A Virtual Patients Ensemble Approach for Predicting Surgical Complications

A Virtual Patients Ensemble Approach for Predicting Surgical Complications | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search A Virtual Patients Ensemble Approach for Predicting Surgical Complications Yair Neuman , Yochai Cohen , Yiftach Neuman doi: https://doi.org/10.1101/2025.09.21.25336262 Yair Neuman 1 The Functor Lab Ben-Gurion University of the Negev , Israel Find this author on Google Scholar Find this author on PubMed Search for this author on this site Yochai Cohen 2 Gilasio Coding, Tel-Aviv , Israel Find this author on Google Scholar Find this author on PubMed Search for this author on this site Yiftach Neuman 3 indivi.ai, Tel-Aviv , Israel Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: yneuman{at}bgu.ac.il Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract AI has shown promise in predicting surgical complications, but most existing models estimate overall risk levels rather than identifying the specific complications an individual patient may develop. We present an AI agent that uses a Virtual Patients Ensemble (VPE) approach to generate individualized predictions of surgical complications from unstructured case descriptions. The agent applies structured reasoning to extract diagnoses, surgical procedures, and risk factors from clinical narratives. From this profile, it generates a cohort of N virtual patients , each a plausible variation of the original case. This ensemble captures uncertainty in patient-specific risk factors and grounds LLM-based clinical reasoning in individualized clinical scenarios. For each virtual patient, the agent predicts the most likely complications, and a final distribution is presented over the virtual patients. The agent was evaluated on 1440 case reports from the PMC-Patients dataset, of which 186 met the inclusion criteria. Predictive performance was compared with both null-hypothesis expectations and baseline LLM predictions. The agent correctly identified 32% of the observed complications, significantly outperforming the null-hypothesis baseline and a baseline prediction generated by the LLM. Unlike risk calculators or machine-learning models trained on population averages, this approach derives predictions directly from a patient’s clinical profile, generating a VPE to predict specific complications rather than general risk levels. The results suggest that ensemble-based, patient-centered simulation can support clinical decision-making by offering interpretable, individualized predictions. Prospective validation is required before integration into practice. We thus provide surgeons with an app for experimenting with the agent and providing feedback for improvement. Introduction Surgeons face significant challenges when anticipating postoperative complications. A patient’s clinical profile often includes a complex interplay of medical history, comorbidities, and procedure-related risks, making individualized prediction difficult. Traditional approaches, such as the ACS Surgical Risk Calculator 1 , have demonstrated value compared with physician intuition 2 , yet these models that focus on risk are constrained by linear assumptions in estimating log odds and limited capacity to capture complex interactions. In recent years, artificial intelligence (AI) has emerged as a promising tool for surgical risk prediction 3 . Most efforts have focused on machine-learning models trained on population-level datasets 4 . While these models achieve reasonable accuracy, their reliance on population averages can limit their ability to provide patient-specific predictions. Systematic reviews of AI for predicting postoperative complications after major surgery highlight both potential and variability in performance 5 , 6 , 7 , with recent work noting “variable rates of precision” 8 and “inconsistent results” 9 . More recently, large language models (LLMs) have shown potential for medical applications 10 , 11 because of their ability to perform context-sensitive 12 and probabilistic reasoning 13 . A few studies demonstrate their capacity to classify complications or apply clinical grading systems. For example 14 , showed that ChatGPT-4 can grade postoperative complications using the Clavien–Dindo classification with promising accuracy. Yet applications in surgery remain at an early stage 15 , 16 . We present an AI agent for predicting surgical complications. Instead of estimating general risk levels or restricting predictions to a predefined and closed set of complications, the agent generates an open-ended distribution of likely complications. This process involves the generation of a Virtual Patients Ensemble (VPE): a cohort of simulated patient profiles derived from a single clinical case. By varying risk factors across simulated patients, the AI agent explores plausible outcome scenarios and identifies complications most likely to occur. Therefore, the AI agent combines the clinical breadth of LLM reasoning with individualized patient simulation, offering interpretable predictions that align with precision medicine. The AI agent The agent, called BIBAS, accepts the patient’s description as input. Prompt 1 simulates a panel of experts to identify a diagnosis leading to a surgical procedure that produces an intra- or postoperative complication. Prompt 1.5 identifies the MeSH Tree Number of the complication. Prompt 2 identifies the text preceding the surgical procedure, and Prompt 3 identifies in this text those of the patient’s risk factors that may increase the likelihood of a surgical complication. Prompts 4A-4C simulate the VPE. The simulation uses the patient’s specific profile, which includes diagnosis, surgical procedure, risk factors, and expected timing of a complication. It clones up to 50 virtual patients from this profile. We generated 50 virtual patients as a pragmatic choice, balancing computational efficiency with performance, but the number of patients is a variable parameter in the model and can be changed by the user when running the simulation. The justification for generating 50 virtual patients appears in the appendix. All virtual patients share the same diagnosis and surgical procedure, but differ in the parameters of the risk factors. For each virtual patient, the system predicts the most likely complications that share the same time frame as the observed complication (i.e., immediate or early complication versus late or long-term). The system then summarizes the distribution of the identified complications across the VPE and presents them in descending order of probability. Justifying the Virtual Patients Ensemble approach The justification for using a VPE is fourfold: 1. Patient-centered modeling rather than population averages Traditional risk calculators and most ML models are trained on population-level data. While they perform well statistically, they struggle with individualization. A VPE starts from the unique patient profile (diagnosis, surgical plan, and risk factors) and generates multiple patient-specific simulations. This approach aligns more closely with precision medicine by focusing on the individual rather than the average . 2. Captures complex interactions between risk factors Surgical outcomes emerge from interactions (e.g., advanced age × infection × comorbidities) that are often oversimplified by linear models and even by some ML approaches. By varying the patient’s risk factor parameters across virtual patients, the VPE explores a wide range of plausible scenarios, enabling the emergence of complication patterns that depend on non-linear interactions. 3. Transparency and clinical interpretability Surgeons may find it difficult to trust or interpret black-box predictions. A VPE produces a distribution of complications across virtual patients—for example, “40% of the ensemble developed surgical site infection.” This frequency-based output is intuitive for clinicians and mirrors their own reasoning when they mentally simulate “patients like this one” to anticipate complications. 4. Bridging LLM knowledge with clinical data Large language models (LLMs) bring broad clinical knowledge but lack grounding in structured, patient-specific variability. A VPE bridges this gap: it anchors the LLM’s knowledge and general reasoning in concrete patient features while allowing exploration of uncertainty across multiple plausible scenarios. The result is a prediction framework that is both knowledge-driven and individualized. Illustrating the agent’s outcomes The patient descriptions include the case of Maria. This case provides the following information: “Maria G., a 74-year-old female, presented to the emergency department with a 2-day history of severe epigastric pain radiating to the back, associated with nausea and multiple episodes of non-bloody vomiting.” CT imaging confirmed the diagnosis of “acute gallstone pancreatitis, with signs of peripancreatic fat stranding and inflammation.” The surgical intervention was “laparoscopic cholecystectomy,” followed on postoperative day 2 by the complication, viz., a “postoperative localized intra-abdominal infection”. BIBAS identified the following risk factors: [Age 74 years, Demographic] [Acute pancreatitis, Anatomical/Pathological] [Elevated white blood cell count, Laboratory/Physiological] [Peripancreatic inflammation, Anatomical/Pathological] And through the VPE simulation, the top predicted complications were: Surgical site infection Bile duct injury Postoperative bleeding Evaluation Sample We used the PMC-Patients dataset 17 and ran the agent on 1440 patient descriptions. The latter were randomly sampled from each of six age groups (see Table 1 in the appendix), with equal weight given to the proportion of males and females. The data set was collected from the case reports in the PMC archive. Case reports naturally emphasize the rare, unusual, or unique 18 , 19 . Therefore, the use of this dataset is particularly challenging, given that we analyze free texts rather than a structured dataset, and texts dealing with case reports where the distribution of complications may deviate significantly from baseline predictions. Procedure All prompts appear in the appendix. GPT-5 was the LLM used for all prompts, and we ran the agent through the API. We applied Prompt 1 and identified cases of a diagnosis leading to surgical intervention and subsequent complications. In this study, we analyzed only surgical procedures that led to a complication. Next, we selected cases where the text preceding the surgical procedure (i.e., the context) was longer than 400 characters. This ensured enough contextual information to identify risk factors. 