The aims of the study were to develop risk-adjusted models and apply them for comparisons of hospital performance to define potentially preventable adverse outcomes (OAs) in Medicare lung resection
The Medicare Limited Data Set for 2010–2012 was used to design predictive risk models for the four OAs of inpatient deaths, prolonged length-of-stay outliers, 90-day postdischarge deaths without hospital readmission, and 90-day readmissions after removal of unrelated readmission events. The probability of adverse events for each hospital was used to compute the hospital-specific standard deviation (SD) tailored to patient risk profiles. Observed versus predicted adverse events divided by the hospital-specific SD identified the z
score for each hospital. Risk-adjusted OA rates were then computed for comparing hospital performance.
A total of 39,405 lung resection
patients from 739 hospitals had 768 inpatient deaths (1.9%), 3147 had prolonged LOS (8.0%), 514 had 90-day postdischarge deaths without readmission (1.3 %), and 7701 had one or more 90-day readmissions (19.5%); 10,924 patients (27.7%) had one or more of these OAs. Twenty-six hospitals were two SDs better than predicted and 34 hospitals were two SDs poorer than predicted. When evaluated by deciles of risk-adjusted OAs, the top performing decile of hospitals had rates of 14.3% and the poorest performing decile had OA rates of 41.0%.
The differences in risk-adjusted comparative outcomes between top- and suboptimal-performing hospitals in lung resections define the potential opportunities for care improvement. Identification of risk factors associated with OAs and causes for readmissions provides direction for specific areas of care redesign for improvement.