Predicting Outcomes Following Endovascular Abdominal Aortic Aneurysm Repair Using Machine Learning.

OBJECTIVE: To develop machine learning (ML) models that predict outcomes following endovascular aneurysm repair (EVAR) for abdominal aortic aneurysm (AAA). BACKGROUND: EVAR carries non-negligible perioperative risks; however, there are no widely used outcome prediction tools. METHODS: The National Surgical Quality Improvement Program targeted database was used to identify patients who underwent EVAR for infrarenal AAA between 2011 and 2021. Input features included 36 preoperative variables. The primary outcome was 30-day major adverse cardiovascular event (composite of myocardial infarction, stroke, or death). Data were split into training (70%) and test (30%) sets. Using 10-fold cross-validation, 6 ML models were trained using preoperative features. The primary model evaluation metric was area under the receiver operating characteristic curve. Model robustness was evaluated with calibration plot and Brier score. Subgroup analysis was performed to assess model performance based on age, sex, race, ethnicity, and prior AAA repair. RESULTS: Overall, 16,282 patients were included. The primary outcome of 30-day major adverse cardiovascular event occurred in 390 (2.4%) patients. Our best-performing prediction model was XGBoost, achieving an area under the receiver operating characteristic curve (95% CI) of 0.95 (0.94-0.96) compared with logistic regression [0.72 [0.70-0.74)]. The calibration plot showed good agreement between predicted and observed event probabilities with a Brier score of 0.06. Model performance remained robust on all subgroup analyses. CONCLUSIONS: Our newer ML models accurately predict 30-day outcomes following EVAR using preoperative data and perform better than logistic regression. Our automated algorithms can guide risk mitigation strategies for patients being considered for EVAR.

Ambulatory Cardiology or General Internal Medicine Assessment Prior to Scheduled Major Vascular Surgery is Associated with Improved Outcomes.

OBJECTIVE: To characterize the association between ambulatory cardiology or general internal medicine (GIM) assessment prior to surgery and outcomes following scheduled major vascular surgery. BACKGROUND: Cardiovascular risk assessment and management prior to high-risk surgery remains an evolving area of care. METHODS: This is population-based retrospective cohort study of all adults who underwent scheduled major vascular surgery in Ontario, Canada, April 1, 2004-March 31, 2019. Patients who had an ambulatory cardiology and/or GIM assessment within 6 months prior to surgery were compared to those who did not. The primary outcome was 30-day mortality. Secondary outcomes included: composite of 30-day mortality, myocardial infarction or stroke; 30-day cardiovascular death; 1-year mortality; composite of 1-year mortality, myocardial infarction or stroke; and 1-year cardiovascular death. Cox proportional hazard regression using inverse probability of treatment weighting (IPTW) was used to mitigate confounding by indication. RESULTS: Among 50,228 patients, 20,484 (40.8%) underwent an ambulatory assessment prior to surgery: 11,074 (54.1%) with cardiology, 8,071 (39.4%) with GIM and 1,339 (6.5%) with both. Compared to patients who did not, those who underwent an assessment had a higher Revised Cardiac Risk Index (N with Index over 2= 4,989[24.4%] vs. 4,587[15.4%], P<0.001) and more frequent pre-operative cardiac testing (N=7,772[37.9%] vs. 6,113[20.6%], P<0.001) but, lower 30-day mortality (N=551[2.7%] vs. 970[3.3%], P<0.001). After application of IPTW, cardiology or GIM assessment prior to surgery remained associated with a lower 30-day mortality (weighted Hazard Ratio [95%CI] = 0.73 [0.65-0.82]) and a lower rate of all secondary outcomes. CONCLUSIONS: Major vascular surgery patients assessed by a cardiology or GIM physician prior to surgery have better outcomes than those who are not. Further research is needed to better understand potential mechanisms of benefit.

Using Machine Learning (XGBoost) to Predict Outcomes After Infrainguinal Bypass for Peripheral Artery Disease.

