Liver surgery is one of the most technically demanding fields in general surgery due to the liver's complex vascular anatomy, fragile parenchyma, and essential physiological function. Historically, operative intervention...Liver surgery is one of the most technically demanding fields in general surgery due to the liver's complex vascular anatomy, fragile parenchyma, and essential physiological function. Historically, operative intervention was associated with prohibitive mortality, mainly because of uncontrollable hemorrhage and limited perioperative knowledge. This narrative review retraces the historical evolution of liver surgery from early experimental attempts to the modern era of minimally invasive procedures. Key milestones include the definition of functional liver anatomy and anatomical hepatectomy, as well as the development of techniques for bleeding control and parenchymal transection. The late 20th century marked the beginning of the modern era, driven by intraoperative ultrasound, advanced imaging, and strategies to preserve functional liver reserve. Liver transplantation further transformed the field, evolving into a standardized therapy with modern immunosuppression, graft optimization, split transplantation, and living donation. More recently, laparoscopic and robotic approaches have enabled increasingly complex resections while minimizing surgical trauma.
IntroductionPrimary closure of contaminated abdominal wall reconstruction (AWR) wounds carries significant wound complication risks. Vacuum-assisted, delayed primary closure (VADPC) improves outcomes but requires a secon...IntroductionPrimary closure of contaminated abdominal wall reconstruction (AWR) wounds carries significant wound complication risks. Vacuum-assisted, delayed primary closure (VADPC) improves outcomes but requires a second anesthetic. The "French fry" technique (FFT) utilizes VA benefits without reoperation. The aim of this study was to evaluate the outcomes of FFT in complex, contaminated AWR.MethodsA single center's open AWRs (CDC class 2-4) managed with FFT were reviewed. Technique: incision is closed except for small gaps to place subcutaneous 1 × 1 × 8cm NPWT foam "French fries" at 5-6 cm intervals under an incisional NPWT (-125 mmHg). FF are removed POD3 at the bedside and precise measurement is taken to assure all FFs are accounted for. Incisional NPWT is continued. Primary outcome included wound complications compared to Carolinas Equation for Determining Associated Risks (CeDAR) predictions.ResultsTwenty-seven AWR patients (BMI 31.6 ± 5.9 kg/m; 40.7% diabetic; 66.7% recurrent hernias; 100% CDC wound class 2-4; massive defect size 384.2 ± 143.8 cm) underwent FFT. Biologic mesh was placed in 74.1% of cases; 18.5% of defects were closed primarily. Fascial closure was achieved in 92.6%. CeDAR-predicted wound complication rate was 53.8%. Observed wound complication rate was 25.9%, including infection (14.8%) requiring intervention, with 7.4% requiring reopening of their wounds. There were no mesh infections and no hernia recurrences over average follow-up of 8.2 ± 19.8 months.ConclusionFFT is a practical, single-stage, strategy for contaminated AWR, reducing complications compared to predicted risk.
BackgroundUnintentional falls are the leading cause of injury among adults 65 years and older. It's believed that adverse benzodiazepine side effects increase the risk and frequency of falls. We hypothesize that geriatri...BackgroundUnintentional falls are the leading cause of injury among adults 65 years and older. It's believed that adverse benzodiazepine side effects increase the risk and frequency of falls. We hypothesize that geriatric patients taking benzodiazepines prior to admission experience worse clinical outcomes compared to those without preadmission use.MethodsUsing our level 1 trauma database, patients 65 years or older admitted following a ground level fall were divided into 2 groups, benzodiazepines positive on admission urine drug screen (+Benzo) and negative (-Benzo). Primary outcomes included in-hospital complications with additional outcomes including ICU admission rates, ICU LOS, overall hospital LOS, injury pattern data, and Injury Severity Score (ISS).Results11 133 patients were included with 764 in the +Benzo group. There was no difference in Injury Severity Score between the two groups (6.1 vs 5.7, = .140). The + Benzo group had a higher percentage of TBI (39.7% vs 29.8%, < .001), rib fractures (12.7% vs 9.1%, = .001), and extremity fractures (49.9% vs 43.2%, <.001) with no difference in ISS (6.1 vs 5.7, = .140). The + Benzo group had higher ICU admission rates (25.8% vs 17.7%, < .001) and experienced more cardiac arrests (2.2% vs 1.0%, = .004), pulmonary embolus (4.6% vs 3.1%, = .032), and pneumonia/VAP (0.65% vs 0.15%, = .008). On multivariable analysis, benzodiazepines prior to admission were an independent predictor of in-hospital complications (OR: 1.60, CI: 1.18-2.17, = .003).DiscussionOur study identifies pre-hospital benzodiazepine use and its association with worse in-hospital outcomes. Our findings can be used to identify high-risk fall patients on admission to mobilize additional resources to reduce the risk of adverse in-hospital outcomes.
BackgroundMotor vehicle collisions (MVCs) are the second leading cause of childhood mortality in the US, and child restraint systems (CRS) remain underutilized. We examine differences in demographics and clinical charact...BackgroundMotor vehicle collisions (MVCs) are the second leading cause of childhood mortality in the US, and child restraint systems (CRS) remain underutilized. We examine differences in demographics and clinical characteristics in pediatric patients presenting to the emergency department (ED) after MVCs with or without CRS.MethodsWe identified MVC trauma patients reported within the National Trauma Data Bank (NTDB) < 8 years of age and height/weight CRS-eligible. Bivariate descriptive analyses, interquartile range for continuous variables and a chi-square test of proportions tested differences of categorical variables of CRS strata. Sensitivity analysis was used in patients that were age appropriate for car and booster seats.ResultsIn all patients, median age for those without CRS was older (6 years, IQR 4-7) than those with CRS (4 years, IQR 2-5, < 0.0001). A lower proportion of Black patients had CRS compared to White (24.0% vs 32.6%, < 0.0001). For those with CRS, the highest proportion were boosters (38.7%). Injury severity score (ISS) was higher in the non-CRS group (8, IQR 4-14) than CRS (5, IQR 2-11, < 0.0001). A larger proportion of CRS patients were discharged home from ED (26.8% vs 18.5%). Racial disparities persisted in older patients, with fewer booster-eligible Black children in CRS (30.0% vs 17.9%, < 0.001).ConclusionsThis study demonstrates that older and Black children were less likely to be in a CRS, and that those who were not in CRS were more severely injured. Our study serves as the foundation for research to mitigate disparities, and outreach related to CRS improvements.
BackgroundThe adoption of robotic surgery has increased in foregut procedures, but its comparative value to laparoscopy in Heller myotomy remains unclear. We sought to evaluate the short-term perioperative outcomes of th...BackgroundThe adoption of robotic surgery has increased in foregut procedures, but its comparative value to laparoscopy in Heller myotomy remains unclear. We sought to evaluate the short-term perioperative outcomes of the robotic vs laparoscopic approach for Heller myotomy.MethodsThe American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database from 2022 to 2023 was queried for patients undergoing Heller myotomy, in which the "robotic" variable was included. Outcomes were evaluated using univariable analysis and multivariable logistic and linear regression.ResultsA total of 712 patients underwent a Heller myotomy; laparoscopy accounted for 63.6% of cases (n = 453) and robotic 36.3% (n = 259). The cohorts were matched aside from higher rates of HTN in the robotic group (42.9% vs 34.7%, < 0.05). Robotic Heller myotomy had significantly ( < 0.05) greater use of intraoperative EGD (22.8% vs 16.8%) and balloon distension studies (5.4% vs 1.5%). The robotic approach was associated with a longer median operative time (167 min vs 119.5 min) and independently associated with longer operative time on linear regression (β 35.39 min, CI 26.19-45.96 min). There were no significant differences in short-term complications including reoperation, readmission, and mortality.DiscussionThese data suggest no significant differences in the short-term outcomes between robotic and laparoscopic Heller myotomy. Robotic surgery was associated with longer operative times and utilization of more EGD and balloon distension studies. Further research is required to evaluate the clinical benefit of robotic surgery in achalasia.
BackgroundIntraoperative visualization of ground glass opacities (GGOs) remains challenging, with past techniques limited by safety concerns and poor margin assessment. We evaluate the utility of pafolacianine, an intrao...BackgroundIntraoperative visualization of ground glass opacities (GGOs) remains challenging, with past techniques limited by safety concerns and poor margin assessment. We evaluate the utility of pafolacianine, an intraoperative molecular imaging (IMI) agent for GGO visualization.MethodsWe conducted a prospective study from an integrated health system of patients undergoing video-assisted thoracoscopic surgery (VATS) lung resections for GGOs from 6/1/2025 to 8/31/2025. Pafolacianine was administered intravenously one hour preoperatively. Lung parenchyma was inspected for fluorescence using Stryker 1788 imaging before resection. Real-time intraoperative and postoperative surveys were completed, and chart review was performed to collect sociodemographic, pathologic, and surgical outcome data.ResultsTwenty-one patients (median age 70 years), predominantly female (62%), Asian (48%), and never-smokers (71%), underwent VATS with a median hospital stay of 1 day. Intraoperative molecular imaging prompted escalation to completion lobectomy in 10% of patients. Among 19 non-metastatic patients, 26 lesions (median diameter 14 mm) were resected, the majority of which were pure GGOs (69%). Intraoperative molecular imaging detected all malignant lesions, with 27% identified exclusively by IMI. Adenocarcinoma accounted for 62% of lesions, while 23% were benign. Intraoperative molecular imaging had a positive predictive value of 0.76 (95% CI, 0.54-0.90) and false discovery rate of 0.24 (95% CI, 0.10-0.46). All margins were negative, and no adverse events were reported.DiscussionThis exploratory real-world study suggests that pafolacianine facilitates optimal intraoperative visualization of challenging GGO lesions, supporting its potential to improve surgical decision-making, tumor detection, and margin assessment during VATS lung resections.
IntroductionObesity is a well-established risk factor for the development, recurrence, and postoperative complications of ventral hernia repair. However, the relationship between obesity and inguinal hernia repair is les...IntroductionObesity is a well-established risk factor for the development, recurrence, and postoperative complications of ventral hernia repair. However, the relationship between obesity and inguinal hernia repair is less well defined, with limited evidence on how body mass index (BMI) influences operative approach and outcomes.MethodsA retrospective review of the 2022-2023 ACS-NSQIP database was performed including all adult patients who underwent inguinal and femoral hernia repairs identified using CPT codes. Bivariate and multivariable regression analyses were used to examine the impact of obesity and BMI on operative approach, case acuity, operative time, and short-term outcomes after inguinal hernia repair.ResultsA total of 73 456 adult patients were identified with a mean age of 61 years. Higher BMI was associated with greater operative complexity, including longer operative times and increased use of robotic and combined approaches. Surgical site occurrences (SSOs) were uncommon overall (0.7%) but increased progressively with higher BMI. In multivariable analysis, the association between BMI and operative characteristics, SSOs, and major complications varied across BMI levels. Higher BMI was independently associated with increased odds of SSOs and major complications, while laparoscopic and robotic approaches were consistently associated with lower odds of adverse outcomes compared with open or combined approaches.ConclusionHigher BMI is associated with increased operative complexity and worse short-term postoperative outcomes following inguinal hernia repair, while minimally invasive approaches were associated with lower odds of obesity-related adverse outcomes. These findings highlight the importance of individualized risk stratification and preoperative optimization in inguinal hernia repairs.
Several advances in surgery originated from observations made by medical students during their training. Although students are usually viewed primarily as learners, their position at the intersection of scientific study...Several advances in surgery originated from observations made by medical students during their training. Although students are usually viewed primarily as learners, their position at the intersection of scientific study and patient care gave a handful a chance to have significant impact. Jean-François Calot described the hepatocystic triangle in his doctoral thesis, establishing an anatomical principle that remains central to safe cholecystectomy. Harvey Cushing and Ernest Amory Codman developed the anesthesia record, a chart that tracks vital signs during surgery and became the foundation of modern perioperative monitoring. Jay McLean identified anticoagulant substances that led to the discovery of heparin, enabling the development of vascular and cardiac surgery. After reasoning that platelets were destroyed in the spleen, Paul Kaznelson proposed splenectomy as treatment for immune thrombocytopenic purpura. Charles Best's observation that ligation of the pancreatic duct in dogs caused atrophy of the exocrine pancreas while preserving the islets led to the isolation of insulin, one of the most important therapeutic discoveries in medicine. Seeking a more efficient method of blood transfusion, Michael DeBakey designed the roller pump, a device that later became essential to cardiopulmonary bypass. As a medical student at Johns Hopkins, David C. Sabiston Jr participated in early work with Mark Ravitch exploring restorative operations for ulcerative colitis. Thomas Fogarty conceived the balloon embolectomy catheter after seeing the difficulty of removing arterial emboli. These episodes illustrate how curiosity and careful observation during medical education can produce insights that shape surgical practice for decades.
BackgroundLarge language models (LLMs) have demonstrated strong performance on general medical and surgical examinations; however, their capacity to accurately interpret and apply subspecialty clinical practice guideline...BackgroundLarge language models (LLMs) have demonstrated strong performance on general medical and surgical examinations; however, their capacity to accurately interpret and apply subspecialty clinical practice guidelines remains incompletely characterized.ObjectiveTo evaluate and compare the accuracy and consistency of two contemporary LLMs-Google Gemini and OpenEvidence-using multiple-choice questions (MCQs) derived directly from the 2022 American Society of Colon and Rectal Surgeons (ASCRS) Clinical Practice Guidelines for anorectal abscess, fistula-in-ano, and rectovaginal fistula.MethodsThirty guideline-based MCQs were developed and independently validated by surgeon reviewers. Each question was presented to both models under identical conditions without additional prompting. Accuracy was calculated with 95% confidence intervals and compared against chance performance (p = .25). Inter-model agreement was assessed using Cohen's kappa coefficient.ResultsBoth Gemini and OpenEvidence correctly answered 29 of 30 questions (96.7%; 95% CI, 0.83-0.999), significantly exceeding chance performance ( < .0001 for both models). Both models missed the same question, yielding perfect inter-model agreement (κ = 1.0).ConclusionIn this focused pilot study restricted to ASCRS anorectal disease guidelines, both LLMs demonstrated near-perfect and statistically equivalent accuracy. These findings suggest that contemporary LLMs can accurately apply subspecialty surgical guidelines within a narrow domain, though broader, multi-guideline evaluations are required before generalization.
UNLABELLED: Venothromboembolism (VTE) is rare in pediatric trauma but can have devastating consequences. While chemical prophylaxis guidelines exist in adult trauma, guidelines for pediatric populations are sparse. Adher...UNLABELLED: Venothromboembolism (VTE) is rare in pediatric trauma but can have devastating consequences. While chemical prophylaxis guidelines exist in adult trauma, guidelines for pediatric populations are sparse. Adherence to the Eastern Association of the Surgery of Trauma (EAST) and Pediatric Trauma Society (PTS) guidelines is unclear. Our objective was to evaluate rate of VTE prophylaxis and adherence to current guidelines in our pediatric trauma population. METHODS: We performed a retrospective cohort study at a single ACS-verified adult Level I/pediatric Level II Trauma Center from the years 2014-2024. Patients aged 7-18 were included. A subgroup based on the EAST/PTS guidelines. VTE prophylaxis, deep vein thrombosis (DVT) or pulmonary embolism (PE), injury severity score (ISS), and pubertal status. RESULTS: During the study period, 1271 patients met inclusion criteria. Four patients (0.31%) experienced VTE. Three patients (0.24%) had a bleeding complication. While 787 patients met criteria for prophylaxis by EAST/PTS guidelines, only 279 patients (35.4%) received prophylaxis. Use of chemical prophylaxis increased after publication of EAST/PTS guidelines (20.6% vs 39.9%, p < 0.001). Of patients who qualified for chemical prophylaxis by the EAST/PTS guidelines, two patients (0.3%) developed VTE, while 1 (0.1%) developed a bleeding complication. One patient who developed a DVT was younger than 15 with an ISS <25 and did not meet criteria by EAST/PTS guidelines. CONCLUSION: Adherence to pediatric VTE prophylaxis guidelines increased over the study period without increase in adverse events. There was no difference in VTE incidence, but VTE was rare.
While perioperative anxiety and depression are well-recognized in surgical patients, there is a long-standing, little-discussed, curious, but potentially concerning concept: that of the patient with a premonition that th...While perioperative anxiety and depression are well-recognized in surgical patients, there is a long-standing, little-discussed, curious, but potentially concerning concept: that of the patient with a premonition that they will not survive the surgery. This phenomenon has seen variable interest since the 1960s, but its presence raises a concern for surgeons and worry about operating. After a brief review of the literature, it is concluded that the current literature is sparse and mainly descriptive in nature; no convincing evidence demonstrates that premonition of death predicts perioperative mortality, and surgical decision-making should not be altered based on the presence of such a premonition alone. However, we urge awareness by surgeons if such an attitude is present and, if so, it should not be ignored or minimized, but evaluated and treated, preferably in concert with a consulting psychiatrist, to optimize reassurance for patient, family, and the surgeon.
Nguyen PD, Grigorian A, Costantini T
… +44 more, Santorelli J, Von Husen L, Dilday J, Inaba K, Schellenberg M, Stephens D, Abou Chaar M, Mahmoud K, Schreiber M, Minoza K, Underwood S, Biffl W, Lu N, Bayat D, Coimbra R, Edwards S, Tillou A, Zheng DJ, Plevin RE, Kim H, Stein D, Smith J, Sonntag CC, Callcut R, Furlong H, Graeff LW, Trevizo E, Cochran-Yu M, Duncan T, Diaz G, Kim D, Neville A, Balan N, Siletz A, Ley E, Park G, Alexander J, Johnson A, Cotton-Betteridge A, Smith M, Alvarez C, Scolaro J, Aryan N, Nahmias J
Morel-Lavallee lesions (MLLs) are traumatic closed degloving injuries with operative intervention (OI), especially excisional debridement, remaining the mainstay of treatment. However, high-quality evidence describing op...Morel-Lavallee lesions (MLLs) are traumatic closed degloving injuries with operative intervention (OI), especially excisional debridement, remaining the mainstay of treatment. However, high-quality evidence describing operative techniques and outcomes remains limited. This study aimed to characterize current OI practices and associated outcomes, especially among patients undergoing excisional debridement, hypothesizing an increased risk of wound complications with this approach for MLLs.MethodsThis prospective (2021-2024), multicenter, observational study was conducted across 15 trauma centers and included patients diagnosed with MLL who underwent OI. The association between operative technique (eg, excisional debridement) and the primary outcome of wound complications was evaluated. Wound complications were defined as cellulitis, deep wound infection, skin necrosis, necrotizing soft tissue infection (NSTI), and/or organized hematoma.ResultsOf 88 MLL patients undergoing OI, 36 (40.9%) had excisional debridement, with 4 (11.1%) limited excision to skin, 20 (55.6%) extended to subcutaneous tissue, and 12 (33.3%) down to muscle. Excisional debridement had higher rates of skin necrosis (22.2% vs. 1.9%, = .002). Although overall rates of wound complications were statistically similar between cohorts (33.3% vs. 17.3%, = .098), univariate analysis demonstrated excisional debridement had increased risk of wound complications (OR 2.52, CI 1.07-5.97, = .035).ConclusionThis prospective multicenter study over three years demonstrated that a fifth of MLLs undergoing excisional debridement developed postoperative skin necrosis. Excisional debridement also had a more than two-fold increased risk of wound complications compared to other OIs. These findings underscore the importance of investigating alternative operative techniques, such as incision and drainage or minimally invasive debridement methods.
BackgroundRib fractures in elderly trauma patients carry increased morbidity and mortality. Prior guidelines recommended indiscriminate intensive care unit (ICU) admission for patients ≥65 years with ≥3 rib fractures. In...BackgroundRib fractures in elderly trauma patients carry increased morbidity and mortality. Prior guidelines recommended indiscriminate intensive care unit (ICU) admission for patients ≥65 years with ≥3 rib fractures. In 2024, our institution implemented the Rib Injury Guideline (RIG) score, a novel triage and disposition tool, through a standardized smartphrase (SP).MethodsWe retrospectively reviewed trauma patients ≥65 years with ≥1 rib fracture admitted between 1/1/2020 and 7/5/2025. Patients admitted before 10/31/2024 (pre-RIG) were compared to those admitted after (post-RIG). Primary outcomes were ICU admission and unanticipated ICU upgrade. Secondary outcomes included mortality, ICU and hospital length of stay (LOS), and compliance with the RIG SP.Results546 patients were included, with a median age of 80 years and median Injury Severity Score (ISS) of 10. There were 462 (85%) patients pre- and 84 (15%) post-RIG. ICU utilization significantly decreased in the post-RIG cohort (61% vs 37%, OR 0.38, < 0.01), with an even greater reduction among patients with multiple rib fractures (67% vs 38%, OR 0.29, < 0.01). Unanticipated ICU upgrade was 1% for both groups ( > 0.99). ICU LOS (median 2 vs 3, = 0.17), hospital LOS (median 6 vs 6, = 0.65), and mortality (4% vs 5%, = 0.55) were similar between cohorts. Compliance with the SP was 82%.DiscussionImplementation of the RIG score with an integrated smartphrase substantially reduced ICU admissions in elderly rib fracture patients without compromising patient safety, as evidenced by the absence of increased rates of ICU upgrades and no impact on mortality.
Inter-surgeon variability in robotic surgery learning curves, as well as the impact of trainee involvement and the presence or absence of a formal robotic training curriculum, remain poorly defined. Cumulative sum (CUSUM...Inter-surgeon variability in robotic surgery learning curves, as well as the impact of trainee involvement and the presence or absence of a formal robotic training curriculum, remain poorly defined. Cumulative sum (CUSUM) analysis is a validated method for evaluating learning curves in robotic surgery. Our study aimed to utilize CUSUM analysis to explore whether surgical trainees affect surgical operating times. We retrospectively analyzed robotic-assisted cholecystectomies performed by 7 surgeons at a single academic institution between 2012 and 2022. A robotic surgery curriculum was implemented in 2016. Cases were ordered chronologically for each surgeon, and CUSUM learning curves were generated using operative time as the outcome. The first peak in the CUSUM curve indicated the completion of the learning phase. Trainee involvement was analyzed by postgraduate year. 707 operations were performed. Five surgeons demonstrated a distinct learning phase, with a learning phase ranging from 20 to 59 cases, whereas two surgeons exhibited baseline proficiency without an identifiable learning phase. Despite consistently high trainee participation (94.5% of cases) and similar distributions of trainee seniority, learning curve variability persisted. Implementation of an institutional robotic training curriculum was not associated with abrupt changes in learning curve trajectories among surgeons. Learning curves in robotic-assisted cholecystectomies are highly variable and surgeon-specific. Implementation of the robotic training program did not influence the overall trajectory of the surgeon's personal learning phase.
Nearly 1 million older adults are hospitalized annually for emergency general surgery (EGS) conditions in the United States. Recent literature has demonstrated the impact of discordant care in EGS, defined as a mismatch...Nearly 1 million older adults are hospitalized annually for emergency general surgery (EGS) conditions in the United States. Recent literature has demonstrated the impact of discordant care in EGS, defined as a mismatch between treatment choice and patients' clinical risk profile. In the present work, we characterize the association between preoperative decision variation and outcomes in elderly EGS patients.MethodsAll geriatric (≥65 years) hospitalizations with EGS conditions (appendicitis, diverticulitis, cholecystitis, hernia, bowel obstruction, perforated peptic ulcer, intestinal ischemia, perforated bowel) were tabulated from the 2022 National Inpatient Sample (NIS). A propensity score (PS) was developed to estimate the likelihood of undergoing operative management. Low and high probability for surgery was defined using a cut-off of PS = 0.5. Model-concordant cases were defined as those whose treatment matched their predicted likelihood, while model-discordant cases did not. Multivariable regression models were developed to determine the association of care variation on outcomes.ResultsOf an estimated 900 730 patients with EGS conditions, 154 300 (17.1%) were classified as model-discordant. Following risk adjustment, model-discordance was associated with greater odds of mortality (AOR: 1.33, 95% CI: 1.25-1.42), as well as gastrointestinal (AOR: 2.07, 95% CI: 1.95-2.19), infectious (AOR: 1.41, 95% CI: 1.36-1.46), and respiratory (AOR: 1.26, 95% CI: 1.20-1.32) complications. Furthermore, model-discordance was associated with increased hospitalization costs (β: +$9 760, 95% CI: +$9 255-10 265) and length of stay duration (β: +2.72 days, 95% CI: +2.59-2.84).ConclusionPatients classified as model-discordant exhibited elevated rates of in-hospital mortality and postoperative complications. Given this variation, our study warrants investigation into guideline adherence and outcomes in this vulnerable population.
Esophageal function tests (esophageal manometry and 24-hour pH monitoring) are essential to diagnose benign esophageal disorders such as gastroesophageal reflux disease and achalasia. Unfortunately, these tests are rarel...Esophageal function tests (esophageal manometry and 24-hour pH monitoring) are essential to diagnose benign esophageal disorders such as gastroesophageal reflux disease and achalasia. Unfortunately, these tests are rarely taught during medical school or surgical training, so that surgeons who treat these diseases must rely on the interpretation and conclusions made by other physicians, mostly gastroenterologists, just reading the report without looking at the tracings themselves. Potentially, this approach may lead to poor outcomes or even malpractice.
IntroductionVisceral aneurysms pose diagnostic and therapeutic challenges in vascular surgery. Large language models (LLMs) may assist in clinical decision-making, but their application requires rigorous validation. Trad...IntroductionVisceral aneurysms pose diagnostic and therapeutic challenges in vascular surgery. Large language models (LLMs) may assist in clinical decision-making, but their application requires rigorous validation. Traditional validation methods are labor-intensive and difficult to scale.ObjectiveWe examined the capability of an LLM in managing visceral aneurysms and explored an automated framework for validating AI-generated clinical responses.MethodsUsing Python with the Pandas library and OpenAI API, we probed the Society for Vascular Surgery (SVS) clinical practice guidelines on visceral aneurysm management. ChatGPT-4o-mini was instructed to review guideline recommendations, generate clinical scenarios, propose management strategies, and evaluate its own responses using a four-tier rubric (1 = completely correct; 2 = partially correct; 3 = partially incorrect; 4 = no correct information). Human evaluators independently assessed the same responses and graded questions as good, fair, or poor and whether they were leading.ResultsEighty visceral aneurysm scenarios were generated and evaluated. ChatGPT-4o-mini self-assessed 89% of responses as correct (scores 1-2), compared to 67% by human evaluators (chi-square, < 0.0001), with the greatest discrepancy in the partially correct category. Most AI-generated questions were of good quality (56%), though 44% were considered leading questions.ConclusionAn automated validation framework for AI-generated clinical responses is feasible. However, the 67% correctness rate and systematic AI self-overestimation indicate that current LLMs remain unsuitable for independent clinical use, reinforcing the need for expert oversight. The integration of Python-driven automation, structured AI inference, and expert review holds promise for increasing the efficiency of evaluating LLMs at-scale across clinical domains.