PSA: You can't use MIMIC data with ChatGPT / GPT / other LLM APIs. More detail here: physionet.org/news/post/415
- MIMIC-IV was published this week. The core dataset has been out for a while, but we've just published the deidentified free-text clinical notes: 300,000+ discharge summaries and 2.5 million radiology reports! nature.com/articles/s4159…
- Evaluation: We use the AUROC, a common measure in machine learning (Johnson 2013, Johnson 2016, Johnson 2017, Johnson 2018).
- The MIMIC-IV-ED corpus was just published, ~450,000 ED stays at the BIDMC and you can link them to the hospital/ICU data in MIMIC-IV: physionet.org/content/mimic-…
- maybe the secret to novelty is to just be absolutely shameless
