Description:

Ryan Ellis, MD

Research Objective: Public reporting of hospital performance can influence where patients seek surgical care. Existing public reporting tools do not generate hospital rankings that incorporate differences in patient age, comorbidities, or surgical indication. This lack of patient-level stratification potentially limits the accuracy of the rankings for some patients. The aim of this study was to evaluate whether hospital performance rankings changed for low- vs. high-risk patients.

Methods: Retrospective cohort study using the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP). All patients undergoing proctectomy (N=10,995), colectomy (N=82,683), or pancreatectomy (N=13,573) were identified from the ACS NSQIP database from 2009 to 2013. Mutually exclusive low-risk and high-risk patient strata were created based on age, functional status, cardiac/respiratory comorbidities, diabetes, and American Society of Anesthesiologists (ASA) classification. Risk-adjusted hospital performance rankings for mortality, Death or Serious Morbidity (DSM), serious morbidity, and Surgical Site Infection (SSI) were derived using hierarchical random slope and intercept models. Hospital rankings were determined in low- and high-risk groups separately and the two ranking systems were compared using Spearman’s correlation (ρ) and weighted kappa (κ) statistics.

Principal Findings: Our analysis included 364 hospitals for proctectomy, 405 hospitals for colectomy, and 296 hospitals for pancreatectomy. The unadjusted hospital-level complication rates by procedure ranged from 1.5% to 3.6% for mortality, 15.3% to 21.8% for DSM, 14.9% to 21.5% for serious morbidity, and 11.2% to 18.4% for SSI. Hospital rankings for proctectomy varied most based on patient risk profile, with 56% of top-quintile hospitals for mortality in low-risk patients falling out of the top quintile when managing high-risk patients (ρ=0.66, κ=0.61). Moreover, 11.1% of hospitals ranked in the best quintile for mortality in low-risk cases shifted to the worst quintile for high-risk patients. For colectomy, the rankings across all outcomes were relatively similar for low- and high-risk patients, with the largest variation occurring for Serious Morbidity (ρ=0.94, κ=0.80). Pancreatectomy had the most stable rankings across all outcomes, with the least variation seen in DSM (ρ=1.00, κ=0.99) and the most variation seen in overall mortality (ρ=0.93, κ=0.84).

Conclusions: Hospital performance rankings varied considerably based on patient risk factors for proctectomy and were relatively stable for colectomy and pancreatectomy. Incorporating patient-specific risk factors into public reporting tools may provide more tailored information to patients about where to seek care for certain procedures. Further identification of which procedures exhibit the largest patient risk effects may help hospitals and practitioners deploy targeted perioperative interventions aimed at improving outcomes for high-risk patients.