Can preventable adverse events be predicted among hospitalized older patients? The development and validation of a predictive model.

OBJECTIVE To develop and validate a predictive model for preventable adverse events (AEs) in hospitalized older patients, using clinically important risk factors that are readily available on admission. DESIGN Data from two retrospective patient record review studies on AEs were used. Risk factors included patient characteristics as well as admission and organizational characteristics. Multilevel logistical regression analysis was used to develop the model. Backward elimination was applied to identify the most parsimonious model. SETTING Twenty-one Dutch hospitals were included in the 2004 sample and 20 Dutch hospitals in the 2008 sample. PARTICIPANTS A total of 3977 patients aged 70 years or over who were admitted to a Dutch hospital in 2004 and 2119 patients aged 70 years or over admitted in 2008. MAIN OUTCOME MEASURES Identified predictors of preventable AEs in older patients. RESULTS In 2004 predictors of preventable AEs in patients aged 70 years or over were increased age (OR 1.04, confidence interval (CI) 1.01-1.06); elective admission (OR 1.65, CI 1.14-2.40) and admission to a surgical department (OR 1.53, CI 1.08-2.16). The area under the receiver operating characteristic curve for the 2004 sample was 0.60 and for 2008, 0.59. CONCLUSIONS This study showed that several expected risk factors for preventable AEs in older patients, including comorbidity, could not predict these events. It was not possible, using in-patient data available on admission and collected during the course of two patient record review studies, to develop a satisfactory predictive model for preventable AEs in older patients.