Daily Anesthesiology Research Analysis
Three studies stand out in anesthesiology and critical care. A randomized trial showed lung ultrasound–guided alveolar recruitment before one-lung ventilation markedly reduced intraoperative hypoxemia. A prospective, real-time multicenter study found machine learning matched clinician performance in predicting trauma patients needing hemorrhage control resuscitation, with combined use improving sensitivity. A large pediatric cohort revealed social determinants drive disparities in ECMO use and o
Summary
Three studies stand out in anesthesiology and critical care. A randomized trial showed lung ultrasound–guided alveolar recruitment before one-lung ventilation markedly reduced intraoperative hypoxemia. A prospective, real-time multicenter study found machine learning matched clinician performance in predicting trauma patients needing hemorrhage control resuscitation, with combined use improving sensitivity. A large pediatric cohort revealed social determinants drive disparities in ECMO use and outcomes.
Research Themes
- Ultrasound-guided intraoperative ventilation optimization
- Real-time machine learning decision support in trauma resuscitation
- Health equity and social determinants in advanced ICU therapies
Selected Articles
1. Effectiveness of ultrasound-guided alveolar recruitment in thoracic surgery with one-lung ventilation: a randomized-controlled trial.
In thoracic surgery with one-lung ventilation, preemptive lung ultrasound–guided alveolar recruitment reduced intraoperative hypoxemia from 14.3% to 1.2% compared with conventional recruitment. LUS guidance also led to less atelectasis before emergence, without increasing postoperative complications.
Impact: This RCT provides actionable evidence that LUS-guided recruitment improves intraoperative oxygenation and reduces atelectasis, offering a practical, bedside strategy to optimize ventilation in thoracic anesthesia.
Clinical Implications: Incorporating LUS-guided recruitment before one-lung ventilation can lower hypoxemia risk and atelectasis without added harm. Teams should standardize LUS protocols and training to implement this approach.
Key Findings
- Intraoperative hypoxemia (SpO2 < 95%) decreased from 14.3% (control) to 1.2% (LUS-guided); RR 0.09 (95% CI 0.01–0.64), P=0.002.
- Severe hypoxemia (SpO2 < 90%) did not differ significantly (1.2% vs 6.0%).
- Pre-emergence atelectasis scores were lower with LUS-guided strategy; postoperative adverse outcomes were similar between groups.
Methodological Strengths
- Randomized controlled design with predefined primary outcome (hypoxemia incidence).
- Objective intraoperative assessments including serial LUS scoring, arterial blood gases, and respiratory parameters.
Limitations
- Single-center study; generalizability may be limited.
- No demonstrated reduction in severe hypoxemia; no long-term clinical outcomes reported.
Future Directions: Multicenter trials to confirm external validity, define standardized LUS-guided ARS protocols, and evaluate impacts on postoperative pulmonary complications and recovery.
OBJECTIVES: Although alveolar recruitment strategy (ARS) before one-lung ventilation (OLV) is beneficial in intraoperative oxygenation, the optimal protocol remains unestablished. As lung ultrasound (LUS) has been used recently, we designed a randomized controlled trial to compare preemptive LUS-guided ARS with conventional ARS in thoracic surgery. METHODS: Patients aged 20-80 years scheduled to undergo lung resection surgery with OLV were randomized into 2 groups: (i) control group receiving conventional ARS and (ii) LUS group receiving LUS-guided ARS. ARS and modified LUS scoring were performed 5 min after intubation and before emergence. Arterial blood samples and respiratory parameters were collected every 30 min during OLV. The primary outcome was the incidence of intraoperative hypoxaemia (SpO2 < 95%). RESULTS: In total, 166 patients were included. The incidence of intraoperative hypoxaemia was 1.2% in the LUS group and 14.3% in the control group [risk ratio (95% CI) 0.09 (0.01-0.64), P = 0.002]. However, the incidence of intraoperative severe hypoxaemia (SpO2 < 90%) was not significantly different [1.2% vs 6.0%, risk ratio (95% CI) 0.20 (00.02-1.72), P = 0.213]. In the LUS before emergence, higher atelectasis score (P = 0.005) and more significant atelectasis (P = 0.031) was observed in the control group. Postoperative adverse outcomes were comparable between both groups. CONCLUSIONS: LUS-guided ARS before OLV was more effective than conventional ARS in preventing intraoperative hypoxaemia during thoracic surgery. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov (NCT03770793, 10 December 2018).
