Key result
Surgical risk models correlate poorly with observed ~40% 30-day mortality in octogenarians undergoing emergency laparotomy.
Why the study?
The performance of established perioperative risk prediction models in elderly patients undergoing emergency laparotomy remains uncertain.
Do perioperative risk prediction models (NELA, ACS-NSQIP, SORT) accurately predict 30-day mortality in patients aged ≥80 years undergoing emergency laparotomy?
Cohort (n=155)
Yes
Do perioperative risk prediction models (NELA, ACS-NSQIP, SORT) accurately predict 30-day mortality in patients aged ≥80 years undergoing emergency laparotomy?
Effect estimate: NELA R²=0.40, AUC=0.8776
p-value: p=<0.0001
Established perioperative risk prediction models correlate poorly with observed 30-day mortality in octogenarian and nonagenarian patients undergoing emergency laparotomy, demonstrating substantial calibration error.
NELA overestimates mortality (58.8% predicted vs 39.7% actual) despite correlation; leaves open calibration for risk stratification in emergency laparotomy.
Background Emergency laparotomy (EL) carries a 30-day mortality (30-dm) of 8–18%, exceeding 22% in the elderly. The performance of established perioperative risk prediction models in elderly patients undergoing EL remains uncertain. This study evaluated the correlation between observed 30-day mortality and that predicted by the National-Emergency-Laparotomy-Audit (NELA), the American-College-of-Surgeons-National-Surgical-Quality-Improvement-Program (ACS-NSQIP), and the Surgical-Outcome-Risk-Tool (SORT) in Romanian patients aged ≥80 years. Methods This was a multicentre retrospective cohort study spanning 2 years and included all adults ≥80 undergoing EL at 4 acute care hospitals in North-Eastern Romania. The primary outcome was the correlation between actual 30-dm and NELA, ACS-NSQIP, and SORT-predicted 30-dm in patients aged ≥80 years. The study adhered to the TRIPOD checklist. Statistical analyses were performed using GraphPad Prism v10.6.1. Results 155 patients were included in the final analysis, representing 19.6% of all EL procedures performed during the study period. Median age was 84 (82,87) years; M/F ratio 0.43; ASA grade 4 (3,4); Clinical-Frailty-Score 6 (5,7), and Charlson-Comorbidity-Index 7 (5,9). Actual 30-dm was 39.7%. Predicted and correlated mortality estimates were: NELA 58.8% (43.8,79.1), R²=0.40, AUC=0.8776, p<0.0001; ACS-NSQIP unadjusted 24.5% (9.2,42.2), R²=0.35, AUC=0.8497, p<0.0001; ACS-NSQIP risk somewhat higher 29.6% (14.0,44.2), R²=0.33, AUC=0.8370, p<0.0001; ACS-NSQIP risk significantly higher 29.6% (15.3,44.2), R²=0.31, AUC=0.8278, p<0.0001; and SORT 26.5% (19.1,44.4), R²=0.14, AUC=0.7197, p<0.0001. Data are presented as median±IQR. Conclusion Risk prediction models correlate poorly with 30-dm and demonstrate substantial calibration error in elderly EL patients.
No takes yet. Share an insight, caveat, or question.
Iordache et al. (2026) conducted a cohort in Emergency laparotomy in elderly patients (n=155). Perioperative risk prediction models (NELA, ACS-NSQIP, SORT) vs. Observed 30-day mortality was evaluated on Correlation between actual 30-day mortality and predicted 30-day mortality (NELA R²=0.40, AUC=0.8776, p=<0.0001). Established risk prediction models (NELA, ACS-NSQIP, and SORT) correlated poorly with the observed 39.7% 30-day mortality in patients aged ≥80 years undergoing emergency laparotomy.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: