Journal of hospital medicine最新文献

Background: Nursing and healthcare-associated pneumonia (NHCAP) is defined as pneumonia occurring in individuals with frequent healthcare contact, such as residents of care facilities or patients with impaired activities of daily living. The effectiveness of broad-spectrum antibiotics in treating NHCAP remains unclear.

Objective: To compare clinical outcomes between broad- and narrow-spectrum antibiotic treatments in patients with NHCAP using a nationwide inpatient database.

Methods: Patients diagnosed with NHCAP between April 2014 and March 2022 were identified from the Diagnosis Procedure Combination inpatient database in Japan. Patients were categorised into those receiving broad-spectrum antibiotics (antipseudomonal penicillins, antipseudomonal cephalosporins, and carbapenems) and those receiving narrow-spectrum antibiotics (third-generation cephalosporins and penicillin plus β-lactamase inhibitor combinations). Instrumental variable analysis using hospital preference for broad-spectrum antibiotics was conducted to compare 30-day in-hospital mortality. A subgroup analysis was performed for patients with ≥3 risk factors for antibiotic-resistant pathogens.

Results: Among 828,283 eligible patients, 24.8% received broad-spectrum antibiotics, while 75.2% received narrow-spectrum antibiotics. Instrumental variable analysis showed that broad-spectrum antibiotic use was not associated with 30-day in-hospital mortality (10.0% vs. 10.0%; risk difference, 0.0%; 95% confidence interval, -0.7% to 0.8%) compared with narrow-spectrum antibiotic use. The subgroup analysis of patients with three or more risk factors for antibiotic-resistant pathogens, broad-spectrum antibiotic use was also not associated with 30-day mortality (10.5% vs. 11.0%; risk difference, -0.6%; 95% confidence interval, -2.5% to 1.3%).

Conclusions: Broad-spectrum antibiotic use was not associated with short-term in-hospital mortality in patients with NHCAP, underscoring the importance of individualized antibiotic selection based on patient-specific risk factors.

Background: Alcohol withdrawal hospitalizations present unmet opportunities to initiate medications for alcohol use disorder (MAUD).

Objective: To examine patient and hospital factors associated with MAUD initiation during hospitalizations.

Methods: We conducted a retrospective cohort study of alcohol withdrawal hospitalizations in a large health system using electronic health record and pharmacy data from 2022 to 2024. The outcome was MAUD initiation defined as inpatient administration of injectable naltrexone or prescriptions for outpatient oral naltrexone, acamprosate, disulfiram, topiramate, or baclofen. We excluded patients who filled MAUD in the prior 90 days. We used a multivariable linear model with socioeconomic, clinical, and hospitalization factors as fixed effects and individual patients and hospital sites as random effects to account for clustering.

Results: Among 5993 alcohol withdrawal hospitalizations across 12 hospitals, 19.8% (range: 5.1%- 43.2%) initiated MAUD. Oral naltrexone was the most common MAUD initiated, while topiramate was the least common (66.9% and 1.2% of MAUD initiated, respectively). Patients less likely to initiate MAUD were age ≥ 65 years (absolute adjusted difference -6.89 percentage points [pp] vs. age 18-29; 95% confidence interval [CI] -12.20, -1.58) and those without cirrhosis and liver enzymes >200 U/L (-3.46 pp vs. liver enzymes ≤ 200 U/L; 95% CI: -6.87, -0.01). Four hospitals initiated MAUD at significantly lower rates, and two at higher rates including the only hospital with an addiction consult service.

Conclusion: MAUD initiation during alcohol withdrawal hospitalizations was low and varied by patient factors and hospital sites. Factors driving this variation can inform interventions to address MAUD underutilization.

