Prediction of interpersonal personality for people living with HIV using machine learning

Abstract

Background: The ultimate goals of HIV treatment and care has changed from viral suppression to improving quality of life due to the development of antiretroviral therapy (ART). Social functioning is an important aspect of quality of life, which is also associated with many health outcomes. Interpersonal personality influences individuals’ tendencies and preferences in the process of interpersonal communication. This study aims to develop and validate a prediction model of interpersonal personality for people living with HIV (PLWH) using machine learning. Methods: We recruited participants from seven HIV/AIDS designated hospitals in China from 2022 to 2023. This study included 3,040 participants and collected 16 demographic and clinical variables according to the social determinants of health (SDOH) framework. A standard questionnaire was used to collect data on interpersonal personality. Five machine learning algorithms were applied to make predictions: random forest (RF), extreme gradient boosting (XGBoost), gradient boosting machine (GBM), bidirectional recurrent neural network (BRNN), and neural network (NNet). Results: XGBoost was found to have the best prediction performance, with a root mean square error (RMSE) value of 2.53. We used XGBoost to analyze the 16 variables and found that the top five most heavily weighted variables were CD4+ T cell count, age, months since ART, months since HIV diagnosis, and viral load, suggesting that these five variables had the greatest impact on PLWH’s interpersonal personality. The performance of the XGBoost model was evaluated, and significant correlations (P < 0.05) were found between the measured and predicted scores. Conclusions: The interpersonal personality of PLWH can be predicted by demographic and clinical variables, with high importance weights for HIV-related clinical variables. It may be influenced and shaped after HIV infection, which highlights the profound impact of HIV infection on PLWH even in the future functional cure era. Medical professionals should be aware that interventions can be designed to buffer the impact of HIV infection on the interpersonal personality of PLWH. Subjective needs and preferences that influenced by interpersonal personality are important for the development of social support interventions for PLWH.