OBJECTIVE: To develop machine learning (ML) algorithms that predict outcomes after infrainguinal bypass. BACKGROUND: Infrainguinal bypass for peripheral artery disease carries significant surgical risks; however, outcome prediction tools remain limited. METHODS: The Vascular Quality Initiative database was used to identify patients who underwent infrainguinal bypass for peripheral artery disease between 2003 and 2023. We identified 97 potential predictor variables from the index hospitalization [68 preoperative (demographic/clinical), 13 intraoperative (procedural), and 16 postoperative (in-hospital course/complications)]. The primary outcome was 1-year major adverse limb event (composite of surgical revision, thrombectomy/thrombolysis, or major amputation) or death. Our data were split into training (70%) and test (30%) sets. Using 10-fold cross-validation, we trained 6 ML models using preoperative features. The primary model evaluation metric was the area under the receiver operating characteristic curve (AUROC). The top-performing algorithm was further trained using intraoperative and postoperative features. Model robustness was evaluated using calibration plots and Brier scores. RESULTS: Overall, 59,784 patients underwent infrainguinal bypass, and 15,942 (26.7%) developed 1-year major adverse limb event/death. The best preoperative prediction model was XGBoost, achieving an AUROC (95% CI) of 0.94 (0.93-0.95). In comparison, logistic regression had an AUROC (95% CI) of 0.61 (0.59-0.63). Our XGBoost model maintained excellent performance at the intraoperative and postoperative stages, with AUROCs (95% CI's) of 0.94 (0.93-0.95) and 0.96 (0.95-0.97), respectively. Calibration plots showed good agreement between predicted and observed event probabilities with Brier scores of 0.08 (preoperative), 0.07 (intraoperative), and 0.05 (postoperative). CONCLUSIONS: ML models can accurately predict outcomes after infrainguinal bypass, outperforming logistic regression.

Using machine learning to predict outcomes following suprainguinal bypass.

OBJECTIVE: Suprainguinal bypass for peripheral artery disease (PAD) carries important surgical risks; however, outcome prediction tools remain limited. We developed machine learning (ML) algorithms that predict outcomes following suprainguinal bypass. METHODS: The Vascular Quality Initiative database was used to identify patients who underwent suprainguinal bypass for PAD between 2003 and 2023. We identified 100 potential predictor variables from the index hospitalization (68 preoperative [demographic/clinical], 13 intraoperative [procedural], and 19 postoperative [in-hospital course/complications]). The primary outcomes were major adverse limb events (MALE; composite of untreated loss of patency, thrombectomy/thrombolysis, surgical revision, or major amputation) or death at 1 year following suprainguinal bypass. Our data were split into training (70%) and test (30%) sets. Using 10-fold cross-validation, we trained six ML models using preoperative features (Extreme Gradient Boosting [XGBoost], random forest, Naïve Bayes classifier, support vector machine, artificial neural network, and logistic regression). The primary model evaluation metric was area under the receiver operating characteristic curve (AUROC). The best performing algorithm was further trained using intra- and postoperative data. Model robustness was evaluated using calibration plots and Brier scores. Performance was assessed on subgroups based on age, sex, race, ethnicity, rurality, median Area Deprivation Index, symptom status, procedure type, prior intervention for PAD, concurrent interventions, and urgency. RESULTS: Overall, 16,832 patients underwent suprainguinal bypass, and 3136 (18.6%) developed 1-year MALE or death. Patients with 1-year MALE or death were older (mean age, 64.9 vs 63.5 years; P < .001) with more comorbidities, had poorer functional status (65.7% vs 80.9% independent at baseline; P < .001), and were more likely to have chronic limb-threatening ischemia (67.4% vs 47.6%; P < .001) than those without an outcome. Despite being at higher cardiovascular risk, they were less likely to receive acetylsalicylic acid or statins preoperatively and at discharge. Our best performing prediction model at the preoperative stage was XGBoost, achieving an AUROC of 0.92 (95% confidence interval [CI], 0.91-0.93). In comparison, logistic regression had an AUROC of 0.67 (95% CI, 0.65-0.69). Our XGBoost model maintained excellent performance at the intra- and postoperative stages, with AUROCs of 0.93 (95% CI, 0.92-0.94) and 0.98 (95% CI, 0.97-0.99), respectively. Calibration plots showed good agreement between predicted and observed event probabilities with Brier scores of 0.12 (preoperative), 0.11 (intraoperative), and 0.10 (postoperative). Of the top 10 predictors, nine were preoperative features including chronic limb-threatening ischemia, previous procedures, comorbidities, and functional status. Model performance remained robust on all subgroup analyses. CONCLUSIONS: We developed ML models that accurately predict outcomes following suprainguinal bypass, performing better than logistic regression. Our algorithms have potential for important utility in guiding perioperative risk mitigation strategies to prevent adverse outcomes following suprainguinal bypass.

