Developing a machine learning model for predicting 30-day major adverse cardiac and cerebrovascular events in patients undergoing noncardiac surgery

Developing a machine learning model for predicting 30-day major adverse cardiac and cerebrovascular events in patients undergoing noncardiac surgery | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (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],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Developing a machine learning model for predicting 30-day major adverse cardiac and cerebrovascular events in patients undergoing noncardiac surgery Jung-Won Suh, Ju-Seung Kwun, Houng-beom Ahn, Si-Hyuck Kang, Sooyoung Yoo, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4524391/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract To reduce unnecessary delays and manage medical costs efficiently for low-risk patients undergoing noncardiac surgery, we developed a predictive model for major adverse cardiac and cerebrovascular events (MACCE) using the OMOP Common Data Model (CDM) and machine learning algorithms. This retrospective study collected data from 46,225 patients at Seoul National University Bundang Hospital and 396,424 patients at Asan Medical Center. Patients aged 65 or older undergoing non-cardiac, non-emergency surgeries with at least 30 days of observation were included. Machine learning models were developed using the OHDSI open-source patient-level prediction package in R version 4.1.0. All models outperformed the Revised Cardiac Risk Index (RCRI), with the random forest model achieving an AUROC of 0.817 in external validation and demonstrating moderate calibration. Key predictors included previous diagnoses and laboratory measurements, highlighting their importance in perioperative risk prediction. Our model shows promise for improving clinical practice and reducing medical costs. Health sciences/Medical research/Outcomes research Health sciences/Health care/Disease prevention/Preventive medicine Health sciences/Health care/Prognosis Perioperative risk evaluation noncardiac surgery prediction models machine learning common data model Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction The annual count of noncardiac surgeries worldwide exceeds 300 million each year, and major adverse cardiac and cerebrovascular events (MACCE) remain a leading contributor to perioperative morbidity and mortality. (1-4) The rising prevalence of perioperative morbidity and mortality, driven by factors such as an aging population, necessitates precise preoperative prediction. (5, 6) However, the predictive accuracy of traditional assessment tools is not consistently high, and various tools are employed at different physicians’ discretion. (7) Traditionally, the Revised Cardiac Risk Index (RCRI), which comprises six equally weighted components, is extensively used to mitigate major perioperative cardiac complications owing to its simplicity and relatively high predictability of in-hospital major adverse cardiac events (MACE) or cardiovascular-related death.(8) However, the index developed over 2 decades ago has certain challenges, including limited external validation and reduced precision in vascular surgery. (9) These factors may modestly impact its effectiveness in predicting clinical outcomes following noncardiac surgeries in practical clinical environments. (10) Subsequent predictive tools such as the American College of Surgeons (ACS), National Surgical Quality Improvement Project (NSQIP), and NSQIP Myocardial Infarction or Cardiac Arrest (MICA), developed after RCRI, also show strong performance in predicting postoperative MICA. However, these tools pose challenges for clinicians in practical clinical use because they rely on subjective predictors, leading to low interrater reliability.(11) Given these challenges, our research leverages the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM), which converts diverse observational databases into a standardized format, enabling access to comprehensive patient data.(12-14) Building upon this standardized data foundation, we aimed to develop a machine learning-based prediction model that harnesses the power of machine learning to identify patterns and relationships within extensive patient data, ultimately improving personalized risk prediction. (15, 16) Focusing on enhancing the accuracy of perioperative event prediction, our research utilized CDM data from real-world electronic health records (EHRs) as a foundation for developing a machine learning-based model to provide a more advanced and precise tool for personalized risk prediction in noncardiac surgeries. Results Study population A total of 46,225 patients were enrolled at the SNUBH and 396,424 were enrolled at the AMC, with an average age of 72.9 years at both hospitals (Table 1). More male than female patients were enrolled at both institutions, with 25,573 males (55.3%) at SNUBH and 232,522 males (58.7%) at AMC. Hypertension was the most common comorbidity, affecting 62.0% and 54.6% of patients at SNUBH and AMC, respectively. However, at SNUBH, cerebrovascular disease was more common (15.9% at SNUBH and 10.2% at AMC). In contrast, at AMC, congestive heart failure (2.1% at SNUBH and 4.9% at AMC) and ischemic heart disease (8.5% at SNUBH and 10.9% at AMC) had a higher representation. Preoperative laboratory results were within normal ranges, with patients with a creatinine level of 2.0 mg/dL or higher accounting for 3,434 (7.4%) at SNUBH and 74,594 (18.8%) at AMC. Regarding medications, the AMC data showed a higher proportion of patients registered with aspirin, P2Y12 inhibitors, beta-blockers, RAS inhibitors, calcium channel blockers, statins, and insulin treatment. There was a significant difference between the two hospitals regarding the type of surgery and post-noncardiac surgery MACCE within 30 days across all categories. Surgeries with a risk exceeding 1% are presented in Table 1; those with unmapped names were classified as unspecified. Post-noncardiac surgery MACCE within 30 days included myocardial infarction, which occurred in 4.9% of patients at SNUBH and 6.3% at AMC, cardiac arrest/shock in 0.6% and 0.3%, heart failure in 0.7% and 0.6%, and stroke in 1.7% and 1.5%, respectively. In-hospital deaths accounted for 0.9% and 3.0% of the deaths at SNUBH and AMC, respectively. Prediction model performance The prediction model discrimination for internal and external validation is presented in Table 2. The numbers of patients included in the training, test, and external validation sets of the SNUBH model who met the inclusion criteria are presented in Supplementary Table 1. When assessed using the RCRI score and compared with five other machine learning prediction models, all machine learning models outperformed the RCRI model with a higher AUROC for MACCE prediction than the RCRI score (AUROC 0.704) (Figure 1A). The RF generally showed the best overall performance in internal and external validations across outcomes with moderate calibration among the five predictive models. The AUROC of this model was 0.897 (0.883–0.911) and 0.817 (0.815–0.819) for internal and external validations, respectively (Figure 1A and Table 2), and the AUPRC was 0.095 (Figure 1B). In addition, it demonstrated outstanding calibration, showing strong alignment with the average predicted probability on the calibration plot. (Figure 1C). Predictors In the prediction model, we assessed the relative importance of various covariates based on their values (Figure 2). Rather than identifying a single outstanding covariate, the analysis grouped covariates into similar thematic clusters. Predominantly, predictors associated with the patient’s underlying medical history were relatively high in the developed prediction model. These include ischemic heart disease, traumatic and non-traumatic brain injury, heart failure, heart disease, and cerebral infarction. The model highlights the importance of the measurement predictors. Preoperative laboratory measurements revealed that hemoglobin, creatinine, albumin, CK-MB, and erythrocyte sedimentation rate played crucial roles. Among the medication predictors, antithrombotic agents and beta-blockers were notably prominent, whereas the significance of the others was less pronounced. Furthermore, although important, the significance of the type of surgery did not appear to be as substantial as expected when compared with other factors in the model. Additionally, we developed prediction models by recombining the data and considering previous diagnoses, medication, type of surgery, and measurement data in various combinations (Supplementary Table 2). However, none of the additional recombination models outperformed the original models. Nevertheless, these models generally exhibited superior predictability compared with RCRI, except for the recombination model that excluded the previous diagnosis group, which yielded results comparable to or slightly inferior to those of RCRI (Supplementary Figure 1-3). Discussion In this study, we developed and evaluated an advanced perioperative risk prediction model using a CDM-based machine learning approach. Our model demonstrated superior predictive accuracy compared with traditional models, such as the RCRI score. This study provides several key insights and implications. Advances in machine learning for extensive dataset analysis have led to increased interest in applying patient-level prediction and offer the potential for medical practice to consider personalized risks as part of clinical decision-making. (17) The adoption of the OMOP CDM has streamlined the transformation of diverse concept domains, encompassing medical conditions, drugs, procedures, and measurements derived from health record systems or reported information into labeled analytic data. This transformation ensures semantic and syntactic interoperability, enhancing the extraction of prediction variables and facilitating seamless integration across various healthcare data sources. (18, 19) In addition, the standardized data across different institutions allowed a fair evaluation of the predictive performance of the models by extensive external validation. 