Development and validation of a scoring system for the prediction of HIV drug resistance in Hubei province, China

Abstract

ObjectiveThe present study aimed to build and validate a new nomogram-based scoring system for the prediction of HIV drug resistance (HIVDR).Design and methodsTotally 618 patients with HIV/AIDS were included. The predictive model was created using a retrospective set (N = 427) and internally validated with the remaining cases (N = 191). Multivariable logistic regression analysis was carried out to fit a model using candidate variables selected by Least absolute shrinkage and selection operator (LASSO) regression. The predictive model was first presented as a nomogram, then transformed into a simple and convenient scoring system and tested in the internal validation set.ResultsThe developed scoring system consisted of age (2 points), duration of ART (5 points), treatment adherence (4 points), CD4 T cells (1 point) and HIV viral load (1 point). With a cutoff value of 7.5 points, the AUC, sensitivity, specificity, PLR and NLR values were 0.812, 82.13%, 64.55%, 2.32 and 0.28, respectively, in the training set. The novel scoring system exhibited a favorable diagnostic performance in both the training and validation sets.ConclusionThe novel scoring system can be used for individualized prediction of HIVDR patients. It has satisfactory accuracy and good calibration, which is beneficial for clinical practice.