Hyperleukocytosis is defined as a white blood cell count exceeding 100,000/µL and is a critical manifestation of acute and chronic leukemia. It can lead to life-threatening complications such as leukostasis, disseminated intravascular coagulation, and tumor lysis syndrome. Leukostasis results from microvascular obstruction by leukemic blasts, causing multisystemic manifestations. Leukostasis is primarily a clinical diagnosis requiring prompt recognition and intervention. Supportive care includes intravenous hydration, uric acid-lowering agents, and cautious blood transfusions to prevent worsening hyperviscosity. Hydroxyurea and cytarabine are used as cytoreductive agents until a definitive diagnosis and management are started. The role of leukapheresis is controversial, and practice patterns differ between clinicians and institutions.

Background: Delayed diagnosis of venous thromboembolism (VTE) is prevalent among hospitalized patients, yet case identification is challenging and feedback limited.

Objective: To develop a large language model (LLM)-based electronic-trigger to identify VTE diagnostic delays.

Methods: All admissions to internal medicine (IM) residents at NYU Langone Health between January 2022 and December 2023 (n = 20,843) were included. Using an open-source LLM, prompts were validated to detect (1) residents considering VTE in admission notes and (2) VTE confirmation in five types of imaging reports (n = 100 for each prompt validation set). The validated prompts were applied to determine discordance between admission note differential omitting VTE and imaging report confirming VTE. Two hospitalists reviewed discordant cases using a validated tool to identify diagnostic delays. Hospitalizations were labeled as diagnostic delays, in-hospital complication, or false-positive. Based on in-hospital complication and false-positive patterns, exclusion criteria were implemented. Positive predictive value (PPV) and negative predictive value (NPV) were calculated.

Results: The LLM prompts correctly classified admission notes and VTE imaging studies with high accuracy (range 98%-100%, n = 699 VTE cases identified). Of the 137 diagnostic delays the LLM-based electronic-trigger identified, 31 were true-positives, 60 in-hospital complications, and 46 false-positives. 4.4% of all VTE hospitalizations had a diagnostic delay. With the exclusion criteria, the PPV was 48% (95% confidence interval [CI], 35%-62%) and NPV was 95% (95% CI, 87%-98%).

Conclusions: We developed the first LLM-based electronic-trigger to identify VTE diagnostic delays, with higher performance than existing non-LLM electronic-triggers. LLM-based approaches can facilitate diagnostic performance feedback and are scalable to other conditions and institutions.

Logic models are an accessible tool to organize the planning, management, and evaluation of programs across healthcare disciplines. This methodological progress note describes the five standard components of a logic model (Inputs, Activities, Outputs, Outcomes, and Impact) and provides guidance for developing a logic model. Logic models offer a visual representation of the work and theory of a program and should be created with input from all interested parties. Logic models are living tools that need to be revised as programs evolve and new information becomes available.

Background: Hospital discharge summaries are critical for ensuring continuity of care, but their quality often varies. Large language models (LLMs) have the potential to standardize and enhance the efficiency of this documentation process.

Objectives: To evaluate the quality of hospital discharge summaries created by an LLM-based hospital course drafting tool created by Epic Systems compared with human-written summaries.

Methods: Retrospective study at a single tertiary-care institution in 2024. The cohort included 100 adult hospitalizations lasting >72 hours across medical and surgical dismissing services. No interventions were performed. Summaries (LLM-generated vs human-written) were independently reviewed using a standardized rubric covering nine domains (e.g., comprehensiveness, clarity, relevance). Scores were normalized and compared. Readability was assessed using Flesch Reading Ease.

Results: LLM-generated summaries outperformed human-written summaries across all criteria (p < .05), with the greatest difference observed in comprehensiveness (LLM median 0.62 vs. human -0.23). Human-written summaries from surgical services scored lower than those from medical services, but LLM performance was consistent across both. Human summaries had higher Flesch Reading Ease scores (33.11 vs. 26.2; p < .05), reflecting simpler language.

Conclusions: LLM-generated summaries demonstrated superior quality, consistency, and clinical utility compared with human-written summaries, highlighting their potential to improve documentation efficiency and standardization.