The Current State of Robot-Assisted Minimally Invasive Esophagectomy (RAMIE): Outcomes from the Upper GI International Robotic Association (UGIRA) Esophageal Registry.

BACKGROUND: Robot-assisted minimally invasive esophagectomy (RAMIE) is increasingly adopted in centers worldwide, with ongoing refinements to enhance results. This study aims to assess the current state of RAMIE worldwide and to identify potential areas for improvement. METHODS: This descriptive study analyzed prospective data from esophageal cancer patients who underwent transthoracic RAMIE in Upper GI International Robotic Association (UGIRA) centers. Main endpoints included textbook outcome rate, surgical techniques, and perioperative outcomes. Analyses were performed separately for intrathoracic (Ivor-Lewis) and cervical anastomosis (McKeown), divided into three time cohorts (2016-2018, 2019-2020, 2021-2023). A sensitivity analysis was conducted with cases after the learning curve (> 70 cases). RESULTS: Across 28 UGIRA centers, 2012 Ivor-Lewis and 1180 McKeown procedures were performed. Over the time cohorts, textbook outcome rates were 39%, 48%, and 49% for Ivor-Lewis, and 49%, 63%, and 61% for McKeown procedures, respectively. Fully robotic procedures accounted for 66%, 51%, and 60% of Ivor-Lewis procedures, and 53%, 81%, and 66% of McKeown procedures. Lymph node yield showed 27, 30, and 30 nodes in Ivor-Lewis procedures, and 26, 26, and 34 nodes in McKeown procedures. Furthermore, high mediastinal lymphadenectomy was performed in 65%, 43%, and 37%, and 70%, 48%, and 64% of Ivor-Lewis and McKeown procedures, respectively. Anastomotic leakage rates were 22%, 22%, and 16% in Ivor-Lewis cases, and 14%, 12%, and 11% in McKeown cases. Hospital stay was 13, 14, and 13 days for Ivor-Lewis procedures, and 12, 9, and 11 days for McKeown procedures. In Ivor-Lewis and McKeown, respectively, the sensitivity analysis revealed textbook outcome rates of 43%, 54%, and 51%, and 47%, 64%, and 64%; anastomotic leakage rates of 28%, 18%, and 15%, and 13%, 11%, and 10%; and hospital stay of 11, 12, and 12 days, and 10, 9, and 9 days. CONCLUSIONS: This study demonstrates favorable outcomes over time in achieving textbook outcome after RAMIE. Areas for improvement include a reduction of anastomotic leakage and shortening of hospital stay.

Proctored adoption of robotic hiatus hernia surgery: outcomes and learning curves in a high-volume UK centre.

BACKGROUND: The adoption of new surgical technologies is inevitably accompanied by a learning curve. With the increasing adoption of robotic techniques in benign foregut surgery, it is imperative to define optimal learning pathways, to ensure a clinically safe introduction of such a technique. The aim of this study was to assess the learning curve for robotic hiatal hernia repair with a pre-defined adoption process and proctoring. METHODS: The learning curve was assessed in four surgeons in a high-volume tertiary referral centre, performing over a 100 hiatal hernia repairs annually. The robotic adoption process included simulation-based training and a multi-day wet lab-based course, followed by robotic operations proctored by robotic upper GI experts. CUSUM analysis was performed to assess changes in operating time in sequential cases. RESULTS: Each surgeon (A, B, C and D) performed between 22 and 32 cases, including a total of 109 patients. Overall, 40 cases were identified as 'complex' (36.7%), including 16 revisional cases (16/109, 14.7%). With CUSUM analysis inflection points for operating time were seen after 7 (surgeon B) to 15 cases (surgeon B). CONCLUSION: The learning curve for robotic laparoscopic fundoplication may be as little as 7-15 cases in the setting of a clearly organized learning pathway with proctoring. By integrating these organized learning pathways learning curves may be shortened, ensuring patient safety, preventing detrimental outcomes due to longer learning curves, and accelerating adoption and integration of novel surgical techniques.

Specialist-led urgent cholecystectomy for acute gallstone disease.

