Indian Journal of Critical Care Medicine最新文献

Objective: Delayed recognition of treatment failure with heated humidified high-flow nasal cannula (HFNC) in pediatric acute respiratory distress can lead to adverse outcomes. This study presents and verifies a pediatric-modified [heart rate, acidosis, consciousness, oxygenation, respiratory rate (HACOR)] (p-HACOR) score as an instrument for the early prediction of HFNC therapy failure in children experiencing acute hypoxemic respiratory distress.

Design: Single-center prospective observational study.

Setting: Pediatric intensive care unit (PICU) of a tertiary care teaching hospital in Pune, India.

Patients: One hundred and forty-six children (aged 1 month-18 years) admitted to PICU with acute hypoxemic respiratory distress necessitating HFNC therapy.

Interventions/measurements: The p-HACOR score-a composite metric incorporating heart rate, acidity, consciousness, oxygenation, and the respiratory rate adjusted for age-specific criteria was assessed at designated intervals: Before initiation and at 1, 6, 12, 24, 36, and 48 hours following commencement of HFNC therapy.

Results: Treatment failure was observed in 23 patients (15.8%), with 83% necessitating intubation within 1-2 hours after therapy commencement. The p-HACOR score exhibited superior predictive ability at 1 hour postinitiation, with 91.3% sensitivity, 83.74% specificity, and 84.93% diagnostic accuracy [area under the curve (AUC) 0.951, p < 0.01]. A score threshold exceeding 10.5 at 1 hour proved to be the most dependable predictor of failure. The p-HACOR score showed a high correlation with recognized clinical metrics, exceeding the prognostic accuracy of individual markers.

Conclusions: The p-HACOR score may be useful as an early indicator of HFNC therapy failure. Its predictive ability was noted to be best at 1 hour postinitiation of HFNC, suggesting its potential for contributing to early decision-making regarding HFNC failure.

How to cite this article: Vora H, Sarangi BU, Walimbe A. Early Prediction of High-flow Nasal Cannula Failure Using Pediatric-modified HACOR Score in Children with Acute Hypoxemic Respiratory Distress: A Prospective Observational Study. Indian J Crit Care Med 2025;29(11):902-906.

Background and aims: This study compares traditional emergency department (ED) triage systems with artificial intelligence (AI)-assisted triage to assess their impact on time to treatment (TTT) and patient outcomes. Emergency departments (EDs) manage high patient volumes and time-critical decisions, necessitating efficient triage. Traditional methods such as the emergency severity index (ESI) and Manchester triage system (MTS) rely on human judgment and may introduce variability. Artificial intelligence (AI)-based systems use machine learning (ML) algorithms to analyze patient data in real time, offering the potential for faster and more consistent decisions.

Patients and methods: We conducted a single-center randomized controlled trial (RCT) in a high-volume tertiary hospital in Pune, India. One hundred and five patients were randomized to traditional triage (Group A) or AI-assisted triage (Group B). The primary outcome was TTT, defined as arrival at first medical intervention. Mean TTT was 31.02 minutes with AI vs 44.12 minutes with traditional triage (p < 0.001); variability was lower with AI (standard deviation 7.75 vs 11.69).

Results: Intensive care unit (ICU) admission rates did not differ. Clinician ratings favored AI in terms of accuracy, workload reduction, and perceived impact. Multiple linear regression estimated an adjusted -13.1-minute effect of AI on TTT, independent of severity (p < 0.001).

Conclusion: Artificial intelligence (AI)-assisted triage improves ED efficiency by reducing TTT without altering ICU admission rates.

How to cite this article: Nadaf H, Jabade MV, Jamadar K, Jogdeo B, Jamdade V. Improving Emergency Response: A Comparative Analysis of Traditional vs Artificial Intelligence-assisted Triage Systems in Health Care and Their Impact. Indian J Crit Care Med 2025;29(11):925-929.

Background and aims: Obtaining informed consent (IC) for tracheostomy is a frequent and essential process in the intensive care unit (ICU). With the increasing use of artificial intelligence (AI) in health care, chatbots such as ChatGPT and Google Gemini (GG) are being explored as potential tools to assist in drafting IC documents.

