration of the Risk Analysis Index (RAI) comparing origin RAI (RAI-A) to the recalibrated RAI (RAI-rev) in VASQIP dataset; external validation of RAI-rev in ACS-NSQIP cohort and subcohorts of men and women.
| Outcome | Sample | Predictor | C-statistic (95% C.I.) | AIC | Max. R2 |
|---|---|---|---|---|---|
| VASQIP Recalibration | |||||
| 30-day Mortality | Total [N=480,731] | RAI-A | 0.842 (0.835–0.848) | 47,002.2 | 0.1990 |
| RAI-rev | 0.864 (0.858–0.869) | 45,104.2 | 0.2330 | ||
| 180-day Mortality | Total [N=480,731] | RAI-A | 0.813 (0.810–0.817) | 120,967.0 | 0.2110 |
| RAI-rev | 0.842 (0.839–0.845) | 114,881.8 | 0.2550 | ||
| 365-day Mortality | Total [N=480,731] | RAI-A | 0.784 (0.781–0.787) | 175,931.5 | 0.1970 |
| RAI-rev | 0.816 (0.814–0.819) | 167,259.0 | 0.2440 | ||
| ACS-NSQIP External Validation | |||||
| 30-day Mortality | Female [N=807,087] | RAI-rev | 0.885 (0.881–0.889) | 55,462.3 | 0.2398 |
| Male [N=584,698] | RAI-rev | 0.845 (0.841– 0.850) | 63,368.8 | 0.1999 | |
| Total [N=1,391,785] | RAI-rev | 0.870 (0.867–0.873) | 118,997.0 | 0.2221 | |
| RAI-C NWIHCS Validation | |||||
| 30-day Mortality | Total [N=6,803] | RAI-C | 0.704 (0.596–0.812) | 356.8 | 0.0590 |
| RAI-C- rev | 0.743 (0.657–0.829) | 353.4 | 0.0690 | ||
| 180-day Mortality | Total [N=6,419] | RAI-C | 0.772 (0.727–0.816) | 1,030.1 | 0.1120 |
| RAI-C-rev | 0.804 (0.766–0.842) | 1,000.7 | 0.1400 | ||
p<0.0001 for all model comparisons between original RAI and RAI-rev in VASQIP and ACS-NSQIP cohorts; p=0.204 for 30-day mortality and p<.001 for 180-day mortality in the NWIHCS sample;
VASQIP: Veterans Affairs Surgical Quality Improvement Program; ACS-NSQIP: American College of Surgeons National Surgical Quality Improvement Program; NWIHCS: Nebraska Western Iowa Health Care System; CI: Confidence interval; AIC: Aikake information criterion