Background: Data on severe and extensive burns in China are limited, as is data on the prevalence of a range of related gastrointestinal (GI) disorders [such as stress ulcers, delayed defecation, opioid-related bowel immotility, and abdominal compartment syndrome (ACS)]. We present a multicentre analysis of coincident GI dysfunction and its effect on burn-related mortality. Methods: This retrospective analysis was conducted on patients with severe [≥ 20% total burn surface area (TBSA)] and extensive (> 50% TBSA or > 25% full-thickness TBSA) burns admitted to three university teaching institutions in China between January 1, 2011 and December 31, 2020. Both 30- and 90-day mortality were assessed by collating demographic data, burn causes, admission TBSA, % full-thickness TBSA, Baux score, Abbreviated Burn Severity Index (ABSI) score, and Sequential Organ Failure Assessment (SOFA) score, shock at admission and the presence of an inhalation injury. GI dysfunction included abdominal distension, nausea/vomiting, diarrhoea/constipation, GI ulcer/haemorrhage, paralytic ileus, feeding intolerance and ACS. Surgeries, length of intensive care unit (ICU) stay, pain control [in morphine milligram equivalents (MME)] and overall length of hospital stay (LOHS) were recorded. Results: We analyzed 328 patients [75.6% male, mean age: (41.6 ± 13.6) years] with a median TBSA of 62.0% (41.0–80.0%); 256 (78.0%) patients presented with extensive burns. The 90-day mortality was 23.2% (76/328), with 64 (84.2%) of these deaths occurring within 30 d and 25 (32.9%) occurring within 7 d. GI dysfunction was experienced by 45.4% of patients and had a significant effect on 90-day mortality [odds ratio ( OR ) = 14.070, 95% confidence interval (CI) 5.886–38.290, P < 0.001]. Multivariate analysis showed that GI dysfunction was associated with admission SOFA score and % full-thickness TBSA. Overall, 88.2% (67/76) of deceased patients had GI dysfunction [hazard ratio ( HR ) for death of GI dysfunction = 5.951], with a survival advantage for functional disorders (diarrhoea, constipation, or nausea/vomiting) over GI ulcer/haemorrhage ( P < 0.001). Conclusion: Patients with severe burns have an unfavourable prognosis, as nearly one-fifth died within 90 d. Half of our patients had comorbidities related to GI dysfunction, among which GI ulcers and haemorrhages were independently correlated with 90-day mortality. More attention should be given to severe burn patients with GI dysfunction.

Perioperative complications are a major global concern, yet manual detection suffers from 27% under‑reporting and frequent misclassification. Clinical LLM deployment is constrained by data sovereignty, compute cost, and limited locally deployable model performance. We show targeted prompt engineering plus Low‑Rank Adaptation (LoRA) fine‑tuning converts smaller open‑source LLMs into expert‑level diagnostic tools. In dual‑center validation, we built a framework simultaneously identifying and grading 22 complication severities. State‑of‑the‑art models outperformed human experts; Chain‑of‑Thought prompting significantly improved general models ( p < 0.001) while preserving reasoning models’ performance. Across documentation length quartiles, AI models maintained F1 > 0.64, whereas human performance declined from 0.73 to 0.45, demonstrating superior robustness to documentation complexity. Our targeted strategy—decomposing detection into focused single‑complication assessments—improved small models, with further gains from LoRA. On external validation (Center 2), the optimized 4B model’s micro‑F1 rose from 0.28 to 0.64, approaching human experts (F1 = 0.69), driven by the targeted strategy (ΔF1 = 0.256, 95% CI [0.181, 0.336]) and LoRA (ΔF1 = 0.103, 95% CI [0.023, 0.186]). Concurrently, the 8B model surpassed human experts (F1 > 0.70). Optimized small models enable expert‑level accuracy with local deployment and preserved data sovereignty, offering a practical path for resource‑limited healthcare.

