Evaluating predictive performance, validity, and applicability of machine learning models for predicting HIV
Williams Kwarah1,2, Frances Baaba da-Costa Vroom3, Duah Dwomoh3
1Department of Biostatistics, School of Public Health, University of Ghana, Accra, Ghana. Kwarah@gmail.com.
BMC Global and Public Health
|July 24, 2025
Summary
Machine learning models show promise for predicting HIV treatment interruption. However, most studies had a high risk of bias, highlighting the need for better validation and data handling in future research.
Area of Science:
- Artificial Intelligence in Public Health
- Predictive Modeling for Infectious Diseases
Background:
- HIV treatment interruption is a major obstacle to global HIV/AIDS control.
- Machine learning (ML) offers potential for predicting treatment interruption using clinical data.
- Understanding ML model development, validation, and application is crucial for research advancement.
Purpose of the Study:
- To systematically review machine learning models for predicting HIV treatment interruption.
- To assess the development, validation, performance, and risk of bias of these models.
Main Methods:
- A comprehensive literature search was conducted across multiple databases (PubMed, Scopus, etc.) from 1990 to September 2024.
- Studies were screened, and data extracted using the CHARMS checklist.
- Risk of bias was assessed using PROBAST, adhering to PRISMA guidelines.
Main Results:
- Nine studies reported 12 ML models, predominantly Random Forest, XGBoost, and AdaBoost.
- All models underwent internal validation, but only two had external validation.
- Models demonstrated moderate discrimination (mean AUC-ROC = 0.668), with 75% showing high risk of bias due to data handling and lack of calibration/DCA.
Conclusions:
- ML models show potential for predicting HIV treatment interruption, especially in resource-limited settings.
- Future research must focus on external validation, robust missing data handling, and decision curve analysis.
- Incorporating sociocultural predictors can enhance model robustness and clinical utility.
Related Concept Videos
01:18Steps in Outbreak Investigation
207
In the ever-evolving field of public health, statistical analysis serves as a cornerstone for understanding and managing disease outbreaks. By leveraging various statistical tools, health professionals can predict potential outbreaks, analyze ongoing situations, and devise effective responses to mitigate impact. For that to happen, there are a few possible stages of the analysis:
207
01:10Retrovirus Life Cycles
46.9K
Retroviruses have a single-stranded RNA genome that undergoes a special form of replication. Once the retrovirus has entered the host cell, an enzyme called reverse transcriptase synthesizes double-stranded DNA from the retroviral RNA genome. This DNA copy of the genome is then integrated into the host’s genome inside the nucleus via an enzyme called integrase. Consequently, the retroviral genome is transcribed into RNA whenever the host’s genome is transcribed, allowing the...
46.9K
01:20Comparing the Survival Analysis of Two or More Groups
289
Survival analysis is a cornerstone of medical research, used to evaluate the time until an event of interest occurs, such as death, disease recurrence, or recovery. Unlike standard statistical methods, survival analysis is particularly adept at handling censored data—instances where the event has not occurred for some participants by the end of the study or remains unobserved. To address these unique challenges, specialized techniques like the Kaplan-Meier estimator, log-rank test, and...
289