BACKGROUND: Despite overwhelming evidence of the clinical and financial benefit of urgent cholecystectomy, there is variable enthusiasm and uptake across the UK. In 2014, following the First National Emergency Laparotomy Audit Organisational Report, we implemented a specialist-led urgent surgery service, whereby all patients with gallstone-related pathologies were admitted under the direct care of specialist upper gastrointestinal surgeons. We have analysed 5 years of data to investigate the results of this service model. METHODS: Computerised operating theatre records were interrogated to identify all patients within a 5-year period undergoing cholecystectomy. Patient demographics, admission details, length of stay, duration of surgery, and complications were analysed. RESULTS: Between 01/01/2016 and 31/12/2020, a total of 4870 cholecystectomies were performed; 1793 (36.8%) were urgent cases and 3077 (63.2%) were elective cases. All cases were started laparoscopically; 25 (0.5%) were converted to open surgery-14 of 1793 (0.78%) urgent cases and 11 of 3077 (0.36%) elective cases. Urgent cholecystectomy took 20 min longer than elective surgery (median 74 versus 52 min). No relevant difference in conversion rate was observed when urgent cholecystectomy was performed within 2 days, between 2 and 4 days, or greater than 4 days from admission (P = 0.197). Median total hospital stay was 4 days. CONCLUSION: Urgent laparoscopic cholecystectomy is safe and feasible in most patients with acute gall bladder disease. Surgery under the direct care of upper gastrointestinal specialist surgeons is associated with a low conversion rate, low complication rate, and short hospital stay. Timing of surgery has no effect on conversion rate or complication rate.

Catalytic Enantioselective Addition of Alkylzirconium Reagents to Aliphatic Aldehydes.

A catalytic methodology for the enantioselective addition of alkylzirconium reagents to aliphatic aldehydes is reported here. The versatile and readily accessible chiral Ph-BINMOL ligand, in the presence of Ti(OiPr)4 and a zinc salt, facilitates the reaction, which proceeds under mild conditions and is compatible with functionalized nucleophiles. The alkylzirconium reagents are conveniently generated in situ by hydrozirconation of alkenes with the Schwartz reagent. This work is a continuation of our previous work on aromatic aldehydes.

Backstepping Mechanism of Kinesin-1.

Kinesin-1 is an ATP-driven molecular motor that transports cellular cargo along microtubules. At low loads, kinesin-1 almost always steps forward, toward microtubule plus ends, but at higher loads, it can also step backward. Backsteps are usually 8 nm but can be larger. These larger backward events of 16 nm, 24 nm, or more are thought to be slips rather than steps because they are too fast to consist of multiple, tightly coupled 8-nm steps. Here, we propose that not only these larger backsteps, but all kinesin-1 backsteps, are slips. We show first that kinesin waits before forward steps for less time than before backsteps and detachments; second, we show that kinesin waits for the same amount of time before backsteps and detachments; and third, we show that by varying the microtubule type, we can change the ratio of backsteps to detachments without affecting forward stepping. Our findings indicate that backsteps and detachments originate from the same state and that this state arises later in the mechanochemical cycle than the state that gives rise to forward steps. To explain our data, we propose that, in each cycle of ATP turnover, forward kinesin steps can only occur before Pi release, whereas backslips and detachments can only occur after Pi release. In the scheme we propose, Pi release gates access to a weak binding K⋅ADP-K⋅ADP state that can slip back along the microtubule, re-engage, release ADP, and try again to take an ATP-driven forward step. We predict that this rescued detachment pathway is key to maintaining kinesin processivity under load.

Stereoselective synthesis of 1,3-disubstituted dihydroisoquinolines vial-phenylalanine-derived dihydroisoquinoline N-oxides.

The preparation of chiral pool-derived nitrone 3 and its use in the protecting-group free, stereoselective synthesis of a range of 1,3-disubstituted tetrahydroisoquinolines is described. Grignard reagent additions to nitrone 3 yielded trans-1,3-disubstituted N-hydroxytetrahydroisoquinolines 6 with good levels of selectivity, while 1,3-dipolar cycloadditions to this nitrone provided access to 3-(2-hydroxyalkyl)isoquinolines 12 as single diastereomers.

