A few weeks ago, I had the opportunity to attend the Council of State and Territorial Epidemiologists (CSTE) Annual Conference right here in Pittsburgh! Getting involved with CSTE this past year, especially through the forecasting and modeling workgroup calls, has been incredibly rewarding, and it was great to see some familiar faces in person.

I highly recommend any student working at the intersection of computer science, machine learning, and public health to attend a CSTE meeting because it helped me (1) internalize why reconciling academic research with public health needs is challenging and (2) brainstorm opportunities to bridge this gap.

Factors Increasing the Gap

1. Unsustainable Effort from Public Health Experts: Public health experts are overwhelmed. Technical solutions often demand more training, attention, and investment from public health experts, even though they have historically seen low returns for their efforts. One thing I heard from attendees was that current epidemiologists have no time to learn these technical skills, like forecasting and modeling, outside of their education curriculum. Until these subjects are part of the core epidemiology curriculum, there will always be a technical gap. More specifically, for research innovations, they’re tired of researchers trying to shoehorn solutions to public health problems. Even well-meaning but unaware technologists asking for time can be cumbersome to public health experts, especially after seeing many substantive research efforts fall apart before practical deployment. Many times, this cost is unknown – when I asked some solution providers at CSTE how they evaluate their technologies, I eventually found that their only source of evaluation would be from health experts – which is a lot to ask from departments that may not fully know their extra labor that may only hypothetically support a future automated solution.

2. Limited Application Infrastructure for Rapid Prototyping: Public health technology includes infrastructure, visualization, and methods. However, commercial tools and discussions focus on creating and maintaining infrastructure. Many of these infrastructure tools are proprietary and can lock in departments for long contracts. The marketing approaches I observed focus on social proof, most commonly sharing “endorsements” from influential public health departments. Yet, I found that some of these tools heavily depend on human support and are pseudo-automatic instead of the automatic solutions they may initially appear to be.
Additionally, I found that, despite the number of visualization and analysis platforms, there were few solutions for public health departments to test and prototype novel research methods or visualization strategies in an automated, department-wide setting. These types of innovations update much more quickly than the existing infrastructure tools. Instead, testing research innovations is often handled by relatively siloed internal teams per department (that potentially partner with local academic institutions). I learned that while there is a push to strengthen connections between these departments, broad ideas that take effort to test are only sometimes shared. In fact, code, which may be easier to tailor to different departments, is rarely ever shared.

3. Overly Ambitious Technical Agenda: Technologists applying the latest innovations to public health may not realize how unique the public health setting is and how poorly new technologies can perform in this setting. Addressing the many “edge cases” in public health (e.g., data types, availability, and usage) needs extensive validation before deployment – which means there is naturally a lag. Right now, many departments are focusing on relationship building and data sharing, which is likely a prerequisite for new technologies. Pitching technologies that cannot succeed without this foundation and will fail in public health settings will continue to widen this gap.

Bridge Building Ideas

1. Meet Public Health Where It Is: Technologists should find ways to listen to public health conversations and understand expert’s needs and the general public health climate. Some suggestions include attending public health conferences [!], joining status update calls for various public health-related communities, and subscribing to newsletters from local health departments to learn more about relevant context, questions, and constraints. This could also involve building trust by focusing on non-infectious diseases.

2. Provide Technical Assistance: Guidance from computer scientists to health departments on system building, forecasting, and evaluating new public health tools may be valuable. Additionally, business students who study negotiation may provide useful guidance about structuring contracts. Enhancing the pipeline between students outside of public health backgrounds can usher in a robust public health prototyping infrastructure.

3. Integrate Technologists Early: Students have the time (and openness of mind) to learn about public health nuances and constraints. When they can explore these in conjunction with their technical curriculum, they may be able to propose realistic solutions. I recommend an experience like this for all first- or second-year Ph.D. students considering this field. This recommendation comes from personal experience – working with the Allegheny County Public Health Department in my first year of the PhD gave me a deeper understanding of public health challenges and how they translate to technical limitations.

4. Supporting Orchestration Codifying an evaluation pipeline for new research insights could provide a standard way for different departments to evaluate methodologies in relevant contexts. Supporting this type of infrastructure generally via tooling or evaluation platforms could tighten the iterations between research and application.

Overall, attending a public health conference was both an enlightening experience and also vastly different from the CS/ML conferences I’ve attended for my work. Bridging this gap between public health needs and academic research is essential for resilient and updated public health technologies. I’m excited to think more about this intersection in my final year of the Ph.D.!

Updates: I’ve been hearing more about the misalignment between health needs and computational research problems. For example, I’ve heard that public health really needs assistive approaches in mitigating food-borne illnesses, which some computational scientists think has less innovation potential (e.g., already explored statistically). This is actually a really great opportunity for computational researchers to dig deeper and structure the limitations public health experts are facing – clearly there are existing limitations, if not directly with the methods, then with how they lend themselves to practical use. I’d love to chat more about this if you find it interesting!