AI tools can be used to inform public health policy and medical decision making. For example, predictive analytics can be used to identify risk factors for disease; and optimization frameworks (whether single stage or repeated) can be used to identify when to screen or treat disease, or which risk groups to target given limited public health resources. We describe several projects and potential project areas below.
Identifying Risk Groups for Infectious Disease Outreach
Treatable infectious diseases are a critical challenge for public health. While treatment regimens may exist, and even be offered at low cost to patients, often individuals do not recognize the need to seek treatment or delay doing so, thereby increasing transmission risk to others. Outreach and education campaigns can encourage undiagnosed patients to seek treatment. However, such programs must be carefully targeted to appropriate demographics (for the specific disease) in resource-constrained settings.
In prior work, we developed an algorithm to optimally allocate limited outreach resources among demographic groups in the population. The algorithm uses a novel multi-agent model of disease spread which both captures the underlying population dynamics and is amenable to optimization. Our algorithm extends, with provable guarantees, to a stochastic setting where we have only a distribution over parameters such as the contact pattern between agents. We evaluate our algorithm on two instances where this distribution is inferred from real world data: tuberculosis in India and gonorrhea in the United States.
Contact Tracing to Identify Undiagnosed Disease
While individuals with symptoms of infectious disease may seek treatment themselves, public health campaigns may fund active-screening programs where health workers seek out undiagnosed cases by canvasing at-risk individuals. Individuals at particularly high risk–those who have come into contact with infected individuals–may be identified through interviews with patients and encouraged to seek screening and treatment.
In an ongoing project, we are developing an algorithm to help inform how contact tracing might be optimized, given a limited budget for disease screening and knowledge of the contact network between individuals in a community. Who should be screened if some individuals are confirmed to be infected while others are merely suspected to have disease? How does this change over time, with the graph structure, and disease progression?
Development of Individual-level Simulations of Disease
While the medical literature offers a rich source of information about disease progression, screening, and treatment, the spread of disease on a population can be a complex process that requires simulation to understand. We develop individual-level simulations of infectious and non-infectious disease to probabilistically infer who will acquire disease, project population disease metrics (such as prevalence and incidence) over long time horizons, and to evaluate the effectiveness (and cost) of interventions. These simulations draw on detailed knowledge about disease progression, patient behavior, and treatment outcomes using information from medical literature and datasets. In prior projects, we have developed individual-level simulations of tuberculosis, and efforts to model chronic disease in elderly persons are ongoing. Ongoing work involves examining how different models (with different runtimes, fidelity to the real world, and noise) can be used in combination to quickly and accurately identify optimal policies for disease control given individual heterogeneity.
Data Analytics for Determining Risk Factors of Disease and Treatment Non-Adherence
Data analytics and machine learning methods can help identify previously unknown risk factors for disease or for treatment non-adherence. These insights can help medical practitioners identify individuals for early screening or to devote dedicated resources to help patients maintain long treatment regimens. Ongoing projects focus on identifying dementia among elderly persons, and identifying tuberculosis patients at risk of treatment default in collaboration with a team in India.
Using Social Networks for Prevention Interventions
One of the fundamental questions facing social science is how social networks and the cognitions people have about their networks affect their mental states and mental health. AI techniques present an opportunity to dynamically model social networks and the messages transmitted across those networks to create predictive models of influence unavailable with standard statistical techniques.
This project focuses on the development of decision support systems for homeless youth drop-in center staff, who need to find the most influential homeless youth to raise awareness about HIV (and other STDs) among their peers, and to drive the homeless youth community towards safer behaviors.
Research has consistently documented levels of cocaine, heroin, methamphetamine, alcohol, and marijuana use and abuse among these adolescents that far exceed that of housed adolescents. This project aims to use algorithms to determine the best group formations to prevent regular use of hard drugs among homeless youth.
Optimization and data analytics for kidney transplantation management
Candidate recipients in resource allocation systems, such as the U.S. Kidney Allocation System and the Public Housing Program, often face long, variable, and uncertain wait times until offered their preferred resources. Moreover, to date, they only have access to crude estimates about wait times that do not account for their individual characteristics and preferences. By observing little information about the wait list status, candidates rely on intuition and a complex series of “what if” calculations to understand what the future might hold. At the same time, accurate wait time estimates are crucial, for example for patients seeking kidney transplantation when they face disease management dilemmas, such as whether to accept or decline a particular offered kidney of marginal quality.
In this project, we develop novel data-driven optimization methodologies that produce accurate, personalized wait time estimates for individual candidates. We calibrate our model using highly detailed historical data from the U.S. Kidney Allocation System and illustrate how it can be used to inform patients seeking kidney transplantation about their wait time, and in turn support medical decision making and improve their welfare.
Our research enables transplant centers to provide personalized wait time estimates to their patients that can help inform their disease management decisions. Similarly, it enables public housing offices to provide personalized wait time estimates to their applicants.