Development and Testing of Step, Error, and Event Frameworks to Evaluate Technical Performance in Peripheral Endovascular Interventions.

OBJECTIVE: Tools for endovascular performance assessment are necessary in competency based education. This study aimed to develop and test a detailed analysis tool to assess steps, errors, and events in peripheral endovascular interventions (PVI). METHODS: A modified Delphi consensus was used to identify steps, errors, and events in iliac-femoral-popliteal endovascular interventions. International experts in vascular surgery, interventional radiology, cardiology, and angiology were identified, based on their scientific track record. In an initial open ended survey round, experts volunteered a comprehensive list of steps, errors, and events. The items were then rated on a five point Likert scale until consensus was reached with a pre-defined threshold (Cronbach's alpha > 0.7) and > 70% expert agreement. An experienced endovascular surgeon applied the finalised frameworks on 10 previously videorecorded elective PVI cases. RESULTS: The expert consensus panel was formed by 28 of 98 invited proceduralists, consisting of three angiologists, seven interventional radiologists, five cardiologists, and 13 vascular surgeons, with 29% from North America and 71% from Europe. The Delphi process was completed after three rounds (Cronbach's alpha; αsteps = 0.79; αerrors = 0.90; αevents = 0.90), with 15, 26, and 18 items included in the final step (73 - 100% agreement), error (73 - 100% agreement), and event (73 - 100% agreement) frameworks, respectively. The median rating time per case was 4.3 hours (interquartile range [IQR] 3.2, 5 hours). A median of 55 steps (IQR 40, 67), 27 errors (IQR 21, 49), and two events (IQR 1, 6) were identified per case. CONCLUSION: An evaluation tool for the procedural steps, errors, and events in iliac-femoral-popliteal endovascular procedures was developed through a modified Delphi consensus and applied to recorded intra-operative data to identify hazardous steps, common errors, and events. Procedural mastery may be promoted by using the frameworks to provide endovascular proceduralists with detailed technical performance feedback.

One-time screening for abdominal aortic aneurysm in Ontario, Canada: a model-based cost-utility analysis.

BACKGROUND: Screening programs for abdominal aortic aneurysm (AAA) are not available in Canada. We sought to determine the effectiveness and costutility of AAA screening in Ontario. METHODS: We compared one-time ultrasonography-based AAA screening for people aged 65 years to no screening using a fully probabilistic Markov model with a lifetime horizon. We estimated life-years, quality-adjusted life-years (QALYs), AAA-related deaths, number needed to screen to prevent 1 AAA-related death and costs (in Canadian dollars) from the perspective of the Ontario Ministry of Health. We retrieved model inputs from literature, Statistics Canada, and the Ontario Case Costing Initiative. RESULTS: Screening reduced AAA-related deaths by 84.9% among males and 81.0% among females. Compared with no screening, screening resulted in 0.04 (18.96 v. 18.92) additional life-years and 0.04 (14.95 v. 14.91) additional QALYs at an incremental cost of $80 per person among males. Among females, screening resulted in 0.02 (21.25 v. 21.23) additional life-years and 0.01 (16.20 v. 16.19) additional QALYs at an incremental cost of $11 per person. At a willingness-to-pay of $50 000 per year, screening was cost-effective in 84% (males) and 90% (females) of model iterations. Screening was increasingly cost-effective with higher AAA prevalence. INTERPRETATION: Screening for AAA among people aged 65 years in Ontario was associated with fewer AAA-related deaths and favourable cost-effectiveness. To maximize QALY gains per dollar spent and AAA-related deaths prevented, AAA screening programs should be designed to ensure that populations with high prevalence of AAA participate.

