Search

BackgroundThe COVID-19 pandemic required clinicians to care for a disease with evolving characteristics while also adhering to care changes (e.g., physical distancing practices) that might lead to diagnostic errors (DEs).ObjectiveTo determine the frequency of DEs and their causes among patients hospitalized under investigation (PUI) for COVID-19.DesignRetrospective cohort.SettingEight medical centers affiliated with the Hospital Medicine ReEngineering Network (HOMERuN).Target populationAdults hospitalized under investigation (PUI) for COVID-19 infection between February and July 2020.MeasurementsWe randomly selected up to 8 cases per site per month for review, with each case reviewed by two clinicians to determine whether a DE (defined as a missed or delayed diagnosis) occurred, and whether any diagnostic process faults took place. We used bivariable statistics to compare patients with and without DE and multivariable models to determine which process faults or patient factors were associated with DEs.ResultsTwo hundred and fifty-seven patient charts underwent review, of which 36 (14%) had a diagnostic error. Patients with and without DE were statistically similar in terms of socioeconomic factors, comorbidities, risk factors for COVID-19, and COVID-19 test turnaround time and eventual positivity. Most common diagnostic process faults contributing to DE were problems with clinical assessment, testing choices, history taking, and physical examination (all p

BackgroundPrior evaluation at our hospital demonstrated that, compared to White patients, Black and Latinx patients with congestive heart failure (CHF) were less likely to be admitted to the cardiology service rather than the general medicine service (GMS). Patients admitted to GMS (compared to cardiology) had inferior rates of cardiology follow-up and 30-day readmission.ObjectiveTo develop and test the feasibility and impacts of using quality improvement (QI) methods, in combination with the Public Health Critical Race Praxis (PHCRP) framework, to engage stakeholders in developing an intervention for ensuring guideline-concordant inpatient CHF care across all patient groups.MethodsWe compared measures for all patients admitted with CHF to GMS between September 2019 and March 2020 (intervention group) to CHF patients admitted to GMS in the previous year (pre-intervention group) and those admitted to cardiology during the pre-intervention and intervention periods (cardiology group). Our primary measures were 30-day readmissions and 14- and 30-day post-discharge cardiology follow-up.ResultsThere were 79 patients admitted with CHF to GMS during the intervention period, all of whom received the intervention. There were similar rates of Black and Latinx patients across the three groups. Compared to pre-intervention, intervention patients had a significantly lower 30-day readmission rate (18.9% vs. 24.8%; p=0.024), though the cardiology group also had a decrease in 30-day readmissions from the pre-intervention to intervention period. Compared to pre-intervention, intervention patients had significantly higher 14-day and 30-day post-discharge follow-up visits scheduled with cardiology (36.7% vs. 24.8%, p=0.005; 55.7% vs. 42.3%, p=0.0029), but no improvement in appointment attendance.ConclusionThis study provides a first test of applying the PHCRP framework within a stakeholder-engaged QI initiative for improving CHF care across races and ethnicities. Our study design cannot evaluate causation. However, the improvements in 30-day readmission, as well as in processes of care that may affect it, provide optimism that inclusion of a racism-conscious framework in QI initiatives is feasible and may enhance QI measures.

Background: The COVID-19 pandemic required clinicians to care for a disease with evolving characteristics while also adhering to care changes (e.g., physical distancing practices) that might lead to diagnostic errors (DEs). Objective: To determine the frequency of DEs and their causes among patients hospitalized under investigation (PUI) for COVID-19. Design: Retrospective cohort. Setting: Eight medical centers affiliated with the Hospital Medicine ReEngineering Network (HOMERuN). Target population: Adults hospitalized under investigation (PUI) for COVID-19 infection between February and July 2020. Measurements: We randomly selected up to 8 cases per site per month for review, with each case reviewed by two clinicians to determine whether a DE (defined as a missed or delayed diagnosis) occurred, and whether any diagnostic process faults took place. We used bivariable statistics to compare patients with and without DE and multivariable models to determine which process faults or patient factors were associated with DEs. Results: Two hundred and fifty-seven patient charts underwent review, of which 36 (14%) had a diagnostic error. Patients with and without DE were statistically similar in terms of socioeconomic factors, comorbidities, risk factors for COVID-19, and COVID-19 test turnaround time and eventual positivity. Most common diagnostic process faults contributing to DE were problems with clinical assessment, testing choices, history taking, and physical examination (all p < 0.01). Diagnostic process faults associated with policies and procedures related to COVID-19 were not associated with DE risk. Fourteen patients (35.9% of patients with errors and 5.4% overall) suffered harm or death due to diagnostic error. Limitations: Results are limited by available documentation and do not capture communication between providers and patients. Conclusion: Among PUI patients, DEs were common and not associated with pandemic-related care changes, suggesting the importance of more general diagnostic process gaps in error propagation. [ABSTRACT FROM AUTHOR]

