Daily Anesthesiology Research Analysis
Three impactful perioperative studies stand out today: a nationwide cohort mediation analysis links pre-operative anemia to cardiac surgery mortality largely via red blood cell transfusion; a randomized trial shows opioid-free anesthesia reduces early adverse events after thoracoscopic lung surgery; and a digital health study demonstrates large language models can detect clinically significant postoperative delirium earlier and more consistently than clinicians.
Summary
Three impactful perioperative studies stand out today: a nationwide cohort mediation analysis links pre-operative anemia to cardiac surgery mortality largely via red blood cell transfusion; a randomized trial shows opioid-free anesthesia reduces early adverse events after thoracoscopic lung surgery; and a digital health study demonstrates large language models can detect clinically significant postoperative delirium earlier and more consistently than clinicians.
Research Themes
- Perioperative risk stratification and blood management
- Opioid-sparing anesthesia strategies
- AI-assisted detection of postoperative delirium
Selected Articles
1. Efficacy of large language models in detecting postoperative delirium from unstructured clinical notes: A retrospective cohort study.
In a retrospective comparison using unstructured notes, Llama-3-70B and GPT-4o identified clinically significant postoperative delirium earlier than physicians with higher sensitivity and near-perfect inter-rater agreement. Despite lower specificity, LLMs shortened time to detection by roughly one day, supporting their use as supervised screening aids.
Impact: Introduces a scalable AI approach that could standardize and accelerate postoperative delirium detection, addressing a persistent perioperative diagnostic gap.
Clinical Implications: Deploying LLM-based screening on perioperative notes could trigger earlier delirium preventive measures, neuropsychiatric consultations, and medication review, potentially reducing complications and length of stay.
Key Findings
- LLMs achieved c-statistics of 0.74 (Llama-3-70B) and 0.76 (GPT-4o) for detecting clinically significant POD.
- Sensitivity was higher for LLMs (0.900 and 0.868) than for physicians (0.723), with lower specificity (0.463 and 0.547 vs 0.814).
- Inter-rater agreement was near-perfect for LLMs (Fleiss’ kappa ≈ 0.85) but only fair for physicians (0.219).
- LLMs detected POD about one day earlier than physicians (median 34.5–37.5 h vs 62.9 h; log-rank P < 0.001).
Methodological Strengths
- Direct head-to-head comparison with clinicians using real-world unstructured notes.
- Robust agreement metrics (Fleiss’ kappa) and time-to-event analysis (Kaplan–Meier, log-rank test).
Limitations
- Retrospective single-system design may limit generalizability and introduces documentation bias.
- Lower specificity raises risk of false positives and alert fatigue without clinical oversight.
Future Directions: Prospective, multi-center trials integrating LLM alerts into perioperative workflows with protocolized responses to assess impact on delirium incidence, complications, and LOS.
Early identification of postoperative delirium (POD) remains challenging. This retrospective observational study compared the performance of large language models (LLMs), Llama-3-70B and GPT-4o, and physicians in predicting clinically significant POD, defined as either requiring antipsychotics or diagnosis of delirium by neurologists following consultation for delirium-related symptoms. The c-statistics of Llama-3-70B and GPT-4o were 0.74 and 0.76, respectively. LLMs showed higher sensitivity (Llama-3-70B, 0.900; GPT-4o, 0.868; physicians, 0.723) and lower specificity (0.463, 0.547, and 0.814, respectively) than physicians. Inter-rater agreement was almost perfect for both Llama-3-70B and GPT-4o (Fleiss' kappa = 0.852 and 0.854, respectively) but fair for physicians (0.219). Both LLMs detected clinically significant POD approximately one day earlier than physicians (Kaplan-Meier analysis, median time to diagnosis: Llama-3-70B, 34.5 h; GPT-4o, 37.5 h; physicians, 62.9 h; log-rank P < 0.001). The integration of LLMs as a complementary screening tool under physician supervision may improve the early, reproducible diagnosis of clinically significant POD.
2. Pre-operative anaemia, red blood cell transfusion and mortality after cardiac surgery: a Netherlands Heart Registration mediation analysis.
In over 71,000 cardiac surgeries, pre-operative anemia was independently associated with higher 120-day mortality, and mediation analysis attributed approximately 59% of this association to in-hospital red blood cell transfusion, rising to 77% in patients ≥70 years. Findings underscore the central role of transfusion in the anemia–mortality pathway and support aggressive patient blood management.
Impact: Quantifies, at scale, how much transfusion mediates the anemia–mortality link after cardiac surgery, providing actionable targets for perioperative optimization and transfusion stewardship.
Clinical Implications: Prioritize pre-operative anemia diagnosis and treatment (iron therapy, erythropoiesis-stimulating approaches where appropriate) and implement restrictive, individualized transfusion strategies—especially in older patients—to mitigate mortality risk.
Key Findings
- Among 71,053 patients, 20.3% had pre-operative anemia; 52.7% of anemic vs 17.5% of non-anemic patients received RBC transfusions (p < 0.001).
- Pre-operative anemia independently associated with 120-day mortality (adjusted OR 1.66, 95% CI 1.47–1.87).
- Mediation analysis: RBC transfusion accounted for 58.9% (95% CI 41.3–76.5%) of the anemia–mortality association; proportion mediated was higher in patients ≥70 years (77.3%).
Methodological Strengths
- Nationwide registry with very large sample size and multivariable adjustment.
- Formal mediation analysis quantifying the indirect effect through transfusion with age-stratified estimates.
Limitations
- Observational design cannot eliminate residual confounding or establish causality.
