Can hospitals "game the system" by avoiding high-risk patients?

Abstract

Background: It has been suggested that implementation of quality-improvement benchmarking programs can lead to risk-avoidance behaviors in some physicians and hospitals in an attempt to improve their rankings, potentially denying patients needed treatment. We hypothesize that avoidance of high-risk patients will not change risk-adjusted rankings.

Study design: We conducted a simulation analysis of 6 complex operations in the Nationwide Inpatient Sample, including abdominal aortic aneurysm repair, aortic valve replacement, coronary artery bypass grafting, percutaneous coronary intervention, esophagectomy, and pancreatic resection. Primary outcomes included in-hospital mortality. Hospitals were ranked into quintiles based on observed-to-expected (O/E) mortality ratios, with their expected mortalities calculated based on models generated from the previous 3 years. Half of the hospitals were then randomly selected to undergo risk avoidance by avoiding 25% of patients with higher than median risks (ie, Charlson, Elixhauser, age, minority, or uninsured status). Their new O/E ratios and hospital-rank categories were compared with their original values.

Results: A total of 2,235,298 patients were analyzed, with an overall observed mortality rate of 1.9%. Median change in O/E ratios across all simulations was zero, and O/E ratios did not change in 97.5% to 99.3% of the hospitals, depending on the risk definitions. Additionally, 70.5% to 98.0% of hospital rankings remained unchanged, 1.3% to 13.1% of hospital rankings improved, and 0.7% to 14.3% of hospital rankings worsened after risk avoidance.

Conclusions: Risk-adjusted rankings of hospitals likely cannot be changed by simply avoiding high-risk patients. In the minority of scenarios in which risk-adjusted rankings changed, they were as likely to improve as worsen after risk avoidance.