17 Therefore, our model supports existing preoperative evaluation guidelines and enables open dissemination that can be extensively validated across OHDSI collaborator networks. The well-structured and labeled dataset improves algorithms in supervised machine learning but sometimes leads to overfitting, which prevents the model’s generalization to fit the observed data well. (20, 21) To overcome the challenge of overfitting, we employed a feature selection method as one of several techniques to identify and prioritize factors essential for the learning process. (22, 23) Additionally, we employed a feature selection method to create new combinations of thematic clusters, including medical conditions, drugs, types of procedures, and measurements, to assess their relative importance in predicting adverse outcomes following surgery. The recombination model, which included past medical conditions and previous laboratory data, exhibited a notably high predictive accuracy. Our model’s ability to discern the varying importance of these factors in real clinical contexts underscores the importance of focusing on patient histories and prior laboratory results during preoperative evaluations. (24) This approach aligns with physicians’ subjective assessments in clinical settings and provides a flexible alternative to traditional methods that may not fully accommodate each patient’s unique circumstances. (25) The practical implications of our research extend to potential time and cost savings in clinical settings. Risk assessments often lead to unnecessary procedures or examinations, such as echocardiography, cardiac computed tomography, and/or cardiac stress tests, being performed on patients. (26) These tests, even when not closely associated with the patient’s post-surgical outcomes, contribute to ongoing wastage in overall medical costs. (27, 28) Our model, with its high predictive accuracy, is poised to reduce the number of unnecessary tests performed and contribute to medical cost savings. Moreover, our model could reduce waiting times for patients as unnecessary consultations and tests may be minimized, ultimately mitigating the challenges posed by healthcare system congestion and assisting patients in undergoing surgery at an appropriate time. In the future, with precise preoperative predictability, we plan to use our model to proactively identify individuals at risk of post-surgical complications and ensure appropriate post-operative management. In an aging population, where surgical mortality and morbidity rates are increasing, (29) this approach can serve as a viable solution to effectively mitigate these challenges. While using the OMOP CDM for data standardization, large healthcare datasets frequently exhibited inherent data inconsistencies and missing values. Variations in data quality and completeness among healthcare institutions may have introduced noise into the model. However, we mitigated this limitation by leveraging data from two of the largest tertiary hospitals in South Korea, where the data quality and quantity were sufficiently substantial to minimize the impact of missing values. This rich dataset enabled the development of a robust machine-learning model, enhancing its potential for accurate perioperative risk prediction. Second, including data from two distinct tertiary hospitals introduced potential heterogeneity in patient populations and clinical practices. This variability could have influenced the model performance, potentially contributing to variations in predictive accuracy, such as relatively low AUPRC values. However, from a machine-learning perspective, analyzing diverse datasets can be advantageous for improving future predictive capabilities. Additionally, the versatility of using frameworks such as the CDM ensures that the scalability of this model is not restricted to specific hospital settings, which can enhance its applicability in various healthcare contexts. In this study, we successfully developed a high-performance machine learning-based preoperative prediction model by using the standardized data format of the OMOP CDM. This approach offers the potential for improved clinical decision-making and extensive external validation across healthcare institutions. In the future, our research has practical implications for potential time and cost savings in clinical settings by reducing unnecessary procedures, tests, and consultations, ultimately addressing healthcare system congestion and improving patient surgical timing. Methods This retrospective, observational network study involved a multidisciplinary team of cardiologists, medical informatics specialists, and data scientists. This study was approved by the Institutional Review Boards (IRB) of Seoul National University Bundang Hospital (IRB No. 2208-772-906) and Asan Medical Center (IRB No. 2022-1547). The requirement for written informed consent was waived owing to the retrospective nature of the study design and the use of de-identified data. Data sources The data sources used in this study were selected and standardized to ensure the integrity and compatibility of the collected information. The EHRs were converted to the OMOP CDM, and source codes were mapped to standard vocabularies, including the systematized nomenclature of medicine clinical terms (SNOMED CT). (18, 30) Data analysis was conducted using the observational health data sciences and informatics (OHDSI) open-source patient-level prediction package, which was purposefully designed for standardized analysis and harmonization with the OMOP CDM. These specialized tools have facilitated efficient data processing and analysis across different datasets within the OHDSI data network. (31) This collaborative aspect enhances the comparability and generalizability of the prediction models, making them applicable to diverse healthcare settings. To develop and evaluate our prediction models, we retrospectively used patient data from two tertiary hospitals, Seoul National University Bundang Hospital (SNUBH) and Asan Medical Center (AMC), which are recognized for their substantial CDM datasets. The SNUBH dataset contains data from 46,225 patients who underwent noncardiac surgery between January 2003 and December 2020, and the AMC dataset includes data from 396,424 patients who underwent noncardiac surgery between January 2010 and December 2020. This extensive dataset included a comprehensive array of demographic information and detailed preoperative characteristics, including diagnosis codes, underlying diseases, laboratory test results, medications, type of surgery, and clinical outcomes from the EHR system (Table 1). Study design and target cohort We conducted a retrospective analysis of patients aged 65 years or older who underwent noncardiac surgeries at two independent tertiary hospitals. Age was determined at the time of surgery. We excluded patients who had undergone cardiac or emergency surgery within 3 days of a hospital visit and those who did not have a sufficient observation period of less than 30 days (Figure 3). The prediction time (t=0) and start date of the time-at-risk window for prediction were set as surgery dates. The end date of the time-at-risk window for clinical outcomes was 30 days after surgery. The data collection period for the predictors was defined as 3–365 days before the start date of the time-at-risk window (Figure 4). We adjusted the observational time frame to collect baseline characteristics and preoperative laboratory measurements within a narrower window of 3–30 days before the onset of the time-at-risk period. This adjustment ensured that the data accurately represented the patient's condition at the beginning of the time-at-risk period. Conclusions In this study, we successfully developed a high-performance machine learning-based preoperative prediction model by using the standardized data format of the OMOP CDM. This approach offers the potential for improved clinical decision-making and extensive external validation across healthcare institutions. In the future, our research has practical implications for potential time and cost savings in clinical settings by reducing unnecessary procedures, tests, and consultations, ultimately addressing healthcare system congestion and improving patient surgical timing. Declarations Acknowledgment This study was funded by the Seoul National University Bundang Hospital Research Fund (Grant No. 14-2022-0023). The funder played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript. Author contributions Conceptualization: Jung-Won Suh Formal analysis: Sooyoung Yoo, Seok Kim, Wongeun Song Investigation: Jung-Won Suh, Ju-Seung Kwun, Houng-Beom Ahn, Si-Hyuck Kang, Sooyoung Yoo Validation: Ji Seon Oh, Junho Hyun, Gakyoung Baek Writing - original draft: Ju-Seung Kwun, Houng-Beom Ahn Writing - review & editing: Jung-Won Suh, Si-Hyuck Kang, Sooyoung Yoo Competing interest All authors declare no financial or non-financial competing interests. Data availability Data underlying this article cannot be shared publicly because of data protection agreements but can be available from the corresponding author on reasonable request. Code availability The Code that supports the findings of this study are available from the corresponding author upon reasonable request. Analysis to process and analyze data was generated with Python 3, R version’s 4.1.0 Pending patents/patent applications Patent applicant (whether author or institution): Seoul National University Bundang Hospital Name of the inventor(s): Jung-Won Suh, Ju-Seung Kwun, Houng-Beom Ahn, Sooyoung Yoo Application number: 10-2024-0011788 Status of the application: provisional application status Specific aspect of manuscript covered in the patent application: Machine Learning Model for Predicting 30-Day MACCE in Noncardiac Surgery Patients References Smilowitz, N. R. et al. Perioperative major adverse cardiovascular and cerebrovascular events associated with noncardiac surgery. JAMA Cardiol . 2 , 181-187 (2017). Sabaté, S. et al. Incidence and predictors of major perioperative adverse cardiac and cerebrovascular events in non-cardiac surgery. Br. J. Anaesth . 107 , 879-890 (2011). Puelacher, C. et al. Perioperative myocardial injury after noncardiac surgery: incidence, mortality, and characterization. Circulation . 137 , 1221-1232 (2018). Writing Committee for the VISION Study Investigators, Devereaux, P. J. et al. Association of postoperative high-sensitivity troponin levels with myocardial injury and 30-day mortality among patients undergoing noncardiac surgery. JAMA . 317 , 1642-1651 (2017). Van Klei, W. A. et al. Role of history and physical examination in preoperative evaluation. Eur. J. Anaesthesiol . 20 , 612-618 (2003). Alkire, B. C. et al. Global access to surgical care: a modelling study. Lancet Glob. Health . 3 , e316-e323 (2015). Cohen, M. E., Bilimoria, K. Y., Ko, C. Y., Richards, K. & Hall, B. L. Effect of subjective preoperative variables on risk-adjusted assessment of hospital morbidity and mortality. Ann. Surg . 249 , 682-689 (2009). Lee, T. H. et al. Derivation and prospective validation of a simple index for prediction of cardiac risk of major noncardiac surgery. Circulation . 100 , 1043-1049 (1999). Gupta, P. K. et al. Development and validation of a risk calculator for prediction of cardiac risk after surgery. Circulation . 124 , 381-387 (2011). Brasher, P. M. & Beattie, W. S. Adjusting clinical prediction rules: an academic exercise or the potential for real world clinical applications in perioperative medicine? Can. J. Anesth. 56 , 190-193 (2009). Bilimoria, K. Y. et al. Development and evaluation of the universal ACS NSQIP surgical risk calculator: a decision aid and informed consent tool for patients and surgeons. J. Am. Coll. Surg . 217 , 833-842.e1-3 (2013). Voss, E. A. et al. Feasibility and utility of applications of the common data model to multiple, disparate observational health databases. J. Am. Med. Inform. Assoc . 22 , 553-564 (2015). Madigan, D. et al. A systematic statistical approach to evaluating evidence from observational studies. Annu. Rev. Stat. Its Appl . 1 , 11-39 (2014). Stang, P. E. et al. Advancing the science for active surveillance: rationale and design for the Observational Medical Outcomes Partnership. Ann. Intern. Med . 153 , 600-606 (2010). Meskó, B. & Görög, M. A short guide for medical professionals in the era of artificial intelligence. npj Digit. Med . 3 , 126 (2020). Rajkomar, A., Dean, J. & Kohane, I. Machine learning in medicine. N. Engl. J. Med . 380 , 1347-1358 (2019). Goldstein, B. A., Navar, A. M., Pencina, M. J. & Ioannidis, J. P. Opportunities and challenges in developing risk prediction models with electronic health records data: a systematic review. J. Am. Med. Inform. Assoc . 24 , 198-208 (2017). Reps, J. M., Schuemie, M. J., Suchard, M. A., Ryan, P. B. & Rijnbeek, P. R. Design and implementation of a standardized framework to generate and evaluate patient-level prediction models using observational healthcare data. J. Am. Med. Inform. Assoc . 25 , 969-975 (2018). Hripcsak, G. et al. Observational health data sciences and informatics (OHDSI): opportunities for observational researchers. Stud. Health Technol. Inform . 216 , 574-578 (2015). Junqué de Fortuny, E., Martens, D. & Provost, F. Predictive modeling with big data: is bigger really better? Big Data . 1 , 215-226 (2013). Ying, X. An overview of overfitting and its solutions. J Phys Conf S . (IOP Publishing, 2019). Hawkins, D. M. The problem of overfitting. J. Chem. Inf. Comput. Sci . 44 , 1-12 (2004). Bagherzadeh-Khiabani, F. et al A tutorial on variable selection for clinical prediction models: feature selection methods in data mining could improve the results. J. Clin. Epidemiol . 71 , 76-85 (2016). Michota, F. A. & Frost, S. D. The preoperative evaluation: use the history and physical rather than routine testing. Cleve Clin. J. Med . 71 , 63-70 (2004). Glance, L. G. et al. Impact of the choice of risk model for identifying low-risk patients using the 2014 American College of Cardiology/American Heart Association Perioperative Guidelines. Anesthesiology . 129 , 889-900 (2018). Johansson, T. et al. Effectiveness of non-cardiac preoperative testing in non-cardiac elective surgery: a systematic review. Br. J. Anaesth . 110 , 926-939 (2013). Bryson, G. L., Wyand, A. & Bragg, P. R. Preoperative testing is inconsistent with published guidelines and rarely changes management. Can. J. Anaesth . 53 , 236-241 (2006). Augoustides, J. G., Neuman, M. D., Al-Ghofaily, L. & Silvay, G. Preoperative cardiac risk assessment for noncardiac surgery: defining costs and risks. J. Cardiothorac. Vasc. Anesth . 27 , 395-399 (2013). Ferrando, A. et al. Guidelines for preoperative assessment: impact on clinical practice and costs. Int. J. Qual. Health Care . 17 , 323-329 (2005). 32. Weiser, T. G. et al. Estimate of the global volume of surgery in 2012: an assessment supporting improved health outcomes. Lancet . 385 , S11 (2015). Overhage, J. M., Ryan, P. B., Reich, C. G., Hartzema, A. G. & Stang, P. E. Validation of a common data model for active safety surveillance research. J. Am. Med. Inform. Assoc . 19 , 54-60 (2012). FitzHenry, F. et al. Creating a common data model for comparative effectiveness with the observational medical outcomes partnership. Appl. Clin. Inform . 6 , 536-547 (2015). Tables Table 1. Baseline characteristics SNUBH AMC P-value Number of populations, N 46,225 396,424 Age, years, mean (SD) 72.9 (5.35) 72.9 (6.08) 0.014 Sex, N (%) < 0.001 Male 25,573 (55.3%) 232,522 (58.7%) Female 20,652 (44.7%) 163,902 (41.3%) BMI, kg/m 2 (SD) 23.7 (3.39) 23.3 (3.70) < 0.001 Underlying disease Hypertension 28,641 (62.0%) 216,440 (54.6%) < 0.001 Diabetes 12,815 (27.7%) 104,269 (26.3%) < 0.001 Dyslipidemia 12,078 (26.1%) 129,601 (32.7%) < 0.001 Congestive heart failure 961 (2.1%) 19,613 (4.9%) < 0.001 Chronic kidney disease 2,588 (5.6%) 45,095 (11.4%) < 0.001 Cerebrovascular disease 7,363 (15.9%) 40,628 (10.2%) < 0.001 Ischemic heart disease 3,939 (8.5%) 43,302 (10.9%) < 0.001 Preoperative lab results White blood cell, 10 3 /μL (SD) 7.0 (2.64) 7.5 (3.56) < 0.001 Hemoglobin, g/dL (SD) 12.8 (1.89) 11.5 (2.24) < 0.001 Platelet, 10 3 / μL (SD) 233.9 (75.86) 214.9 (91.24) < 0.001 Sodium, mmol/L (SD) 139.9 (3.42) 138.2 (4.45) < 0.001 Potassium, mmol/L (SD) 4.3 (0.46) 4.2 (0.52) < 0.001 BUN, mg/dL (SD) 18.3 (9.67) 23.1 (17.06) < 0.001 Creatinine, mg/dL (SD) 1.1 (0.95) 1.3 (1.43) 2.0mg/Dl, N (%) 3,434 (7.4%) 74,594 (18.8%) < 0.001 Total cholesterol, mg/dL (SD) 169.0 (41.55) 146.2 (45.78) < 0.001 LDL, mg/dL (SD) 92.2 (30.90) 91.3 (36.14) 0.165 Albumin, g/dL (SD) 4.0 (0.53) 3.2 (0.71) < 0.001 AST, IU/L (SD) 27.7 (20.58) 31.7 (30.72) < 0.001 ALT, IU/L (SD) 24.1 (21.70) 25.4 (28.08) < 0.001 Glucose, mg/dL (SD) 122.2 (43.59) 133.0 (55.15) < 0.001 PT, INR (SD) 1.0 (0.19) 1.1 (0.31) < 0.001 aPTT, sec (SD) 36.6 (5.83) 31.1 (8.13) < 0.001 Medications Aspirin 11,900 (25.7%) 139,029 (35.1%) < 0.001 P2Y12 inhibitor 6,263 (13.5%) 71,436 (18.0%) < 0.001 Beta blocker 9,678 (20.9%) 179,264 (45.2%) < 0.001 RAS inhibitor 12,357 (26.7%) 161,016 (40.6%) < 0.001 Calcium channel blocker 15,771 (34.1%) 227,915 (57.5%) < 0.001 Statin 11,734 (25.4%) 129,236 (32.6%) < 0.001 Insulin treatment 8,603 (18.6%) 138,392 (34.9%) < 0.001 Type of surgery Intermediate risk: 1-5% Intraperitoneal: splenectomy, hiatal hernia repair, cholecystectomy 1,429 (3.1%) 29,072 (7.3%) < 0.001 Carotid symptomatic (CEA or CAS) 17 (0.0%) 655 (0.2%) < 0.001 Peripheral arterial angioplasty 12 (0.0%) 24,861 (6.3%) < 0.001 Head and neck surgery 2,090 (4.5%) 60,919 (15.4%) < 0.001 Neurological or orthopedic: major (hip and spine surgery) 3,309 (7.2%) 16,878 (4.3%) < 0.001 Urological or gynecological: major 243 (0.5%) 5,350 (1.3%) < 0.001 Renal transplant 23 (0.0%) 2,684 (0.7%) < 0.001 Intra-thoracic: non-major 1,596 (3.5%) 50,247 (12.7%) 5% Aortic and major vascular surgery 2,028 (4.4%) 53,886 (13.6%) < 0.001 Open lower limb revascularization or amputation or thromboembolectomy 250 (0.5%) 7,786 (2.0%) < 0.001 Duodeno-pancreatic surgery 247 (0.5%) 7,216 (1.8%) < 0.001 Liver section, bile duct surgery 373 (0.8%) 19,183 (4.8%) < 0.001 Esophagectomy 75 (0.2%) 7,795 (2.0%) < 0.001 Repair of perforated bowel 1,557 (3.4%) 105,525 (26.6%) < 0.001 Adrenal resection 66 (0.1%) 1,191 (0.3%) < 0.001 Pneumonectomy 1,026 (2.2%) 13,751 (3.5%) < 0.001 Pulmonary or liver transplant 27 (0.1%) 8,656 (2.2%) < 0.001 Unspecified 494 (1.1%) 74,067 (18.7%) < 0.001 Outcome MACE 2,256 (4.9%) 24,853 (6.3%) < 0.001 Myocardial infarction 907 (0.6%) 5,603 (0.3%) < 0.001 Cardiac arrest/shock 35 (0.1%) 168 (0.0%) 0.002 Heart failure 308 (0.7%) 2,310 (0.6%) 0.029 Stroke 799 (1.7%) 6,017 (1.5%) 0.001 Death (In-hospital) 419 (0.9%) 11,875 (3.0%) < 0.001 *Values are given as the mean ± standard deviation or number with percentage (%) Table 2. Predictability of five machine learning prediction models Prediction model SNUBH AMC AUROC (95% CI) Train Test External Validation MACCE Random forest 0.985 (0.980-0.984) 0.897 (0.883-0.911) 0.817 (0.815-0.819) Gradient boosting machine 0.935 (0.996-0.997) 0.898 (0.885-0.912) 0.825 (0.823-0.828) Lasso Logistic Regression 0.906 (0.952-0.956) 0.892 (0.878-0.906) 0.812 (0.810-0.815) AdaBoost 0.907 (0.902-0.907) 0.887 (0.873-0.902) 0.785 (0.782-0.788) DecisionTree 0.895 (0.875-0.883) 0.776 (0.750-0803) 0.663 (0.659-0.667) Additional Declarations (Not answered) Supplementary Files SupplementaryTablePreopFinal.pdf Supplementary Table SupplementaryFigurePreopFinal.pdf Supplementary Figure Supplementaryfiles.pdf Supplementary Table, Figure Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4524391","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":325312104,"identity":"266141cd-7ac4-4118-b577-16091cdbb472","order_by":0,"name":"Jung-Won Suh","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAv0lEQVRIiWNgGAWjYFACHgYDBgMbIIOx8QAJWgrSQFoaiNfCwPDhMJhJnBb5GbkHinkMztutbT8MtKXGJpqgFoMbeQnGPAa3k7edSQRqOZaW20BQi0SOgXEOUIvZAaAWxobDhLXIzwBrOZdsdv4hkVoYboC1HLAzu0GsLQZn3hgY/zFITjC7AbQlgRi/yLfnmBnO+GNnb3Y+/eGDDzU2RDiMgYHNAEgkglUmEKEcBJgfAAl7IhWPglEwCkbBSAQAU6lFsVPBuAUAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-0397-6071","institution":"Seoul National University Bundang Hospital","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Jung-Won","middleName":"","lastName":"Suh","suffix":""},{"id":325312106,"identity":"73f6be50-6d07-47cf-b26b-132211226a29","order_by":1,"name":"Ju-Seung Kwun","email":"","orcid":"","institution":"Seoul National University Bundang Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Ju-Seung","middleName":"","lastName":"Kwun","suffix":""},{"id":325312109,"identity":"9f66b372-f54a-406c-b587-f8313d2838cf","order_by":2,"name":"Houng-beom Ahn","email":"","orcid":"","institution":"Seoul National University Bundang Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Houng-beom","middleName":"","lastName":"Ahn","suffix":""},{"id":325312111,"identity":"de526cab-0790-4309-b358-0a09a65d5205","order_by":3,"name":"Si-Hyuck Kang","email":"","orcid":"","institution":"Seoul National University Bundang Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Si-Hyuck","middleName":"","lastName":"Kang","suffix":""},{"id":325312112,"identity":"2d045354-6a65-470e-ab91-4a10d7228802","order_by":4,"name":"Sooyoung Yoo","email":"","orcid":"","institution":"Seoul National University Bundang Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Sooyoung","middleName":"","lastName":"Yoo","suffix":""},{"id":325312113,"identity":"bdeed35a-d742-482e-9dea-55c5a942239f","order_by":5,"name":"Seok Kim","email":"","orcid":"","institution":"Seoul National