Sanjiv Mehta

BACKGROUND: Existing PICU early warning systems lack sufficient accuracy and timeliness for effective preparation. Machine learning approaches may improve prediction of critical deterioration events (CDEs), but their operational utility relative to existing tools remains unclear. OBJECTIVES: To develop a machine learning model for early detection of CDEs and evaluate operational utility against existing tools using a novel alert burden analysis. DERIVATION COHORT: PICU admissions (ages 0–24 yr, stay ≥ 24 hr) at a quaternary children’s hospital from 2014 to 2020 ( n = 12,771 patients; 21,141 admissions). CDEs (6% of patients) included cardiopulmonary resuscitation, extracorporeal membrane oxygenation initiation, dilute epinephrine administration, or unplanned intubation. VALIDATION COHORT: Temporally distinct PICU admissions from 2021 to 2022 ( n = 5144 patients; 6929 admissions; 6% CDE rate). PREDICTION MODEL: An ensemble of extreme gradient-boosted models (PICU Warning INdex [P-WIN]) trained to predict CDEs at 1–12-hour horizons using 550 features derived from demographics, medications, laboratory results, and vital signs. RESULTS: P-WIN demonstrated excellent discrimination at 2-hour (area under the receiver operating characteristic curve [AUROC], 0.95 [95% CI, 0.94–0.96] and area under the precision-recall curve [AUPRC], 0.76 [95% CI, 0.72–0.80]) and 12-hour horizons (AUROC, 0.93 [95% CI, 0.92–0.94] and AUPRC, 0.68 [95% CI, 0.64–0.72]). To alert before 80% of events, P-WIN generated 0.20 alerts per patient-day at a median 10.17 hours before CDE. Compared with the existing rule-based PICU Warning Tool (alerting before 38% of events), P-WIN generated one-third the alert burden at equivalent sensitivity (0.03 vs. 0.10 alerts per patient-day). CONCLUSIONS: P-WIN accurately predicted PICU CDEs up to 12 hours in advance with low alert burden, providing a viable opportunity for shifting care from reactive rescue to proactive, resource-intensive preparation and prevention.

OBJECTIVES: To evaluate the feasibility of a large language model (LLM)-based chatbot for answering parental questions in the PICU and inform design of a randomized controlled trial (RCT). DESIGN: Prospective single-arm feasibility study conducted from August 2024 to December 2024. SETTING: Quaternary PICU. SUBJECTS: Fourteen parents of children admitted to the PICU. INTERVENTIONS: Parents engaged in 10-minute sessions with a HIPAA-compliant GPT-4o- (Generative Pretrained Transformer 4o, OpenAI, San Francisco, CA) based chatbot prompted with patient-specific electronic health record (EHR) data. MEASUREMENTS AND MAIN RESULTS: Feasibility was assessed through four criteria: parental engagement and satisfaction, provider perceptions, accuracy and safety, and recruitment. Of 16 eligible parents, 14 enrolled and completed all procedures (87.5% recruitment rate). Parents asked a median of six questions (range, 3–13) with 96% positive real-time satisfaction ratings. Post-interaction surveys demonstrated high perceived value (median, 5.0/6.0 across all domains; Net Promoter Score [NPS] +57). Of 1225 chatbot-generated sentences evaluated, 99.3% were accurate with all eight errors classified as minor (inter-rater reliability: Gwet’s AC2, a chance-corrected inter-rater agreement coefficient, = 0.98; 95% CI, 0.97–0.99). Healthcare providers rated response quality highly (median, 5.0/6.0), although physicians expressed greater comfort with bedside use of the tool than nurses (5.0 vs. 4.0; p = 0.004). Sample size calculations using NPS as the primary endpoint suggest enrolling 135 participants would provide adequate power for a future RCT. CONCLUSIONS: An EHR-informed LLM chatbot demonstrated high parental engagement and satisfaction, positive provider perception, and high accuracy and safety, supporting progression to a RCT.

Objective: Fluid overload (FO) after pediatric septic shock resuscitation increases the risk of secondary organ failure and long-term morbidity. We hypothesized that early hemodynamic point-of-care ultrasound (hPOCUS) use in pediatric septic shock would be associated with decreased FO. Methods: Retrospective, observational study between 2015 and 2018 in a large academic Pediatric Intensive Care Unit (PICU) of children < 18 years receiving ≥ 40 mL/kg of bolus fluids or vasoactive infusion for treatment of septic shock. %FO and severe FO (%FO ≥ 15%) at 72 h after shock onset were compared between children with hPOCUS exposure < 6 h after shock onset and those without. %FO was calculated: [(fluid in – out)/weight on PICU admission]*100. The association between hPOCUS < 6 h after septic shock onset and %FO and severe FO were evaluated. Results: Of 591 children included, 115 (19.4%) had hPOCUS within 6 h. Children with early hPOCUS vs. without had higher PIM-3; median 4.3, IQR 2.9–9.4 vs. 3.2, IQR 0.9–5.7, p = 0.0012. %FO was not significantly different in the early hPOCUS group, median difference -1.6%, 95% CI -3.2 to 0.03, p = 0.06. After controlling for confounders, %FO was significantly lower in the early hPOCUS group, ß-coefficient= -2.76, 95% CI -4.7 to -0.6, p = 0.012. Those with early hPOCUS had lower occurrence of severe FO 11.3% vs. 22.9%, p = 0.006; adjusted OR 0.41, 95% CI 0.22 to 0.76, p = 0.005. Conclusions: Early hPOCUS assessment during septic shock resuscitation was independently associated with decreased FO. Prospective research is needed to optimize hPOCUS use in pediatric septic shock management.

