AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Medical Dead-ends and Learning to Identify High-Risk States and Treatments

Mehdi Fatemi, Taylor W. Killian, Jayakumar Subramanian, Marzyeh Ghassemi

NeurIPS 2021 · Reinforcement learning › Visual and representation RL · approach: Inferring decision-making from behavior · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use DeD to flag potentially detrimental administered treatmentsSeptic patients in MIMIC-III · MIMIC-III septic patientsdetrimental administered treatments identified 24 hours before death: more than $12\%$; $0.6\%$ false positive rateImproved
Use DeD to identify unavoidable dead-end trajectoriesSeptic patients in MIMIC-III · MIMIC-III septic patientsnon-surviving cases identified up to 48 hours before expiration: $2.7\%$; $0.2\%$ false positive rateImproved
Use red-flag membership as a mortality-risk signalDeD on septic patient trajectories · MIMIC-III survivors and nonsurvivorsred flags maintained for more than eight hours: 0.5\% of survivors; 32.6\% of nonsurvivorsMixed
Use median Q values to flag states rather than maximum Q valuesDeD · MIMIC-III septic patientsrobustness of state flagging to extreme approximation errorImproved
Use looser thresholds for yellow flagsDeD · MIMIC-III septic patientssensitivity of early deterioration flags: $\delta_{D}=-0.15$ and $\delta_{R}=0.85$Mixed
Use a separate rescue-value network alongside the dead-end-value networkDeD · MIMIC-III septic patientsconfirmation of presumed dead-end statesImproved
Use value-function trends to detect deteriorationDeD · MIMIC-III septic patientslead time before clinical interventions: 4 to 8 hoursImproved
Compare administered treatments with the 20% most secure treatmentsDeD · MIMIC-III first-flag analysissecurity of administered treatments: 20% most secure treatments (5 out of 25)Mixed
Replace DeD's double-DQN learner with a contemporary offline RL methodDeD · MIMIC-III septic patientsprediction qualityNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy