MS Vaughan Sarrazin, Iowa City VA Medical Center and University of Iowa College of Medicine; SM Fuehrer, MBA, Cleveland VA Medical Center; DL Harper, DO, Cleveland Health Quality Choice; GE Rosenthal, MD, Iowa City VA Medical Center, VA Quality Scholars Program, and University of Iowa College of Medicine

Objectives: Compare in-hospital mortality and length of stay (LOS) in a VA hospital and private sector hospitals serving the same health care market, adjusting for admission severity of illness.

Methods: The medical records of 1960 consecutive patients admitted to a large university-affiliated VA hospital in 1994-95 with 9 diagnoses (myocardial infarction, pneumonia, stroke, congestive heart failure, gastrointestinal hemorrhage, obstructive lung disease, bowel resection, peripheral vascular surgery, and coronary artery bypass surgery) were abstracted by trained reviewers. Comparable data from 157,147 patients in 30 private sector hospitals were obtained from a regional initiative to measure hospital performance. For each patient, the probability of death (0 to 100%) and expected LOS were estimated from validated disease-specific multivariable models based on demographic and clinical findings from the first 48 hours of hospitalization. For the 9 mortality models, discrimination was relatively similar in VA and private sector patients (mean c-statistics, 0.84 vs. 0.87, respectively); for the LOS models, model performance was somewhat lower in VA patients (mean R-square values, 0.14 vs. 0.23). Predicted probabilities of death and LOS were entered into logistic or linear regression models that included an indicator variable for VA hospitalization. LOS comparisons were based on log transformed data. Nested analyses were used to account for hospital-level variation.

Results: Compared to private sector patients, VA patients were younger (mean ages, 66 vs.68 years; p<.001) and were more likely (p<.001) to be male (98% vs. 49%) and nonwhite (35% vs. 22%). Unadjusted mortality was similar in VA and private sector patients (5.0% vs. 5.6%, respectively; p=.26), although mean LOS was longer (12.7 vs. 7.0 days; p<.001). In analyses adjusting for severity of illness, the odds of death in VA patients was similar (OR 1.09; 95% CI; 0.86-1.39; p=.46). However, VA patients died later during hospitalization (p<.001); for example, 37% of VA deaths occurred after 21 days, compared to 8% of private sector deaths. The odds of death in VA patients were actually lower in analyses limited to deaths during the first 2 (OR 0.39; 95% CI; 0.18-0.84; p=.02), 7 (OR 0.59; 95% CI; 0.39-0.89; p=.01), 14 (OR 0.65; 95% CI; 0.47-0.90; p=.01), or 21 (OR 0.74; 95% CI; 0.56-0.99; p=.04) hospital days. In linear regression analyses, adjusted LOS was longer (p<.001) in VA patients for each of the 9 diagnoses; for 6 of the diagnoses, percent difference in LOS were greater than 30%.

Conclusions: If generalizable to VA hospitals in other regional markets, our findings suggest that an important indicator of performance—in-hospital mortality—may be similar or lower in VA, relative to private sector, hospitals. The longer LOS of VA patients may reflect differences in utilization patterns or accessibility of post acute care resources and may be an important source of bias in comparisons of VA and private sector hospital mortality.

Impact: The findings provide insight into the quality and efficiency of a VA hospitals. While the process of care in VA hospitals may meet or exceed community standards, the longer LOS of VA patients may represent opportunities for substantial cost savings.