Data-driven model visualizing patient care in an HIV/AIDS clinic.

Boosting HIV Care: How a Predictive Model Can Help Keep Patients on Track

Reese Marlowe monogram
Reese Marlowe in Health & Wellness October 2026 • 4 min read.

"Discover how a new model is using data to improve retention in HIV/AIDS treatment programs and help achieve crucial health goals."


In the ongoing fight against HIV/AIDS, keeping individuals engaged in consistent clinical follow-up is crucial. With global targets set to control HIV/AIDS, ensuring people living with HIV/AIDS (PLWA) remain in care is both a necessity and a significant challenge. Consistent care helps manage the infection, prevent transmission, and improve overall health outcomes.

Retention in care is a multi-faceted issue, influenced by individual patient circumstances, healthcare service delivery, and broader environmental factors. To tackle this challenge, researchers have been exploring innovative approaches to identify and address the barriers that prevent PLWA from staying connected to the healthcare system.

A recent study published in Cadernos de Saúde Pública introduces a predictive model designed to improve retention in specialized HIV/AIDS care. This model uses data-driven insights to identify individuals at risk of dropping out of care, allowing healthcare providers to tailor interventions and support to those who need it most. Let's dive into how this model works and what it means for the future of HIV/AIDS care.

AI Search Multiple angles on this topic

HIV’s Ongoing Health Impact

HIV attacks the immune system and, without treatment, can progress to AIDS, a late stage in which the immune system is badly damaged. The CDC reports that there is currently no effective cure and that people who acquire HIV have it for life, although proper medical care can control the virus. HIV-1 accounts for the majority of HIV infections globally, according to the cited overview.

The Need for Consistent Care

HIV care generally depends on ongoing medical management and consistent engagement with healthcare services. A predictive model could potentially help care teams recognize patients who may need additional support, but its usefulness would depend on accurate data, appropriate clinical oversight, and validation in real-world settings. Such a model would complement, rather than replace, individualized care.

From Discovery to Long-Term Management

The history of HIV care reflects a shift from understanding a newly recognized virus to managing a lifelong condition. Foundational discoveries established the relationship between HIV, immune-system damage, and AIDS, while later medical progress made sustained control possible. A predictive-care approach builds on that broader history by focusing on continuity and prevention of avoidable gaps in care.

What is the Predictive Model for HIV/AIDS Care?

Data-driven model visualizing patient care in an HIV/AIDS clinic.

The predictive model developed in this study is a statistical tool that uses a decision tree algorithm. By analyzing various factors, the model aims to predict whether a PLWA is likely to remain engaged in care. The data used to build this model came from a database of 260 individuals enrolled in a specialized HIV/AIDS treatment service.

Researchers analyzed a range of variables, including:

  • Sociodemographic factors: Age, marital status, and location of residence.
  • Lifestyle and health behaviors: Sexual orientation and alcohol use.
  • Clinical variables: Use of antiretroviral therapy (ART), viral load test results, hospitalization history, and number of ART pills taken daily.
AI Search Multiple angles on this topic

Why Ongoing Monitoring Matters

The Mayo Clinic reports that HIV makes the body less able to clear infections and diseases. If untreated over time, HIV can cause acquired immunodeficiency syndrome, or AIDS. These risks help explain why research into tools that support continued monitoring and engagement in care remains relevant.

Limits of Prediction

A predictive model may identify patterns associated with missed care, but a prediction is not a certainty about an individual patient. Models can also perform unevenly when data are incomplete, biased, or drawn from populations unlike the people being assessed. For that reason, predictions should be treated as prompts for supportive human follow-up rather than automatic judgments.

Models Alongside Established Care

Predictive modeling offers a different function from routine HIV care: it aims to anticipate possible barriers before they result in disengagement. Standard clinical care remains focused on assessing patients and managing their health needs directly. The two approaches may be complementary, but their relative value would need to be evaluated through careful comparison and outcome testing.

The decision tree model identified nine key variables that significantly impact retention in care. These variables were used to create 24 decision rules, resulting in a model with 80.4% accuracy in predicting patient retention. This level of accuracy makes the model a valuable tool for healthcare providers looking to improve their retention strategies.

Why This Model Matters

The development and implementation of predictive models like this one represent a significant step forward in the fight against HIV/AIDS. By proactively identifying individuals at risk of dropping out of care, healthcare providers can offer targeted support and interventions. This approach not only improves individual health outcomes but also contributes to broader public health goals, such as reducing HIV transmission and controlling the epidemic. Further research and wider adoption of these models are essential to maximize their impact and achieve a future free from HIV/AIDS.

AI Search Multiple angles on this topic

Support, Not Substitution

A predictive model could strengthen HIV care by helping teams direct attention toward patients who may need timely assistance. Its role should remain supportive, with clinicians interpreting model outputs in the context of each person’s circumstances. The central goal is better continuity of care, not replacing professional judgment or patient choice.

Toward More Responsive Care

Future work may examine whether predictive tools can identify care barriers early enough for practical interventions to help. Progress will likely depend on reliable data, transparent evaluation, and testing across different patient populations and care settings. Any expansion should be guided by evidence that the tools improve care without creating new harms.

The Systems Around Patients

Keeping patients on track is shaped not only by medical recommendations but also by the accessibility, coordination, and responsiveness of the care system. A model may highlight risk, yet healthcare teams still need the resources and processes to respond effectively. This makes implementation and equity as important as predictive accuracy.

