Practical Modeling Concepts for Public Health: Activity 1, Scenario 2

Activity 1 asks you to decide which analytic approach is best for a given scenario and then determine how to communicate with the modeling and analytics team to get that output. For Case Study 2, you reviewed and answered questions about what type of output would be best and why, when given the options of a qualitative assessment, nowcast, short-term forecast, and scenario model. You were asked, what information should you be prepared to provide to the modeling and analytics team? What other epidemiological information would be useful and how would you explain it? And what sources of uncertainty might affect the outputs the team develops? Case Study 2 summarizes a major hurricane response effort and an evacuation center measles outbreak. As the incident manager planning staffing and operations, you're overseeing an evacuation center with more than 2,000 individuals being temporarily housed. You are also aware of a measles outbreak ongoing in the state. Symptomatic individuals have presented to the evacuation center health care team and have been confirmed to have measles. So you want to know how many measles cases you may expect within the evacuation center over the coming weeks. You contact your state health department. Mike, a former state epidemiologist, is here to help us decide which approach is best. Thanks, Sarah. I think in this case, short-term forecast would be best. Why short-term forecast? Because we are interested in estimating the number of cases we might expect in the evacuation center. Short-term forecasting works in this example because we have a defined population in a congregate setting, an initial number of confirmed cases, known vaccination rates, and a pathogen with well-characterized transmission parameters. In 2024, there was a measles outbreak in a large congregate shelter. This slide shows some of the real short-term forecasts from that shelter-based measles outbreak and shows how forecasts become more accurate with more data. As data came in during this outbreak, projections were adjusted for the median final outbreak size. The final forecasted estimate for this outbreak after 52 cases were observed was 58 cases. Seeing these projections progress and refine over time helped outbreak responders set expectations and communicate potential outbreak size and duration to leadership. To create short-term forecasts for a case study like this, information relevant for the modeling and analytics team includes current vaccination status of individuals in the evacuation center, at minimum, coverage rates for the community evacuating to the shelter, number of people sheltering in the evacuation center, and a timeline of confirmed cases so far, formatted so that analysts can track revisions and reporting delays. Additional relevant epidemiologic information that modelers might use are transmission dynamics of measles in congregate settings, which they might find from literature if there are similar published case studies, and a description of the evacuation center to evaluate population contact and mixing, like if there is a shared sleeping area, single bathroom, and ventilation system. This concludes Activity 1, Case Study 2.