PURPOSE: The potential impact of drug-related problems (DRP) on morbidity and mortality is a serious concern in hospitalized patients. This study aimed to design a risk score to identify patients most at risk of a DRP. METHODS: Data from patients admitted to a tertiary university hospital between January and August 2009 were used to design the risk score (training set). DRP were detected through a pharmacy warning system integrated in the computerized medical history. The variables associated with developing a DRP were identified through a binary multivariate logistic regression analysis and were used to compute the DRP risk score, which was subsequently validated in patients admitted between September and December 2009 (validation set). RESULTS: Of the 8713 patients included in the training set, at least one DRP was detected in 2425 (27.8%). Prescription of a higher number of drugs, higher comorbidity, advanced age, certain groups of the Anatomical Therapeutic Chemical classification system, and some major diagnostic categories were associated with risk of DRP. These variables were used to compute the DRP risk score. The area under the receiver operator characteristic curve was 0.778 (95%CI [0.768, 0.789]). Of the 4058 admissions included in the validation set, at least one DRP was detected in 876 (21.6%). The area under the receiver operator characteristic curve was 0.776 (95%CI [0.759, 0.792]). CONCLUSIONS: Knowledge of the variables associated with DRP could aid their early detection in at-risk patients. The use of an application that can be continually updated in daily clinical practice helps to optimize resources.
PURPOSE: The potential impact of drug-related problems (DRP) on morbidity and mortality is a serious concern in hospitalized patients. This study aimed to design a risk score to identify patients most at risk of a DRP. METHODS: Data from patients admitted to a tertiary university hospital between January and August 2009 were used to design the risk score (training set). DRP were detected through a pharmacy warning system integrated in the computerized medical history. The variables associated with developing a DRP were identified through a binary multivariate logistic regression analysis and were used to compute the DRP risk score, which was subsequently validated in patients admitted between September and December 2009 (validation set). RESULTS: Of the 8713 patients included in the training set, at least one DRP was detected in 2425 (27.8%). Prescription of a higher number of drugs, higher comorbidity, advanced age, certain groups of the Anatomical Therapeutic Chemical classification system, and some major diagnostic categories were associated with risk of DRP. These variables were used to compute the DRP risk score. The area under the receiver operator characteristic curve was 0.778 (95%CI [0.768, 0.789]). Of the 4058 admissions included in the validation set, at least one DRP was detected in 876 (21.6%). The area under the receiver operator characteristic curve was 0.776 (95%CI [0.759, 0.792]). CONCLUSIONS: Knowledge of the variables associated with DRP could aid their early detection in at-risk patients. The use of an application that can be continually updated in daily clinical practice helps to optimize resources.