Using Machine Learning to Predict Young People's Internet Health and Social Service Information Seeking.

Area of Science:

• Digital health
• Machine learning applications in public health
• Adolescent health information seeking behavior

Background:

• Youth at risk for HIV (YARH) actively seek health information online.
• Previous research focused on descriptive associations, lacking predictive model strength estimation.
• Machine learning offers new opportunities for designing targeted digital health interventions.

Purpose of the Study:

• To develop and evaluate machine learning models predicting health information seeking among YARH.
• To identify key predictors of general health, sexual and reproductive health (SRH), and social service information seeking.
• To inform the design of digital health interventions for YARH.

Main Methods:

• Applied elastic net and lasso regression models to self-reported internet use data from 1287 YARH (aged 14-24).
• Utilized sociodemographic characteristics, minority identity, healthcare access, sexual behavior, substance use, and mental health as predictor variables.
• Assessed model performance using Area Under the ROC Curve (AUC), with AUC ≥ 0.75 indicating strong predictive power.

Main Results:

• Approximately half of YARH sought general health and SRH information; 26% sought social service information.
• Models demonstrated strong predictive accuracy (AUC ≥ 0.75).
• Higher education and sexual minority status predicted increased information seeking. Transgender/non-binary identities and intimate partner violence experiences were also significant predictors.

Conclusions:

• Machine learning models can effectively predict health information seeking behaviors in YARH.
• Findings underscore the need to address health disparities within digital health intervention design.
• Predictive models can guide targeted health information dissemination strategies for YARH.