Hospital volume-outcome relationships for robot-assisted surgeries: a population-based analysis.

BACKGROUND: Associations between procedure volumes and outcomes can inform minimum volume standards and the regionalization of health services. Robot-assisted surgery continues to expand globally; however, data are limited regarding which hospitals should be using the technology. STUDY DESIGN: Using administrative health data for all residents of Ontario, Canada, this retrospective cohort study included adult patients who underwent a robot-assisted radical prostatectomy (RARP), total robotic hysterectomy (TRH), robot-assisted partial nephrectomy (RAPN), or robotic portal lobectomy using 4 arms (RPL-4) between January 2010 and September 2021. Associations between yearly hospital volumes and 90-day major complications were evaluated using multivariable logistic regression models adjusted for patient characteristics and clustering at the level of the hospital. RESULTS: A total of 10,879 patients were included, with 7567, 1776, 724, and 812 undergoing a RARP, TRH, RAPN, and RPL-4, respectively. Yearly hospital volume was not associated with 90-day complications for any procedure. Doubling of yearly volume was associated with a 17-min decrease in operative time for RARP (95% confidence interval [CI] - 23 to - 10), 8-min decrease for RAPN (95% CI - 14 to - 2), 24-min decrease for RPL-4 (95% CI - 29 to - 19), and no significant change for TRH (- 7 min; 95% CI - 17 to 3). CONCLUSION: The risk of 90-day major complications does not appear to be higher in low volume hospitals; however, they may not be as efficient with operating room utilization. Careful case selection may have contributed to the lack of an observed association between volumes and complications.

Hospital learning curves for robot-assisted surgeries: a population-based analysis.

BACKGROUND: Robot-assisted surgery has been rapidly adopted. It is important to define the learning curve to inform credentialling requirements, training programs, identify fast and slow learners, and protect patients. This study aimed to characterize the hospital learning curve for common robot-assisted procedures. STUDY DESIGN: This cohort study, using administrative health data for Ontario, Canada, included adult patients who underwent a robot-assisted radical prostatectomy (RARP), total robotic hysterectomy (TRH), robot-assisted partial nephrectomy (RAPN), or robotic portal lobectomy using four arms (RPL-4) between 2010 and 2021. The association between cumulative hospital volume of a robot-assisted procedure and major complications was evaluated using multivariable logistic models adjusted for patient characteristics and clustering at the hospital level. RESULTS: A total of 6814 patients were included, with 5230, 543, 465, and 576 patients in the RARP, TRH, RAPN, and RPL-4 cohorts, respectively. There was no association between cumulative hospital volume and major complications. Visual inspection of learning curves demonstrated a transient worsening of outcomes followed by subsequent improvements with experience. Operative time decreased for all procedures with increasing volume and reached plateaus after approximately 300 RARPs, 75 TRHs, and 150 RPL-4s. The odds of a prolonged length of stay decreased with increasing volume for patients undergoing a RARP (OR 0.87; 95% CI 0.82-0.92) or RPL-4 (OR 0.77; 95% CI 0.68-0.87). CONCLUSION: Hospitals may adopt robot-assisted surgery without significantly increasing the risk of major complications for patients early in the learning curve and with an expectation of increasing efficiency.

