Correlation of Proposed Surgical Volume Standards for Complex Cancer Surgery with Hospital Mortality

Abstract

Minimum case volume thresholds for complex cancer operations have been proposed by the Leapfrog Group. There has been no formal study of how these standards correlate with actual hospital mortality. The National Cancer Database was used to identify patients undergoing operations for esophageal, lung, pancreatic, and rectal cancer between 2013 and 2015. Recommended annual hospital case volume was used to divide hospitals into those meeting a minimum volume threshold (VT) and those below it. Hospitals in the highest quartile of adjusted hospital mortality were designated as poor performing hospitals (PPHs). Sensitivity, specificity, negative predictive value, and positive predictive value of current minimum VTs to predict PPHs were calculated. The proportion of hospitals meeting minimum VTs varied from 7% for esophagectomy to 27% for rectal operations. Proposed minimum VTs had a sensitivity of 69% to 93%, specificity of 7% to 27%, and area under the curve of 0.59 to 0.65 for identifying PPHs. Although the negative predictive value varied from 72% to 79%, the positive predictive value was only 24% to 26%. Optimal minimum VTs to identify PPHs were lower than those currently proposed-esophagus was 4 vs 20, lung was 21 vs 40, pancreas was 7 vs 20, and rectum was 8 vs 16. Even under these idealized volume cutoffs, the best performing procedure-specific model (esophagus) had an area under the curve of 0.68. Although proposed minimum VTs are reasonably good at identifying PPHs, they misclassify 3 of 4 hospitals below the minimum VT as PPHs and 1 of 4 PPHs as meeting the minimum VT. Use of case volume cutoffs alone does not correlate well with actual hospital mortality.