Dataset Information

Generating Accurate Electronic Health Assessment from Medical Graph.

ABSTRACT: One of the fundamental goals of artificial intelligence is to build computer-based expert systems. Inferring clinical diagnoses to generate a clinical assessment during a patient encounter is a crucial step towards building a medical diagnostic system. Previous works were mainly based on either medical domain-specific knowledge, or patients' prior diagnoses and clinical encounters. In this paper, we propose a novel model for automated clinical assessment generation (MCAG). MCAG is built on an innovative graph neural network, where rich clinical knowledge is incorporated into an end-to-end corpus-learning system. Our evaluation results against physician generated gold standard show that MCAG significantly improves the BLEU and rouge score compared with competitive baseline models. Further, physicians' evaluation showed that MCAG could generate high-quality assessments.

SUBMITTER: Yang Z

PROVIDER: S-EPMC7821471 | biostudies-literature | 2020 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Generating Accurate Electronic Health Assessment from Medical Graph.

Yang Zhichao Z Yu Hong H

Proceedings of the Conference on Empirical Methods in Natural Language Processing. Conference on Empirical Methods in Natural Language Processing 20201101

One of the fundamental goals of artificial intelligence is to build computer-based expert systems. Inferring clinical diagnoses to generate a clinical assessment during a patient encounter is a crucial step towards building a medical diagnostic system. Previous works were mainly based on either medical domain-specific knowledge, or patients' prior diagnoses and clinical encounters. In this paper, we propose a novel model for automated clinical assessment generation (MCAG). MCAG is built on an in ...[more]

PMID: 33491009

Similar Datasets

Project description:BACKGROUND:Technological advances in personal informatics allow people to track their own health in a variety of ways, representing a dramatic change in individuals' control of their own wellness. However, research regarding patient interpretation of traditional medical tests highlights the risks in making complex medical data available to a general audience. OBJECTIVE:This study aimed to explore how people interpret medical test results, examined in the context of a mobile blood testing system developed to enable self-care and health management. METHODS:In a preliminary investigation and main study, we presented 27 and 303 adults, respectively, with hypothetical results from several blood tests via one of the several mobile interface designs: a number representing the raw measurement of the tested biomarker, natural language text indicating whether the biomarker's level was low or high, or a one-dimensional chart illustrating this level along a low-healthy axis. We measured respondents' correctness in evaluating these results and their confidence in their interpretations. Participants also told us about any follow-up actions they would take based on the result and how they envisioned, generally, using our proposed personal health system. RESULTS:We find that a majority of participants (242/328, 73.8%) were accurate in their interpretations of their diagnostic results. However, 135 of 328 participants (41.1%) expressed uncertainty and confusion about their ability to correctly interpret these results. We also find that demographics and interface design can impact interpretation accuracy, including false confidence, which we define as a respondent having above average confidence despite interpreting a result inaccurately. Specifically, participants who saw a natural language design were the least likely (421.47 times, P=.02) to exhibit false confidence, and women who saw a graph design were less likely (8.67 times, P=.04) to have false confidence. On the other hand, false confidence was more likely among participants who self-identified as Asian (25.30 times, P=.02), white (13.99 times, P=.01), and Hispanic (6.19 times, P=.04). Finally, with the natural language design, participants who were more educated were, for each one-unit increase in education level, more likely (3.06 times, P=.02) to have false confidence. CONCLUSIONS:Our findings illustrate both promises and challenges of interpreting medical data outside of a clinical setting and suggest instances where personal informatics may be inappropriate. In surfacing these tensions, we outline concrete interface design strategies that are more sensitive to users' capabilities and conditions.

Project description:BackgroundElectronic clinical quality measures (eCQMs) from electronic health records (EHRs) are a key component of quality improvement (QI) initiatives in small-to-medium size primary care practices, but using eCQMs for QI can be challenging. Organizational strategies are needed to effectively operationalize eCQMs for QI in these practice settings.ObjectiveThis study aimed to characterize strategies that seven regional cooperatives participating in the EvidenceNOW initiative developed to generate and report EHR-based eCQMs for QI in small-to-medium size practices.MethodsA qualitative study comprised of 17 interviews with representatives from all seven EvidenceNOW cooperatives was conducted. Interviewees included administrators were with both strategic and cooperative-level operational responsibilities and external practice facilitators were with hands-on experience helping practices use EHRs and eCQMs. A subteam conducted 1-hour semistructured telephone interviews with administrators and practice facilitators, then analyzed interview transcripts using immersion crystallization. The analysis and a conceptual model were vetted and approved by the larger group of coauthors.ResultsCooperative strategies consisted of efforts in four key domains. First, cooperative adaptation shaped overall strategies for calculating eCQMs whether using EHRs, a centralized source, or a "hybrid strategy" of the two. Second, the eCQM generation described how EHR data were extracted, validated, and reported for calculating eCQMs. Third, practice facilitation characterized how facilitators with backgrounds in health information technology (IT) delivered services and solutions for data capture and quality and practice support. Fourth, performance reporting strategies and tools informed QI efforts and how cooperatives could alter their approaches to eCQMs.ConclusionCooperatives ultimately generated and reported eCQMs using hybrid strategies because they determined neither EHRs alone nor centralized sources alone could operationalize eCQMs for QI. This required cooperatives to devise solutions and utilize resources that often are unavailable to typical small-to-medium-sized practices. The experiences from EvidenceNOW cooperatives provide insights into how organizations can plan for challenges and operationalize EHR-based eCQMs.

Project description:Post-market medical device surveillance is a challenge facing manufacturers, regulatory agencies, and health care providers. Electronic health records are valuable sources of real-world evidence for assessing device safety and tracking device-related patient outcomes over time. However, distilling this evidence remains challenging, as information is fractured across clinical notes and structured records. Modern machine learning methods for machine reading promise to unlock increasingly complex information from text, but face barriers due to their reliance on large and expensive hand-labeled training sets. To address these challenges, we developed and validated state-of-the-art deep learning methods that identify patient outcomes from clinical notes without requiring hand-labeled training data. Using hip replacements-one of the most common implantable devices-as a test case, our methods accurately extracted implant details and reports of complications and pain from electronic health records with up to 96.3% precision, 98.5% recall, and 97.4% F1, improved classification performance by 12.8-53.9% over rule-based methods, and detected over six times as many complication events compared to using structured data alone. Using these additional events to assess complication-free survivorship of different implant systems, we found significant variation between implants, including for risk of revision surgery, which could not be detected using coded data alone. Patients with revision surgeries had more hip pain mentions in the post-hip replacement, pre-revision period compared to patients with no evidence of revision surgery (mean hip pain mentions 4.97 vs. 3.23; t?=?5.14; p?<?0.001). Some implant models were associated with higher or lower rates of hip pain mentions. Our methods complement existing surveillance mechanisms by requiring orders of magnitude less hand-labeled training data, offering a scalable solution for national medical device surveillance using electronic health records.

Dataset Information

Generating Accurate Electronic Health Assessment from Medical Graph.

Publications

Generating Accurate Electronic Health Assessment from Medical Graph.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets