Functional Multivariable Logistic Regression With an Application to HIV Viral Suppression Prediction.

Area of Science:

• Biostatistics
• Health Informatics
• Epidemiology

Background:

• Accurate prediction of human immunodeficiency virus (HIV) suppression status is crucial for patient management.
• Electronic Health Records (EHR) offer rich longitudinal data for predictive modeling.
• Existing models may not fully capture the complex longitudinal nature of patient data.

Purpose of the Study:

• To develop and evaluate a novel functional multivariable logistic regression model.
• To simultaneously analyze longitudinal binary and continuous processes from EHR data for HIV suppression prediction.
• To enhance the prediction of viral suppression status in people living with HIV.

Main Methods:

• Functional principal components analysis (FPCA) to model longitudinal binary and continuous variables.
• Logistic regression incorporating FPCA scores for prediction.
• Penalized splines for estimation, group-lasso for variable selection, and multivariate FPCA for score revision.

Main Results:

• The proposed model effectively accounts for simultaneous longitudinal binary and continuous data.
• Simulation studies demonstrated the method's validity and performance.
• Application to EHR data from South Carolina showed promise in predicting HIV viral suppression.

Conclusions:

• The functional multivariable logistic regression model offers a robust approach for predicting HIV suppression status.
• This method leverages complex longitudinal EHR data for improved clinical insights.
• The findings support the use of advanced statistical techniques in HIV care and management.