A predictive model of an effectiveness of antiviral drugs treated in PLWHA under accountability of The Office of Disease Prevention and Control 6 Khon Kaen

Abstract

Objective: To find a predictive model of effectiveness of antiviral drug used among people infected with HIV/AIDS patients.

          Methods:  The retrospective cohort study was conducted using NAP (National AIDS Program) data in the part of under responsibility of the Office of Disease Prevention and Control, Region 6, Khon Kaen province. The data for creating the model were collected from 1^st April, 2011 to 31^st May, 2012. The data for validating the model were collected from 1^st June, 2012 to 5^th July, 2013.The multiple logistic regression was employed to find the model. The ability of a discrimination of the model was assessed via an area under the receiver operating characteristics (ROC) curve.

          Results: After treated with antiviral drugs for one year, approximately 73.8% of 317 patients (95%CI=68.6-78.6) had a plasma HIV RNA viral load below 50 copies/mL, and about 71.6% of them (95%CI=66.3-76.5) had CD4 cell counts of > 100 cell/mm.³ The variables in the predictive model included sex, adherence , antiviral drug formula, and CD4 levels before treatment. The model had the area under ROC curve equaled 64.3 (95%CI: 57.7-71.0). At the probability cut-off point > 0.71, the model had sensitivity 76.1% (95%CI: 71.4-80.8), specificity 41.0% (95%CI: 35.6-46.4), accuracy 66.9% (95%CI: 61.4-72.0), positive predictive value 78.4% (95%CI: 73.9-82.9), negative predictive value 37.8% (95%CI: 32.4-43.1), LR⁺ 1.3, and LR^- 0.6. The validated model had the area under the ROC curve equaled to 60.1 (95%CI: 57.7-71.0). At the probability cut-off point of > 0.71, the model had sensitivity 31.76% (95%CI: 25.7-37.8), specificity 83.5% (95%CI: 78.7-88.4), accuracy of 49.8% (95%CI: 43.1-56.5), positive predictive value of 78.3% (95%CI: 73.0-83.7), negative predictive value of 39.5% (95%CI: 33.2-45.9), LR⁺ 1.2, and LR^- 0.3.

          Conclusions: The predictive model had low ability in discrimination. However, the result showed that people infected with HIV/AIDS patients with low adherence and had low CD4 counts before treatment were at risk of having higher levels of plasma viral load after treatment. In order to reduce mortality, risks of complications and have living longer among patients, providers and patients should be concerned on the issues and give attention in treatment participation.