University Bundang Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Seok","middleName":"","lastName":"Kim","suffix":""},{"id":325312114,"identity":"56127d09-ce81-49b1-8771-1c84f0b97a8b","order_by":6,"name":"Wongeun Song","email":"","orcid":"https://orcid.org/0000-0003-4314-3995","institution":"Seoul National University Bundang Hospital","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Wongeun","middleName":"","lastName":"Song","suffix":""},{"id":325312115,"identity":"36180425-262a-46d4-b59f-c7e382ef246b","order_by":7,"name":"Junho Hyun","email":"","orcid":"","institution":"Asan Medical Center","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Junho","middleName":"","lastName":"Hyun","suffix":""},{"id":325312116,"identity":"edbf3e51-7be7-4390-9233-1aaf5ff5852a","order_by":8,"name":"Ji Seon Oh","email":"","orcid":"","institution":"Asan Medical Center","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Ji","middleName":"Seon","lastName":"Oh","suffix":""},{"id":325312117,"identity":"669b1c70-3232-4b75-bb0c-bf234983325d","order_by":9,"name":"Gakyoung Baek","email":"","orcid":"","institution":"Asan Medical Center","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Gakyoung","middleName":"","lastName":"Baek","suffix":""}],"badges":[],"createdAt":"2024-06-04 01:10:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4524391/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4524391/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":62188882,"identity":"124f621a-9761-43d0-a4bc-b74c77d42dd2","added_by":"auto","created_at":"2024-08-10 12:18:37","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":155212,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSNUBH prediction model based on validation data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) AUROC for predicting MACCE, demonstrating the model's ability to distinguish between patients with and without MACCE\u003c/p\u003e\n\u003cp\u003e(B) AUPRC for predicting MACCE, illustrating the precision-recall trade-off of the model.\u003c/p\u003e\n\u003cp\u003e(C) Calibration plot for MACCE, assessing the agreement between predicted probabilities and observed outcomes\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4524391/v1/fc1e096cd22e464a713a4593.png"},{"id":62188881,"identity":"8f06c791-1b35-4108-94e6-7438a606ea5a","added_by":"auto","created_at":"2024-08-10 12:18:37","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":150121,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eImportance of Covariates in the Prediction Model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e* The model identified key predictors grouped by themes. Medical history factors like ischemic heart disease and brain injury were highly influential. Crucial preoperative lab measurements included hemoglobin, creatinine, albumin, CK-MB, and ESR. Antithrombotic agents and beta-blockers were significant among medications, while the type of surgery had less impact than expected.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4524391/v1/8aba950c8d9a4751fa42552d.png"},{"id":62188885,"identity":"c46a35da-6c77-4f5f-8d25-e346141227de","added_by":"auto","created_at":"2024-08-10 12:18:37","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":111305,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTwo tertiary hospital cohort design\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e* Patients who underwent cardiac or emergency surgery within 3 days of a hospital visit or had an observation period of less than 30 days were excluded. Age was assessed at the time of surgery.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4524391/v1/c667906ec2e829e43736c100.png"},{"id":62188886,"identity":"a816f43a-fd01-4571-b572-3210066eddac","added_by":"auto","created_at":"2024-08-10 12:18:37","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":70185,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eData Collection for Predictors\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e* Predictor data were collected from 3–365 days before the start of the time-at-risk window. Baseline characteristics and preoperative laboratory measurements were adjusted to a narrower window of 3–30 days before the time-at-risk period to ensure accurate representation of the patient's condition.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-4524391/v1/56fff4c7271a12eb55d2033b.png"},{"id":62281962,"identity":"f3e8ba0c-eb5e-484b-806c-f42fad5ff1de","added_by":"auto","created_at":"2024-08-12 12:38:37","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1087341,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4524391/v1/b5acd15c-ca86-44a0-be27-18659265f15e.pdf"},{"id":62188884,"identity":"64dcf3c6-204c-42d2-8807-f91fb0dc1616","added_by":"auto","created_at":"2024-08-10 12:18:37","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":321764,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Table\u003c/p\u003e","description":"","filename":"SupplementaryTablePreopFinal.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4524391/v1/1aed10e40314372f3e6ce5af.pdf"},{"id":62190310,"identity":"8440a69f-766c-4943-94fc-8674fe999f68","added_by":"auto","created_at":"2024-08-10 12:26:37","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":674753,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Figure\u003c/p\u003e","description":"","filename":"SupplementaryFigurePreopFinal.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4524391/v1/03f5981199c43602d77e9453.pdf"},{"id":62188887,"identity":"2f455fe3-73ab-4f6a-a638-8cde59cd9c88","added_by":"auto","created_at":"2024-08-10 12:18:37","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":656850,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Table, Figure\u003c/p\u003e","description":"","filename":"Supplementaryfiles.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4524391/v1/67967a633d2b1c9bf47a553b.pdf"}],"financialInterests":"(Not answered)","formattedTitle":"Developing a machine learning model for predicting 30-day major adverse cardiac and cerebrovascular events in patients undergoing noncardiac surgery","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe annual count of noncardiac surgeries worldwide exceeds 300 million each year, and major adverse cardiac and cerebrovascular events (MACCE) remain a leading contributor to perioperative morbidity and mortality.\u0026nbsp;(1-4)\u0026nbsp;The rising prevalence of perioperative morbidity and mortality, driven by factors such as an aging population, necessitates precise preoperative prediction.\u0026nbsp;(5, 6)\u0026nbsp;However, the predictive accuracy of traditional assessment tools is not consistently high, and various tools are employed at different physicians’ discretion.\u0026nbsp;(7)\u003c/p\u003e\n\u003cp\u003eTraditionally, the Revised Cardiac Risk Index (RCRI), which comprises six equally weighted components, is extensively used to mitigate major perioperative cardiac complications owing to its simplicity and relatively high predictability of in-hospital major adverse cardiac events (MACE) or cardiovascular-related death.(8)\u0026nbsp;However, the index developed over 2 decades ago has certain challenges, including limited external validation and reduced precision in vascular surgery.\u0026nbsp;(9)\u0026nbsp;These factors may modestly impact its effectiveness in predicting clinical outcomes following noncardiac surgeries in practical clinical environments.\u0026nbsp;(10)\u0026nbsp;Subsequent predictive tools such as the American College of Surgeons (ACS), National Surgical Quality Improvement Project (NSQIP), and NSQIP Myocardial Infarction or Cardiac Arrest (MICA), developed after RCRI, also show strong performance in predicting postoperative MICA. However, these tools pose challenges for clinicians in practical clinical use because they rely on subjective predictors, leading to low interrater reliability.(11)\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGiven these challenges, our research leverages the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM), which converts diverse observational databases into a standardized format, enabling access to comprehensive patient data.(12-14)\u0026nbsp;Building upon this standardized data foundation, we aimed to develop a machine learning-based prediction model that harnesses the power of machine learning to identify patterns and relationships within extensive patient data, ultimately improving personalized risk prediction.\u0026nbsp;(15, 16)\u003c/p\u003e\n\u003cp\u003eFocusing on enhancing the accuracy of perioperative event prediction, our research utilized CDM data from real-world electronic health records (EHRs) as a foundation for developing a machine learning-based model to provide a more advanced and precise tool for personalized risk prediction in noncardiac surgeries.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003eStudy population\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 46,225 patients were enrolled at the SNUBH and 396,424 were enrolled at the AMC, with an average age of 72.9 years at both hospitals (Table 1). More male than female patients were enrolled at both institutions, with 25,573 males (55.3%) at SNUBH and 232,522 males (58.7%) at AMC.\u003c/p\u003e\n\u003cp\u003eHypertension was the most common comorbidity, affecting 62.0% and 54.6% of patients at SNUBH and AMC, respectively. However, at SNUBH, cerebrovascular disease was more common (15.9% at SNUBH and 10.2% at AMC). In contrast, at AMC, congestive heart failure (2.1% at SNUBH and 4.9% at AMC) and ischemic heart disease (8.5% at SNUBH and 10.9% at AMC) had a higher representation. Preoperative laboratory results were within normal ranges, with patients with a creatinine level of 2.0 mg/dL or higher accounting for 3,434 (7.4%) at SNUBH and 74,594 (18.8%) at AMC. Regarding medications, the AMC data showed a higher proportion of patients registered with aspirin, P2Y12 inhibitors, beta-blockers, RAS inhibitors, calcium channel blockers, statins, and insulin treatment.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThere was a significant difference between the two hospitals regarding the type of surgery and post-noncardiac surgery MACCE within 30 days across all categories. Surgeries with a risk exceeding 1% are presented in Table 1; those with unmapped names were classified as unspecified. Post-noncardiac surgery MACCE within 30 days included myocardial infarction, which occurred in 4.9% of patients at SNUBH and 6.3% at AMC, cardiac arrest/shock in 0.6% and 0.3%, heart failure in 0.7% and 0.6%, and stroke in 1.7% and 1.5%, respectively. In-hospital deaths accounted for 0.9% and 3.0% of the deaths at SNUBH and AMC, respectively.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003ePrediction model performance\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe prediction model discrimination for internal and external validation is presented in Table 2. The numbers of patients included in the training, test, and external validation sets of the SNUBH model who met the inclusion criteria are presented in Supplementary Table\u0026nbsp;1. When assessed using the RCRI score and compared with five other machine learning prediction models, all machine learning models outperformed the RCRI model with a higher AUROC for MACCE prediction than the RCRI score (AUROC 0.704) (Figure\u0026nbsp;1A). The RF generally showed the best overall performance in internal and external validations across outcomes with moderate calibration among the five predictive models. The AUROC of this model was 0.897 (0.883\u0026ndash;0.911) and 0.817 (0.815\u0026ndash;0.819) for internal and external validations, respectively (Figure\u0026nbsp;1A and Table 2), and the AUPRC was 0.095 (Figure\u0026nbsp;1B). In addition, it demonstrated outstanding calibration, showing strong alignment with the average predicted probability on the calibration plot. (Figure\u0026nbsp;1C).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003ePredictors\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn the prediction model, we assessed the relative importance of various covariates based on their values (Figure 2). Rather than identifying a single outstanding covariate, the analysis grouped covariates into similar thematic clusters. Predominantly, predictors associated with the patient\u0026rsquo;s underlying medical history were relatively high in the developed prediction model. These include ischemic heart disease, traumatic and non-traumatic brain injury, heart failure, heart disease, and cerebral infarction. The model highlights the importance of the measurement predictors. Preoperative laboratory measurements revealed that hemoglobin, creatinine, albumin, CK-MB, and erythrocyte sedimentation rate played crucial roles. Among the medication predictors, antithrombotic agents and beta-blockers were notably prominent, whereas the significance of the others was less pronounced. Furthermore, although important, the significance of the type of surgery did not appear to be as substantial as expected when compared with other factors in the model.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAdditionally, we developed prediction models by recombining the data and considering previous diagnoses, medication, type of surgery, and measurement data in various combinations (Supplementary Table 2). However, none of the additional recombination models outperformed the original models. Nevertheless, these models generally exhibited superior predictability compared with RCRI, except for the recombination model that excluded the previous diagnosis group, which yielded results comparable to or slightly inferior to those of RCRI (Supplementary Figure 1-3).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we developed and evaluated an advanced perioperative risk prediction model using a CDM-based machine learning approach. Our model demonstrated superior predictive accuracy compared with traditional models, such as the RCRI score. This study provides several key insights and implications.\u003c/p\u003e\n\u003cp\u003eAdvances in machine learning for extensive dataset analysis have led to increased interest in applying patient-level prediction and offer the potential for medical practice to consider personalized risks as part of clinical decision-making.\u0026nbsp;(17)\u0026nbsp;The adoption of the OMOP CDM has streamlined the transformation of diverse concept domains, encompassing medical conditions, drugs, procedures, and measurements derived from health record systems or reported information into labeled analytic data. This transformation ensures semantic and syntactic interoperability, enhancing the extraction of prediction variables and facilitating seamless integration across various healthcare data sources.\u0026nbsp;(18, 19)\u0026nbsp;In addition, the standardized data across different institutions allowed a fair evaluation of the predictive performance of the models by extensive external validation.\u003csup\u003e17\u003c/sup\u003e Therefore, our model supports existing preoperative evaluation guidelines and enables open dissemination that can be extensively validated across OHDSI collaborator networks.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe well-structured and labeled dataset improves algorithms in supervised machine learning but sometimes leads to overfitting, which prevents the model\u0026rsquo;s generalization to fit the observed data well.\u0026nbsp;(20, 21)\u0026nbsp;To overcome the challenge of overfitting, we employed a feature selection method as one of several techniques to identify and prioritize factors essential for the learning process.\u0026nbsp;(22, 23)\u0026nbsp;Additionally, we employed a feature selection method to create new combinations of thematic clusters, including medical conditions, drugs, types of procedures, and measurements, to assess their relative importance in predicting adverse outcomes following surgery. The recombination model, which included past medical conditions and previous laboratory data, exhibited a notably high predictive accuracy. Our model\u0026rsquo;s ability to discern the varying importance of these factors in real clinical contexts underscores the importance of focusing on patient histories and prior laboratory results during preoperative evaluations.\u0026nbsp;(24)\u0026nbsp;This approach aligns with physicians\u0026rsquo; subjective assessments in clinical settings and provides a flexible alternative to traditional methods that may not fully accommodate each patient\u0026rsquo;s unique circumstances.\u0026nbsp;(25)\u003c/p\u003e\n\u003cp\u003eThe practical implications of our research extend to potential time and cost savings in clinical settings. Risk assessments often lead to unnecessary procedures or examinations, such as echocardiography, cardiac computed tomography, and/or cardiac stress tests, being performed on patients. (26) These tests, even when not closely associated with the patient\u0026rsquo;s post-surgical outcomes, contribute to ongoing wastage in overall medical costs. (27, 28) Our model, with its high predictive accuracy, is poised to reduce the number of unnecessary tests performed and contribute to medical cost savings. Moreover, our model could reduce waiting times for patients as unnecessary consultations and tests may be minimized, ultimately mitigating the challenges posed by healthcare system congestion and assisting patients in undergoing surgery at an appropriate time. In the future, with precise preoperative predictability, we plan to use our model to proactively identify individuals at risk of post-surgical complications and ensure appropriate post-operative management. In an aging population, where surgical mortality and morbidity rates are increasing, (29) this approach can serve as a viable solution to effectively mitigate these challenges.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWhile using the OMOP CDM for data standardization, large healthcare datasets frequently exhibited inherent data inconsistencies and missing values. Variations in data quality and completeness among healthcare institutions may have introduced noise into the model. However, we mitigated this limitation by leveraging data from two of the largest tertiary hospitals in South Korea, where the data quality and quantity were sufficiently substantial to minimize the impact of missing values. This rich dataset enabled the development of a robust machine-learning model, enhancing its potential for accurate perioperative risk prediction.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSecond, including data from two distinct tertiary hospitals introduced potential heterogeneity in patient populations and clinical practices. This variability could have influenced the model performance, potentially contributing to variations in predictive accuracy, such as relatively low AUPRC values. However, from a machine-learning perspective, analyzing diverse datasets can be advantageous for improving future predictive capabilities. Additionally, the versatility of using frameworks such as the CDM ensures that the scalability of this model is not restricted to specific hospital settings, which can enhance its applicability in various healthcare contexts.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn this study, we successfully developed a high-performance machine learning-based preoperative prediction model by using the standardized data format of the OMOP CDM. This approach offers the potential for improved clinical decision-making and extensive external validation across healthcare institutions. In the future, our research has practical implications for potential time and cost savings in clinical settings by reducing unnecessary procedures, tests, and consultations, ultimately addressing healthcare system congestion and improving patient surgical timing.\u0026nbsp;\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eThis retrospective, observational network study involved a multidisciplinary team of cardiologists, medical informatics specialists, and data scientists. This study was approved by the Institutional Review Boards (IRB) of Seoul National University Bundang Hospital (IRB No. 2208-772-906) and Asan Medical Center (IRB No. 2022-1547). The requirement for written informed consent was waived owing to the retrospective nature of the study design and the use of de-identified data.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eData sources\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data sources used in this study were selected and standardized to ensure the integrity and compatibility of the collected information. The EHRs were converted to the OMOP CDM, and source codes were mapped to standard vocabularies, including the systematized nomenclature of medicine clinical terms (SNOMED CT).\u0026nbsp;(18, 30)\u0026nbsp;Data analysis was conducted using the observational health data sciences and informatics (OHDSI) open-source patient-level prediction package, which was purposefully designed for standardized analysis and harmonization with the OMOP CDM. These specialized tools have facilitated efficient data processing and analysis across different datasets within the OHDSI data network.