Data science for mental health crises

Please join PyData Pittsburgh for the talk Data Science for Mental Health Crises by Pim Welle, Chief Data Scientist at the Allegheny County Department of Human Services! We'll be gathering in person at the Code & Supply Workspace.

Please note that the date for this event has changed. The event will now take place on Wednesday, November 20, not Wednesday, November 13.

About the talk

In this talk, you'll hear how cutting-edge data science techniques can be used in government settings. By making use of machine learning / predictive analytics, survival modeling, and non-experimental causal inference, the Allegheny County Department of Human Services (DHS) did a complete profile of its mental health system, including a deep dive into involuntary commitments (sometimes called 302s).

The results of the analysis were staggering. This work led to a report and upcoming academic publication, which found that (1) 302s are very common, affecting 350 per 100k individuals (roughly in line with prison sentencing), (2) folks have very poor outcomes post 302, with 20% of the population passing away in 5 years and (3) the 302 population is very expensive - we spend 25% of our Medicaid behavioral health funding on 2% of the population. Moreover, using machine learning, we can detect the individuals who are likely to have poor outcomes from the moment they step foot in an inpatient unit. Lastly, the upcoming academic work highlights non-experimental causal inference techniques to show whether the commitment itself is helping or hurting downstream outcomes.

The resulting analysis is a use case that demonstrates how statistics, machine learning, and causal inference can all come together to help understand how our public systems are working.

How to find us

Attendees are welcome to use the parking lot associated with the Code & Supply building off St. Clair Street. The front door on Friendship Avenue will be open but is stairs-only. There's an elevator by the parking lot entrance. Head to the third floor and look for signs pointing to the presentation room, where the event will be held. All doors should be unlocked and open, so you're welcome to come right in!

Learn more about PyData Pittsburgh and subscribe to announcements and event updates at https://news.pypgh.org.