• English 
  • Spanish 

Statistical model to determine variables of influence for adjustment in Spanish public hospitals

In this work we analyze the variability observed in hospital outcomes as measured by a relevant set of indicators related to hospital performance and quality of care (average stay, mortality rate, etc.). A multivariate statistical model for categorical data is introduced (a probabilistic Bayesian network) to find relevant associations between these indicators and a minimum set of administrative and clinical discharge variables (complexity, severity, diagnosis category, age, gender, etc.). To this aim we use a total of ~3.5 million records, corresponding to the discharges occurred in the Spanish National Health (SNH) system in 2004. The model allows obtaining the most significant variables influencing the hospital outcome indicators, thus explaining part of the variability observed among different hospitals in terms of clinical and administrative casuistry. Thus, the outcome indicators can be adjusted for an unbiased estimation of hospital care quality.