Using AI to predict COVID-19 patient outcomes
Episode description
Given the strain on hospital resources caused by the pandemic, many informaticists have focused on the ability to try and predict patient populations. In January, researchers at the Regenstrief Institute and Indiana University found that machine learning models trained using statewide health information exchange data can actually predict a patient's likelihood of being hospitalized with COVID-19.
Joining Healthcare IT News Senior Editor Kat Jericch to discuss the study's implications are two of its lead authors, Dr. Shaun Grannis and Suranga Kasturi.
Talking points:
- How tools like this might be useful for health systems and hospitals
- Connecting system-generated data with public health
- How COVID-19 has shined a light on cracks in different systems
- The Indiana Health Information Exchange as a data repository
- Seeing data-sharing blossom during the pandemic
- Biases in the model and how they can be addressed
- How integrated data can be a powerful tool to shape policy
More about this episode:
Regenstrief launches initiative to disseminate SDOH data
HIE-trained AI models can forecast individual COVID-19 hospitalization
Data from 175K COVID-19 patients fuels predictive severity model
Predicting COVID-19 hotspots: Kaiser Permanente tool uses EHR data to forecast surges
Even innocuous-seeming data can reproduce bias in AI