FAMU Epidemiologist Talks Predictive Models, And Why They Should Be Taken With A Grain Of Salt
Last week, Tallahassee Memorial Healthcare published predictive models regarding COVID-19 hospitalizations in the region. The “most likely” model predicts 361 daily hospital admission, peaking at the end of May. Its best case, nearly 230 admissions per day in mid-June. (Both models are pictured at the bottom of this story)
Dr. C. Perry Brown, professor of epidemiology at Florida A&M University’s Institute of Public Health, is on a board that assists the Florida Department of Health with its modeling. He spoke with WFSU's Ryan Dailey about modeling – and why every predictive model should be taken with a grain of salt.
Questions, with timestamps:
0:20 - You’ve looked over the TMH models – what can you tell us about how those were drawn up?
2:03 - Is there a reason why, as a hospital, you would want to find a worst case?
2:36 - Could you explain your involvement with the Department of Health, and how their modeling process works?
4:00 - You made the statement, “All models are wrong” – what do you mean by that?