Feature Robustness in Non-stationary Health Records: Caveats to Deployable Model Performance in Common Clinical Machine Learning Tasks

By Bret A. Nestor, Matthew B. A. McDermott, Willie Boag, Gabriela Berner, Tristan Naumann, Michael C. Hughes, Anna Goldenberg, Marzyeh Ghassemi. 2019

131 citations recorded by the source corpus.

When training clinical prediction models from electronic health records (EHRs), a key concern should be a model's ability to sustain performance over time when deployed, even as care practices, database systems, and population demographics evolve. Due to de-identification requirements, however, current experimental practices for public EHR benchmarks (such as the MIMIC-III critical care dataset) are time agnostic, assigning care records to train or test sets without regard for the actual dates of care. As a result, current benchmarks cannot assess how well models trained on one year generalise to another. In this work, we obtain a Limited Data Use Agreement to access year of care for each record in MIMIC and show that all tested state-of-the-art models decay in prediction quality when trained on historical data and tested on future data, particularly in response to a system-wide record-keeping change in 2008 (0.29 drop in AUROC for mortality prediction, 0.10 drop in AUROC for length-of-stay prediction with a random forest classifier). We further develop a simple yet effective mitigation strategy: by aggregating raw features into expert-defined clinical concepts, we see only a 0.06 drop in AUROC for mortality prediction and a 0.03 drop in AUROC for length-of-stay prediction. We demonstrate that this aggregation strategy outperforms other automatic feature preprocessing techniques aimed at increasing robustness to data drift. We release our aggregated representations and code to encourage more deployable clinical prediction models.

Subfields: Artificial Intelligence, Health Information Management, Evaluation & Benchmarks, Medical & Health AI, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Health Information Management

In Evaluation & Benchmarks

In Medical & Health AI

In Signal Processing

Citations in the corpus

WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020, neutral): “Finally, changes over time in care settings can also produce distribution shifts and drops in algorithmic performance: Nestor et al. (2019) shows that switching between two electronic health record (EHR) systems produced a drop in algorithmic performance.”
Shifting machine learning for healthcare from development to deployment and from models to data (2022, background): “…of the effects of temporal shifts on model performance over time showed that, whereas all models experienced a moderate decline over time, the most significant drop in performance occurred owing to a shift in clinical practice, when EHRs transitioned systems 164 (from CareVue to MetaVision).”
A Review of Challenges and Opportunities in Machine Learning for Health. (2018, background): “Many factors impact generalizability, including local hospital practices, different patient populations, available equipment, and even the specific kind of EHR each uses—transitions from one EHR to another create non-obvious feature mapping problems 81,82 .”
Clinical artificial intelligence quality improvement: towards continual monitoring and updating of AI algorithms in healthcare (2022, background): “Indeed, many have documented ML performance decay due to patient case mix, clinical practice patterns, treatment options, and more 9 – 11 .”
Fairness Violations and Mitigation under Covariate Shift (2019, background): “[41] find that models for critical care tasks degraded in performance over time resulting from changes in the instrumentation of the electronic health records.”
Ethical Machine Learning in Healthcare (2020, methodology): “Even within a single hospital, models trained on data from an initial EHR system data deteriorated significantly when tested on data from a new EHR system (132).”
Secure and Robust Machine Learning for Healthcare: A Survey (2020, methodology): “In [40], features robustness using EHRs data for the year of care for each record is examined for two tasks, i.”
The myth of generalisability in clinical research and machine learning in health care (2020, background): “25,26 Furthermore, techniques from continual learning 27 offer enormous potential to create more advanced machine learning systems that continuously update based on new data.”