Establishment of a risk prediction for tuberculosis infection model in HIV-infected/AIDS patients

Abstract

Objective To investigate the infectious status of Mycobacterium tuberculosis (TB) in human immunodeficiency virus (HIV)-infected/acquired immune deficiency syndrome (AIDS) patients, and to establish a prediction model for infection risk. Methods A retrospective cohort study was conducted on 816 HIV-infected/AIDS patients who were registered in our hospital from January 2017 to August 2022. Then with the aid of envelopes, 571 cases were assigned into a testing group and 245 cases into a validation group at a ratio of 7 ∶3. Relevant indicators that may cause TB infection were collected in these patients. The patients were also divided into tuberculosis group and non-tuberculosis group. The relevant factors were compared between the 2 groups. After screening variables with LASSO, multivariate logistic regression analysis was performed and the obtained factors were used to establish a nomogram model, which was further verified. Results Among the 571 HIV-infected/AIDS patients in the model group, 59 (10.33%) were diagnosed as TB infection. Multivariate logistic regression analysis showed that age, drug abuse history, BCG vaccination history, recent CD4+ T cell level, tuberculosis contact history, and cough history in recent 1 year were independent influencing factors of TB infection in the patients (P<0.05). Then a nomograph model for predicting the risk of TB infection was constructed in the HIV-infected/AIDS patients. ROC curve analysis indicated that the area under the curve (AUC) of the predictive value of TB infection risk in HIV infected/AIDS patients was 0.823, and the 95%CI was 0.761~0.884; the sensitivity was 77.97%, the specificity was 96.48%, the positive predictive value was 71.88%, and the negative predictive value was 97.44% in the model group. The AUC value of the validation group was 0.817, the 95%CI was 0.723~0.912, the sensitivity was 76.92%, the specificity was 96.35%, the positive predictive value was 71.43%, and the negative predictive value was 97.24%. The calibration curve results suggested that the prediction models of the model group and the validation group were basically matched with the ideal model, indicating that the model accuracy is good. Conclusion TB infection of HIV-infected/AIDS patients is mainly affected by age, drug abuse history, BCG vaccination history, recent CD4+ T cell level, tuberculosis contact history and cough history in the past 1 year. Our prediction model established based on the above factors has high accuracy and discrimination in predicting TB infection risk of these patients.

Keywords

• human immunodeficiency virus
• acquired immune deficiency syndrome
• mycobacterium tuberculosis
• multivariate analysis
• prediction model