Faculty Bibliography

BACKGROUND:Patient experience is a key component of care quality, linked to outcomes like safety and medication adherence. In high-acuity inpatient settings with rotating teams, opportunities exist to better integrate patients' identities and values into the hospital experience. AIM/OBJECTIVE:To design, implement, and evaluate an EHR-integrated tool that enables hospitalized patients to share personal, non-medical information to help care teams deliver more personalized care. SETTING/METHODS:Fourteen adult inpatient units across a large academic medical center (NYU Langone Health, New York, NY). PARTICIPANTS/METHODS:Hospitalized adults and their interprofessional care teams, including physicians, nurses, advanced practice providers, and allied health professionals. PROGRAM DESCRIPTION/METHODS:About Me is an Epic-integrated digital tool that allows patients or caregivers to share brief, non-clinical details-such as hobbies, values, family, and communication preferences-visible to the entire care team. Patients complete entries via the NYU Langone Health App or with nurse assistance; nurses validate content for appropriateness. PROGRAM EVALUATION/RESULTS:Over a 2-month period, 1001 of 5452 eligible patients (18.4%) submitted entries. Entries were brief but meaningful. Five major themes emerged: hobbies, family, identity, career, and animals. DISCUSSION/CONCLUSIONS:Implementation of About Me was feasible and acceptable. Even brief entries offered meaningful context that may strengthen rapport and promote more personalized, compassionate inpatient care.

BACKGROUND:Pediatric skin-limited discoid lupus erythematosus (DLE-only) is rare, with limited data on risk factors for progression to systemic lupus erythematosus (SLE). OBJECTIVE:To assess incidence, risk factors, and phenotype of pediatric DLE-only progression to SLE. METHODS:In this 17-site retrospective cohort of pediatric DLE, the primary outcome was time to SLE diagnosis (ACR classification criterion ≥4). Kaplan-Meier estimates for 1-, 2-, and 5-year progression to SLE were generated. Cox proportional hazards modeling identified baseline predictors of progression to SLE. RESULTS:The 1-year progression rate from DLE-only to SLE was 14.4% (95% CI: 9.6-18.9%). Progression to SLE was most strongly associated with baseline ANA positivity (HR 3.71) and older age (HR 1.11/yr). Antiphospholipid antibodies and cytopenias also predicted progression to SLE in multivariable analysis. The SLE phenotype was relatively mild, with most patients developing mucocutaneous and laboratory criteria (22/236, 9%) and few developing other end-organ disease (7/236, 3%). LIMITATIONS/CONCLUSIONS:Retrospective design, missing data. CONCLUSION/CONCLUSIONS:ANA-positive DLE-only patients warrant close monitoring for progression SLE, especially within the first year. Severe end-organ disease in DLE patients who progress to SLE is uncommon. Future studies should test whether early recognition and intervention in DLE-only slows progression to SLE.

BACKGROUND:Poor documentation quality can significantly affect healthcare operations, but the feedback process for clinicians to improve clinical notes is time-consuming and often insufficient. Large language models (LLMs) such as Generative Pre-trained Transformer 4 (GPT-4) have the potential to streamline this process. OBJECTIVES/OBJECTIVE:To determine whether an LLM can generate feedback to improve the medical contingency and discharge planning (MCDP) component of clinical documentation that is non-inferior to feedback by physicians. METHODS:A cross-sectional study of GPT-4 feedback and physician feedback on inpatient progress notes was conducted. A random sample of 64 inpatient progress notes identified by the validated AI Audit Tool as having a low likelihood of containing MCDP was included from adult general medicine patients hospitalized at New York University Langone Health (NYULH) in December 2023. Both GPT-4 model and attending physicians generated feedback on these inpatient progress notes. A/B testing was then conducted on the measures of understandability, usefulness, acceptability, and impartiality. Evaluations employed 5-point Likert scales that were converted to 10-point bidirectional interval scales for interpretability, ranging from -10 (human suggestions significantly better) to +10 (GPT-4 suggestions significantly better), with a non-inferiority threshold set to -1 for the primary endpoint. RESULTS:64 inpatient progress notes were included, representing 55% female patients with a median age of 73. GPT-4 feedback was non-inferior to physician feedback in all measures: understandability (mean 1.27, 95% CI 0.73 to 1.8, P < 0.001), usefulness (mean 2.09, 95% CI 1.27 to 2.91, P < 0.001), acceptability (mean 2.07, 95% CI 1.33 to 2.81, P < 0.001), and impartiality (mean -0.20, 95% CI -0.52 to 0.12, P < 0.001). CONCLUSIONS:This study shows that an LLM can be leveraged to generate note quality feedback that is non-inferior to expert clinician feedback.

There is limited data on which hospitalist switch day is optimal for hospital operations and throughput. A quality improvement intervention was implemented, changing the hospitalist switch day from Monday to Tuesday. Retrospective observational analysis revealed an increase in Monday discharges (1.3%, p = .01), a decrease in Tuesday discharges (-1.6%, p < .005), and a significant reduction in 30-day unplanned readmission rates (-1.5%, p = .003), with no significant changes in the average length of stay. Additional studies are needed to further verify these findings in different hospital settings and to consider other switch day patterns.