- Transfusion practices and thresholds may vary across centers and over time.
Future Directions: Prospective trials of pre-operative anemia correction and transfusion-threshold strategies, with subgroup analyses in older adults, to test causality and optimize patient blood management.
INTRODUCTION: Pre-operative anaemia is an established risk factor for mortality after cardiac surgery. The extent to which this risk is mediated by complications related to red blood cell transfusion remains uncertain, particularly across different age groups. METHODS: This nationwide cohort study included adult cardiac surgery patients from the Netherlands Heart Registration. Pre-operative anaemia was defined according to World Health Organization criteria and red blood cell transfusion as any red blood cell transfusion during hospital stay. The main study endpoint was 120-day mortality. We used multivariable logistic regression to quantify the associations between pre-operative anaemia, red blood cell transfusion and 120-day mortality. Subsequently, a mediation analysis was conducted to quantify how much of the effect of pre-operative anaemia on postoperative mortality is because of red blood cell transfusion. RESULTS: Of 71,053 patients, 14,452 (20.3%) had pre-operative anaemia. Of these, 7621 (52.7%) received red blood cell transfusion during hospital stay, compared with 9930 (17.5%) of patients without anaemia (p < 0.001). Observed 120-day mortality was 612 (4.2%) and 901 (1.6%), respectively. In multivariable regression analysis, pre-operative anaemia remained independently associated with postoperative mortality (adjusted odds ratio 1.66, 95%CI 1.47-1.87), with consistent effects across age groups. Mediation analysis showed that red blood cell transfusion accounted for 58.9% (95%CI 41.3-76.5%) of the association between pre-operative anaemia and 120-day mortality. The proportion mediated was larger in patients aged ≥ 70 years (77.3%, 95%CI 43.1-100%) compared with younger patients (39.3%, 95%CI 21.4-57.2%).
3. Efficacy of opioid-free anesthesia in reducing postoperative adverse events for thoracoscopic pulmonary surgery: a randomized controlled trial.
In a double-blind RCT of thoracoscopic lung surgery, opioid-free anesthesia with paravertebral block and dexmedetomidine reduced early hypotension, overall adverse reactions, PONV at 24 hours, vasopressor use, and opioid consumption, while accelerating first feeding. Single-center design and modest sample size temper generalizability.
Impact: Provides randomized evidence supporting an opioid-sparing anesthetic strategy that improves early postoperative recovery metrics in thoracic surgery.
Clinical Implications: Consider OFA protocols (regional anesthesia plus alpha-2 agonist) to reduce early adverse events and opioid exposure after thoracoscopic pulmonary procedures, with monitoring for hemodynamics and individualized analgesia.
Key Findings
- At 2 h postop, hypotension and overall adverse reactions were lower with OFA (8.2% vs 33.3%, P=0.002; 18.4% vs 39.6%, P=0.018).
- At 24 h, PONV was 0% with OFA vs 10.4% in controls (P=0.027); dizziness 8.2% vs 22.9% (P=0.041); overall adverse reactions 10.2% vs 27.1% (P=0.029).
- OFA reduced vasopressor use (P<0.001), 24-hour opioid consumption (P=0.033), analgesic pump presses (P=0.004), and hastened first feeding (9.5±3.8 h vs 11.4±4.7 h, P=0.035).
Methodological Strengths
- Prospective randomized double-blind design with preregistration (NCT04507165).
- Standardized regional anesthesia (paravertebral block) across groups.
Limitations
- Single-center, relatively small sample size limits external validity.
- Primary outcomes focused on early adverse reactions; longer-term pain and functional outcomes were not reported.
Future Directions: Multi-center trials powered for clinical endpoints (e.g., quality of recovery, length of stay, persistent postoperative pain) and safety across diverse thoracic populations.
BACKGROUND: This study compared the effects of opiate-free anesthesia (OFA) using thoracic paravertebral block combined with dexmedetomidine versus traditional anesthesia with sufentanil and remifentanil on postoperative adverse reactions after thoracoscopic lung surgery. METHODS: A prospective, randomized, double-blind trial included 97 patients(49 in OFA group,48 in S group).All received preoperative thoracic paravertebral block.The primary outcome was the incidence of opioid-related adverse reactions (hypoxemia, hypotension, nausea, vomiting, dizziness) at 2,4,6,24,and 48 h postoperatively.Secondary outcomes included intraoperative events, pain scores, opioid consumption, time to first flatus/defecation, and cardiovascular complications. RESULTS: At 2 h postoperatively, hypotension incidence was lower in the OFA group(8.2%vs 33.3%,P = 0.002),with fewer overall adverse reactions (18.4%vs 39.6%,P = 0.018).At 24 h, PONV was 0%in OFA vs. 10.4%in S group(P = 0.027),dizziness was 8.2%vs 22.9%(P = 0.041),and overall adverse reactions were 10.2%vs 27.1%(P = 0.029).OFA had fewer vasopressor applications (P < 0.001), lower 24-hour opioid use (P = 0.033), fewer analgesic pump presses(P = 0.004),and earlier first feeding (9.5 ± 3.8 h vs. 11.4 ± 4.7 h, P = 0.035). CONCLUSION: OFA significantly reduced opioid-related adverse reactions, decreased vasopressor and opioid use, and improved recovery.However, the single-center, small-sample design limits generalizability, and multicenter validation is needed. TRIAL REGISTRATION: This randomized controlled trial was registered on clinicaltrials.gov (ID: NCT04507165; URL: https://clinicaltrials.gov/study/NCT04507165 , 11/08/2020).