186 patient descriptions were selected for the final analysis. A case report may include more than one surgical intervention and more than one complication resulting from a given intervention. For brevity, we analyzed the first surgical intervention mentioned in the text, followed by a single complication, and focused on the first complication resulting from the surgery. Analysis and Results We performed three tests to evaluate the performance of BIBAS. Primary Evaluation We looked for a match between the MeSH tree number of the observed complication and the tree numbers of each of the ten highest-ranked complications predicted by the agent. We focused on ten complications because the patient descriptions may be biased toward rarer complications. We tested whether one of the ten predicted complications was the same as the observed complication (i.e., a HIT). There were 59 cases of HIT (32%) with 95% CI [0.25, 0.39], and 40 of these cases (68%) were in the top three predicted complications. We compared this hit rate to a random guess and a baseline LLM prediction. To compare this hit rate to a random guess, we identified 269 unique complications predicted by the agent. The chance of a match between a randomly selected complication from this list and the observed complication was therefore 0.004. The chance of a match between the observed complication and one of the ten predicted complications was p = 0.04 with 95% CI [0.02, 0.079]. The difference between the proportion of complications identified by BIBAS and the one hypothesized under the null hypothesis was found to be statistically significant (z = 7.03, p < 0.001). The effect size (Cohen’s h = 0.80) was found to exceed Cohen’s convention for a medium effect size (h = 0.5). We used Prompt 5 to identify the most likely complication to follow the surgical procedure (MOST_LIKELY). This predicted complication served as a baseline prediction generated by the LLM. Next, we looked for a match between this LLM prediction and what was observed. There were 24 cases of HIT (13%). This means that the proportion of cases identified by BIBAS is about 2.5 times greater than the proportion identified by the LLM baseline prediction. The difference between the hit rate of BIBAS and the baseline prediction generated by the LLM was statistically significant (z = 4.39, p < 0.001, Cohen’s h = 0.46) Secondary Evaluation 1 We measured how similar the meaning of each predicted complication was to the meaning of the observed complication. To address this challenge, we represented the meaning of a complication by using its text embedding. More specifically, we used OpenAI’s text-embedding-3-large model, which includes 3072 dimensions 20 . Next, we measured the cosine similarity between each of the ten highest predicted complications and the observed complications and selected the predicted complication that scored highest (HIGHEST_PREDICTED). This prediction is BIBAS’s best guess in terms of semantic similarity. Next, we measured the cosine similarity between the embedding of MOST_LIKELY and the observed complication. We hypothesized that, if BIBAS performs better than a general prediction generated by the LLM, then its best prediction, in terms of semantic similarity to the observed (i.e., HIGHEST_PREDICTED), should be closer to the meaning of the observed complication than the meaning of the MOST_LIKELY. The average HIGHEST_PREDICTED score was 0.49(0.12), with 95% CI [0.48–0.51], and for the MOST_LIKELY it was 0.41 (0.11) with a 95% CI [0.39–0.42]. We computed the difference between the two scores, expecting it to be significantly higher than 0. To test this hypothesis, we used the one-sample Wilcoxon signed-rank test. In 75% of the cases, the difference in the similarity score was positive and in favor of HIGHEST_PREDICTED. The test was found to be statistically significant (V = 13830, p < 0.001) and with a medium-large effect size (rank-biserial correlation = 0.59). Secondary Evaluation 2 Finally, through Prompt 6, we tested whether the HIGHEST_PREDICTED and the observed complication share the same complication category. We used 13 categories of complications. Running Prompt 6, we found a match in 138 out of 186 cases (74%). Given the expected proportion of complications sharing the same category under the null hypothesis (i.e., 0.08), the difference was found to be statistically significant (z = 34.016, p < 0.001) and with a very large effect size (Cohen’s h = 1.51). When we compared the proportion of surgical categories shared by MOST_LIKELY and the observed (i.e., 36%) to the shared categories with HIGEST_PREDICTED (i.e., 74%), the difference was found to be statistically significant (z = 7.36, p < 0.001, Cohen’s h = 0.78). Conclusions This study introduces a novel framework for predicting surgical complications using a VPE. To our knowledge, this is the first approach that combines large language model reasoning with an ensemble of patient-specific simulations to move beyond general +risk estimation and predict concrete complications . Unlike existing methods that rely on population-level averages or static ML models, the VPE framework bootstraps directly from a patient’s clinical description, generates a cohort of individualized virtual patients, and identifies complication patterns emerging from complex interactions of risk factors. The significance of this contribution lies in its dual novelty: (1) reframing surgical complication prediction as an individualized, scenario-based simulation problem rather than a population-based probability estimate, and (2) operationalizing LLM knowledge in a structured, clinically transparent way. Nevertheless, the agent requires human validation, which was beyond the budget scope of the current paper, and further validation in prospective and structured clinical datasets is required. Moreover, running the simulation is computationally expensive, so we provide a “lightweight” prompt in the appendix that can be run directly from the UI. This version performs well, although it does not generate the VPE, but it works as if it does (see the appendix). We also provide a link to an app running the agent from the API 21 . In sum, the agent and the approach have the potential to augment surgical decision-making and broaden the toolkit of predictive methods available to clinicians. In its present state, the agent can be used by surgeons as an assisting tool for mapping likely complications. Data Availability All data produced in the present work are contained in the manuscript https://doi.org/10.6084/m9.figshare.30075007 Declaration of interests Y.N. Y.C. and Yi. N. declare they have no conflicts of interest. Funding No funding Data sharing The dataset generated and analyzed for this study is freely available at: https://doi.org/10.6084/m9.figshare.30075007 Authors’ contribution Y.N. proposed the original approach. Y.N., Y.C. Yi. N. developed the methods. Y.C. Yi. N. wrote the code for the analysis. Y.N., Y.C., Yi. N. wrote the manuscript. Yi. N. designed the app. Acknowledgments The authors would like to thank Prof. Irun Cohen, MD, for his helpful feedback and Dr. Dan Assaf, MD, for constructive talks. Footnotes yohai{at}gilasio.com yiftachn{at}gmail.com References 1. ↵ Bilimoria , K. Y. , Liu , Y. , Paruch , J. L. , Zhou , L. , Kmiecik , T. E. , Ko , C. Y. , & Cohen , M. E. ( 2013 ). Development and evaluation of the universal ACS NSQIP surgical risk calculator: a decision aid and informed consent tool for patients and surgeons . Journal of the American College of Surgeons , 217 ( 5 ), 833 – 842 . OpenUrl CrossRef PubMed 2. ↵ Marwaha , J. S. , et al. , ( 2023 ). Quantifying the prognostic value of preoperative surgeon intuition: comparing surgeon intuition and clinical risk prediction as derived from the American College of Surgeons NSQIP risk calculator . Journal of the American College of Surgeons , 236 ( 6 ), 1093 – 1103 . OpenUrl PubMed 3. ↵ Assaf , D. , Gutman , Y. A. , Neuman , Y. , Segal , G. , Amit , S. , Gefen-Halevi , S. , … & Tirosh , A. ( 2020 ). Utilization of machine-learning models to accurately predict the risk for critical COVID-19 . Internal and emergency medicine , 15 ( 8 ), 1435 – 1443 . OpenUrl 4. ↵ Bihorac , A. , Ozrazgat-Baslanti , T. , Ebadi , A. , Motaei , A. , Madkour , M. , Pardalos , P. M. , … & Momcilovic , P. ( 2019 ). MySurgeryRisk: development and validation of a machine-learning risk algorithm for major complications and death after surgery . Annals of Surgery , 269 ( 4 ), 652 – 662 . OpenUrl CrossRef PubMed 5. ↵ Hopkins , B. , Mazmudar , A. , Driscoll , C. , Svet , M. , Goergen , J. , Kelsten , M. , Shlobin , N. , Kesavabhotla , K. , Smith , Z. , & Dahdaleh , N. ( 2020 ). Using artificial intelligence (AI) to predict postoperative surgical site infection: A retrospective cohort of 4046 posterior spinal fusions . Clinical Neurology and Neurosurgery , 192 . doi: 10.1016/j.clineuro.2020.105718 . OpenUrl CrossRef 6. ↵ Stam , W. T. , Goedknegt , L. K. , Ingwersen , E. W. , Schoonmade , L. J. , Bruns , E. R. , & Daams , F. ( 2022 ). The prediction of surgical complications using artificial intelligence in patients undergoing major abdominal surgery: a systematic review . Surgery , 171 ( 4 ), 1014 – 1021 . OpenUrl PubMed 7. ↵ Mohammadi , I. , Firouzabadi , S. , Hosseinpour , M. , Akhlaghpasand , M. , Hajikarimloo , B. , Zeraatian-Nejad , S. , & Nia , S. ( 2024 ). Using artificial intelligence to predict post-operative outcomes in congenital heart surgeries: a systematic review . BMC Cardiovascular Disorders , 24 . doi: 10.1186/s12872-024-04336-6 . OpenUrl CrossRef 8. ↵ Hassan , A. M. , Rajesh , A. , Asaad , M. , Nelson , J. A. , Coert , J. H. , Mehrara , B. J. , & Butler , C. E. ( 2023 ). Artificial intelligence and machine learning in prediction of surgical complications: current state, applications, and implications . The American Surgeon , 89 ( 1 ), 25 – 30 . OpenUrl PubMed 9. ↵ Devault-Tousignant , C. , Harvie , M. , Bissada , É. , Christopoulos , A. , Tabet , P. , Guertin , L. , Bahig , H. , & Ayad , T. ( 2024 ). The use of artificial intelligence in reconstructive surgery for head and neck cancer: a systematic review . European archives of oto-rhino-laryngology: official journal of the European Federation of Oto-Rhino-Laryngological Societies (EUFOS): affiliated with the German Society for Oto-Rhino-Laryngology - Head and Neck Surgery . doi: 10.1007/s00405-024-08663-4 . OpenUrl CrossRef 10. ↵ Eriksen , A. V. , Möller , S. , & Ryg , J. ( 2024 ). Use of GPT-4 to diagnose complex clinical cases . Nejm Ai , 1 ( 1 ), AIp2300031 . OpenUrl 11. ↵ Meng , X. , Yan , X. , Zhang , K. , Liu , D. , Cui , X. , Yang , Y. , … & Tang , Y. D. ( 2024 ). The application of large language models in medicine: A scoping review . Iscience , 27 ( 5 ). 12. ↵ Neuman , Y. ( 2024 ). AI for Understanding Context . N.Y .: Springer . 13. ↵ Paruchuri , A. , Garrison , J. , Liao , S. , Hernandez , J. , Sunshine , J. , Althoff , T. , … & McDuff , D. ( 2024 ). What are the odds? Language models are capable of probabilistic reasoning . arXiv preprint arXiv: 2406.12830 . 14. ↵ Staubli , S. M. , Walker , H. L. , Saner , F. , Salinas , C. H. , Broering , D. C. , Malagò , M. , … & Raptis , D. A. ( 2025 ). Decoding the Clavien-Dindo classification: Artificial intelligence (AI) as a novel tool to grade postoperative complications . Annals of Surgery , 281 ( 2 ), 273 – 279 . OpenUrl 15. ↵ Varghese , C. , Harrison , E. M. , O’Grady , G. , & Topol , E. J. ( 2024 ). Artificial intelligence in surgery . Nature Medicine , 30 ( 5 ), 1257 – 1268 . OpenUrl PubMed 16. ↵ Hsueh , J. Y. , Nethala , D. , Singh , S. , Linehan , W. M. , & Ball , M. W. ( 2024 , September ). Investigating the clinical reasoning abilities of large language model GPT-4: an analysis of postoperative complications from renal surgeries . In Urologic Oncology: Seminars and Original Investigations (Vol. 42 , No. 9 , pp. 292 – e1 ). Elsevier . OpenUrl 17. ↵ Zhao , Z. , Jin , Q. , Chen , F. , Peng , T. , & Yu , S. ( 2023 ). A large-scale dataset of patient summaries for retrieval-based clinical decision support systems . Scientific data , 10 ( 1 ), 909 . OpenUrl PubMed 18. ↵ Jenicek , M. ( 2001 ). Clinical Case Reporting in Evidence-Based Medicine . Oxford : Butterworth-Heinemann . 19. ↵ Nissen , T. , & Wynn , R. ( 2014 ). The clinical case report: a review of its merits and limitations . BMC research notes , 7 ( 1 ), 264 . OpenUrl PubMed 20. ↵ https://platform.openai.com/docs/models/text-embedding-3-large 21. ↵ https://apps.indivi.ai/s/bibas-calculator View the discussion thread. Back to top Previous Next Posted September 22, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following A Virtual Patients Ensemble Approach for Predicting Surgical Complications Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share A Virtual Patients Ensemble Approach for Predicting Surgical Complications Yair Neuman , Yochai Cohen , Yiftach Neuman medRxiv 2025.09.21.25336262; doi: https://doi.org/10.1101/2025.09.21.25336262 Share This Article: Copy Citation Tools A Virtual Patients Ensemble Approach for Predicting Surgical Complications Yair Neuman , Yochai Cohen , Yiftach Neuman medRxiv 2025.09.21.25336262; doi: https://doi.org/10.1101/2025.09.21.25336262 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Surgery Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4435) Dentistry and Oral Medicine (444) Dermatology (382) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1509) Epidemiology (15229) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6600) Geriatric Medicine (668) Health Economics (997) Health Informatics (4536) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (541) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15916) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (146) Nephrology (667) Neurology (6599) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1144) Occupational and Environmental Health (957) Oncology (3332) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (691) Primary Care Research (711) Psychiatry and Clinical Psychology (5447) Public and Global Health (9232) Radiology and Imaging (2198) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a00c37c45ba52bef',t:'MTc3OTYyNDg0Mg=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();