Catalytic Enantioselective Addition of Organozirconium Reagents to Aldehydes.

A catalytic enantioselective addition reaction of alkylzirconium species to aromatic aldehydes is reported. The reaction, facilitated by a chiral nonracemic diol ligand complex with Ti(OiPr)4, proceeds under mild and convenient conditions, and no premade organometallic reagents are required since the alkylzirconium nucleophiles are generated in situ by hydrozirconation of alkenes with the Schwartz reagent. The methodology is compatible with functionalized nucleophiles and a broad range of aromatic aldehydes.

FAST in the deployed military setting.

INTRODUCTION: Focused assessment with sonography in trauma (FAST) is historically an effective method of assessing the patient in the trauma bay in order to aid decision-making and optimise patient outcomes. However, in the UK civilian practice, the use of FAST may decline given a recent change in National Institute for Health and Care Excellence guidance as a result of improvement in CT availability and resuscitation techniques. METHOD: In the Role 3 Medical Treatment Facility, Camp Bastion, 187 patients with trauma who received FAST in the trauma bay in 2014 were reviewed to determine the accuracy of FAST in the deployed environment. RESULTS: The data demonstrates the sensitivity and specificity of FAST to be 75% and 99.3%, respectively. CONCLUSIONS: This study demonstrates that FAST is accurate on operations. FAST is provided by the integrated radiologist as part of damage control radiology, which gives the team leader rapid diagnostic information to improve decision-making and ultimately patient outcomes. CT is heavily utilised in civilian practice; however, the military operates in a different environment often with multiple casualties and limited access to CT, as a result, portable ultrasound will continue to be a valuable tool on operations if used properly. The next challenge is to develop and maintain this high diagnostic accuracy in future deployments where the memories of our prior success may fade.

Web-Based Video Assessments of Operative Performance for Remote Telementoring.

INTRODUCTION: Performance-based feedback is critical to surgical skills acquisition. Barriers of geography and time limit trainees' access to expert mentorship. In this study, we hypothesized that telementoring using an asynchronous, web-based video interface would allow trainees to receive systematic feedback from expert mentors despite these barriers. MATERIALS AND METHODS: Between October 2014 and October 2016, 18 surgeons in Brazil, Dominican Republic, Haiti, and Paraguay underwent in-person training in Lichtenstein for hernioplasty or laparoscopic total extraperitoneal inguinal hernia repair. After initial training, surgeons submitted 6- to 12-month interval operative videos for expert review. Expert surgeons reviewed each video using the Surgus web platform with performance metrics adapted from the Operative Performance Rating Scale (OPRS). The time required to perform video review, number of freeform comments, mean OPRS scores, and variance of OPRS scores among telementors was assessed. RESULTS: A total of 18 surgeons submitted 20 operative videos, and three expert surgeons reviewed each video using the Surgus platform. The median time to perform video review was 20 minutes. Median number of freeform verbal comments was eight. Mean OPRS overall performance scores were 3.9 ± 0.9 (scale of five). Mean variance in scoring among telementors for overall performance was 0.25 (maximum 5.29), suggesting a high degree of concordance. CONCLUSIONS: Video-based assessments had a high degree of concordance among expert raters. Asynchronous performance reviews by telementors offer opportunities for longitudinal feedback that overcome geographical, material, and temporal disparities. This platform offers a means of sharing expertise in surgical training, continuing education, credentialing, and global health.

Early postoperative weight loss predicts maximal weight loss after sleeve gastrectomy and Roux-en-Y gastric bypass.