\u0026nbsp;(31)\u0026nbsp;This collaborative aspect enhances the comparability and generalizability of the prediction models, making them applicable to diverse healthcare settings.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo develop and evaluate our prediction models, we retrospectively used patient data from two tertiary hospitals, Seoul National University Bundang Hospital (SNUBH) and Asan Medical Center (AMC), which are recognized for their substantial CDM datasets. The SNUBH dataset contains data from 46,225 patients who underwent noncardiac surgery between January 2003 and December 2020, and the AMC dataset includes data from 396,424 patients who underwent noncardiac surgery between January 2010 and December 2020.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThis extensive dataset included a comprehensive array of demographic information and detailed preoperative characteristics, including diagnosis codes, underlying diseases, laboratory test results, medications, type of surgery, and clinical outcomes from the EHR system\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e(Table 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eStudy design and target cohort\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe conducted a retrospective analysis of patients aged 65 years or older who underwent noncardiac surgeries at two independent tertiary hospitals. Age was determined at the time of surgery. We excluded patients who had undergone cardiac or emergency surgery within 3 days of a hospital visit and those who did not have a sufficient observation period of less than 30 days (Figure 3). The prediction time (t=0) and start date of the time-at-risk window for prediction were set as surgery dates. The end date of the time-at-risk window for clinical outcomes was 30 days after surgery. The data collection period for the predictors was defined as 3\u0026ndash;365 days before the start date of the time-at-risk window (Figure 4). We adjusted the observational time frame to collect baseline characteristics and preoperative laboratory measurements within a narrower window of 3\u0026ndash;30 days before the onset of the time-at-risk period. This adjustment ensured that the data accurately represented the patient\u0026apos;s condition at the beginning of the time-at-risk period.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this study, we successfully developed a high-performance machine learning-based preoperative prediction model by using the standardized data format of the OMOP CDM. This approach offers the potential for improved clinical decision-making and extensive external validation across healthcare institutions. In the future, our research has practical implications for potential time and cost savings in clinical settings by reducing unnecessary procedures, tests, and consultations, ultimately addressing healthcare system congestion and improving patient surgical timing.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgment\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was funded by the Seoul National University Bundang Hospital Research Fund (Grant No. 14-2022-0023). The funder played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConceptualization: Jung-Won Suh\u003c/p\u003e\n\u003cp\u003eFormal analysis: Sooyoung Yoo, Seok Kim, Wongeun Song\u003c/p\u003e\n\u003cp\u003eInvestigation: Jung-Won Suh,\u0026nbsp;Ju-Seung Kwun, Houng-Beom Ahn,\u0026nbsp;Si-Hyuck Kang,\u0026nbsp;Sooyoung Yoo\u003c/p\u003e\n\u003cp\u003eValidation: Ji Seon Oh, Junho Hyun, Gakyoung Baek\u003c/p\u003e\n\u003cp\u003eWriting - original draft: Ju-Seung Kwun, Houng-Beom Ahn\u003c/p\u003e\n\u003cp\u003eWriting - review \u0026amp; editing: Jung-Won Suh, Si-Hyuck Kang, Sooyoung Yoo\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors declare no financial or non-financial competing interests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData\u0026nbsp;underlying this article cannot be shared publicly because of data protection agreements but can be available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Code that supports the findings of this study are available from the corresponding author upon reasonable request. Analysis to process and analyze data was generated with Python 3, R version’s 4.1.0\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePending patents/patent applications\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePatent applicant (whether author or institution): Seoul National University Bundang Hospital\u003c/p\u003e\n\u003cp\u003eName of the inventor(s): Jung-Won Suh, Ju-Seung Kwun, Houng-Beom Ahn, Sooyoung Yoo\u003c/p\u003e\n\u003cp\u003eApplication number: 10-2024-0011788\u003c/p\u003e\n\u003cp\u003eStatus of the application: \u0026nbsp;provisional application status\u003c/p\u003e\n\u003cp\u003eSpecific aspect of manuscript covered in the patent application: Machine Learning Model for Predicting 30-Day MACCE in Noncardiac Surgery Patients\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSmilowitz, N. R. et al. Perioperative major adverse cardiovascular and cerebrovascular events associated with noncardiac surgery. \u003cem\u003eJAMA Cardiol\u003c/em\u003e. \u003cstrong\u003e2\u003c/strong\u003e, 181-187 (2017).\u003c/li\u003e\n\u003cli\u003eSabat\u0026eacute;, S. et al. Incidence and predictors of major perioperative adverse cardiac and cerebrovascular events in non-cardiac surgery. \u003cem\u003eBr. J. Anaesth\u003c/em\u003e. \u003cstrong\u003e107\u003c/strong\u003e, 879-890 (2011).\u003c/li\u003e\n\u003cli\u003ePuelacher, C. et al. Perioperative myocardial injury after noncardiac surgery: incidence, mortality, and characterization. \u003cem\u003eCirculation\u003c/em\u003e. \u003cstrong\u003e137\u003c/strong\u003e, 1221-1232 (2018). \u003c/li\u003e\n\u003cli\u003eWriting Committee for the VISION Study Investigators, Devereaux, P. J. et al. Association of postoperative high-sensitivity troponin levels with myocardial injury and 30-day mortality among patients undergoing noncardiac surgery. \u003cem\u003eJAMA\u003c/em\u003e. \u003cstrong\u003e317\u003c/strong\u003e, 1642-1651 (2017).\u003c/li\u003e\n\u003cli\u003eVan Klei, W. A. et al. Role of history and physical examination in preoperative evaluation. \u003cem\u003eEur. J. Anaesthesiol\u003c/em\u003e. \u003cstrong\u003e20\u003c/strong\u003e, 612-618 (2003). \u003c/li\u003e\n\u003cli\u003eAlkire, B. C. et al. Global access to surgical care: a modelling study. \u003cem\u003eLancet Glob. Health\u003c/em\u003e. \u003cstrong\u003e3\u003c/strong\u003e, e316-e323 (2015). \u003c/li\u003e\n\u003cli\u003eCohen, M. E., Bilimoria, K. Y., Ko, C. Y., Richards, K. \u0026amp; Hall, B. L. Effect of subjective preoperative variables on risk-adjusted assessment of hospital morbidity and mortality. \u003cem\u003eAnn. Surg\u003c/em\u003e. \u003cstrong\u003e249\u003c/strong\u003e, 682-689 (2009). \u003c/li\u003e\n\u003cli\u003eLee, T. H. et al. Derivation and prospective validation of a simple index for prediction of cardiac risk of major noncardiac surgery. \u003cem\u003eCirculation\u003c/em\u003e. \u003cstrong\u003e100\u003c/strong\u003e, 1043-1049 (1999). \u003c/li\u003e\n\u003cli\u003eGupta, P. K. et al. Development and validation of a risk calculator for prediction of cardiac risk after surgery. \u003cem\u003eCirculation\u003c/em\u003e. \u003cstrong\u003e124\u003c/strong\u003e, 381-387 (2011). \u003c/li\u003e\n\u003cli\u003eBrasher, P. M. \u0026amp; Beattie, W. S. Adjusting clinical prediction rules: an academic exercise or the potential for real world clinical applications in perioperative medicine? \u003cem\u003eCan. J. Anesth.\u003c/em\u003e \u003cstrong\u003e56\u003c/strong\u003e, 190-193 (2009).\u003c/li\u003e\n\u003cli\u003eBilimoria, K. Y. et al. Development and evaluation of the universal ACS NSQIP surgical risk calculator: a decision aid and informed consent tool for patients and surgeons. \u003cem\u003eJ. Am. Coll. Surg\u003c/em\u003e. \u003cstrong\u003e217\u003c/strong\u003e, 833-842.e1-3 (2013).\u003c/li\u003e\n\u003cli\u003eVoss, E. A. et al. Feasibility and utility of applications of the common data model to multiple, disparate observational health databases. \u003cem\u003eJ. Am. Med. Inform. Assoc\u003c/em\u003e. \u003cstrong\u003e22\u003c/strong\u003e, 553-564 (2015).\u003c/li\u003e\n\u003cli\u003eMadigan, D. et al. A systematic statistical approach to evaluating evidence from observational studies. \u003cem\u003eAnnu. Rev. Stat. Its Appl\u003c/em\u003e. \u003cstrong\u003e1\u003c/strong\u003e, 11-39 (2014). \u003c/li\u003e\n\u003cli\u003eStang, P. E. et al. Advancing the science for active surveillance: rationale and design for the Observational Medical Outcomes Partnership. \u003cem\u003eAnn. Intern. Med\u003c/em\u003e. \u003cstrong\u003e153\u003c/strong\u003e, 600-606 (2010). \u003c/li\u003e\n\u003cli\u003eMesk\u0026oacute;, B. \u0026amp; G\u0026ouml;r\u0026ouml;g, M. A short guide for medical professionals in the era of artificial intelligence. \u003cem\u003enpj Digit. Med\u003c/em\u003e. \u003cstrong\u003e3\u003c/strong\u003e, 126 (2020).\u003c/li\u003e\n\u003cli\u003eRajkomar, A., Dean, J. \u0026amp; Kohane, I. Machine learning in medicine. \u003cem\u003eN. Engl. J. Med\u003c/em\u003e. \u003cstrong\u003e380\u003c/strong\u003e, 1347-1358 (2019).\u003c/li\u003e\n\u003cli\u003eGoldstein, B. A., Navar, A. M., Pencina, M. J. \u0026amp; Ioannidis, J. P. Opportunities and challenges in developing risk prediction models with electronic health records data: a systematic review. \u003cem\u003eJ. Am. Med. Inform. Assoc\u003c/em\u003e. \u003cstrong\u003e24\u003c/strong\u003e, 198-208 (2017). \u003c/li\u003e\n\u003cli\u003eReps, J. M., Schuemie, M. J., Suchard, M. A., Ryan, P. B. \u0026amp; Rijnbeek, P. R. Design and implementation of a standardized framework to generate and evaluate patient-level prediction models using observational healthcare data. \u003cem\u003eJ. Am. Med. Inform. Assoc\u003c/em\u003e. \u003cstrong\u003e25\u003c/strong\u003e, 969-975 (2018). \u003c/li\u003e\n\u003cli\u003eHripcsak, G. et al. Observational health data sciences and informatics (OHDSI): opportunities for observational researchers. \u003cem\u003eStud. Health Technol. Inform\u003c/em\u003e. \u003cstrong\u003e216\u003c/strong\u003e, 574-578 (2015).\u003c/li\u003e\n\u003cli\u003eJunqu\u0026eacute; de Fortuny, E., Martens, D. \u0026amp; Provost, F. Predictive modeling with big data: is bigger really better? \u003cem\u003eBig Data\u003c/em\u003e. \u003cstrong\u003e1\u003c/strong\u003e, 215-226 (2013). \u003c/li\u003e\n\u003cli\u003eYing, X. An overview of overfitting and its solutions. \u003cem\u003eJ Phys Conf S\u003c/em\u003e. (IOP Publishing, 2019).