skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Query-Focused EHR Summarization to Aid Imaging Diagnosis
Electronic Health Records (EHRs) provide vital contextual information to radiologists and other physicians when making a diagnosis. Unfortunately, because a given patient’s record may contain hundreds of notes and reports, identifying relevant information within these in the short time typically allotted to a case is very difficult. We propose and evaluate models that extract relevant text snippets from patient records to provide a rough case summary intended to aid physicians considering one or more diagnoses. This is hard because direct supervision (i.e., physician annotations of snippets relevant to specific diagnoses in medical records) is prohibitively expensive to collect at scale. We propose a distantly supervised strategy in which we use groups of International Classification of Diseases (ICD) codes observed in ‘future’ records as noisy proxies for ‘downstream’ diagnoses. Using this we train a transformer-based neural model to perform extractive summarization conditioned on potential diagnoses. This model defines an attention mechanism that is conditioned on potential diagnoses (queries) provided by the diagnosing physician. We train (via distant supervision) and evaluate variants of this model on EHR data from Brigham and Women’s Hospital in Boston and MIMIC-III (the latter to facilitate reproducibility). Evaluations performed by radiologists demonstrate that these distantly supervised models yield better extractive summaries than do unsupervised approaches. Such models may aid diagnosis by identifying sentences in past patient reports that are clinically relevant to a potential diagnosis. Code is available at https://github.com/dmcinerney/ehr-extraction-models.  more » « less
Award ID(s):
1901117
PAR ID:
10251771
Author(s) / Creator(s):
; ; ; ; ;
Editor(s):
Doshi-Velez, Finale; Fackler, Jim; Jung, Ken; Kale, David; Ranganath, Rajesh; Wallace, Byron; Wiens, Jenna
Date Published:
Journal Name:
Proceedings of the 5th Machine Learning for Healthcare Conference
Volume:
126
Page Range / eLocation ID:
632--659
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; ; ; ; ; (, JMIR Medical Informatics)
    null (Ed.)
    Background Increased work through electronic health record (EHR) messaging is frequently cited as a factor of physician burnout. However, studies to date have relied on anecdotal or self-reported measures, which limit the ability to match EHR use patterns with continuous stress patterns throughout the day. Objective The aim of this study is to collect EHR use and physiologic stress data through unobtrusive means that provide objective and continuous measures, cluster distinct patterns of EHR inbox work, identify physicians’ daily physiologic stress patterns, and evaluate the association between EHR inbox work patterns and physician physiologic stress. Methods Physicians were recruited from 5 medical centers. Participants (N=47) were given wrist-worn devices (Garmin Vivosmart 3) with heart rate sensors to wear for 7 days. The devices measured physiological stress throughout the day based on heart rate variability (HRV). Perceived stress was also measured with self-reports through experience sampling and a one-time survey. From the EHR system logs, the time attributed to different activities was quantified. By using a clustering algorithm, distinct inbox work patterns were identified and their associated stress measures were compared. The effects of EHR use on physician stress were examined using a generalized linear mixed effects model. Results Physicians spent an average of 1.08 hours doing EHR inbox work out of an average total EHR time of 3.5 hours. Patient messages accounted for most of the inbox work time (mean 37%, SD 11%). A total of 3 patterns of inbox work emerged: inbox work mostly outside work hours, inbox work mostly during work hours, and inbox work extending after hours that were mostly contiguous to work hours. Across these 3 groups, physiologic stress patterns showed 3 periods in which stress increased: in the first hour of work, early in the afternoon, and in the evening. Physicians in group 1 had the longest average stress duration during work hours (80 out of 243 min of valid HRV data; P=.02), as measured by physiological sensors. Inbox work duration, the rate of EHR window switching (moving from one screen to another), the proportion of inbox work done outside of work hours, inbox work batching, and the day of the week were each independently associated with daily stress duration (marginal R2=15%). Individual-level random effects were significant and explained most of the variation in stress (conditional R2=98%). Conclusions This study is among the first to demonstrate associations between electronic inbox work and physiological stress. We identified 3 potentially modifiable factors associated with stress: EHR window switching, inbox work duration, and inbox work outside work hours. Organizations seeking to reduce physician stress may consider system-based changes to reduce EHR window switching or inbox work duration or the incorporation of inbox management time into work hours. 
    more » « less
  2. ; ; ; ; (, Journal of Medical Internet Research)
    Background Online physician reviews are an important source of information for prospective patients. In addition, they represent an untapped resource for studying the effects of gender on the doctor-patient relationship. Understanding gender differences in online reviews is important because it may impact the value of those reviews to patients. Documenting gender differences in patient experience may also help to improve the doctor-patient relationship. This is the first large-scale study of physician reviews to extensively investigate gender bias in online reviews or offer recommendations for improvements to online review systems to correct for gender bias and aid patients in selecting a physician. Objective This study examines 154,305 reviews from across the United States for all medical specialties. Our analysis includes a qualitative and quantitative examination of review content and physician rating with regard to doctor and reviewer gender. Methods A total of 154,305 reviews were sampled from Google Place reviews. Reviewer and doctor gender were inferred from names. Reviews were coded for overall patient experience (negative or positive) by collapsing a 5-star scale and coded for general categories (process, positive/negative soft skills), which were further subdivided into themes. Computational text processing methods were employed to apply this codebook to the entire data set, rendering it tractable to quantitative methods. Specifically, we estimated binary regression models to examine relationships between physician rating, patient experience themes, physician gender, and reviewer gender). Results Female reviewers wrote 60% more reviews than men. Male reviewers were more likely to give negative reviews (odds ratio [OR] 1.15, 95% CI 1.10-1.19; P<.001). Reviews of female physicians were considerably more negative than those of male physicians (OR 1.99, 95% CI 1.94-2.14; P<.001). Soft skills were more likely to be mentioned in the reviews written by female reviewers and about female physicians. Negative reviews of female doctors were more likely to mention candor (OR 1.61, 95% CI 1.42-1.82; P<.001) and amicability (OR 1.63, 95% CI 1.47-1.90; P<.001). Disrespect was associated with both female physicians (OR 1.42, 95% CI 1.35-1.51; P<.001) and female reviewers (OR 1.27, 95% CI 1.19-1.35; P<.001). Female patients were less likely to report disrespect from female doctors than expected from the base ORs (OR 1.19, 95% CI 1.04-1.32; P=.008), but this effect overrode only the effect for female reviewers. Conclusions This work reinforces findings in the extensive literature on gender differences and gender bias in patient-physician interaction. Its novel contribution lies in highlighting gender differences in online reviews. These reviews inform patients’ choice of doctor and thus affect both patients and physicians. The evidence of gender bias documented here suggests review sites may be improved by providing information about gender differences, controlling for gender when presenting composite ratings for physicians, and helping users write less biased reviews. 
    more » « less
  3. ; ; (, Proceedings of the AAAI Conference on Artificial Intelligence)
    With the wide application of electronic health records (EHR) in healthcare facilities, health event prediction with deep learning has gained more and more attention. A common feature of EHR data used for deep-learning-based predictions is historical diagnoses. Existing work mainly regards a diagnosis as an independent disease and does not consider clinical relations among diseases in a visit. Many machine learning approaches assume disease representations are static in different visits of a patient. However, in real practice, multiple diseases that are frequently diagnosed at the same time reflect hidden patterns that are conducive to prognosis. Moreover, the development of a disease is not static since some diseases can emerge or disappear and show various symptoms in different visits of a patient. To effectively utilize this combinational disease information and explore the dynamics of diseases, we propose a novel context-aware learning framework using transition functions on dynamic disease graphs. Specifically, we construct a global disease co-occurrence graph with multiple node properties for disease combinations. We design dynamic subgraphs for each patient's visit to leverage global and local contexts. We further define three diagnosis roles in each visit based on the variation of node properties to model disease transition processes. Experimental results on two real-world EHR datasets show that the proposed model outperforms state of the art in predicting health events. 
    more » « less
  4. ; ; ; ; ; ; ; ; ; ; et al (, European Heart Journal: Acute Cardiovascular Care)
    Abstract AimsMyocardial infarction and heart failure are major cardiovascular diseases that affect millions of people in the USA with morbidity and mortality being highest among patients who develop cardiogenic shock. Early recognition of cardiogenic shock allows prompt implementation of treatment measures. Our objective is to develop a new dynamic risk score, called CShock, to improve early detection of cardiogenic shock in the cardiac intensive care unit (ICU). Methods and resultsWe developed and externally validated a deep learning-based risk stratification tool, called CShock, for patients admitted into the cardiac ICU with acute decompensated heart failure and/or myocardial infarction to predict the onset of cardiogenic shock. We prepared a cardiac ICU dataset using the Medical Information Mart for Intensive Care-III database by annotating with physician-adjudicated outcomes. This dataset which consisted of 1500 patients with 204 having cardiogenic/mixed shock was then used to train CShock. The features used to train the model for CShock included patient demographics, cardiac ICU admission diagnoses, routinely measured laboratory values and vital signs, and relevant features manually extracted from echocardiogram and left heart catheterization reports. We externally validated the risk model on the New York University (NYU) Langone Health cardiac ICU database which was also annotated with physician-adjudicated outcomes. The external validation cohort consisted of 131 patients with 25 patients experiencing cardiogenic/mixed shock. CShock achieved an area under the receiver operator characteristic curve (AUROC) of 0.821 (95% CI 0.792–0.850). CShock was externally validated in the more contemporary NYU cohort and achieved an AUROC of 0.800 (95% CI 0.717–0.884), demonstrating its generalizability in other cardiac ICUs. Having an elevated heart rate is most predictive of cardiogenic shock development based on Shapley values. The other top 10 predictors are having an admission diagnosis of myocardial infarction with ST-segment elevation, having an admission diagnosis of acute decompensated heart failure, Braden Scale, Glasgow Coma Scale, blood urea nitrogen, systolic blood pressure, serum chloride, serum sodium, and arterial blood pH. ConclusionThe novel CShock score has the potential to provide automated detection and early warning for cardiogenic shock and improve the outcomes for millions of patients who suffer from myocardial infarction and heart failure. 
    more » « less
  5. ; ; (, IEEE Transactions on Cybernetics)
    Electronic health records (EHRs) have been heavily used in modern healthcare systems for recording patients' admission information to health facilities. Many data-driven approaches employ temporal features in EHR for predicting specific diseases, readmission times, and diagnoses of patients. However, most existing predictive models cannot fully utilize EHR data, due to an inherent lack of labels in supervised training for some temporal events. Moreover, it is hard for the existing methods to simultaneously provide generic and personalized interpretability. To address these challenges, we propose Sherbet, a self-supervised graph learning framework with hyperbolic embeddings for temporal health event prediction. We first propose a hyperbolic embedding method with information flow to pretrain medical code representations in a hierarchical structure. We incorporate these pretrained representations into a graph neural network (GNN) to detect disease complications and design a multilevel attention method to compute the contributions of particular diseases and admissions, thus enhancing personalized interpretability. We present a new hierarchy-enhanced historical prediction proxy task in our self-supervised learning framework to fully utilize EHR data and exploit medical domain knowledge. We conduct a comprehensive set of experiments on widely used publicly available EHR datasets to verify the effectiveness of our model. Our results demonstrate the proposed model's strengths in both predictive tasks and interpretable abilities. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email