BACKGROUND: Previous studies show that 'poor responders' to Roux-en-Y gastric bypass (RYGBP) may be identified on the basis of early postoperative weight loss. Early identification of poor responders could allow earlier provision of postoperative behavioural and/or intensive lifestyle interventions and enhance their maximal weight loss. Our aim was to investigate whether early postoperative weight loss predicts the maximal weight loss response after RYGBP and sleeve gastrectomy (SG). METHODS: We undertook a retrospective cross-sectional study of 1,456 adults who underwent either RYGBP (n = 918) or SG (n = 538) as a primary procedure in one of two European centres. Postoperative weight loss was expressed as weight loss velocity (WLV) and percentage weight loss. Linear regression analyses were performed to determine the association of early postoperative weight loss with maximal %WL, including adjustment for baseline variables. RESULTS: There was marked variability in maximal %WL following both RYGBP (mean 32.9 %, range 4.1-60.9 %) and SG (mean 26.2 %, range 1.1-58.3 %). WLV 3-6 months postoperatively was more strongly associated with maximal %WL (r (2) = 0.32 for RYGBP and r (2) = 0.26 for SG, P < 0.001 for both) than either WLV 0-6 weeks or 6 weeks to 3 months postoperatively (r (2) = 0.14 and 0.10 for RYGBP, respectively; r (2) = 0.18 and 0.21 for SG, respectively; P < 0.001 for all). Multiple linear regression analysis, including baseline variables of age, sex, preoperative BMI, type 2 diabetes, ethnicity, and bariatric centre, revealed that 3-6 month WLV was an independent predictor of maximal %WL in both SG and RYGBP groups (standardised ß-coefficients 0.51 and 0.52, respectively; P < 0.001 for both). CONCLUSIONS: There is a marked variability in weight loss response following RYGBP and SG. Early postoperative weight loss can be used to identify patients whose predicted weight loss trajectories are suboptimal. Early targeting of poor responders with more intensive postoperative lifestyle and behavioural support could potentially enhance their weight loss response.

Dyslipidaemia and the military patient.

Dyslipidaemias refer to abnormal levels of circulating lipids and high cholesterol and is related to cardiovascular death. This paper examines the types and prevalence of dyslipidaemia with specific reference to a military population and describes who to target in screening strategies used to detect people with abnormal lipid profiles. The diagnostic limits for a diagnosis of dyslipidaemia are explored. Finally, medical management of hyperlipidaemia is discussed and how this may affect military medical grading.

Management of a Nipple Injury Causing Nipple Bifurcation and Infection.

Military physicians working in resource-constrained environments, including reduced access to care, must evaluate for any circumstances that would necessitate an aeromedical evacuation to a higher level of care; this is particularly important in foreign countries. Due to these circumstances, military physicians must be resourceful while maintaining expertise to stabilize and treat any medical issue whether or not evacuation is requested. A 26-year-old female requested urgent care evaluation after she endured an accidental traumatic nipple stud removal at an austere military base located in Egypt. Upon initial evaluation at the base clinic, a left nipple bifurcation was identified. The case was complicated by a subsequent spontaneously draining bacterial infection at the base of the bifurcation, which resolved without further complication following antibiotic course and daily monitoring. Furthermore, adhesive strips were applied daily to approximate the nipple, which was critical to the restoration of gross anatomy. As such, the nipple bifurcation was held in place utilizing adhesive strips, thereby producing counterforce to facilitate healing by primary intention for two weeks resulting in a favorable cosmetic outcome. Ultimately, the patient's clinical course resulted in a well-healed and well-approximated linear scar on her left nipple. Long-term distal milk duct function is unable to be assessed outside of lactation periods or advanced imaging, but given the location and extent of the injury, a plastic surgery evaluation and imaging were deferred. This case report outlines the management considerations, literature search, and treatment course of a rare, unreported injury type in an austere, military environment along with consideration of future research.

Three-arm robotic cholecystectomy: a novel, cost-effective method of delivering and learning robotic surgery in upper GI surgery.

Cholecystectomy is one of the commonest performed surgeries worldwide. With the introduction of robotic surgery, the numbers of robot-assisted cholecystectomies has risen over the past decade. Despite the proven use of this procedure as a training operation for those surgeons adopting robotics, the consumable cost of routine robotic cholecystectomy can be difficult to justify in the absence of evidence favouring or disputing this approach. Here, we describe a novel method for performing a robot-assisted cholecystectomy using a "three-arm" technique on the newer, 4th generation, da Vinci system. Whilst maintaining the ability to perform precision dissection, this method reduces the consumable cost by 46%. The initial series of 109 procedures proves this procedure to be safe, feasible, trainable and time efficient.