2. Comparison of machine learning and human prediction to identify trauma patients in need of hemorrhage control resuscitation (ShockMatrix study): a prospective observational study.
In eight trauma centers, real-time clinician predictions and a machine learning model showed comparable accuracy for identifying patients needing hemorrhage control resuscitation, with combined use achieving sensitivity 83% and specificity 73%. Agreement between human and model was moderate (κ=0.51), supporting ML as a complementary decision aid.
Impact: This study provides prospective, real-world validation of ML for critical trauma triage and shows additive value when combined with clinician judgment, a key step toward safe clinical integration.
Clinical Implications: Implement ML-assisted prealert triage to augment clinician assessment for early activation of hemorrhage control pathways; integrate into smartphone workflows with training and governance to monitor impact.
Key Findings
- Among 1292 trauma patients, 13% required hemorrhage control resuscitation (HCR).
- Human prediction: PLR 3.74 (95% CI 3.20–4.36), NLR 0.36 (0.29–0.46); ML prediction: PLR 4.01 (3.43–4.70), NLR 0.35 (0.38–0.44).
- Combined human+ML approach achieved sensitivity 83% (77–88%) and specificity 73% (70–75%); Cohen’s kappa for agreement was 0.51.
Methodological Strengths
- Prospective, multicenter, real-time design with standardized predictor inputs and prespecified performance metrics.
- Use of net clinical benefit and likelihood ratios, and direct comparison with clinician predictions via a unified workflow.
Limitations
- Observational design without randomized implementation; no direct assessment of patient outcomes after using predictions.
- External generalizability beyond participating European centers and predictor set may be limited.
Future Directions: Cluster trials to test ML-assisted triage on clinical outcomes and workflow efficiency; external validation across regions; calibration drift monitoring and human–AI teaming strategies.
BACKGROUND: Machine learning could improve the timely identification of trauma patients in need of hemorrhage control resuscitation (HCR), but the real-life performance remains unknown. The ShockMatrix study aimed to compare the predictive performance of a machine learning algorithm with that of clinicians in identifying the need for HCR. METHODS: Prospective, observational study in eight level-1 trauma centers. Upon receiving a prealert call, trauma clinicians in the resuscitation room entered nine predictor variables into a dedicated smartphone app and provided a subjective prediction of the need for HCR. These predictors matched those used in the machine learning model. The primary outcome, need for HCR, was defined as: transfusion in the resuscitation room, transfusion of more than four red blood cell units in 6 h of admission, any hemorrhage control procedure within 6 h, or death from hemorrhage within 24 h. The human and machine learning performances were assessed by sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and net clinical benefit. Human and machine learning agreement was assessed with Cohen's kappa coefficient. FINDINGS: Between August 2022 and June 2024, out of 5550 potential eligible patients, 1292 were ultimately included in the analyses. The need for HCR occurred in 170/1292 patients (13%). The results showed a positive likelihood ratio of 3.74 (95% confidence interval [CI]: 3.20-4.36) and a negative likelihood ratio of 0.36 (95% CI: 0.29-0.46) for the human prediction and a positive likelihood ratio of 4.01 (95% CI: 3.43-4.70) and negative likelihood ratio of 0.35 (95% CI: 0.38-0.44) for the machine learning prediction. The combined use of human and machine learning prediction yielded a sensitivity of 83% (95% CI: 77-88%) and a specificity of 73% (95% CI: 70-75%). The Cohen's kappa coefficient showed an agreement of 0.51 (95% CI: 0.48-0.55). INTERPRETATION: The prospective ShockMatrix temporal validation study suggests a comparable human and machine learning performance to predict the need for HCR using real-life and real-time information with a moderate level of agreement between the two. Machine learning enhanced decision awareness could potentially improve the detection of patients in need of HCR if used by clinicians. FUNDING: The study received no funding.
