UC Berkeley School of Public Health researchers Sandra McCoy, Jingshen Wang, and Will Dow have been awarded a $2.8 million award from the National Institutes of Health to conduct research on predicting those at risk of disengaging from HIV care in Tanzania.
The new work will build on previous research on offering financial support to HIV patients in Tanzania, which can strengthen adherence to antiretroviral treatment. Past findings, conducted between April and June 2019, determined that out of 530 participants, financial assistance improved HIV treatment retention rates by 83% for those given a smaller cash incentive and 86% for those given larger incentives. For the control group not given incentives, 73% remained in care.
The study will provide adherence counseling for three months, coupled with short-term economic support, to address both behavioral and structural barriers patients face when coming to care.
“The goal of enhanced adherence counseling is to assess possible barriers to care adherence at the individual, household, and community levels, in a nonjudgemental way,” said Sandra McCoy, Berkeley Public Health associate professor in residence in epidemiology and biostatistics. “It does not focus solely on knowledge of HIV and antiretroviral treatment but expands the discussion to include the psychological, emotional and socio-economic factors that may contribute to poor adherence like stigma, substance use, poverty, and mental health.”
The multi-session program with trained counselors and counselor-nurses includes development of a customized adherence plan with concrete objectives. McCoy said the pairing of economic support with adherence counseling is intended to partially mitigate financial barriers to care and also motivate clinic attendance.
The research team will also incorporate machine learning into their new research to determine if it can predict whether a patient will drop out of care. Such foresight may give health care providers an opportunity for timely intervention.
A machine learning algorithm designed by Jingshen Wang, Berkeley Public Health assistant professor of biostatistics, will track each time a patient logs an appointment, along with their medication prescriptions and medical record information—such as weight, sex, marital status, age and residence—and use these records, as well as frequency of appointment data and prescription refills, to predict when patients might need interventions. Using this data, a physician can then step in and inquire about any barriers to that person’s care.
“Usually a doctor would do this kind of diagnostic based on their own knowledge and experience. Machine learning can build this algorithm and learn this kind of outcome, so if we are given a [patient] history, it can help doctors to predict if this patient will withdraw in the future. It will identify patients at high risk in a timely manner, so we can catch it and keep them in care, and save their lives, ” said Wang.
For McCoy, continuing this research means delving into new frontiers through multidisciplinary approaches.
“This kind of approach can reveal predictors and patterns that we’re not even aware of, and we can use that information to better identify people who are in need of more support,” McCoy said. “It might be the case that we learn about highly predictive characteristics or attributes that we didn’t even expect. And that’s what we’re really hoping to learn because we want to unlock new knowledge; we want to unlock new ways of thinking about people’s engagement in HIV care.”
Will Dow, professor of health policy and management at Berkeley Public Health, is excited to continue the collaboration with McCoy’s team to better understand how to effectively use financial incentives to improve adherence to HIV treatment.
“As we have seen with COVID-19, there are many opinions about when and how to use financial incentives to encourage appropriate preventive care and treatment, but there is no substitute for rigorous experimental evidence in the field,” Dow said. “Professor McCoy has developed a fantastic set of community-based collaborators in Tanzania who have been key to developing an acceptable program design, and with the addition to the team of cutting-edge Berkeley Public Health biostatistics faculty such as Jingshen Wang, I am optimistic that the project will produce high impact results.”