Background: Late recognition of in‐hospital deterioration is a source of preventable harm. Emergency transfers (ET), when hospitalized patients require intensive care unit (ICU) interventions within 1 h of ICU transfer, are a proximal measure of late recognition associated with increased mortality and length of stay (LOS).Objective: To apply diagnostic process improvement frameworks to identify missed opportunities for improvement in diagnosis (MOID) in ETs and evaluate their association with outcomes.Design, Settings, and Participants: A single‐center retrospective cohort study of ETs, January 2015 to June 2019. ET criteria include intubation, vasopressor initiation, or 60 mL/kg fluid resuscitation 1 h before to 1 h after ICU transfer. The primary exposure was the presence of MOID, determined using SaferDx. Cases were screened by an ICU and non‐ICU physician. Final determinations were made by an interdisciplinary group. Diagnostic process improvement opportunities were identified.Main Outcome and Measures: Primary outcomes were in‐hospital mortality and posttransfer LOS, analyzed by multivariable regression adjusting for age, service, deterioration category, and pretransfer LOS.Results: MOID was identified in 37 of 129 ETs (29%, 95% confidence interval [CI] 21%–37%). Cases with MOID differed in originating service, but not demographically. Recognizing the urgency of an identified condition was the most common diagnostic process opportunity. ET cases with MOID had higher odds of mortality (odds ratio 5.5; 95% CI 1.5−20.6; p = .01) and longer posttransfer LOS (rate ratio 1.7; 95% CI 1.1–2.6; p = .02).Conclusion: MOID are common in ETs and are associated with increased mortality risk and posttransfer LOS. Diagnostic improvement strategies should be leveraged to support earlier recognition of clinical deterioration.

Emergency transfers (ETs), deterioration events with late recognition requiring ICU interventions within 1 hour of transfer, are associated with adverse outcomes. We leveraged electronic health record (EHR) data to assess the association between ETs and outcomes. We also evaluated the association between intervention timing (urgency) and outcomes. We conducted a propensity-score-matched study of hospitalized children requiring ICU transfer between 2015 and 2019 at a single institution. The primary exposure was ET, automatically classified using Epic Clarity Data stored in our enterprise data warehouse endotracheal tube in lines/drains/airway flowsheet, vasopressor in medication administration record, and/or ≥60 ml/kg intravenous fluids in intake/output flowsheets recorded within 1 hour of transfer. Urgent intervention was defined as interventions within 12 hours of transfer. Of 2037 index transfers, 129 (6.3%) met ET criteria. In the propensity-score-matched cohort (127 ET, 374 matched controls), ET was associated with higher in-hospital mortality (13% vs 6.1%; odds ratio, 2.47; 95% confidence interval [95% CI], 1.24–4.9, P = .01), longer ICU length of stay (subdistribution hazard ratio of ICU discharge 0.74; 95% CI, 0.61–0.91, P < .01), and longer posttransfer length of stay (SHR of hospital discharge 0.71; 95% CI, 0.56–0.90, P < .01). Increased intervention urgency was associated with increased mortality risk: 4.1% no intervention, 6.4% urgent intervention, and 10% emergent intervention. An EHR measure of deterioration with late recognition is associated with increased mortality and length of stay. Mortality risk increased with intervention urgency. Leveraging EHR automation facilitates generalizability, multicenter collaboratives, and metric consistency.

Introduction: Inpatient rounding serves numerous roles. The American Academy of Pediatrics recommends a family-centered bedside model. Residents identified physical examination teaching during and satisfaction with rounds as areas for improvement. A resident group developed a project utilizing quality improvement (QI) methodology to address these concerns. We aimed to increase the frequency of bedside physical examination teaching most or every day on a single inpatient unit by 20% over 1 year, with secondary goals to increase the percentage of interns spending one hour or more at bedside per day by 10% and intern satisfaction by 15%, without impacting rounding duration. Methods: We developed an organizational structure to complete a long-term resident-led project. Interventions included daily bedside examination teaching on rounds, afternoon examinations, goal communication, topic recording, and a teaching “tip sheet.” Using an institutional QI framework, we utilized iterative plan-do-study-act cycles to implement interventions and surveys to assess outcomes, with rounding efficiency as a balancing measure. Results: The survey response rate was 57%. Bedside teaching frequency increased from a mean of 10% to 61%, perceived time at the bedside increased from 37% to 59%, and rounding satisfaction improved from a rating of 6.7/10 to 7.4/10. Efficiency was not impacted. Conclusions: We improved inpatient rounds bedside physical examination teaching and satisfaction without sacrificing efficiency. This project demonstrates the feasibility and success of a resident-driven education initiative to successfully motivate fellow residents and colleagues across disciplines to enact change. The organizational structure may serve as a model for resident-led QI projects across institutions.