Dataset Information

Combining chest X-rays and electronic health record (EHR) data using machine learning to diagnose acute respiratory failure.

ABSTRACT:

Objective

When patients develop acute respiratory failure (ARF), accurately identifying the underlying etiology is essential for determining the best treatment. However, differentiating between common medical diagnoses can be challenging in clinical practice. Machine learning models could improve medical diagnosis by aiding in the diagnostic evaluation of these patients.

Materials and methods

Machine learning models were trained to predict the common causes of ARF (pneumonia, heart failure, and/or chronic obstructive pulmonary disease [COPD]). Models were trained using chest radiographs and clinical data from the electronic health record (EHR) and applied to an internal and external cohort.

Results

The internal cohort of 1618 patients included 508 (31%) with pneumonia, 363 (22%) with heart failure, and 137 (8%) with COPD based on physician chart review. A model combining chest radiographs and EHR data outperformed models based on each modality alone. Models had similar or better performance compared to a randomly selected physician reviewer. For pneumonia, the combined model area under the receiver operating characteristic curve (AUROC) was 0.79 (0.77-0.79), image model AUROC was 0.74 (0.72-0.75), and EHR model AUROC was 0.74 (0.70-0.76). For heart failure, combined: 0.83 (0.77-0.84), image: 0.80 (0.71-0.81), and EHR: 0.79 (0.75-0.82). For COPD, combined: AUROC = 0.88 (0.83-0.91), image: 0.83 (0.77-0.89), and EHR: 0.80 (0.76-0.84). In the external cohort, performance was consistent for heart failure and increased for COPD, but declined slightly for pneumonia.

Conclusions

Machine learning models combining chest radiographs and EHR data can accurately differentiate between common causes of ARF. Further work is needed to determine how these models could act as a diagnostic aid to clinicians in clinical settings.

SUBMITTER: Jabbour S

PROVIDER: S-EPMC9093032 | biostudies-literature | 2022 May

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Combining chest X-rays and electronic health record (EHR) data using machine learning to diagnose acute respiratory failure.

Jabbour Sarah S Fouhey David D Kazerooni Ella E Wiens Jenna J Sjoding Michael W MW

Journal of the American Medical Informatics Association : JAMIA 20220501 6

<h4>Objective</h4>When patients develop acute respiratory failure (ARF), accurately identifying the underlying etiology is essential for determining the best treatment. However, differentiating between common medical diagnoses can be challenging in clinical practice. Machine learning models could improve medical diagnosis by aiding in the diagnostic evaluation of these patients.<h4>Materials and methods</h4>Machine learning models were trained to predict the common causes of ARF (pneumonia, hear ...[more]

PMID: 35271711

Similar Datasets

Project description:BackgroundThe electronic medical record (EMR) contains a rich source of information that could be harnessed for epidemic surveillance. We asked if structured EMR data could be coupled with computerized processing of free-text clinical entries to enhance detection of acute respiratory infections (ARI).MethodologyA manual review of EMR records related to 15,377 outpatient visits uncovered 280 reference cases of ARI. We used logistic regression with backward elimination to determine which among candidate structured EMR parameters (diagnostic codes, vital signs and orders for tests, imaging and medications) contributed to the detection of those reference cases. We also developed a computerized free-text search to identify clinical notes documenting at least two non-negated ARI symptoms. We then used heuristics to build case-detection algorithms that best combined the retained structured EMR parameters with the results of the text analysis.Principal findingsAn adjusted grouping of diagnostic codes identified reference ARI patients with a sensitivity of 79%, a specificity of 96% and a positive predictive value (PPV) of 32%. Of the 21 additional structured clinical parameters considered, two contributed significantly to ARI detection: new prescriptions for cough remedies and elevations in body temperature to at least 38°C. Together with the diagnostic codes, these parameters increased detection sensitivity to 87%, but specificity and PPV declined to 95% and 25%, respectively. Adding text analysis increased sensitivity to 99%, but PPV dropped further to 14%. Algorithms that required satisfying both a query of structured EMR parameters as well as text analysis disclosed PPVs of 52-68% and retained sensitivities of 69-73%.ConclusionStructured EMR parameters and free-text analyses can be combined into algorithms that can detect ARI cases with new levels of sensitivity or precision. These results highlight potential paths by which repurposed EMR information could facilitate the discovery of epidemics before they cause mass casualties.

