Communications of the ACM
ML Flags Remedies That Might Do More Harm Than Good
The new machine learning system could give physicians risk scores for different treatments.
Credit: Christine Daniloff/MIT
Scientists at the Massachusetts Institute of Technology (MIT), Microsoft Research, and Adobe India designed a machine learning (ML) model to identify medical treatments that could pose greater potential danger than alternatives, and to alert doctors when sepsis patients may require a change in treatment.
The researchers trained a reinforcement learning model on limited data from a hospital intensive care unit (ICU) to identify treatments to avoid.
The Dead-end Discovery model indicated about 12% of treatments administered to sepsis patients in an ICU were harmful, with about 3% of patients entering a medical “dead end” in their treatment as long as 48 hours prior to their deaths.
MIT's Taylor Killian said the model is “almost eight hours ahead of a doctor's recognition of a patient's deterioration,” which he described as “powerful because in these really sensitive situations, every minute counts, and being aware of how the patient is evolving, and the risk of administering certain treatment at any given time, is really important."
From MIT News
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found