Preliminary prediction model for identifying patients with the possibility of pharmacotherapy improvement

Abstract

Objective

To develop a prediction model for identifying patients with the possibility of improving pharmacotherapy during the process of pharmaceutical validation of the prescription.

Method

Cross-sectional study over two months, performed in the Internal Medicine and Infectious Disease divisions. Detecting opportunities for improving quality of pharmacotherapy is done by means of a pharmacist's validation of the prescription. Based on the information we obtained through this process, we performed a multivariate logistic regression analysis using as prognostic factors the demographic, pharmacotherapy and clinical variables related to identifying any drug-related problems (DRPs) in the patient. The model's prediction validity was assessed using the diagnostic performance curve and calculating the area under it.

Results

The final prediction model included the variables age, cardiovascular drugs (digoxin) and drugs for which a dosage adjustment is recommended in the case of organ failures. Analysis of the ROC curve showed an estimated area under the curve AUCROC) of 84.0% (95% CI: 80.5–87.1), a sensitivity value of 28% (95% CI: 24.07–32.19), a specificity value of 99.10% (95% CI: 97.80–99.73), a positive predictive value of 77.78% and a negative predictive value of 92.41%.

Conclusion

The resulting prediction model enables population-based detection of pharmacotherapy safety risks in adult patients admitted to the selected hospital units. The predictive variables used by the model are commonly used in daily practice.