Twenty- to fifty-thousand patients die annually within 72 h of interhospital transfer (early death after transfer; EDAT). The characteristics and trajectories of these patients are ill-defined. In this retrospective cohort study, we characterized EDAT at three representative major referral centers. Primary outcomes included the presence and timing of goals of care (GOC) and/or prognostic discussions. Among 190 medical patients experiencing EDAT, 95 (50.0%) were >65 years, 115 (60.5%) male, and 137 (72.6%) White; 140 (73.7%) patients traveled >50 miles from home, and 174 (91.6%) were referred for specialty care. Whereas GOC were documented pretransfer for 40 patients (21.1%) and unknown for 97 patients (51%); 152 (80.0%) had posttransfer discussions, often within 24 h of death (125; 82.2%). Transfer >50 miles was associated with death ≤24 h after transfer and with posttransfer changes in code status. Further research is needed to evaluate disparities and describe the potential burdens of transfer at end-of-life. Infrequent pretransfer discussions of GOC suggest potential targets for improvement., (© 2024 Society of Hospital Medicine.)

Background: Large language models (LLMs) have shown promising performance in various healthcare domains, but their effectiveness in identifying specific clinical conditions in real medical records is less explored. This study evaluates LLMs for detecting signs of cognitive decline in real electronic health record (EHR) clinical notes, comparing their error profiles with traditional models. The insights gained will inform strategies for performance enhancement., Methods: This study, conducted at Mass General Brigham in Boston, MA, analysed clinical notes from the four years prior to a 2019 diagnosis of mild cognitive impairment in patients aged 50 and older. We developed prompts for two LLMs, Llama 2 and GPT-4, on Health Insurance Portability and Accountability Act (HIPAA)-compliant cloud-computing platforms using multiple approaches (e.g., hard prompting, retrieval augmented generation, and error analysis-based instructions) to select the optimal LLM-based method. Baseline models included a hierarchical attention-based neural network and XGBoost. Subsequently, we constructed an ensemble of the three models using a majority vote approach. Confusion-matrix-based scores were used for model evaluation., Findings: We used a randomly annotated sample of 4949 note sections from 1969 patients (women: 1046 [53.1%]; age: mean, 76.0 [SD, 13.3] years), filtered with keywords related to cognitive functions, for model development. For testing, a random annotated sample of 1996 note sections from 1161 patients (women: 619 [53.3%]; age: mean, 76.5 [SD, 10.2] years) without keyword filtering was utilised. GPT-4 demonstrated superior accuracy and efficiency compared to Llama 2, but did not outperform traditional models. The ensemble model outperformed the individual models in terms of all evaluation metrics with statistical significance (p < 0.01), achieving a precision of 90.2% [95% CI: 81.9%-96.8%], a recall of 94.2% [95% CI: 87.9%-98.7%], and an F1-score of 92.1% [95% CI: 86.8%-96.4%]. Notably, the ensemble model showed a significant improvement in precision, increasing from a range of 70%-79% to above 90%, compared to the best-performing single model. Error analysis revealed that 63 samples were incorrectly predicted by at least one model; however, only 2 cases (3.2%) were mutual errors across all models, indicating diverse error profiles among them., Interpretation: LLMs and traditional machine learning models trained using local EHR data exhibited diverse error profiles. The ensemble of these models was found to be complementary, enhancing diagnostic performance. Future research should investigate integrating LLMs with smaller, localised models and incorporating medical data and domain knowledge to enhance performance on specific tasks., Funding: This research was supported by the National Institute on Aging grants (R44AG081006, R01AG080429) and National Library of Medicine grant (R01LM014239)., Competing Interests: Declaration of interests None., (Copyright © 2024 The Author(s). Published by Elsevier B.V. All rights reserved.)

The practice of transferring patients between acute care hospitals is variable and largely nonstandardized. Although often-cited reasons for transfer include providing patients access to specialty services only available at the receiving institution, little is known about whether and when patients receive such specialty care during the transfer continuum. We performed a retrospective analysis using 2013 100% Master Beneficiary Summary and Inpatient claims files from Centers for Medicare and Medicaid Services. Beneficiaries were included if they were aged =65 years, continuously enrolled in Medicare A and B, with an acute care hospitalization claim, and transferred to another acute care hospital with a primary diagnosis of acute myocardial infarction, gastrointestinal bleed, renal failure, or hip fracture/dislocation. Associated specialty procedure codes (International Classification of Diseases, Ninth Revision, Clinical Modification) were identified for each diagnosis. We performed descriptive analyses to compare receipt of specialty procedural services between transferring and receiving hospitals, stratified by diagnosis. Across the 19,613 included beneficiaries, receipt of associated specialty procedures was more common at the receiving than the transferring hospital, with the exception of patients with a diagnosis of gastrointestinal bleed. Depending on primary diagnosis, between 32.4% and 89.1% of patients did not receive any associated specialty procedure at the receiving hospital. Our results demonstrate variable receipt of specialty procedural care across the transfer continuum, implying the likelihood of alternate drivers of interhospital transfer other than solely receipt of specialty procedural care., (© 2017 Society of Hospital Medicine.)