\u003c/li\u003e\n\u003cli\u003eHawkins, D. M. The problem of overfitting. \u003cem\u003eJ. Chem. Inf. Comput. Sci\u003c/em\u003e. \u003cstrong\u003e44\u003c/strong\u003e, 1-12 (2004). \u003c/li\u003e\n\u003cli\u003eBagherzadeh-Khiabani, F. et al A tutorial on variable selection for clinical prediction models: feature selection methods in data mining could improve the results. \u003cem\u003eJ. Clin. Epidemiol\u003c/em\u003e. \u003cstrong\u003e71\u003c/strong\u003e, 76-85 (2016). \u003c/li\u003e\n\u003cli\u003eMichota, F. A. \u0026amp; Frost, S. D. The preoperative evaluation: use the history and physical rather than routine testing. \u003cem\u003eCleve Clin. J. Med\u003c/em\u003e. \u003cstrong\u003e71\u003c/strong\u003e, 63-70 (2004). \u003c/li\u003e\n\u003cli\u003eGlance, L. G. et al. Impact of the choice of risk model for identifying low-risk patients using the 2014 American College of Cardiology/American Heart Association Perioperative Guidelines. \u003cem\u003eAnesthesiology\u003c/em\u003e. \u003cstrong\u003e129\u003c/strong\u003e, 889-900 (2018). \u003c/li\u003e\n\u003cli\u003eJohansson, T. et al. Effectiveness of non-cardiac preoperative testing in non-cardiac elective surgery: a systematic review. \u003cem\u003eBr. J. Anaesth\u003c/em\u003e. \u003cstrong\u003e110\u003c/strong\u003e, 926-939 (2013). \u003c/li\u003e\n\u003cli\u003eBryson, G. L., Wyand, A. \u0026amp; Bragg, P. R. Preoperative testing is inconsistent with published guidelines and rarely changes management. \u003cem\u003eCan. J. Anaesth\u003c/em\u003e. \u003cstrong\u003e53\u003c/strong\u003e, 236-241 (2006). \u003c/li\u003e\n\u003cli\u003eAugoustides, J. G., Neuman, M. D., Al-Ghofaily, L. \u0026amp; Silvay, G. Preoperative cardiac risk assessment for noncardiac surgery: defining costs and risks. \u003cem\u003eJ. Cardiothorac. Vasc. Anesth\u003c/em\u003e. \u003cstrong\u003e27\u003c/strong\u003e, 395-399 (2013).\u003c/li\u003e\n\u003cli\u003eFerrando, A. et al. Guidelines for preoperative assessment: impact on clinical practice and costs. \u003cem\u003eInt. J. Qual. Health Care\u003c/em\u003e. \u003cstrong\u003e17\u003c/strong\u003e, 323-329 (2005). 32. Weiser, T. G. et al. Estimate of the global volume of surgery in 2012: an assessment supporting improved health outcomes. \u003cem\u003eLancet\u003c/em\u003e. \u003cstrong\u003e385\u003c/strong\u003e, S11 (2015). \u003c/li\u003e\n\u003cli\u003eOverhage, J. M., Ryan, P. B., Reich, C. G., Hartzema, A. G. \u0026amp; Stang, P. E. Validation of a common data model for active safety surveillance research. \u003cem\u003eJ. Am. Med. Inform. Assoc\u003c/em\u003e. \u003cstrong\u003e19\u003c/strong\u003e, 54-60 (2012). \u003c/li\u003e\n\u003cli\u003eFitzHenry, F. et al. Creating a common data model for comparative effectiveness with the observational medical outcomes partnership. \u003cem\u003eAppl. Clin. Inform\u003c/em\u003e. \u003cstrong\u003e6\u003c/strong\u003e, 536-547 (2015).\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable 1.\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;Baseline characteristics\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"595\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003eSNUBH\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003eAMC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003eP-value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eNumber of populations, N\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e46,225\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e396,424\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eAge, years, mean (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e72.9 (5.35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e72.9 (6.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e0.014\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eSex, N (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; Male\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e25,573 (55.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e232,522 (58.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp; Female\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e20,652 (44.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e163,902 (41.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eBMI, kg/m\u003csup\u003e2\u003c/sup\u003e (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e23.7 (3.39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e23.3 (3.70)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eUnderlying disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eHypertension\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e28,641 (62.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e216,440 (54.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eDiabetes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e12,815 (27.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e104,269 (26.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eDyslipidemia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e12,078 (26.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e129,601 (32.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eCongestive heart failure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e961 (2.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e19,613 (4.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eChronic kidney disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e2,588 (5.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e45,095 (11.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eCerebrovascular disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e7,363 (15.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e40,628 (10.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eIschemic heart disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e3,939 (8.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e43,302 (10.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003ePreoperative lab results\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eWhite blood cell, 10\u003csup\u003e3\u003c/sup\u003e/\u0026mu;L (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e7.0 (2.64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e7.5 (3.56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eHemoglobin, g/dL (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e12.8 (1.89)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e11.5 (2.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003ePlatelet, 10\u003csup\u003e3\u003c/sup\u003e/ \u0026mu;L (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e233.9 (75.86)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e214.9 (91.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eSodium, mmol/L (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e139.9 (3.42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e138.2 (4.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003ePotassium, mmol/L (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e4.3 (0.46)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e4.2 (0.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eBUN, mg/dL (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e18.3 (9.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e23.1 (17.06)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eCreatinine, mg/dL (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e1.1 (0.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e1.3 (1.43)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eCreatinine level \u0026gt;2.0mg/Dl, N (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e3,434 (7.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e74,594 (18.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eTotal cholesterol, mg/dL (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e169.0 (41.55)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e146.2 (45.78)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eLDL, mg/dL\u0026nbsp;(SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e92.2 (30.90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e91.3 (36.14)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e0.165\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eAlbumin, g/dL (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e4.0 (0.53)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e3.2 (0.71)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eAST, IU/L\u0026nbsp;(SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e27.7 (20.58)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e31.7 (30.72)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eALT, IU/L\u0026nbsp;(SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e24.1 (21.70)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e25.4 (28.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eGlucose, mg/dL (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e122.2 (43.59)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e133.0 (55.15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003ePT, INR\u0026nbsp;(SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e1.0 (0.19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e1.1 (0.31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eaPTT, sec\u0026nbsp;(SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e36.6 (5.83)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e31.1 (8.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eMedications\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eAspirin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e11,900 (25.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e139,029 (35.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eP2Y12 inhibitor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e6,263 (13.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e71,436 (18.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eBeta blocker\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e9,678 (20.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e179,264 (45.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eRAS inhibitor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e12,357 (26.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e161,016 (40.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eCalcium channel blocker\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e15,771 (34.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e227,915 (57.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eStatin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e11,734 (25.