Although HIV was used as a case study, the UMass Amherst researchers say their assessment tool has applications for other diseases
AMHERST, Mass. — Researchers at the University of Massachusetts Amherst have quantified the impacts of a constellation of social factors on the spread of HIV. Their study, published in Health Care Management Science , found that a hypothetical 100% effective intervention addressing barriers to HIV treatment and care from depression, homelessness, individual and neighborhood poverty, education disparities, lack of insurance and unemployment could reduce the national HIV incidence by 29% over 10 years. The mathematical model, a novel integration of machine learning, probability theory and simulation, is positioned to be an important tool for decision-makers to optimize social programs and will have applications for other diseases.
Although 100% suppression of the virus — and consequently a 0% spread of the disease — is feasible through medication, there were still 31,800 new cases in 2022. Furthermore, the lifetime HIV cost per person in America is $420,285.
"HIV strikes me as something that we should be able to eliminate, but it's really the social vulnerability that is driving the epidemic," says Chaitra Gopalappa , associate professor of mechanical and industrial engineering at UMass Amherst and corresponding author on the paper. For instance, 44% of people with HIV have some kind of disability and 43% have a household income at or below the poverty line.
While previous work has focused on behavioral risk factors of HIV spread — sexual behavior and needle sharing — less has been done to quantify the association of social factors with HIV risk behaviors.
"Just having behavioral interventions is not going to be sufficient, so what are those additional interventions that are needed?" says Gopalappa. "What our work did was to develop a model that helps us determine what the joint social burden is and how is that related to behavioral mechanisms that increase the risk of HIV?"
Gopalappa, with doctoral candidate Amir Khosheghbal and Manning College of Information and Computer Science professor Peter Haas , first identified the proportion of people with diagnosed HIV (PWDH) who are also affected by depression, homelessness, individual and neighborhood poverty, education disparities, lack of insurance or unemployment. They found that 78% of PWDH are impacted by at least one of these social factors: 58% of PWDH have one or two of these social factors and another 20% have more than two. (It is important to note that the model looked at association and not causality and assumed an idealistic hypothetical intervention that is 100% effective.)
The model takes into consideration that different social factors present different-sized barriers to HIV care and treatment. Compare unemployment and lacking insurance. More PWDH are unemployed than lack insurance — 14% versus 3% of PWDH. However, previous research indicates that insurance has a greater effect on HIV care access than unemployment.
There is a certain chance that an individual faces one or both of these social factors. The disparity gap in HIV care for someone with only one social factor may be different than with the gap for someone with both. The complexity of estimating these disparity gaps increases as more social factors come into play. The model, a novel integration of probability theory with machine learning, is built to quantify this. Followed by a simulation to do a what-if analysis to estimate the impact of a hypothetical intervention that can completely close these disparity gaps i.e., all persons will receive the same higher level of HIV care as that of people with no social burden.
"The likely interventions that we need, like food and housing assistance, are likely to involve some cost but we're going to be reducing the percent of cases," says Gopalappa. "The cost of treatment itself is very high. Could investing in prevention avert those future costs for treatment?"
With this perspective, this model can help decision-makers optimize the most cost-effective combination of social programs. If each intervention is like a lever, you may not need to pull every lever to 100% if you can find the optimal combination, she explains. "This tool helps analyze those resource allocation decisions."
She also highlights that the impacts of social programs extend beyond HIV. "Diseases don't occur in silos," she says. "What we're seeing is that the same social determinants are associated with other diseases from HIV and STIs to mental health, cardiovascular disease and diabetes to maternal morbidity and mortality. The decision-making process for each of these diseases occurs by different entities, but it is in the same people. So, how do we integrate all of these interventions for decision analysis? This is one part of a long-term goal to develop a tool that helps resource allocation strategies."
This research was supported by the U.S. National Science Foundation .
