Prediction of drug-related problems in diabetic outpatients in a number of hospitals, using a modeling approach

OBJECTIVE: Drug-related problems (DRPs) are considered a serious, expensive, and important undesirable complication of health care. However, as current health care resources are limited, pharmacist DRP services cannot be provided to all patients. Using a modeling approach, we aimed to identify risk factors for DRPs so that patients for DRP-reduction services can be better identified. METHODS: Patients with diabetes from outpatient clinics from five key university-affiliated and public hospitals in Jordan were assessed for DRPs (drug without an indication, untreated indication, and drug efficacy problems). Potential risk factors for DRPs were assessed. A logistic regression model was used to identify risk factors using a randomly selected, independent, nonoverlapping development (75%) subsample from full dataset. The remaining validation subsample (25%) was reserved to assess the discriminative ability of the model. RESULTS: A total of 1,494 patients were recruited. Of them, 81.2% had at least one DRP. Using the development subsample (n=1,085), independent risk factors for DRPs identified were male gender, number of medications, prescribed gastrointestinal medication, and nonadherence to self-care and non-pharmacological recommendations. Validation results (n=403) showed an area under the receiver operating characteristic curve of 0.679 (95% confidence interval=0.629–0.720); the model sensitivity and specificity values were 65.4% and 63.0%, respectively. CONCLUSION: Within the outpatient setting, the results of this study predicted DRPs with acceptable accuracy and validity. Such an approach will help in identifying patients needing pharmacist DRP services, which is an important first step in appropriate intervention to address DRPs.