Machine learning for patient risk stratification: standing on, or looking over, the shoulders of clinicians?

Abstract

Machine learning can help clinicians to make individualized patient predictions only if researchers demonstrate models that contribute novel insights, rather than learning the most likely next step in a set of actions a clinician will take. We trained deep learning models using only clinician-initiated, administrative data for 42.9 million admissions using three subsets of data: demographic data only, demographic data and information available at admission, and the previous data plus charges recorded during the first day of admission. Models trained on charges during the first day of admission achieve performance close to published full EMR-based benchmarks for inpatient outcomes: inhospital mortality (0.89 AUC), prolonged length of stay (0.82 AUC), and 30-day readmission rate (0.71 AUC). Similar performance between models trained with only clinician-initiated data and those trained with full EMR data purporting to include information about patient state and physiology should raise concern in the deployment of these models. Furthermore, these models exhibited significant declines in performance when evaluated over only myocardial infarction (MI) patients relative to models trained over MI patients alone, highlighting the importance of physician diagnosis in the prognostic performance of these models. These results provide a benchmark for predictive accuracy trained only on prior clinical actions and indicate that models with similar performance may derive their signal by looking over clinician’s shoulders—using clinical behavior as the expression of preexisting intuition and suspicion to generate a prediction. For models to guide clinicians in individual decisions, performance exceeding these benchmarks is necessary.

Prognostication for prelabor rupture of membranes and the time of delivery in nationwide insured women: development, validation, and deployment

“…We may or may not model a human pathophysiology using this algorithm, but, we definitely model the doctor’s behavior on coding a diagnosis or procedure. 40 By providing an interface to the internal properties of this model, a human user may assess each prediction case-by-case. More details related to this phenomenon are described in this exploration (Supplementary Information).…”

Prediction of Impending Septic Shock in Children With Sepsis

“…This enables clinicians to act only on highly reliable predictions or to take different courses of action informed with the likelihood that a patient will develop septic shock. The goal is to give clinicians information that is novel, rather than merely to confirm what may already be apparent ( 37 ). Inquiry into how clinicians incorporate information from risk scores into clinical decision-making is needed.…”

External Validation Of A Novel Signature Of Illness In Continuous Cardiorespiratory Monitoring To Detect Early Respiratory Deterioration Of ICU Patients

“…13 As a result, these models may reflect the decisions of clinicians more so than changes in patient physiology. 14 That is, if the Electronic Health Record (EHR) shows that a clinician ordered a stat blood gas and chest X-ray, is it really a prediction to say that respiratory failure is imminent? If the physician thought of it first, do these analytics really represent the leading indicators of a patient's illness, or are they just the lagging indicators of clinicians' actions?…”