IMPORTANCE/UNASSIGNED:Hospital course (HC) summarization represents an increasingly onerous discharge summary component for physicians. Literature supports large language models (LLMs) for HC summarization, but whether physicians can effectively partner with electronic health record-embedded LLMs to draft HCs is unknown. OBJECTIVES/UNASSIGNED:To compare the editing effort required by time-constrained resident physicians to improve LLM- vs physician-generated HCs toward a novel 4Cs (complete, concise, cohesive, and confabulation-free) HC. DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:Quality improvement study using a convenience sample of 10 internal medicine resident editors, 8 hospitalist evaluators, and randomly selected general medicine admissions in December 2023 lasting 4 to 8 days at New York University Langone Health. EXPOSURES/UNASSIGNED:Residents and hospitalists reviewed randomly assigned patient medical records for 10 minutes. Residents blinded to author type who edited each HC pair (physician and LLM) for quality in 3 minutes, followed by comparative ratings by attending hospitalists. MAIN OUTCOMES AND MEASURES/UNASSIGNED:Editing effort was quantified by analyzing the edits that occurred on the HC pairs after controlling for length (percentage edited) and the degree to which the original HCs' meaning was altered (semantic change). Hospitalists compared edited HC pairs with A/B testing on the 4Cs (5-point Likert scales converted to 10-point bidirectional scales). RESULTS/UNASSIGNED:Among 100 admissions, compared with physician HCs, residents edited a smaller percentage of LLM HCs (LLM mean [SD], 31.5% [16.6%] vs physicians, 44.8% [20.0%]; P < .001). Additionally, LLM HCs required less semantic change (LLM mean [SD], 2.4% [1.6%] vs physicians, 4.9% [3.5%]; P < .001). Attending physicians deemed LLM HCs to be more complete (mean [SD] difference LLM vs physicians on 10-point bidirectional scale, 3.00 [5.28]; P < .001), similarly concise (mean [SD], -1.02 [6.08]; P = .20), and cohesive (mean [SD], 0.70 [6.14]; P = .60), but with more confabulations (mean [SD], -0.98 [3.53]; P = .002). The composite scores were similar (mean [SD] difference LLM vs physician on 40-point bidirectional scale, 1.70 [14.24]; P = .46). CONCLUSIONS AND RELEVANCE/UNASSIGNED:Electronic health record-embedded LLM HCs required less editing than physician-generated HCs to approach a quality standard, resulting in HCs that were comparably or more complete, concise, and cohesive, but contained more confabulations. Despite the potential influence of artificial time constraints, this study supports the feasibility of a physician-LLM partnership for writing HCs and provides a basis for monitoring LLM HCs in clinical practice.

BACKGROUND:Discharging clinically ready patients before noon on their discharge day may influence overall discharge process quality, emergency department (ED) boarding times, and length of stay (LOS). This study evaluated the effectiveness of a discharge before noon (DBN) initiative. METHODS:Many DBN components were refined or added during a pilot, including incorporating the DBN process into daily rounds, an electronic tracking system, and other elements for possible DBN patients such as a car service when appropriate and expedited lab results and physical therapy consults. DBN was evaluated through a retrospective pre-post study (12-month periods). Study patients were from Maimonides Medical Center's medicine units. Kaplan-Meier estimates and a log-rank test characterized and compared the discharge time probabilities in pre-DBN and post-DBN groups. Log-logistic accelerated failure time (AFT) analysis assessed the influence of DBN on discharge time. Secondary analyses examined the relationship between LOS and readmission within 30 days for any cause and DBN. RESULTS:The percentage of patients discharged before noon increased from 5.0% to 11.4% pre/post-DBN (p < 0.001). The AFT analysis estimated that post-DBN patients had discharge times 41.5% earlier (p < 0.001). DBN as an independent factor was not associated with LOS or subsequent readmissions within 30 days for any cause. Despite an increase in the percentage of patients admitted during the daytime (8:00 a.m. to 5:00 p.m.), the median ED boarding time increased by 41 minutes in post-DBN patients (p < 0.001). CONCLUSION:The DBN initiative was associated with an increased percentage of patients discharged before noon. Further research is needed to identify strategies that reliably improve discharge timeliness while reducing ED boarding.

Electronic health records have become an integral part of modern health care, but their implementation has led to unintended consequences, such as poor note quality. This case study explores how NYU Langone Health leveraged artificial intelligence (AI) to address the challenge to improve the content and quality of medical documentation. By quickly and accurately analyzing large volumes of clinical documentation and providing feedback to organizational leadership and individually to providers, AI can help support a culture of continuous note quality improvement, allowing organizations to enhance a critical component of patient care.

BACKGROUND:Hospital acquired anemia is common during admission and can result in increased transfusion and length of stay. Recumbent posture is known to lead to lower hemoglobin measurements. We tested to see if an initiative promoting evening lab draws would lead to higher hemoglobin measurements due to more time in upright posture during the day and evening. METHODS:We included patients hospitalized on 2 medical units, beginning March 26, 2020 and discharged prior to January 25, 2021. On one of the units, we implemented an initiative to have routine laboratory draws in the evening rather than the morning starting on August 26, 2020. There were 1217 patients on the control unit and 1265 on the intervention unit during the entire study period. First we used a linear mixed-effects model to see if timing of blood draw was associated with hemoglobin level in the pre-intervention period. We then compared levels of hemoglobin before and after the intervention using a difference-in-difference analysis. RESULTS:In the pre-intervention period, evening blood draws were associated with higher hemoglobin compared to morning (0.28; 95% CI, 0.22-0.35). Evening blood draws increased with the intervention (10.3% vs 47.9%, P > 0.001). However, the intervention floor was not associated with hemoglobin levels in difference-in-difference analysis (coefficient of -0.15; 95% CI, -0.51-0.21). CONCLUSIONS:While evening blood draws were associated with higher hemoglobin levels, an intervention that successfully changed timing of routine labs to the evening did not lead to an increase in hemoglobin levels.