Rationale of calculus
A total of 349 HIV-infected patients who started antiretroviral therapy in 2012-2015 were evaluated retrospectively at a University Hospital in Barcelona (Spain).
Adherence was assessed every two months by pharmacy refills at the hospital pharmacy and self-report. A patient not attending <90% to pick up prescription drugs from the hospital pharmacy and / or ART interruption for more than a week was considered non-adherent.
Univariate and multivariate logistic regression (LR) was used to test the association of risk factors with non-adherence. The probabilities obtained from the multivariate model were considered as a non-adherence score, ranging from 0 to 100.
The best cut-off point was obtained using the bootstrapping methodology.
The C statistic was evaluated as a model adjustment measure. The sample size had enough statistical power (about 90%) to detect as significant a C-statistic of at least 0.6.
The C statistic for the score was 0.91 (95% CI: 0.87 to 0.94). The optimal cutoff point was set at> 29. This cutoff has: sensitivity 0.81, specificity 0.87, positive predictive value 0.74, and negative predictive value 0.94.
PRESENTED AT THE 9TH IAS CONFERENCE ON HIV SCIENCE - PARIS, FRANCE. 23-26 July 2017
Development of a risk score based on patient's individual factors for predicting non-adherence to antiretroviral therapy
Authors: H. Knobel1, A. Guelar1, S. Calzado1, A. Retamero2, M. De Antonio2, J. Vila3, I. Subirana3, R. Guerri1, G. Casamayor1, E. Lerma1, J. Villar1, A. Gonzalez1, S. Luque2
Affiliation: 1Dept. Infectious Diseases, 2 Pharmacy Dept. Hospital del Mar, 3CIBER Epidemiologia y Salud Publica, Barcelona, Spain
Address: Paseo Marítimo 25-29 08003 Barcelona, Spain. Contact: Hernando Knobel (Head HIV Unit) firstname.lastname@example.org