Defining Volume Thresholds for Colorectal Cancer Surgery Using Stratum Specific Likelihood Ratios

Authors:

Abstract
BACKGROUND: Higher procedure volume correlates with improved surgical outcomes, yet recent literature has questioned whether this relationship persists in modern times. Reports of an attenuated volume-outcome association may reflect the limitations of a single, static, cutoff rather than true weakening of the underlying relationships. Stratum-specific likelihood ratios generate multiple empirically derived volume strata that may detect persistent gradients that are obscured by binary threshold approaches. OBJECTIVES: (1) To define hospital-level volume thresholds for colorectal cancer resections using stratum-specific likelihood ratios, (2) assess these thresholds across multiple outcome measures, procedure types, and tumor locations, and (3) examine whether the volume-outcome relationships persist across temporal cohorts. DESIGN: Retrospective cohort study. SETTING: National Cancer Database from 2004-2020. PATIENTS: Patients undergoing colorectal resection for cancer. Patients with metastatic disease were excluded. MAIN OUTCOME MEASURES: Primary outcome was 90-day mortality. Secondary outcomes included 30-day mortality and textbook outcome (length of stay ≤7 days, negative surgical margin, survival >90 days, and no 30-day readmission). RESULTS: Overall, 1,052,875 patients underwent colorectal resection at 1344 hospitals. Stratum-specific likelihood ratios identified 6 volume strata (≤42, 43-70, 71-108, 109-280, 281-439, and ≥440 colorectal resections per year) with progressively decreasing 90-day mortality across groups (≤42: 5.9% to ≥440: 0.6%). The threshold of 71 resections per year was identified as the optimal threshold. Consistent gradients were also observed across 30-day mortality and textbook outcome and persisted across procedure types and tumor locations. Volume-outcome associations remained significant across all temporal cohorts, including the most recent period (2016-2020; p Video Abstract.