Dataset Information

An Electronic Medical Record-Based Discharge Disposition Tool Gets Bundle Busted: Decaying Relevance of Clinical Data Accuracy in Machine Learning.

ABSTRACT: Background:Determining discharge disposition after total joint arthroplasty (TJA) has been a challenge. Advances in machine learning (ML) have produced computer models that learn by example to generate predictions on future events. We hypothesized a trained ML algorithm's diagnostic accuracy will be better than that of current predictive tools to predict discharge disposition after primary TJA. Methods:This study was a retrospective cohort study from a single, tertiary referral center for primary TJA. We trained and validated an artificial neural network (ANN) based on 4368 distinct surgical encounters between 1/1/2013 and 6/28/2016. The ANN's ability to identify discharge disposition was then tested on 1452 distinct surgical encounters between 1/3/17 and 11/30/17. Results:The area under the curve and accuracy achieved during model validation were 0.973 and 91.7%, respectively, with 25% of patients being discharged to skilled nursing facilities (SNFs). Within our testing data set, 6.7% of patients went to SNFs. The performance in the testing set included an area under the curve of 0.804, accuracy of 61.3%, sensitivity of 28.9%, and specificity of 93.8%. Conclusions:This is the first prediction tool using an electronic medical record-integrated ANN to predict discharge disposition after TJA based on locally generated data. Dramatically reduced numbers of patients discharged to SNFs due to implementation of a bundled payment model lead to poor recall in the testing model. This model serves as a proof of concept for developing an ML prediction tool using a relatively small data set and subsequent integration into the electronic medical record.

SUBMITTER: Greenstein AS

PROVIDER: S-EPMC7567055 | biostudies-literature | 2020 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

An Electronic Medical Record-Based Discharge Disposition Tool Gets Bundle Busted: Decaying Relevance of Clinical Data Accuracy in Machine Learning.

Greenstein Alexander S AS Teitel Jack J Mitten David J DJ Ricciardi Benjamin F BF Myers Thomas G TG

Arthroplasty today 20201014 4

<h4>Background</h4>Determining discharge disposition after total joint arthroplasty (TJA) has been a challenge. Advances in machine learning (ML) have produced computer models that learn by example to generate predictions on future events. We hypothesized a trained ML algorithm's diagnostic accuracy will be better than that of current predictive tools to predict discharge disposition after primary TJA.<h4>Methods</h4>This study was a retrospective cohort study from a single, tertiary referral ce ...[more]

PMID: 33088883

Similar Datasets

Project description:BackgroundObtaining accurate clinical information about recent acute care visits is extremely important for outpatient providers. However, documents used to communicate this information are often difficult to use. This puts patients at risk of adverse events. Elderly patients who are seen by more providers and have more care transitions are especially vulnerable.ObjectiveThis study aimed to (1) identify the information about elderly patients' recent acute care visits needed to coordinate their care, (2) use this information to assess discharge summaries, and (3) provide recommendations to help improve the quality of electronic health record (EHR)-generated discharge summaries, thereby increasing patient safety.MethodsA literature review, clinician interviews, and a survey of outpatient providers were used to identify and categorize data needed to coordinate care for recently discharged elderly patients. Based upon those data, 2 guidelines for creating useful discharge summaries were created. The new guidelines, along with 17 previously developed medical documentation usability heuristics, were applied to assess 4 simulated elderly patient discharge summaries.ResultsThe initial research effort yielded a list of 29 items that should always be included in elderly patient discharge summaries and a list of 7 "helpful, but not always necessary" items. Evaluation of 4 deidentified elderly patient discharge summaries revealed that none of the documents contained all 36 necessary items; between 14 and 18 were missing. The documents each had several other issues, and they differed significantly in organization, layout, and formatting.ConclusionsVariations in content and structure of discharge summaries in the United States make them unnecessarily difficult to use. Standardization would benefit both patients, by lowering the risk of care transition-related adverse events, and outpatient providers, by helping reduce frustration that can contribute to burnout. In the short term, acute care providers can help improve the quality of their discharge summaries by working with EHR vendors to follow recommendations based upon this study. Meanwhile, additional human factors work should determine the most effective way to organize and present information in discharge summaries, to facilitate effective standardization.

Project description:ObjectiveAdoption of clinical decision support (CDS) tools by clinicians is often limited by workflow barriers. We sought to assess characteristics associated with clinician use of an electronic health record-embedded clinical decision support system (CDSS).MethodsIn a prospective study on emergency department (ED) activation of a CDSS tool across 14 hospitals between 9/1/14 to 4/30/15, the CDSS was deployed at 10 active sites with an on-site champion, education sessions, iterative feedback, and up to 3 gift cards/clinician as an incentive. The tool was also deployed at 4 passive sites that received only an introductory educational session. Activation of the CDSS - which calculated the Pulmonary Embolism Severity Index (PESI) score and provided guidance - and associated clinical data were collected prospectively. We used multivariable logistic regression with random effects at provider/facility levels to assess the association between activation of the CDSS tool and characteristics at: 1) patient level (PESI score), 2) provider level (demographics and clinical load at time of activation opportunity), and 3) facility level (active vs. passive site, facility ED volume, and ED acuity at time of activation opportunity).ResultsOut of 662 eligible patient encounters, the CDSS was activated in 55%: active sites: 68% (346/512); passive sites 13% (20/150). In bivariate analysis, active sites had an increase in activation rates based on the number of prior gift cards the physician had received (96% if 3 prior cards versus 60% if 0, p<0.0001). At passive sites, physicians < age 40 had higher rates of activation (p=0.03). In multivariable analysis, active site status, low ED volume at the time of diagnosis and PESI scores I or II (compared to III or higher) were associated with higher likelihood of CDSS activation.ConclusionsPerforming on-site tool promotion significantly increased odds of CDSS activation. Optimizing CDSS adoption requires active education.

Dataset Information

An Electronic Medical Record-Based Discharge Disposition Tool Gets Bundle Busted: Decaying Relevance of Clinical Data Accuracy in Machine Learning.

Publications

An Electronic Medical Record-Based Discharge Disposition Tool Gets Bundle Busted: Decaying Relevance of Clinical Data Accuracy in Machine Learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets