Application of Artificial Intelligence of Machine learning in Assessing Stroke Among HIV Patients on Protease Inhibitors-ART: A Bayesian Network Approach

Background In our investigation, we aim to utilize Bayesian network models in the field of machine learning to assess the likelihood of CVD in adults who are HIV positive and receiving protease inhibitors-antiretroviral therapy (PIs-ART). It is imperative to comprehend the risk factors and prognosis of stroke in order to effectively manage individuals infected with HIV, particularly those who have not yet initiated HAART. Methods This retrospective cohort study investigates stroke prevalence among HIV patients on Protease Inhibitors-ART at Zambia’s Adult Infectious Disease Center from 2009 to 2019. Data from 2867 patients’ EHRs were analyzed for demographic, clinical, and mortality information. Demographic and clinical data were obtained from an anonymous electronic case system. We utilized descriptive analysis along with logistic regression and Bayesian Network Model models to elucidate the characteristics and predictors of stroke among HAART-naive PLWH. Results This study analyzed data from 2867 HIV patients on Protease Inhibitors-ART to assess stroke prevalence and associated risk factors. Of the cohort, 105 individuals had stroke (prevalence: 3.7%), primarily ischemic infarction (56.2%). Most patients were aged 30-55 years (64.4%) and male (90.2%). Common comorbidities included diabetes (3.8%), hypertension (12.2%), and opportunistic infections like CMV (27.9%) and PCP (36.1%). Mortality rate was 6.6%. Bayesian network modeling predicted post-stroke outcomes, identifying age, CD4 count, lipid profile, comorbidities, and previous cardiovascular events as significant predictors. These findings highlight the complex interplay of risk factors in stroke occurrence among HIV patients on ART. Conclusions Our findings highlight the significance of early screening for stroke, timely intervention for risk factors across various age groups, and management of CD4 count among HAART-naive PLWH in order to alleviate the burden of stroke. These insights are crucial for informing targeted interventions aimed at reducing the occurrence and mortality associated with stroke in this population.