Methods: In this cross-sectional study, IC drafts for tracheostomy were generated by ChatGPT and GG. Fifteen experienced intensivists independently evaluated these drafts for accuracy, completeness, readability, and sentiment. Readability was measured using the Flesch Reading Ease (FRE) score, while sentiment analysis assessed the emotional tone of the text.

Results: No statistically significant differences were observed in terms of accuracy or completeness between the two chatbots. The inter-rater reliability was assessed using the intraclass correlation (ICC). The ICC for completeness and accuracy ratings between ChatGPT and GG were 0.85 (95% CI: 0.75-0.92) and 0.80 (95% CI: 0.68-0.89), respectively, suggesting excellent to good inter-rater reliability between the two Chatbots. However, ChatGPT drafts had higher FRE scores (76.46 vs 60.04), indicating better readability. Sentiment analysis revealed that both drafts were predominantly neutral, with GG incorporating slightly more positive expressions.

Conclusion: Both ChatGPT and GG can generate clinically appropriate IC content for tracheostomy. ChatGPT appears to have an advantage in producing more readable and patient-friendly material, highlighting its potential utility in clinical communication.

How to cite this article: Chilkoti GT, Jain S, Gondode PG. Comparison of Artificial Intelligence Chatbots (ChatGPT vs Google Gemini) for Informed Consent Quality: A Cross-sectional Evaluation. Indian J Crit Care Med 2025;29(11):967-969.

Background and aims: Acute kidney injury (AKI) occurs in 15-50% of trauma patients and worsens outcomes. Early identification of high-risk individuals may enable reno-protective strategies. Urinary tissue inhibitor of metalloproteinases-2 (TIMP-2) and insulin-like growth factor-binding protein 7 (IGFBP7) have shown promise in predicting AKI in various settings, but their utility in critically ill trauma patients remains unclear. This study aimed to evaluate the combined urinary biomarker product (TIMP-2) × (IGFBP7) for early AKI prediction in this group.

Patients and methods: This prospective observational study was conducted at a tertiary care center in India. Critically ill trauma patients aged 18-65 years admitted to the critical care unit were enrolled and followed for AKI development (KDIGO criteria). Injury Severity Score (ISS) and APACHE II scores were recorded at admission. Urine samples collected at admission and 24 hours post-admission were analyzed for TIMP-2 and IGFBP7 using ELISA-based kits. Associations between biomarker levels, AKI occurrence, and need for renal replacement therapy (RRT) were assessed, along with clinical and hemodynamic parameters.

Results: Seventy-nine patients were included; 14 (17.7%) developed AKI. Median APACHE II scores were 11 (7-18) vs 9 (7-11), and ISS scores 34 (25-34) vs 25 (25-34) in AKI and non-AKI groups, respectively. The ROC-AUC for (TIMP-2) × (IGFBP7) was 0.49 at admission and 0.57 at 24 hours. A 24-hour cut-off of 0.008 (ng/mL)²/1,000 yielded 85.7% sensitivity, 27.7% specificity, NPV 90%, and PPV 20.3%.

Conclusion: Urinary (TIMP-2) × (IGFBP7) measured at admission did not show any significant discriminating power. However Urinary (TIMP-2) × (IGFBP7) measured 24 hours after admission may help identify critically ill trauma patients at risk of AKI.

How to cite this article: Parthiban M, Cherian A, Kundra P, Priyamvada PS, Bobby Z, Senthilnathan M. Role of Urinary Biomarkers TIMP-2 and IGFBP7 in Predicting Acute Kidney Injury in Critically Ill Trauma Patients: A Prospective Observational Study. Indian J Crit Care Med 2025;29(11):936-941.

Background and aims: Chest X-ray (CXR) and computed tomography (CT) are established imaging modalities for patients in respiratory distress, and lung ultrasound (LUS) has emerged as an efficient point-of-care alternative. This study aimed to evaluate the diagnostic performance of LUS and CXR in critically ill patients, utilizing CT thorax as the reference standard.