Project description:BackgroundElectronic health record (EHR) transitions are inherently disruptive to healthcare workers who must rapidly learn a new EHR and adapt to altered clinical workflows. Healthcare workers' perceptions of EHR usability and their EHR use patterns following transitions are poorly understood. The Department of Veterans Affairs (VA) is currently replacing its homegrown EHR with a commercial Cerner EHR, presenting a unique opportunity to examine EHR use trends and usability perceptions.ObjectiveTo assess EHR usability and uptake up to 1-year post-transition at the first VA EHR transition site using a novel longitudinal, mixed methods approach.DesignA concurrent mixed methods strategy using EHR use metrics and qualitative interview data.Participants141 clinicians with data from select EHR use metrics in Cerner Lights On Network®. Interviews with 25 healthcare workers in various clinical and administrative roles.ApproachWe assessed changes in total EHR time, documentation time, and order time per patient post-transition. Interview transcripts (n = 90) were coded and analyzed for content specific to EHR usability.Key resultsTotal EHR time, documentation time, and order time all decreased precipitously within the first four months after go-live and demonstrated gradual improvements over 12 months. Interview participants expressed ongoing concerns with the EHR's usability and functionality up to a year after go-live such as tasks taking longer than the old system and inefficiencies related to inadequate training and inherent features of the new system. These sentiments did not seem to reflect the observed improvements in EHR use metrics.ConclusionsThe integration of quantitative and qualitative data yielded a complex picture of EHR usability. Participants described persistent challenges with EHR usability 1 year after go-live contrasting with observed improvements in EHR use metrics. Combining findings across methods can provide a clearer, contextualized understanding of EHR adoption and use patterns during EHR transitions.

Project description:AimsUnderstanding atypical forms of diabetes (AD) may advance precision medicine, but methods to identify such patients are needed. We propose an electronic health record (EHR)-based algorithmic approach to identify patients who may have AD, specifically those with insulin-sufficient, non-metabolic diabetes, in order to improve feasibility of identifying these patients through detailed chart review.MethodsPatients with likely T2D were selected using a validated machine-learning (ML) algorithm applied to EHR data. "Typical" T2D cases were removed by excluding individuals with obesity, evidence of dyslipidemia, antibody-positive diabetes, or cystic fibrosis. To filter out likely type 1 diabetes (T1D) cases, we applied six additional "branch algorithms," relying on various clinical characteristics, which resulted in six overlapping cohorts. Diabetes type was classified by manual chart review as atypical, not atypical, or indeterminate due to missing information.ResultsOf 114,975 biobank participants, the algorithms collectively identified 119 (0.1%) potential AD cases, of which 16 (0.014%) were confirmed after expert review. The branch algorithm that excluded T1D based on outpatient insulin use had the highest percentage yield of AD (13 of 27; 48.2% yield). Together, the 16 AD cases had significantly lower BMI and higher HDL than either unselected T1D or T2D cases identified by ML algorithms (P<0.05). Compared to the ML T1D group, the AD group had a significantly higher T2D polygenic score (P<0.01) and lower hemoglobin A1c (P<0.01).ConclusionOur EHR-based algorithms followed by manual chart review identified collectively 16 individuals with AD, representing 0.22% of biobank enrollees with T2D. With a maximum yield of 48% cases after manual chart review, our algorithms have the potential to drastically improve efficiency of AD identification. Recognizing patients with AD may inform on the heterogeneity of T2D and facilitate enrollment in studies like the Rare and Atypical Diabetes Network (RADIANT).