The missing toxic link: Exposure of non-target native marsupials to second-generation anticoagulant rodenticides (SGARs) suggest a potential route of transfer into apex predators.

Anticoagulant rodenticides (ARs) are used globally to control rodent pests. Second-generation anticoagulant rodenticides (SGARs) persist in the liver and pose a significant risk of bioaccumulation and secondary poisoning in predators, including species that do not generally consume rodents. As such, there is a clear need to understand the consumption of ARs, particularly SGARs, by non-target consumers to determine the movement of these anticoagulants through ecosystems. We collected and analysed the livers from deceased common brushtail possums (Trichosurus vulpecula) and common ringtail possums (Pseudocheirus peregrinus), native Australian marsupials that constitute the main diet of the powerful owl (Ninox strenua), an Australian apex predator significantly exposed to SGAR poisoning. ARs were detected in 91 % of brushtail possums and 40 % of ringtail possums. Most of the detections were attributed to SGARs, while first-generation anticoagulant rodenticides (FGARs) were rarely detected. SGAR concentrations were likely lethal or toxic in 42 % of brushtail possums and 4 % of ringtail possums with no effect of age, sex, or weight detected in either species. There was also no effect of the landscape type possums were from, suggesting SGAR exposure is ubiquitous across landscapes. The rate of exposure detected in these possums provides insight into the pathway through which ARs are transferred to one of their key predators, the powerful owl. With SGARs entering food-webs through non-target species, the potential for bioaccumulation and broader secondary poisoning of predators is significantly greater and highlights an urgent need for routine rodenticide testing in non-target consumers that present as ill or found deceased. To limit their impact on ecosystem stability the use of SGARs should be significantly regulated by governing agencies.

Learning curves for adoption of robotic bariatric surgery: a systematic review of safety, efficiency and clinical outcomes.

Robotic bariatric surgery may overcome challenges associated with laparoscopy, potentially achieving technically superior results. This review aims to summarise current literature reporting on learning curves for surgeons newly adopting robotic bariatrics and implications for safety, efficiency and outcomes. A systematic review was performed in line with the PRISMA guidelines. Electronic databases PubMed and MEDLINE were searched and articles reporting on learning curves in robotic bariatric surgery were identified. Studies that reported changes in outcome over time, or learning curves for surgeons newly adopting robotic bariatric surgery were included in this review. Eleven studies reporting on 1237 patients were included in this review. Most surgeons reported prior bariatric surgical experience. Differences were noted regarding the approach and adoption of robotics. Ten studies found significant reduction in operative time, with the shortest learning curve of 11 cases. Reporting of clinical outcomes was limited. Three studies reported statistically significant improvement in outcomes after the learning curve. Long-term outcomes were in line with current literature, though none assessed differences between learning curve groups. Reported learning curves in robotic bariatric surgery is variable, with limited reporting of clinical outcomes. With appropriate mentorship, surgeons can improve efficiency, safety and clinical outcomes, maximising the benefits of minimally invasive surgery.

Kinesin's moonwalk.

Kinesin-1 is a single-molecule walking machine, driven by ATP turnover. Recent optical trapping experiments show that pulling backwards on a walking kinesin-1 molecule causes the mechanical walking action to reverse, while the coupled chemical cycle of ATP turnover continues, apparently, to run forwards -- kinesin can moonwalk. Individual forward- and back-steps are fast, and each appears to be a single event, complete in a few tens of microseconds, with no substeps. Between steps, kinesin pauses, waiting for the next ATP to arrive. Several lines of evidence indicate that during these between-step dwells, only one of the two heads is strongly attached to the microtubule. The position of the other head during the dwells is less certain, and more controversial.

Silent killers? The widespread exposure of predatory nocturnal birds to anticoagulant rodenticides.