Patient and methods: A prospective observational study was conducted in a tertiary care intensive care unit (ICU) involving mechanically ventilated adult patients requiring CT thorax. Before the CT, patients underwent portable CXR and LUS. Diagnostic performance metrics were calculated using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and AUC-ROC for pneumothorax, pleural effusion, consolidation, collapse, and pulmonary edema for CXR and LUS compared to CT. Inter-rater agreement was assessed using Cohen's κ.

Results: A total of 110 patients were included in this study. Overall, the performance of LUS was significantly superior to CXR for diagnosing various pathologies. Lung ultrasound exhibited 100% sensitivity and 97% specificity, outperforming CXR (66.7% sensitivity) for the detection of pneumothorax. For pleural effusion, LUS achieved 100% sensitivity and specificity, reflecting perfect concordance with CT, especially for moderate and severe pleural effusions, while CXR had an accuracy of only 68.18%. Lung consolidation and collapse were identified with 100% sensitivity and PPV by LUS. Furthermore, the accuracy of LUS in diagnosing pulmonary edema was 96.4%, with a specificity of 95.3% and an NPV of 100%.

Conclusion: Across five pathologies that were evaluated in this study, LUS consistently outperformed CXR in diagnostic accuracy and concordance with CT thorax.

How to cite this article: Al Tayar AS, Salem HA, Abdelshafey EE, Rashwan MA, Khalil MF, Alwafi HA, et al. Comparative Analysis of Lung Ultrasound, Computed Tomography, and X-ray in the Diagnosis of Common Pathologies among Critically Ill Mechanically Ventilated Patients-A Prospective Observational Study. Indian J Crit Care Med 2025;29(10):807-814.

How to cite this article: Delazari LEB, Ratti LSR, da Silva AC, Sibinelli M, Heidemann AM, Montedioca HLM, et al. Author Response: Due to Limitations in the Feasibility of the Perme Score it should not be Used to Classify Muscle Weakness in Intensive Care Unit Patients. Indian J Crit Care Med 2025;29(10):884-886.

How to cite this article: Vaithialingam B. Google's AI Search Engine Misinterprets the Glucose Conversion Factor for Lactate. Indian J Crit Care Med 2025;29(10):892-893.

Background and aims: Caregivers of intensive care unit (ICU) patients with advanced chronic illness face significant psychological stress and information gaps regarding palliative care. Generative artificial intelligence (AI) tools like ChatGPT and Google Gemini may offer scalable, personalized educational support. This study aimed to compare the quality of academic content generated by these AI chatbots in terms of readability, sentiment, understandability, actionability, and expert-rated accuracy and completeness.

Materials and methods: On December 10, 2024, ChatGPT and Google Gemini (free browser versions) were queried with a standardized prompt on ICU palliative care for caregivers. Outputs were evaluated for readability (eight validated indices), sentiment polarity (online tool), understandability and actionability, patient education materials assessment tool for printable materials (PEMAT-P), and expert panel (n = 7) assessments of accuracy and completeness using a 5-point Likert scale. Statistical comparisons were conducted using paired t-tests.

Results: ChatGPT and Gemini produced content of comparable readability (12.96 vs 13.21). Sentiment scores favored ChatGPT (+7.8 vs -39.1). Both chatbots achieved similar scores in understandability and actionability (91.7 and 80%, respectively). Accuracy scores were similar (p = 0.36), while completeness was significantly higher for ChatGPT (95% CI 0.08-1.07; p = 0.025).

Conclusion: Both AI tools generated content suitable for caregiver education in ICUs. ChatGPT demonstrated more neutral sentiment and greater completeness, supporting a potential role in critical care communication and education of caregivers in the ICU.

How to cite this article: Singh R, Gondode PG, Duggal S, Nayak SS. Comparative Evaluation of Artificial Intelligence Chatbots in Delivering Palliative Care Education to Intensive Care Unit Caregivers- A Cross-platform Analysis: A Brief Communication. Indian J Crit Care Med 2025;29(10):865-867.

How to cite this article: Hanumanthappa MK, Sharda SC. Comment on "A Prospective Study to Assess the Profile and Outcome of Acute Paraquat Poisoning in a Tertiary Care Hospital of West Bengal". Indian J Crit Care Med 2025;29(10):887.