Care Centered on People

Patients remain more than the patterns captured in a dataset. Effective use of prediction should lead to respectful conversations and practical support tailored to individual needs. Human judgment, trust, and patient participation are essential if a model is to contribute meaningfully to ongoing HIV care.

How other HIV-care models perform

  • 75.2%; 54.7%; 30%Models using electronic medical records and changes across past visits averaged 75.2% accuracy for predicting disengagement over the next 6 months; their mean sensitivity was 54.7% when targeting the 30% at highest risk.[1]
  • 70%–91%; 0.56; 0.99Across eight models predicting loss from care in Nigeria, reported accuracy ranged from 70% to 91%. Other models reported AUC values of 0.56 in Kenya and 0.99 in Ethiopia for identifying viral-load suppression.[2]

What prediction can-and cannot-tell care teams

A model can flag patterns linked to treatment interruption without explaining which factors drive an individual person’s risk. The reported collinearity and non-linear predictors make those individual-level explanations harder to interpret.[3]

A risk estimate alone does not ensure better retention: the potential benefit depends on whether suitable support is available to address the underlying reasons someone may disengage.[2]

How these approaches compare

Reported performance varies across HIV-care prediction efforts: accuracy figures span a range, while AUC results also differ substantially across settings and tasks. These metrics describe different aspects of model performance, so they do not establish that one approach is best in every context.[2]

Adherence clubs and predictive tools represent different approaches to retention: the club findings describe continued care and viral suppression among eligible participants, rather than model accuracy.[4]

Text-message interventions have not produced uniform retention results: reports from Malawi and Kenya found no improvement, while a Mozambique trial found improvement for urban but not rural patients.[4]

What remains uncertain about keeping people engaged

  • Which health-system, structural and psychosocial barriers matter most in a particular care setting, and can a model distinguish among them?[1] [5]
  • When care is interrupted, which forms of re-engagement support—remote contact, in-person tracing or a combination—best help people reconnect?[6]

Sources

  1. Machine learning with routine electronic medical record data to identify people at high risk of disengagement from HIV care in Tanzania (pmc.ncbi.nlm.nih.gov)
  2. Triaging Clients at Risk of Disengagement from HIV Care: Application of a Predictive Model to Clinical Trial Data in South Africa (pmc.ncbi.nlm.nih.gov)
  3. Historical visit attendance as predictor of treatment interruption in South African HIV patients: Extension of a validated machine learning model (pmc.ncbi.nlm.nih.gov)
  4. Impact of SMS and peer navigation on retention in HIV care among adults in South Africa: results of a three‐arm cluster randomized controlled trial (pmc.ncbi.nlm.nih.gov)
  5. Interventions to reengage people living with HIV who are lost to follow-up from HIV treatment programs: A systematic review and meta-analysis (pmc.ncbi.nlm.nih.gov)
  6. HIV - Technical work (www.who.int)

About this Article -

Written with AI assistance from published research, and reviewed by the Mystum team. See our About page for more information.

Everything You Need To Know

1

What is the core functionality of the predictive model used in HIV/AIDS care, and what kind of data does it analyze?

The predictive model is a statistical tool utilizing a decision tree algorithm to forecast whether a Person Living With HIV/AIDS (PLWA) will remain engaged in care. It analyzes factors like sociodemographics (age, marital status, residence), lifestyle/health behaviors (sexual orientation, alcohol use), and clinical variables (antiretroviral therapy (ART) use, viral load, hospitalization history, ART pill count). This model identified nine key variables, creating 24 decision rules with 80.4% accuracy in predicting patient retention.

2

Why is the development of a predictive model significant for improving HIV/AIDS care and achieving epidemic control targets?

This predictive model matters because it allows healthcare providers to proactively identify Persons Living With HIV/AIDS (PLWA) at risk of disengaging from care. By offering targeted support based on the model's predictions, providers can improve individual health outcomes and contribute to broader public health goals, such as reducing HIV transmission and controlling the epidemic. Wider adoption and further research into these models are crucial for maximizing their impact.

3

What specific variables are analyzed by the predictive model to assess the risk of patients dropping out of HIV/AIDS care?

The model analyzes a range of variables related to Persons Living With HIV/AIDS (PLWA). These include sociodemographic factors such as age, marital status, and location of residence; lifestyle and health behaviors such as sexual orientation and alcohol use; and clinical variables such as the use of antiretroviral therapy (ART), viral load test results, hospitalization history, and the number of ART pills taken daily. The decision tree model uses these variables to predict retention in care.

4

How accurate is the predictive model in forecasting patient retention in HIV/AIDS care, and what does this level of accuracy imply for healthcare providers?

The study identified nine key variables that significantly impact retention in specialized HIV/AIDS care. These variables were used to create 24 decision rules. The model achieved 80.4% accuracy in predicting patient retention. This level of accuracy means the model can reliably identify individuals at risk of dropping out of care, allowing healthcare providers to allocate resources and interventions effectively.

5

What are some limitations of the predictive model described, and what additional factors might influence patient retention in HIV/AIDS care that were not directly addressed?

While the study focuses on specific variables and outcomes related to Persons Living With HIV/AIDS (PLWA), it is important to consider broader determinants of health that may influence retention in care. Factors such as access to transportation, social support networks, mental health services, and socioeconomic status are not explicitly addressed, but they can significantly impact a patient's ability to stay engaged with healthcare services. Future research could explore how these factors interact with the variables included in the predictive model to provide a more comprehensive understanding of retention in care.

Newsletter Subscribe

Subscribe to get the latest articles and insights directly in your inbox.