Anticoagulant rodenticides (ARs) influence predator populations and threaten the stability of ecosystems. Understanding the prevalence and impact of rodenticides in predators is crucial to inform conservation planning and policy. We collected dead birds of four nocturnal predatory species across differing landscapes: forests, agricultural, urban. Liver samples were analysed for eight ARs: three First Generation ARs (FGARs) and five SGARs (Second Generation ARs). We investigated interspecific differences in liver concentrations and whether landscape composition influenced this. FGARs were rarely detected, except pindone at low concentrations in powerful owls Ninox strenua. SGARs, however, were detected in every species and 92 % of birds analysed. Concentrations of SGARs were at levels where potential toxicological or lethal impacts would have occurred in 33 % of powerful owls, 68 % of tawny frogmouths Podargus strigoides, 42 % of southern boobooks N. bookbook and 80 % of barn owls Tyto javanica. When multiple SGARs were detected, the likelihood of potentially lethal concentrations of rodenticides increased. There was no association between landscape composition and SGAR exposure, or the presence of multiple SGARs, suggesting rodenticide poisoning is ubiquitous across all landscapes sampled. This widespread human-driven contamination in wildlife is a major threat to wildlife health. Given the high prevalence and concentrations of SGARs in these birds across all landscape types, we support the formal consideration of SGARs as a threatening process. Furthermore, given species that do not primarily eat rodents (tawny frogmouths, powerful owls) have comparable liver rodenticide concentrations to rodent predators (southern boobook, eastern barn owl), it appears there is broader contamination of the food-web than anticipated. We provide evidence that SGARs have the potential to pose a threat to the survival of avian predator populations. Given the functional importance of predators in ecosystems, combined with the animal welfare impacts of these chemicals, we propose governments should regulate the use of SGARs.

Intra-operative ultrasound assessment of the biliary tree during robotic cholecystectomy.

Image-guided assessment of bile ducts and associated anatomy during laparoscopic cholecystectomy can be achieved with intra-operative cholangiography (IOC) or laparoscopic ultrasound (LUS). Rates of robotically assisted cholecystectomy (RC) are increasing and herein we describe the technique of intra-corporeal biliary ultrasound during RC using the Da Vinci system. For intraoperative evaluation of the biliary tree during RC, in cases of suspected choledocholithiasis, the L51K Ultrasound Probe (Hitachi, Tokyo, Japan) is used. The extrahepatic biliary tree is scanned along its length, capitalising on the benefits of the full range of motion offered by the articulated robotic instruments and integrated ultrasonic image display using TileProTM software. Additionally, this technique avoids the additional time and efforts required to undock and re-dock the robot that would otherwise be required for selective IOC or LUS. The average time taken to perform a comprehensive evaluation of the biliary tree, from the hepatic ducts to the ampulla of Vater, is 164.1 s. This assessment is supplemented by Doppler ultrasound, which is used to fully delineate anatomy of the porta hepatis, and accurate measurements of the biliary tree and any ductal stones can be taken, allowing for contemporaneous decision making and management of ductal pathologies. Biliary tract ultrasound has been shown to be equal to IOC in its ability to diagnose choledocholithiasis, but with the additional benefits of being quicker and having higher completion rates. We have described our practice of using biliary ultrasound during robotically assisted cholecystectomy, which is ergonomically superior to LUS, accurate and reproducible.

Assessing the Accuracy and Reliability of AI-Generated Medical Responses: An Evaluation of the Chat-GPT Model.