Opioid consumption for analgesia in burn patients is enormous. Non-opioid analgesics for burn pain management may result in opioid sparing, reducing opioid-related adverse reactions and drug tolerance or addiction. A dual-center, randomized controlled trial assessed Esketamine for the perioperative period in patients with severe [20-50 % total body surface area (TBSA)] and extensive (≥ 50 % TBSA) burns, comparing analgesia with standard anesthesia. Sixty patients were randomly allocated (1:1 ratio) to two arms. In the Treatment Arm, patients received intra-operative Esketamine and postoperative intravenous primary intelligent analgesia pump with Esketamine. Patients in the Control Arm received the same intervention as Treatment Arm without Esketamine. The primary endpoint was subjective analgesic efficacy (SAE) evaluated on Day 28 or the day before hospital discharge. Secondary outcomes included the postoperative Numeric Pain Rating (NPR) Scale at rest (NPRr) and during movement (NPRm) and opioid consumption. Gastrointestinal dysfunction Scores (GIDS) and serum markers of intestinal injury [intestinal fatty acid-binding protein 2 (iFabp2) and apolipoproteinA2 (ApoA2)] were measured in the 1st and 4th post-injury weeks. Depression and sleep quality were assessed by relevant questionnaires. Fifty-five patients were included in the analysis. Esketamine-treated Arm recorded a better analgesic efficacy than the Control Arm (proportion of patients with Grade 1 or 2 SAE scores, 67.9 % vs. 40.7 %, p = 0.022). Esketamine-treated patients had lower NPRm values (p = 0.033) and lower daily opioid consumption (p = 0.033) when compared with Controls. Esketamine-treated patients showed comparable gastrointestinal recovery to those in the Control Arm. The overall sleep quality might be improved in the Treatment Arm. Esketamine use is safe for perioperative primary intelligent analgesia of severe burns, resulting in improved resting pain control and lower opioid requirements. The trial was registered at the Chinese Clinical Trial Registry (www.chictr.org.cn/) (ChiCTR2000034069).

Perioperative complications represent a major global health concern affecting millions of surgical patients annually, yet manual detection methods suffer from significant under-reporting (27%) and misclassification rates. Clinical deployment of large language models (LLMs) for automated complication detection faces substantial barriers including data sovereignty concerns, computational costs, and performance limitations in locally deployable models. Here we show that targeted prompt engineering combined with Low-Rank Adaptation (LoRA) fine-tuning can transform smaller open-source large language models (LLMs) into expert-level diagnostic tools for perioperative complication detection. We conducted a dual-center validation study and developed a comprehensive framework enabling simultaneous identification and severity grading of 22 distinct perioperative complications. Initial evaluations revealed that state-of-the-art models consistently outperformed human experts, particularly reasoning models. AI models maintained consistent performance across varying document complexity, whereas human performance declined with increasing clinical documentation length. Our targeted strategy, which decomposed comprehensive detection into focused single-complication assessments, significantly improved smaller model capabilities. Combined with LoRA fine-tuning, the 4B parameter model’s F1 score increased from 0.18 to 0.55, approaching human expert performance (F1=0.526), while the 8B model achieved F1>0.61. These results demonstrate that optimized smaller models can achieve expert-level diagnostic accuracy while enabling local deployment with preserved data sovereignty, offering a practical solution for resource-limited healthcare institutions to implement automated perioperative complication screening.

Aims: To measure the possible association between flurbiprofen-containing patient-controlled intravenous analgesia (PCIA) and an increased risk of postoperative nausea and vomiting (PONV) following gynecological laparoscopic surgery. Methods: The retrospective study was performed on 2,430 patients who underwent a gynecological laparoscopic surgery between April 2021 through May 2022. Perioperative data were extracted from electronic medical records. Multivariable logistic regression was used to identify independent risk factors for PONV, supported by propensity score matching (PSM) and subgroup analyses. A directed acyclic graph (DAG) was constructed to guide confounder selection. Results: Among 2,430 eligible patients, PONV occurred in 27.7% (554/2000) of patients with flurbiprofen-containing PCIA versus 23.0% (99/430) without (absolute risk difference: 4.7%; number needed to harm [NNH] = 21). Multivariable analysis confirmed an independent association between flurbiprofen use and higher PONV risk (adjusted odds ratio [OR] = 1.414, 95% confidence interval [CI] = 1.042–1.918, p = 0.026). Subgroup analysis showed a particularly elevated risk in patients >60 years and those receiving hydromorphone-based PCIA. Notably, despite lower opioid consumption, patients with flurbiprofen had higher PONV incidence. Conclusion: Flurbiprofen-containing PCIA was associated with a modest yet clinically meaningful increase in PONV, especially when combined with hydromorphone. These findings challenge the assumption that NSAIDs are universally protective against PONV and underscore the importance of tailoring multimodal analgesia strategies according to patient risk profiles.