Der 11. dggö Talk wird die Reihe zu KI im Gesundheitswesen, die mitorganisiert wurde durch den Ausschuss Gesundheitsökonometrie, abschließen.

Noemi Kreif (University of York) spricht am 28.02.2024, 17-18 Uhr zum Thema „Machine learning in health economics and outcomes research: opportunities and some key challenges“.

Hier ist der Link zum Zoom-Meeting: https://uni-due.zoom-x.de/j/62465328399?pwd=RlJIRy9hejlPUksrMXhmWGZxN29Fdz09

In the first half of the talk she will outline the current landscape of using machine learning (ML) for health economics and outcomes research (HEOR). Noemi will focus on supervised ML, and its current application in key HEOR tasks. In the second, main part of the talk she will focus on three key challenges that researchers using ML in HEOR need to tackle: causality, interpretability, and estimation of uncertainty. Noemi will discuss the risks of the naive use of machine learning prediction models to inform treatment decisions, using published examples from clinical medicine. Next, she will bring examples from my own research on data-driven health policy targeting rules, showing the importance of careful adjustment for confounding via causal machine learning. A case study will demonstrate the potential for even state of the art methods to provide counterintuitive and even harmful recommendations, and she will discuss the utility of interpretable ML models in avoiding misleading results. Noemi will discuss the necessity of estimating uncertainty of ML predictions, for these to be used in decision analysis. Finally, she will try to generate an interactive discussion around the varying levels of openness of HEOR stakeholders to start using ML.

Dies ist der Letzte der drei dggö-Talks über KI, ML und das Gesundheitswesen. Für diejenigen, die das Thema vertiefen möchten, findet am 6./7. Juni in Potsdam ein Workshop zum Thema "Applied Economics in Digital Health" statt. Den Call for Papers (Deadline 15. März) finden Sie hier.

*** English version ***

The 11th dggö Talk will conclude the series on AI in healthcare, which was co-organised by the Health Econometrics Committee.

Noemi Kreif (University of York) 28 February 2024, 5-6 pm, will speak on the topic "Machine learning in health economics and outcomes research: opportunities and some key challenges."

Here is the link to the zoom meeting: https://uni-due.zoom-x.de/j/62465328399?pwd=RlJIRy9hejlPUksrMXhmWGZxN29Fdz09

In the first half of the talk she will outline the current landscape of using machine learning (ML) for health economics and outcomes research (HEOR). Noemi will focus on supervised ML, and its current application in key HEOR tasks. In the second, main part of the talk she will focus on three key challenges that researchers using ML in HEOR need to tackle: causality, interpretability, and estimation of uncertainty. Noemi will discuss the risks of the naive use of machine learning prediction models to inform treatment decisions, using published examples from clinical medicine. Next, she will bring examples from my own research on data-driven health policy targeting rules, showing the importance of careful adjustment for confounding via causal machine learning. A case study will demonstrate the potential for even state of the art methods to provide counterintuitive and even harmful recommendations, and she will discuss the utility of interpretable ML models in avoiding misleading results. Noemi will discuss the necessity of estimating uncertainty of ML predictions, for these to be used in decision analysis. Finally, she will try to generate an interactive discussion around the varying levels of openness of HEOR stakeholders to start using ML.

This will be the last of the three dggö Talks on AI, ML and health care. For those interested to dig deeper into the topic, there will be a workshop on “Applied Economics in Digital Health” on 6th/7th June in Potsdam. The Call for Papers (deadline 15th March) can be accessed here.