Background: Natural language processing models such as ChatGPT can generate text-based content and are poised to become a major information source in medicine and beyond. The accuracy and completeness of ChatGPT for medical queries is not known. Methods: Thirty-three physicians across 17 specialties generated 284 medical questions that they subjectively classified as easy, medium, or hard with either binary (yes/no) or descriptive answers. The physicians then graded ChatGPT-generated answers to these questions for accuracy (6-point Likert scale; range 1 - completely incorrect to 6 - completely correct) and completeness (3-point Likert scale; range 1 - incomplete to 3 - complete plus additional context). Scores were summarized with descriptive statistics and compared using Mann-Whitney U or Kruskal-Wallis testing. Results: Across all questions (n=284), median accuracy score was 5.5 (between almost completely and completely correct) with mean score of 4.8 (between mostly and almost completely correct). Median completeness score was 3 (complete and comprehensive) with mean score of 2.5. For questions rated easy, medium, and hard, median accuracy scores were 6, 5.5, and 5 (mean 5.0, 4.7, and 4.6; p=0.05). Accuracy scores for binary and descriptive questions were similar (median 6 vs. 5; mean 4.9 vs. 4.7; p=0.07). Of 36 questions with scores of 1-2, 34 were re-queried/re-graded 8-17 days later with substantial improvement (median 2 vs. 4; p<0.01). Conclusions: ChatGPT generated largely accurate information to diverse medical queries as judged by academic physician specialists although with important limitations. Further research and model development are needed to correct inaccuracies and for validation.

Accuracy and Reliability of Chatbot Responses to Physician Questions.

Importance: Natural language processing tools, such as ChatGPT (generative pretrained transformer, hereafter referred to as chatbot), have the potential to radically enhance the accessibility of medical information for health professionals and patients. Assessing the safety and efficacy of these tools in answering physician-generated questions is critical to determining their suitability in clinical settings, facilitating complex decision-making, and optimizing health care efficiency. Objective: To assess the accuracy and comprehensiveness of chatbot-generated responses to physician-developed medical queries, highlighting the reliability and limitations of artificial intelligence-generated medical information. Design, Setting, and Participants: Thirty-three physicians across 17 specialties generated 284 medical questions that they subjectively classified as easy, medium, or hard with either binary (yes or no) or descriptive answers. The physicians then graded the chatbot-generated answers to these questions for accuracy (6-point Likert scale with 1 being completely incorrect and 6 being completely correct) and completeness (3-point Likert scale, with 1 being incomplete and 3 being complete plus additional context). Scores were summarized with descriptive statistics and compared using the Mann-Whitney U test or the Kruskal-Wallis test. The study (including data analysis) was conducted from January to May 2023. Main Outcomes and Measures: Accuracy, completeness, and consistency over time and between 2 different versions (GPT-3.5 and GPT-4) of chatbot-generated medical responses. Results: Across all questions (n = 284) generated by 33 physicians (31 faculty members and 2 recent graduates from residency or fellowship programs) across 17 specialties, the median accuracy score was 5.5 (IQR, 4.0-6.0) (between almost completely and complete correct) with a mean (SD) score of 4.8 (1.6) (between mostly and almost completely correct). The median completeness score was 3.0 (IQR, 2.0-3.0) (complete and comprehensive) with a mean (SD) score of 2.5 (0.7). For questions rated easy, medium, and hard, the median accuracy scores were 6.0 (IQR, 5.0-6.0), 5.5 (IQR, 5.0-6.0), and 5.0 (IQR, 4.0-6.0), respectively (mean [SD] scores were 5.0 [1.5], 4.7 [1.7], and 4.6 [1.6], respectively; P = .05). Accuracy scores for binary and descriptive questions were similar (median score, 6.0 [IQR, 4.0-6.0] vs 5.0 [IQR, 3.4-6.0]; mean [SD] score, 4.9 [1.6] vs 4.7 [1.6]; P = .07). Of 36 questions with scores of 1.0 to 2.0, 34 were requeried or regraded 8 to 17 days later with substantial improvement (median score 2.0 [IQR, 1.0-3.0] vs 4.0 [IQR, 2.0-5.3]; P < .01). A subset of questions, regardless of initial scores (version 3.5), were regenerated and rescored using version 4 with improvement (mean accuracy [SD] score, 5.2 [1.5] vs 5.7 [0.8]; median score, 6.0 [IQR, 5.0-6.0] for original and 6.0 [IQR, 6.0-6.0] for rescored; P = .002). Conclusions and Relevance: In this cross-sectional study, chatbot generated largely accurate information to diverse medical queries as judged by academic physician specialists with improvement over time, although it had important limitations. Further research and model development are needed to correct inaccuracies and for validation.