# Personalized strategies for HIV treatment maintenance: an application of novel machine learning methods to HIV care in East Africa

> **NIH NIH F31** · UNIVERSITY OF CALIFORNIA BERKELEY · 2020 · $38,995

## Abstract

PROJECT SUMMARY/ABSTRACT
Background: Despite the efficacy of antiretroviral therapy (ART), public health efforts to treat persons living with
HIV must address issues with patient retention in order to achieve lasting epidemic control. People living in
African countries with high rates of HIV infection face different barriers to retention, including structural barriers
(e.g. transport to clinic), psychosocial barriers (e.g., stigma), and others (e.g., long waiting times). This diversity
of barriers contains a critical implication: no single behavioral intervention will help all patients remain in care.
Given the absence of a “one-size-fits-all” solution to HIV treatment maintenance, Drs. Petersen and Geng (the
current project's sponsors) co-lead a nearly completed NIH-funded trial (ADAPT-R; NCT02338739) that identifies
adaptive strategies for patient retention. In ADAPT-R, HIV-positive patients initiating ART are randomized to a
low-intensity intervention to prevent retention lapses. Patients are re-randomized to a high-intensity intervention
to facilitate their return to care, only if initial retention is poor. By responding to a single aspect of individual
patient behavior (days in treatment), this adaptive strategy is actually more efficient (patients succeeding avoid
more costly interventions) and effective (patients not well-retained receive more intensive support) than
intervention assignment strategies in traditional trials that assume “one size fits all.” Approach: The current
proposal uses recent advances in statistics, machine learning, and causal inference to develop an analysis plan
that not only leverages, but also innovates, the parent trial. This project proposes to use ADAPT-R data to design
algorithms that assign each patient his/her personalized interventions based on patient characteristics measured
at baseline (e.g., demographics, distance from clinic, stigma) and over time (e.g., lapses in care, updated clinical
data measures, past interventions). By making optimal use of all patient characteristics measured in ADAPT-R
(i.e., not just early lapses in care) to assign interventions, it is hypothesized that these algorithms will be most
efficient and effective at retaining patients, compared to assignment methods used in ADAPT-R and “one-size-
fits-all” traditional trials. Specific Aims: This F31 Diversity grant aims to: 1) design and test an algorithm that
assigns each patient a personalized low-intensity intervention for remaining in HIV care; 2) design and test an
algorithm that assigns each patient a personalized low- and high-intensity intervention, in sequence, for
remaining in HIV care. Impact: This project provides an opportunity to learn more effective ways of administering
existing interventions to improve HIV retention in rural Africa. At the broadest level, this work aims to advance
the toolkit for developing Precision Public Health strategies. Fellowship information: This project is the
dissertation of Ms. Lina Monto...

## Key facts

- **NIH application ID:** 9952104
- **Project number:** 5F31AI140962-02
- **Recipient organization:** UNIVERSITY OF CALIFORNIA BERKELEY
- **Principal Investigator:** Lina Montoya
- **Activity code:** F31 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2020
- **Award amount:** $38,995
- **Award type:** 5
- **Project period:** 2019-02-06 → 2020-12-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/9952104

## Citation

> US National Institutes of Health, RePORTER application 9952104, Personalized strategies for HIV treatment maintenance: an application of novel machine learning methods to HIV care in East Africa (5F31AI140962-02). Retrieved via AI Analytics 2026-05-23 from https://api.ai-analytics.org/grant/nih/9952104. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*