Detection of Gastrointestinal Bleeding With Large Language Models to Aid Quality Improvement and Appropriate Reimbursement.

BACKGROUND & AIMS: Early identification and accurate characterization of overt gastrointestinal bleeding (GIB) enables opportunities to optimize patient management and ensures appropriately risk-adjusted coding for claims-based quality measures and reimbursement. Recent advancements in generative artificial intelligence, particularly large language models (LLMs), create opportunities to support accurate identification of clinical conditions. In this study, we present the first LLM-based pipeline for identification of overt GIB in the electronic health record (EHR). We demonstrate 2 clinically relevant applications: the automated detection of recurrent bleeding and appropriate reimbursement coding for patients with GIB. METHODS: Development of the LLM-based pipeline was performed on 17,712 nursing notes from 1108 patients who were hospitalized with acute GIB and underwent endoscopy in the hospital from 2014 to 2023. The pipeline was used to train an EHR-based machine learning model for detection of recurrent bleeding on 546 patients presenting to 2 hospitals and externally validated on 562 patients presenting to 4 different hospitals. The pipeline was used to develop an algorithm for appropriate reimbursement coding on 7956 patients who underwent endoscopy in the hospital from 2019 to 2023. RESULTS: The LLM-based pipeline accurately detected melena (positive predictive value, 0.972; sensitivity, 0.900), hematochezia (positive predictive value, 0.900; sensitivity, 0.908), and hematemesis (positive predictive value, 0.859; sensitivity, 0.932). The EHR-based machine learning model identified recurrent bleeding with area under the curve of 0.986, sensitivity of 98.4%, and specificity of 97.5%. The reimbursement coding algorithm resulted in an average per-patient reimbursement increase of $1299 to $3247 with a total difference of $697,460 to $1,743,649. CONCLUSIONS: An LLM-based pipeline can robustly detect overt GIB in the EHR with clinically relevant applications in detection of recurrent bleeding and appropriate reimbursement coding.

Continued midazolam versus diphenhydramine in difficult-to-sedate patients: a randomized double-blind trial.

BACKGROUND AND AIMS: Current guidelines recommend diphenhydramine in patients undergoing endoscopy who are not adequately sedated with a benzodiazepine and opioid combination. Because this practice has not been adequately assessed, we performed a randomized, double-blind trial comparing diphenhydramine with continued midazolam in such patients. METHODS: Patients undergoing elective colonoscopy with moderate sedation were eligible. Sedation was measured with the Modified Observer's Assessment of Alertness/Sedation (MOAA/S) score with adequate sedation defined as 3 on a 0- to 5-point scale. Patients not adequately sedated with midazolam 5 mg and fentanyl 100 µg were randomly assigned to diphenhydramine 25 mg versus continued midazolam 1.5 mg. Adequacy of sedation was assessed 3 minutes after each study medication dose. If MOAA/S was 4 to 5, study medication was repeated, to a maximum of 3 doses. The primary endpoint was adequate sedation. RESULTS: The planned enrollment of 200 patients (100 in each study group) was attained. Adequate sedation was achieved less often with diphenhydramine than midazolam (27% vs 65%, difference = -38%; 95% CI, -50% to -24%; P < .0001). After study medications were completed, more patients required additional medication for sedation or analgesia with diphenhydramine versus midazolam (84% vs 68%, P = .008), whereas the time to discharge from the recovery unit was similar (134 vs 129 minutes). Treatment effect was consistent across subgroups including age ≤55, substance abuse, benzodiazepine use, opioid use, and psychiatric medication use. CONCLUSIONS: Endoscopists performing moderate sedation should continue midazolam rather than switching to diphenhydramine in patients who do not achieve adequate sedation with usual doses of midazolam and an opioid. (Clinical trial registration number: NCT01769586.).

Intermittent vs continuous proton pump inhibitor therapy for high-risk bleeding ulcers: a systematic review and meta-analysis.

IMPORTANCE: Current guidelines recommend an intravenous bolus dose of a proton pump inhibitor (PPI) followed by continuous PPI infusion after endoscopic therapy in patients with high-risk bleeding ulcers. Substitution of intermittent PPI therapy, if similarly effective as bolus plus continuous-infusion PPI therapy, would decrease the PPI dose, costs, and resource use. OBJECTIVE: To compare intermittent PPI therapy with the currently recommended bolus plus continuous-infusion PPI regimen for reduction of ulcer rebleeding. DATA SOURCES: Searches included MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials databases through December 2013; US and European gastroenterology meeting abstracts from 2009 to 2013; and bibliographies of systematic reviews. STUDY SELECTION: Randomized trials of patients with endoscopically treated high-risk bleeding ulcers (active bleeding, nonbleeding visible vessels, and adherent clots) comparing intermittent doses of PPIs and the currently recommended regimen (80-mg intravenous bolus dose of a PPI followed by an infusion of 8 mg/h for 72 hours). DATA EXTRACTION AND SYNTHESIS: Duplicate independent data extraction and risk-of-bias assessment were performed. Data were pooled using a fixed-effects model or a random effects model if statistical heterogeneity was present. MAIN OUTCOMES AND MEASURES: The primary outcome was rebleeding within 7 days; additional predefined outcomes included rebleeding within 3 and 30 days, need for urgent intervention, mortality, red blood cell transfusion, and length of hospital stay. The primary hypothesis, defined before initiation of the literature review, was that intermittent use of PPIs was noninferior to bolus plus continuous infusion of PPIs, with the noninferiority margin predefined as an absolute risk difference of 3%. RESULTS: The risk ratio of rebleeding within 7 days for intermittent vs bolus plus continuous infusion of PPIs was 0.72 (upper boundary of 1-sided 95% CI, 0.97) and the absolute risk difference was -2.64% (upper boundary of 1-sided 95% CI, -0.28%, which is well below the predefined noninferiority margin of 3%). Risk ratios for rebleeding within 30 days and 3 days, mortality, and urgent interventions were less than 1 and mean differences for blood transfusion and hospital length of stay were less than 0, indicating that no summary estimate showed an increased risk with intermittent therapy. The upper boundaries of 95% CIs for absolute risk differences were less than 1.50% for all predefined rebleeding outcomes. CONCLUSIONS AND RELEVANCE: Intermittent PPI therapy is comparable to the current guideline-recommended regimen of intravenous bolus plus a continuous infusion of PPIs in patients with endoscopically treated high-risk bleeding ulcers. Guidelines should be revised to recommend intermittent PPI therapy.