3. Paediatric extracorporeal membrane oxygenation use by social determinants: a multicentre retrospective cohort study.
In a 47-hospital US cohort of 309,937 high-risk PICU admissions, children from lower-opportunity neighborhoods, minoritized racial/ethnic groups, and those with public insurance had higher adjusted risk of dying without ECMO compared with receiving ECMO. Regional differences were evident, underscoring inequities in access to advanced ICU therapies.
Impact: This large multicenter study quantifies inequities in access to ECMO by social determinants, providing targets for policy, referral, and system-level interventions to improve equity in life-saving critical care.
Clinical Implications: Hospitals and systems should apply equity frameworks to ECMO referral and activation, monitor COI-informed quality metrics, and address insurance and regional barriers to ensure equitable access.
Key Findings
- Among 309,937 high-risk PICU patients, 2.8% received ECMO, 4.0% died without ECMO, and 93.2% survived without ECMO.
- Each 10-point decrease in Child Opportunity Index was associated with a 5% higher adjusted relative risk of dying without ECMO (aRRR 1.05, 95% CI 1.01–1.09).
- Higher risk of dying without ECMO was observed in Asian and ‘other’ races, Hispanic ethnicity, and public insurance; regional differences also existed.
Methodological Strengths
- Very large multicenter dataset with standardized administrative and clinical variables across 47 hospitals.
- Multivariable multinomial regression isolating associations across three clinically relevant outcome categories.
Limitations
- Retrospective design with potential residual confounding and unmeasured clinical eligibility nuances.
- Findings reflect US pediatric centers and may not generalize internationally.
Future Directions: Prospective equity-focused interventions (standardized referral criteria, regional networks), evaluation of outcomes after policy changes, and qualitative studies to identify barriers at family and system levels.
BACKGROUND: Social determinants of health have upstream effects on health-care access and decision making to influence outcomes. We aimed to study the use of extracorporeal membrane oxygenation (ECMO) in children according to social determinants of health. METHODS: This retrospective, multicentre cohort study used data from 47 children's hospitals in the USA that contributed to the Pediatric Health Information System. Children (aged <18 years) admitted to an intensive care unit in one of the study hospitals between Oct 1, 2015, and March 31, 2021, with extreme or major mortality risk and cardiac or respiratory diagnoses, were eligible for the study. Social determinants of health considered were Child Opportunity Index (COI; a multidimensional metric of neighbourhood conditions), race, ethnicity, type of health insurance, distance from home to hospital, and hospital region. We calculated relative risk ratios (RRR) using multivariable multinomial regression models to compare the outcome of ECMO use according to three categories: patients who received ECMO, patients who survived without ECMO, and patients who died without ECMO (ie, those who might have benefited from ECMO). FINDINGS: Of 829 445 children admitted to paediatric intensive care units during the study period, 309 937 (37·4%) met the inclusion criteria and were included in the study. 288 717 (93·2%) of 309 937 patients survived without ECMO, 12 542 (4·0%) died without ECMO, and 8678 (2·8%) received ECMO. Patients who received ECMO were younger and more likely to have a cardiac diagnosis than those who died without ECMO. A 5% greater adjusted risk of dying without ECMO (adjusted RRR [aRRR] 1·05 [95% CI 1·01-1·09]) was seen for every 10-point decrease in COI score. A greater risk of dying without ECMO than of receiving ECMO was observed in patients of Asian (aRRR 1·36 [95% CI 1·04-1·78]) or other (1·54 [1·09-2·18]) race, Hispanic ethnicity (1·70 [1·31-2·22]), and with public health insurance (1·33 [1·16-1·52]). The risk of dying without ECMO differed by distance from hospital (aRRR per 50 miles increase 0·98 [95% CI 0·96-0·99]), whereas patients in hospitals in the south (2·34 [1·02-5·38]) and west (3·74 [1·44-9·67]) had a greater risk of dying without ECMO than those in the midwest; only those in the west also had a greater risk of survival without ECMO (3·72 [1·40-9·90]). INTERPRETATION: There are disparities in ECMO use according to social determinants of health, with lower use among children from under-resourced neighbourhoods, from minoritised racial and ethnic backgrounds, and those with public health insurance. Interventions to promote equitable ECMO use can be derived using health equity frameworks. FUNDING: None.