Project description:BackgroundThe predominant implementation paradigm of electronic health record (EHR) systems in low- and middle-income countries (LMICs) relies on standalone system installations at facilities. This implementation approach exacerbates the digital divide, with facilities in areas with inadequate electrical and network infrastructure often left behind. Mobile health (mHealth) technologies have been implemented to extend the reach of digital health, but these systems largely add to the problem of siloed patient data, with few seamlessly interoperating with the EHR systems that are now scaled nationally in many LMICs. Robust mHealth applications that effectively extend EHR systems are needed to improve access, improve quality of care, and ameliorate the digital divide.ObjectiveWe report on the development and scaled implementation of mUzima, an mHealth extension of the most broadly deployed EHR system in LMICs (OpenMRS).MethodsThe "Guidelines for reporting of health interventions using mobile phones: mobile (mHealth) evidence reporting assessment (mERA)" checklist was employed to report on the mUzima application. The World Health Organization (WHO) Principles for Digital Development framework was used as a secondary reference framework. Details of mUzima's architecture, core features, functionalities, and its implementation status are provided to highlight elements that can be adapted in other systems.ResultsmUzima is an open-source, highly configurable Android application with robust features including offline management, deduplication, relationship management, security, cohort management, and error resolution, among many others. mUzima allows providers with lower-end Android smartphones (version 4.4 and above) who work remotely to access historical patient data, collect new data, view media, leverage decision support, conduct store-and-forward teleconsultation, and geolocate clients. The application is supported by an active community of developers and users, with feature priorities vetted by the community. mUzima has been implemented nationally in Kenya, is widely used in Rwanda, and is gaining scale in Uganda and Mozambique. It is disease-agnostic, with current use cases in HIV, cancer, chronic disease, and COVID-19 management, among other conditions. mUzima meets all WHO's Principles of Digital Development, and its scaled implementation success has led to its recognition as a digital global public good and its listing in the WHO Digital Health Atlas.ConclusionsGreater emphasis should be placed on mHealth applications that robustly extend reach of EHR systems within resource-limited settings, as opposed to siloed mHealth applications. This is particularly important given that health information exchange infrastructure is yet to mature in many LMICs. The mUzima application demonstrates how this can be done at scale, as evidenced by its adoption across multiple countries and for numerous care domains.

Project description:AimsThe diagnosis of heart failure (HF) is an important problem in primary care. We previously demonstrated a 74% increase in registered HF diagnoses in primary care electronic health records (EHRs) following an extended audit procedure. What remains unclear is the accuracy of registered HF pre-audit and which EHR variables are most important in the extended audit strategy. This study aims to describe the diagnostic HF classification sequence at different stages, assess general practitioner (GP) HF misclassification, and test the predictive performance of an optimized audit.Methods and resultsThis is a secondary analysis of the OSCAR-HF study, a prospective observational trial including 51 participating GPs. OSCAR used an extended audit based on typical HF risk factors, signs, symptoms, and medications in GPs' EHR. This resulted in a list of possible HF patients, which participating GPs had to classify as HF or non-HF. We compared registered HF diagnoses before and after GPs' assessment. For our analysis of audit performance, we used GPs' assessment of HF as primary outcome and audit queries as dichotomous predictor variables for a gradient boosted machine (GBM) decision tree algorithm and logistic regression model. Of the 18 011 patients eligible for the audit intervention, 4678 (26.0%) were identified as possible HF patients and submitted for GPs' assessment in the audit stage. There were 310 patients with registered HF before GP assessment, of whom 146 (47.1%) were judged not to have HF by their GP (over-registration). There were 538 patients with registered HF after GP assessment, of whom 374 (69.5%) did not have registered HF before GP assessment (under-registration). The GBM and logistic regression model had a comparable predictive performance (area under the curve of 0.70 [95% confidence interval 0.65-0.77] and 0.69 [95% confidence interval 0.64-0.75], respectively). This was not significantly impacted by reducing the set of predictor variables to the 10 most important variables identified in the GBM model (free-text and coded cardiomyopathy, ischaemic heart disease and atrial fibrillation, digoxin, mineralocorticoid receptor antagonists, and combinations of renin-angiotensin system inhibitors and beta-blockers with diuretics). This optimized query set was enough to identify 86% (n = 461/538) of GPs' self-assessed HF population with a 33% reduction (n = 1537/4678) in screening caseload.ConclusionsDiagnostic coding of HF in primary care health records is inaccurate with a high degree of under-registration and over-registration. An optimized query set enabled identification of more than 80% of GPs' self-assessed HF population.

Dataset Information

Combining chest X-rays and electronic health record (EHR) data using machine learning to diagnose acute respiratory failure.

Objective

Materials and methods

Results

Conclusions

Publications

Combining chest X-rays and electronic health record (EHR) data using machine learning to diagnose acute respiratory failure.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets