Accuracy and consistency of publicly available Large Language Models as clinical decision support tools for the management of colon cancer.

BACKGROUND: Large Language Models (LLM; e.g., ChatGPT) may be used to assist clinicians and form the basis of future clinical decision support (CDS) for colon cancer. The objectives of this study were to (1) evaluate the response accuracy of two LLM-powered interfaces in identifying guideline-based care in simulated clinical scenarios and (2) define response variation between and within LLMs. METHODS: Clinical scenarios with "next steps in management" queries were developed based on National Comprehensive Cancer Network guidelines. Prompts were entered into OpenAI ChatGPT and Microsoft Copilot in independent sessions, yielding four responses per scenario. Responses were compared to clinician-developed responses and assessed for accuracy, consistency, and verbosity. RESULTS: Across 108 responses to 27 prompts, both platforms yielded completely correct responses to 36% of scenarios (n = 39). For ChatGPT, 39% (n = 21) were missing information and 24% (n = 14) contained inaccurate/misleading information. Copilot performed similarly, with 37% (n = 20) having missing information and 28% (n = 15) containing inaccurate/misleading information (p = 0.96). Clinician responses were significantly shorter (34 ± 15.5 words) than both ChatGPT (251 ± 86 words) and Copilot (271 ± 67 words; both p < 0.01). CONCLUSIONS: Publicly available LLM applications often provide verbose responses with vague or inaccurate information regarding colon cancer management. Significant optimization is required before use in formal CDS.

Percutaneous Cystic Duct Interventions and Drain Internalization for Calculous Cholecystitis in Patients Ineligible for Surgery.

PURPOSE: To evaluate the feasibility, effectiveness, and outcomes of percutaneous cholecystostomy drain internalization in patients with calculous cholecystitis who were not surgical candidates. MATERIALS AND METHODS: Percutaneous cystic duct interventions were attempted in 17 patients (with the intent to place dual cholecystoduodenal stents) who were deemed unfit for surgery and had previously undergone percutaneous cholecystostomies for acute calculous cholecystitis. Baseline demographics, technical success, time from percutaneous cholecystostomy to internalization (dual cholecystoduodenal stent placement), stent patency duration, and adverse event rates were evaluated. RESULTS: Fifteen (88%) of 17 procedures to cross the cystic duct were technically successful. Of these 17 patients, 13 (76%) underwent successful placement of dual cholecystoduodenal stents. Two of these 13 patients (who had successful dual cholecystoduodenal stent placement) needed repeat percutaneous cholecystostomy drains (1 patient had stent migration leading to recurrent cholecystitis, and the other had a perihepatic biloma). The 1-year patency rate was 77% (95% CI, 47%-100%). CONCLUSIONS: Dual cholecystoduodenal stent placement in nonsurgical patients is a technically feasible treatment option with the goal to remove percutaneous cholecystostomy drains.

Neoadjuvant therapy use and association with postoperative outcomes and overall survival in patients with extrahepatic cholangiocarcinoma.

BACKGROUND AND OBJECTIVES: Evidence for neoadjuvant therapy (NAT) in extrahepatic cholangiocarcinoma (eCCA) is limited. Our objectives were to: (1) characterize treatment trends, (2) identify factors associated with receipt of NAT, and (3) evaluate associations between NAT and postoperative outcomes. METHODS: Retrospective cohort study of the National Cancer Database (2004-2017). Multivariable logistic regression assessed associations between NAT and postoperative outcomes. Stratified analysis evaluated differences between surgery first, neoadjuvant chemotherapy, and neoadjuvant chemoradiation (CRT). RESULTS: Among 8040 patients, 417 (5.2%) received NAT. NAT increased during the study period 2.9%-8.4% (p < 0.001). Factors associated with receipt of NAT included age <50 (vs. >75, odds ratio [OR] 4.32, p < 0.001) and stage 3 disease (vs. 1, OR 1.68, p = 0.01). Compared with surgery first, patients who received NAT had higher odds of R0 resection (OR 1.49, p = 0.01) and lower 30-day mortality (OR 0.51, p = 0.04). On stratified analysis, neoadjuvant chemotherapy was not associated with differences in any outcomes. However, neoadjuvant CRT was associated with improvement in R0 resection (OR 3.52, <0.001) and median survival (47.8 vs. 25.3 months, log-rank < 0.001) compared to surgery first. CONCLUSIONS: NAT, particularly neoadjuvant CRT, was associated with improved postoperative outcomes. These data suggest expanding the use of neoadjuvant CRT for eCCA.

Evaluation of emergency department treat-and-release encounters after major gastrointestinal surgery.

BACKGROUND AND OBJECTIVES: Emergency department (ED) utilization after gastrointestinal cancer operations is poorly characterized. Our study objectives were to determine the incidence of, reasons for, and predictors of ED treat-and-release encounters after gastrointestinal cancer operations. METHODS: Patients who underwent elective esophageal, hepatobiliary, gastric, pancreatic, small intestinal, or colorectal operations for cancer were identified in the 2015-2017 Healthcare Cost and Utilization Project State Inpatient and State Emergency Department Databases for New York, Maryland, and Florida. The primary outcomes were the incidence of ED treat-and-release encounters and readmissions within 30 days of discharge. RESULTS: Among 51 527 patients at 406 hospitals, 4047 (7.9%) had an ED treat-and-release encounter, and 5573 (10.8%) had an ED encounter with readmission. In total, 40.7% of ED encounters were treat-and-release encounters. ED treat-and-release encounters were most frequently for pain (12.0%), device/ostomy complaints (11.7%), or wound complaints (11.4%). ED treat-and-release encounters predictors included non-Hispanic Black race/ethnicity (odds ratio [OR] 1.24, 95% confidence interval [CI] 1.12-1.37) and Medicare (OR 1.27, 95% CI 1.16-1.40) or Medicaid (OR 1.82, 95% CI 1.62-2.40) coverage. CONCLUSIONS: ED treat-and-release encounters are common after major gastrointestinal operations, making up nearly half of postdischarge ED encounters. The reasons for ED treat-and-release encounters differ from those for ED encounters with readmissions.

Discrimination, Abuse, Harassment, and Burnout in Surgical Residency Training.

BACKGROUND: Physicians, particularly trainees and those in surgical subspecialties, are at risk for burnout. Mistreatment (i.e., discrimination, verbal or physical abuse, and sexual harassment) may contribute to burnout and suicidal thoughts. METHODS: A cross-sectional national survey of general surgery residents administered with the 2018 American Board of Surgery In-Training Examination assessed mistreatment, burnout (evaluated with the use of the modified Maslach Burnout Inventory), and suicidal thoughts during the past year. We used multivariable logistic-regression models to assess the association of mistreatment with burnout and suicidal thoughts. The survey asked residents to report their gender. RESULTS: Among 7409 residents (99.3% of the eligible residents) from all 262 surgical residency programs, 31.9% reported discrimination based on their self-identified gender, 16.6% reported racial discrimination, 30.3% reported verbal or physical abuse (or both), and 10.3% reported sexual harassment. Rates of all mistreatment measures were higher among women; 65.1% of the women reported gender discrimination and 19.9% reported sexual harassment. Patients and patients' families were the most frequent sources of gender discrimination (as reported by 43.6% of residents) and racial discrimination (47.4%), whereas attending surgeons were the most frequent sources of sexual harassment (27.2%) and abuse (51.9%). Proportion of residents reporting mistreatment varied considerably among residency programs (e.g., ranging from 0 to 66.7% for verbal abuse). Weekly burnout symptoms were reported by 38.5% of residents, and 4.5% reported having had suicidal thoughts during the past year. Residents who reported exposure to discrimination, abuse, or harassment at least a few times per month were more likely than residents with no reported mistreatment exposures to have symptoms of burnout (odds ratio, 2.94; 95% confidence interval [CI], 2.58 to 3.36) and suicidal thoughts (odds ratio, 3.07; 95% CI, 2.25 to 4.19). Although models that were not adjusted for mistreatment showed that women were more likely than men to report burnout symptoms (42.4% vs. 35.9%; odds ratio, 1.33; 95% CI, 1.20 to 1.48), the difference was no longer evident after the models were adjusted for mistreatment (odds ratio, 0.90; 95% CI, 0.80 to 1.00). CONCLUSIONS: Mistreatment occurs frequently among general surgery residents, especially women, and is associated with burnout and suicidal thoughts.

Unbundling Bundles: Evaluating the Association of Individual Colorectal Surgical Site Infection Reduction Bundle Elements on Infection Rates in a Statewide Collaborative.

BACKGROUND: Surgical site infection reduction bundles are effective but can be complex and resource intensive. Understanding which bundle elements are associated with reduced surgical site infections may guide concise bundle implementation. OBJECTIVE: The purpose of this study was to evaluate the association of individual surgical site infection reduction bundle elements with infection rates. DESIGN: This was a post-hoc analysis of a prospective cohort study. SETTING: This study took place at Illinois Surgical Quality Improvement Collaborative hospitals. PATIENTS: Patients who had elective colorectal resections at participating hospitals from 2016 to 2017. INTERVENTIONS: The intervention was a 16-element colorectal surgical site infection reduction bundle. MAIN OUTCOME MEASURES: Surgical site infection rates were compared among patients by adherence with each bundle element using χ 2 tests and multivariable logistic regression. Principal component analysis identified composites of correlated bundle elements. Coincidence analysis identified combinations of bundle elements or principal component composites associated with the absence of surgical site infection. RESULTS: Among 2722 patients, 192 (7.1%) developed a surgical site infection. Infections were less likely when oral antibiotics (OR 0.63 [95% CI 0.41-0.97]), wound protectors (OR 0.55 [95% CI 0.37-0.81]), and occlusive dressings (OR 0.71 [95% CI 0.51-1.00]) were used. Bundle elements were reduced into 5 principal component composites. Adherence with the combination of oral antibiotics, wound protector, or redosing intravenous antibiotic prophylaxis plus chlorhexidine-alcohol intraoperative skin preparation was associated with the absence of infection (consistency = 0.94, coverage = 0.96). Four of the 5 principal component composites in various combinations were associated with the absence of surgical site infection, whereas the composite consisting of occlusive dressing placement, postoperative dressing removal, and daily postoperative chlorhexidine incisional cleansing had no association with the outcome. LIMITATIONS: The inclusion of hospitals engaged in quality improvement initiatives may limit the generalizability of these data. CONCLUSION: Bundle elements had varying association with infection reduction. Implementation of colorectal surgical site infection reduction bundles should focus on the specific elements associated with low surgical site infections. See Video Abstract at http://links.lww.com/DCR/B808 . DESEMPAQUETANDO PAQUETES EVALUACIN DE LA ASOCIACIN DE ELEMENTOS INDIVIDUALES DEL PAQUETE DE REDUCCIN DE INFECCIONES DEL SITIO QUIRRGICO COLORRECTAL CON LAS TASAS DE INFECCIN EN UNA COLABORACIN ESTATAL: ANTECEDENTES:Los paquetes de reducción de infecciones del sitio quirúrgico son efectivos pero pueden ser complejos y requieren muchos recursos. Comprender qué elementos del paquete están asociados con la reducción de las infecciones del sitio quirúrgico puede guiar la implementación concisa del paquete.OBJETIVO:Evaluar la asociación de los elementos individuales del paquete de reducción de infecciones del sitio quirúrgico con las tasas de infección.DISEÑO:Análisis post-hoc de un estudio de cohorte prospectivo.ESCENARIO:Hospitales colaborativos para la mejora de la calidad quirúrgica de Illinois.PACIENTES:Resecciones colorrectales electivas en los hospitales participantes entre 2016 y 2017.INTERVENCIONES:Paquete de reducción de infección del sitio quirúrgico colorrectal de 16 elementos.PRINCIPALES MEDIDAS DE RESULTADO:Se compararon las tasas de infección del sitio quirúrgico entre los pacientes según la adherencia con cada elemento del paquete mediante pruebas de Chi cuadrado y regresión logística multivariable. El análisis de componentes principales identificó compuestos de elementos de paquete correlacionados. El análisis de coincidencia identificó combinaciones de elementos del haz o compuestos de componentes principales asociados con la ausencia de infección del sitio quirúrgico.RESULTADOS:Entre 2722 pacientes, 192 (7,1%) desarrollaron una infección del sitio quirúrgico. Las infecciones fueron menos probables cuando se administraron antibióticos orales (OR 0,63 (IC 95% 0,41-0,97)), protectores de heridas (OR 0,55 (IC 95% 0,37-0,81)) y vendajes oclusivos (OR 0.71 (IC 95% 0,51-1,00]) fueron usados. Los elementos del paquete se redujeron a 5 grupos de componentes principales. La adherencia a la combinación de (1) antibióticos orales, (2) protector de heridas o (3) redosificación de profilaxis antibiótica intravenosa más preparación de la piel intraoperatoria con clorhexidina-alcohol se asoció con la ausencia de infección (consistencia = 0,94, cobertura = 0,96). Cuatro de los cinco grupos de componentes principales en varias combinaciones se asociaron con la ausencia de infección del sitio quirúrgico, mientras que el grupo que consiste en la colocación del apósito oclusivo, la remosión del apósito en posoperatorio y la limpieza incisional posoperatoria diaria con clorhexidina no tuvo asociación con el resultado.LIMITACIONES:La inclusión de hospitales que participan en iniciativas de mejora de la calidad puede limitar la generalización de estos datos.CONCLUSIONES:Los elementos del paquete tuvieron una asociación variable con la reducción de la infección. La implementación de paquetes de reducción de infecciones del sitio quirúrgico colorrectal debe centrarse en los elementos específicos asociados con pocas infecciones del sitio quirúrgico. Consulte Video Resumen en http://links.lww.com/DCR/B808 . (Traducción-Juan Carlos Reyes ).

Management of colorectal cancer during the COVID-19 pandemic: Recommendations from a statewide multidisciplinary cancer collaborative.

COVID-19 has resulted in significant disruptions in cancer care. The Illinois Cancer Collaborative (ILCC), a statewide multidisciplinary cancer collaborative, has developed expert recommendations for triage and management of colorectal cancer when disruptions occur in usual care. Such recommendations would be applicable to future outbreaks of COVID-19 or other large-scale disruptions in cancer care.

The quality of screening colonoscopy in rural and underserved areas.

BACKGROUND: Screening colonoscopy effectiveness depends on procedure quality; however, knowledge about colonoscopy quality in rural and underserved areas is limited. This study aimed to describe the characteristics and quality of colonoscopy and to examine predictors of colonoscopy quality at rural and underserved hospitals. METHODS: Adults undergoing colonoscopy from April 2017 to March 2019 at rural or underserved hospitals across the Illinois Surgical Quality Improvement Collaborative were prospectively identified. The primary outcome was colorectal adenoma detection, and secondary outcomes included bowel preparation adequacy, cecum photodocumentation, and withdrawal time. Performance was benchmarked against multisociety guidelines, and multivariable logistic regression was used to examine patient, physician, and procedure characteristics associated with adenoma detection. RESULTS: In total, 4217 colonoscopy procedures were performed at 8 hospitals, including 1865 screening examinations performed by 19 surgeons, 9 gastroenterologists, and 2 family practitioners. Physician screening volume ranged from 2 to 218 procedures (median 50; IQR 23-74). Adenoma detection occurred in 26.6% of screening procedures (target: ≥ 25%), 90.7% had adequate bowel preparation (target: ≥ 85%), 93.1% had cecum photodocumentation (target: ≥ 95%), and mean withdrawal time was 8.1 min (target: ≥ 6). Physician specialty was associated with adenoma detection (gastroenterologists: 36.9% vs. surgeons: 22.5%; OR 2.30, 95% CI 1.40-3.77), but adequate bowel preparation (OR 1.15, 95% CI 0.76-1.73) and cecum photodocumentation (OR 1.56, 95% CI 0.91-2.69) were not. CONCLUSION: Colonoscopies performed at rural and underserved hospitals meet many quality metrics; however, quality varied widely. As physicians are scarce in rural and underserved areas, individualized interventions to improve colonoscopy quality are needed.

A National Survey of Motor Vehicle Crashes Among General Surgery Residents.

OBJECTIVES: Evaluate the frequency of self-reported, post-call hazardous driving events in a national cohort of general surgery residents and determine the associations between duty hour policy violations, psychiatric well-being, and hazardous driving events. SUMMARY OF BACKGROUND DATA: MVCs are a leading cause of resident mortality. Extended work shifts and poor psychiatric well-being are risk factors for MVCs, placing general surgery residents at risk. METHODS: General surgery residents from US programs were surveyed after the 2017 American Board of Surgery In-Training Examination. Outcomes included self-reported nodding off while driving, near-miss MVCs, and MVCs. Group-adjusted cluster Chi-square and hierarchical regression models with program-level intercepts measured associations between resident- and program-level factors and outcomes. RESULTS: Among 7391 general surgery residents from 260 programs (response rate 99.3%), 34.7% reported nodding off while driving, 26.6% a near-miss MVC, and 5.0% an MVC over the preceding 6 months. More frequent 80-hour rule violations were associated with all hazardous driving events: nodding off while driving {59.8% with ≥5 months with violations vs 27.2% with 0, adjusted odds ratio (AOR) 2.86 [95% confidence interval (CI) 2.21-3.69]}, near-miss MVCs, [53.6% vs 19.2%, AOR 3.28 (95% CI 2.53-4.24)], and MVCs [14.0% vs 3.5%, AOR 2.46 (95% CI 1.65-3.67)]. Similarly, poor psychiatric well-being was associated with all 3 outcomes [eg, 8.0% with poor psychiatric well-being reported MVCs vs 2.6% without, odds ratio 2.55 (95% CI 2.00-3.24)]. CONCLUSIONS: Hazardous driving events are prevalent among general surgery residents and associated with frequent duty hour violations and poor psychiatric well-being. Greater adherence to duty hour standards and efforts to improve well-being may improve driving safety.

Association of Surgical Resident Wellness With Medical Errors and Patient Outcomes.

OBJECTIVES: The aims of this study were to: (1) measure the prevalence of self-reported medical error among general surgery trainees, (2) assess the association between general surgery resident wellness (ie, burnout and poor psychiatric well-being) and self-reported medical error, and (3) examine the association between program-level wellness and objectively measured patient outcomes. SUMMARY OF BACKGROUND DATA: Poor wellness is prevalent among surgical trainees but the impact on medical error and objective patient outcomes (eg, morbidity or mortality) is unclear as existing studies are limited to physician and patient self-report of events and errors, small cohorts, or examine few outcomes. METHODS: A cross-sectional survey was administered immediately following the January 2017 American Board of Surgery In-training Examination to clinically active general surgery residents to assess resident wellness and self-reported error. Postoperative patient outcomes were ascertained using a validated national clinical data registry. Associations were examined using multivariable logistic regression models. RESULTS: Over a 6-month period, 22.5% of residents reported committing a near miss medical error, and 6.9% reported committing a harmful medical error. Residents were more likely to report a harmful medical error if they reported frequent burnout symptoms [odds ratio 2.71 (95% confidence interval 2.16-3.41)] or poor psychiatric well-being [odds ratio 2.36 (95% confidence interval 1.92-2.90)]. However, there were no significant associations between program-level resident wellness and any of the independently, objectively measured postoperative American College of Surgeons National Surgical Quality improvement Program outcomes examined. CONCLUSIONS: Although surgical residents with poor wellness were more likely to self-report a harmful medical error, there was not a higher rate of objectively reported outcomes for surgical patients treated at hospitals with higher rates of burnout or poor psychiatric well-being.