Introductions - Michael DeWitt

Hi everyone,

My name is Michael DeWitt and I am a research data scientist at the Wake Forest University School of Medicine in Winston-Salem, NC and a founding member of the Center for the Study of Microbial Ecology and Emerging Diseases. I’m interested in infectious disease modelling and dynamics–especially surveillance and detection with incomplete and biased data in which Bayesian frameworks work beautifully. A constant in my thinking is “how does this relate to public health intervention/action/insight” and how our models should be applied under different contexts of outbreaks (e.g., early, later). I typically work in R and Stan, though I find myself doing more in Julia these days.

I came to infectious disease modelling after a bit of a meandering path. I worked in industry for ten years starting during my undergraduate degree (in chemical engineering). During my time in industry I worked on a lot of applied statistics, forecasting, operations research, and data analysis–which ranged from a variety of high velocity, low signal data to sparse and biased data. We also got to run real controlled experiments, which was fun. It was an important lesson to learn how my analysis would be used by machine operators and management to inform their action (and how different audiences were interested in different elements of the analysis).

More recently, I worked as a data scientist in a not-for-profit regional health system (prior to my current role). Because of the way health care is distributed in the United States, during the pandemic we as a system (of 5 hospitals) had to decide how much staffing and materials we needed in order to support our community during the pandemic. I was responsible for putting together all of the long range projections and short term forecasts for the system (and this is where I started working with @samabbott in episoon and later EpiNow2). It was nerve-wracking to present these data and analysis to the leadership of the system, knowing that if I miss high and forecast too many hospitalizations we might run a financial deficit (which could risk us closing our doors) or under-forecasting and not having enough beds/staff to receive patients. Further, because most of the actual public health decisions (e.g., mask mandates, social mobility restrictions) are made at the county level in North Carolina, I was doing presentations to those government bodies. Needless to say, these experiences have shaped my outlook and thinking regarding translating ID modelling to decision-makers.

I also have a masters in statistics, enjoy R package development and exploring workflow tools.

I sincerely look forward to engaging more with everyone!