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e129,236 (32.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eInsulin treatment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e8,603 (18.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e138,392 (34.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eType of surgery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eIntermediate risk: 1-5%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eIntraperitoneal: splenectomy, hiatal hernia repair, cholecystectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e1,429 (3.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e29,072 (7.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eCarotid symptomatic (CEA or CAS)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e17 (0.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e655 (0.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003ePeripheral arterial angioplasty\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e12 (0.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e24,861 (6.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eHead and neck surgery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e2,090 (4.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e60,919 (15.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eNeurological or orthopedic: major (hip and spine surgery)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e3,309 (7.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e16,878 (4.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eUrological or gynecological: major\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e243 (0.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e5,350 (1.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eRenal transplant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e23 (0.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e2,684 (0.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eIntra-thoracic: non-major\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e1,596 (3.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e50,247 (12.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eHigh risk: \u0026gt;5%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eAortic and major vascular surgery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e2,028 (4.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e53,886 (13.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eOpen lower limb revascularization or amputation or thromboembolectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e250 (0.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e7,786 (2.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eDuodeno-pancreatic surgery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e247 (0.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e7,216 (1.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eLiver section, bile duct surgery\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e373 (0.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e19,183 (4.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eEsophagectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e75 (0.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e7,795 (2.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eRepair of perforated bowel\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e1,557 (3.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e105,525 (26.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eAdrenal resection\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e66 (0.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e1,191 (0.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003ePneumonectomy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e1,026 (2.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e13,751 (3.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003ePulmonary or liver transplant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e27 (0.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e8,656 (2.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eUnspecified\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e494 (1.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e74,067 (18.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eOutcome\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eMACE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e2,256 (4.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e24,853 (6.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eMyocardial infarction\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e907 (0.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e5,603 (0.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eCardiac arrest/shock\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e35 (0.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e168 (0.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eHeart failure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e308 (0.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e2,310 (0.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e0.029\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eStroke\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e799 (1.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e6,017 (1.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"41.2751677852349%\" valign=\"top\"\u003e\n \u003cp\u003eDeath (In-hospital)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"23.825503355704697%\" valign=\"top\"\u003e\n \u003cp\u003e419 (0.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"20.63758389261745%\" valign=\"top\"\u003e\n \u003cp\u003e11,875 (3.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"14.261744966442953%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;*Values are given as the mean \u0026plusmn; standard deviation or number with percentage (%)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eTable 2.\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;Predictability of five machine learning prediction models\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"596\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003ePrediction model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\"\u003e\n \u003cp\u003eSNUBH\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eAMC\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAUROC (95% CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eTrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eTest\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eExternal\u003c/p\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\"\u003e\n \u003cp\u003e\u003cstrong\u003eMACCE\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eRandom forest\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.985 (0.980-0.984)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.897 (0.883-0.911)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.817 (0.815-0.819)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eGradient boosting machine\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.935 (0.996-0.997)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.898 (0.885-0.912)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.825 (0.823-0.828)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLasso Logistic Regression\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.906 (0.952-0.956)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.892 (0.878-0.906)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.812 (0.810-0.815)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAdaBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.907 (0.902-0.907)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.887 (0.873-0.902)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.785 (0.782-0.788)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eDecisionTree\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.895 (0.875-0.883)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.776 (0.750-0803)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.663 (0.659-0.667)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Perioperative risk evaluation, noncardiac surgery, prediction models, machine learning, common data model","lastPublishedDoi":"10.21203/rs.3.rs-4524391/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4524391/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTo reduce unnecessary delays and manage medical costs efficiently for low-risk patients undergoing noncardiac surgery, we developed a predictive model for major adverse cardiac and cerebrovascular events (MACCE) using the OMOP Common Data Model (CDM) and machine learning algorithms. This retrospective study collected data from 46,225 patients at Seoul National University Bundang Hospital and 396,424 patients at Asan Medical Center. Patients aged 65 or older undergoing non-cardiac, non-emergency surgeries with at least 30 days of observation were included. Machine learning models were developed using the OHDSI open-source patient-level prediction package in R version 4.1.0. All models outperformed the Revised Cardiac Risk Index (RCRI), with the random forest model achieving an AUROC of 0.817 in external validation and demonstrating moderate calibration. Key predictors included previous diagnoses and laboratory measurements, highlighting their importance in perioperative risk prediction. Our model shows promise for improving clinical practice and reducing medical costs.\u003c/p\u003e","manuscriptTitle":"Developing a machine learning model for predicting 30-day major adverse cardiac and cerebrovascular events in patients undergoing noncardiac surgery","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-08-10 12:18:32","doi":"10.21203/rs.3.rs-4524391/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"cfc3caf3-10ed-4bb4-9fcc-37324bea56f1","owner":[],"postedDate":"August 10th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":34414291,"name":"Health sciences/Medical research/Outcomes research"},{"id":34414292,"name":"Health sciences/Health care/Disease prevention/Preventive medicine"},{"id":34414293,"name":"Health sciences/Health care/Prognosis"}],"tags":[],"updatedAt":"2024-08-12T12:30:31+00:00","versionOfRecord":[],"versionCreatedAt":"2024-08-10 12:18:32","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4524391","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4524391","identity":"rs-4524391","version":["v1"]},"buildId":"cBFmMYwuxLRRLfASyISRj","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}