[Development and validation of a prediction model to identify HIV+ patients with drug-related problems. A prediction study]

Abstract

Objective: To develop and validate a prediction model for the detection of drug-related problems (DRP) in patients on antiretroviral treatment during their regular monitoring in Pharmaceutical Care Clinics before dispensing drugs. MEHOD: Open multicentre prospective study. HIV patients with and without DRP were included. Demographic, clinical and pharmacotherapy variables (related and unrelated to antiretroviral treatment) were included in the model design. To find prognostic factors for DRP, a binary logistic regression model was created after performing a univariate analysis that identified independent variables related to DRP these variables were introduced in the multivariate model for the final selection. Model validity was determined by the shrinkage method and the discriminatory power by Harrell's C-index.

Results: 733 patients were included. The variables "adherence", "prescription of drugs needing dosage adjustment", and "total number of drugs prescribed (apart from the antiretroviral treatment)" were independently related to the appearance of DRP. Probabilities predicted by the model, customising the coefficients using the uniform shrinkage method, showed a R(2)=0.962 for the construction sample and a R(2)=0.872 for the validation sample. The discriminatory capacity of the model was 0.816 for the construction sample and 0.779 for the validation sample.

Conclusions: The developed and validated model enables detection of patients on antiretroviral treatment who are at a higher risk for experiencing a DRP. The prediction variables employed are commonly used in patient's pharmacotherapy record, allowing the model to be used in routine clinical practice.