Ever since they worked together to explore the use of algorithms originally designed to pick the hot slot machines in casinos to detect undiagnosed cases of HIV, Dr. Gregg Gonsalves, PhD, Yale School of Public Health associate professor of epidemiology (microbial diseases), and Dr. Joshua Warren, PhD, YSPH associate professor of biostatistics, have had a strong working relationship. Now, the pair is teaming up again, leveraging Warren’s statistical expertise and Gonsalves’ knowledge of the epidemiology of substance use and infectious diseases to investigate innovative new ways to identify and mitigate outbreaks of HIV among people who use drugs.
The research, funded by an approximately three-year, $3.5 million R01 grant from the National Institutes of Health’s Office of AIDS Research and the National Institute on Drug Abuse, focuses on the converging public health crisis of substance use and infectious diseases in the United States, Gonsalves said.
While the number of new HIV infections among people who use drugs has been on the decline in the U.S., new diagnoses increased by 11% nationally from 2016 to 2018, with more pronounced increases among younger adults and non-Hispanic whites. Since 2015, over a dozen HIV outbreaks have been recorded among people who use drugs. Fueled by the ongoing opioid epidemic, other infections associated with drug use have also been on the rise in the last decade, creating a public health crisis and a significant clinical and financial burden for states.
Recognizing the devastating impact of HIV, the U.S. government launched the Ending the HIV Epidemic initiative in 2019 with a goal of ending the U.S. HIV epidemic by 2030. The current dual threat of HIV and other infections associated with drug use threatens to derail the effort, Gonsalves said.
The newly funded research project focuses on three distinct areas:
- Temporal Pattern-Recognition and Outbreak Vulnerability: Traditional HIV surveillance methods relying on linear regression and static socioeconomic indicators have been inadequate. The YSPH research team plans to use distributed lag interaction models (DLMs) to link current HIV case counts with historical data on related variables, like other infectious diseases and overdose rates. This approach aims to identify early indicators of HIV outbreaks based on temporal patterns.
- Refined Outbreak Detection: The research will adapt "quickest change detection" techniques from the field of engineering to public health. These methods, used in industrial settings to detect abrupt changes in system processes that impact quality control, will be tailored to identify emerging HIV outbreaks more quickly and accurately, while reducing the incidence of ‘false alarms.’
- Adaptive Case-Finding Strategies: The researchers intend to create enhanced algorithms to assist with real-time decision-making during an outbreak response. This includes refining methods to balance the exploration of new data with the exploitation of existing knowledge. The new algorithms will be designed to deploy mobile units and other resources more efficiently for HIV intervention and other infectious disease testing.
The project will also have built-in flexibility, leveraging the recent success of adaptive algorithms used during the COVID-19 pandemic to improve rapid response to potential outbreaks in dynamic HIV and substance use environments.
In launching the study, Gonsalves and Warren have assembled a team of researchers from Yale, Stanford, Oxford, and 13 state and municipal health departments around the country to gather data, develop algorithms, and test the new interventions. The project will also bring in experts in other disciplines at YSPH and employ students as part of their public health training. Gonsalves and Warren are part of YSPH’s esteemed Public Health Modeling Unit and are bringing in their unit colleagues—Professor David Paltiel, an expert in decision science and operations research; and Associate Professor Forrest W. Crawford, an expert biostatistician and mathematician—to assist with the study.
To expedite the research and ensure the development of the most practical and potentially effective interventions, Gonsalves and Warren organized the project so that front-line local responders and practitioners were involved in the planning up front rather than later in the process.
“To maximize the impact of our work, we met with health officials, local clinicians, and community-based organizations to understand the needs of those on the ground as they seek to prevent and manage HIV outbreaks and epidemics in their locales,” Gonsalves said. “We are hoping that the tools developed with NIH’s support will be able give health departments what they need to ensure that HIV, and infections like hepatitis C, can be stopped in their tracks across the country, where there have been over a dozen outbreaks over the last several years among people who use drugs.”
Investigators plan to conduct a state-by-state analysis of leading HIV outbreak indicators and develop their new algorithms in the project’s first year. Subsequent years will focus on evaluating algorithms using historical data and real-time field testing. The team intends to share its algorithms, software, and source codes with other researchers and partner agencies to allow for further insights and evolution of the project.