OBJECTIVES: We examine the relationship of several hospital-level factors to inpatient safety using a 12-year (2007-2018) panel of all acute-care general hospitals in the state of Florida.

METHODS: We analyze four key determinants of patient-safety outcomes: expenditure on peer review of medical quality, expenditure on the continuing education of nurse staff, expenditure on data processing (i.e. IT expenditure), and a measure of per-patient staffing. Our outcome measure is the composite patient safety indicator (PSI-90) from the Agency for Healthcare Research and Quality (AHRQ). The composite is calculated as the weighted sum of 10 risk-adjusted rates of potentially preventable adverse events. Each risk-adjusted rate is normalized to (i.e. divided by) a nationally-representative population rate. All four of our key determinants of patient safety are potentially endogenous. We attempt to mitigate the distortions of endogeneity by: 1) exploiting the panel structure of our data, controlling for unobserved and time-invariant hospital-level factors and 2) employing a proxy-variable approach, including several hospital-level covariates into the regression of patient safety to control for unobserved determinants of PSI-90. These proxy variables represent broadly the context of the hospital’s safety culture (Kaplan et al. 2010) that are largely unmeasurable but that likely influence one or more of our key determinants and our outcome variable. We define four such safety culture domains: leadership, organizational type, organizational structure, and market condition.

RESULTS: We find that a standard-deviation increase in expenditure on the education and training of nursing staff, and on data processing, produce improvements in patient safety of 4 to 6 percent. Other determinants are not statistically significant.

CONCLUSIONS: Hospitals seeking t improve patient safety should focus expenditures in the education and training of nursing staff, and on data processing improvements.