Recommendations for Recovery of the COVID-19 Pandemic-related Diagnostic, Screening, and Procedure Backlog in Ontario: A Survey of Healthcare Leaders.

PURPOSE: The COVID-19 pandemic has resulted in a significant diagnostic, screening, and procedure backlog in Ontario. Engagement of key stakeholders in healthcare leadership positions is urgently needed to inform a comprehensive provincial recovery strategy. METHODS: A list of 20 policy recommendations addressing the diagnostic, screening and procedure backlog in Ontario were transformed into a national online survey. Policy recommendations were rated on a 7-point Likert scale (strongly agree to strongly disagree) and organized into those retained (≥75% strongly agree to somewhat agree), discarded (≥80% somewhat disagree to strongly disagree), and no consensus reached. Survey participants included a diverse sample of healthcare leaders with the potential to impact policy reform. RESULTS: Of 56 healthcare leaders invited to participate, there were 34 unique responses (61% response rate). Participants were from diverse clinical backgrounds, including surgical subspecialties, medicine, nursing, and healthcare administration and held institutional or provincial leadership positions. A total of 11 of 20 policy recommendations reached the threshold for consensus agreement with the remaining 9 having no consensus reached. CONCLUSION: Consensus agreement was reached among Canadian healthcare leaders on 11 policy recommendations to address the diagnostic, screening, and procedure backlog in Ontario. Recommendations included strategies to address patient information needs on expected wait times, expand health and human resource capacity, and streamline efficiencies to increase operating room output. No consensus was reached on the optimal funding strategy within the public system in Ontario or the appropriateness of implementing private funding models.

Characterizing the Impact of Procedure Funding on the Covid-19 Generated Procedure Gap in Ontario: A Population-Based Analysis.

BACKGROUND: Surgical procedures in Canada were historically funded through global hospital budgets. Activity-based funding models were developed to improve access, equity, timeliness, and value of care for priority areas. COVID-19 upended health priorities and resulted in unprecedented disruptions to surgical care, which created a significant procedure gap. We hypothesized that activity-based funding models influenced the magnitude and trajectory of this procedure gap. METHODS: Population-based analysis of procedure rates comparing the pandemic (March 1, 2020-December 31, 2021) to a prepandemic baseline (January 1, 2017-February 29, 2020) in Ontario, Canada. Poisson generalized estimating equation models were used to predict expected rates in the pandemic based on the prepandemic baseline. Analyses were stratified by procedure type (outpatient, inpatient), body region, and funding category (activity-based funding programs vs. global budget). RESULTS: In all, 281,328 fewer scheduled procedures were performed during the COVID-19 period compared with the prepandemic baseline (Rate Ratio 0.78; 95% CI 0.77-0.80). Inpatient procedures saw a larger reduction (24.8%) in volume compared with outpatient procedures (20.5%). An increase in the proportion of procedures funded through activity-based programs was seen during the pandemic (52%) relative to the prepandemic baseline (50%). Body systems funded predominantly through global hospital budgets (eg, gynecology, otologic surgery) saw the least months at or above baseline volumes, whereas those with multiple activity-based funding options (eg, musculoskeletal, abdominal) saw the most months at or above baseline volumes. CONCLUSIONS: Those needing procedures funded through global hospital budgets may have been disproportionately disadvantaged by pandemic-related health care disruptions.

Regional variation in lower extremity revascularization and amputation for peripheral artery disease.

OBJECTIVE: The aim of this study was to quantify the recent and historical extent of regional variation in revascularization and amputation for peripheral artery disease (PAD). METHODS: This was a repeated cross-sectional analysis of all Ontarians aged 40 years or greater between 2002 and 2019. The co-primary outcomes were revascularization (endovascular or open) and major (above-ankle) amputation for PAD. For each of 14 health care administrative regions, rates per 100,000 person-years (PY) were calculated for 6-year time periods from the fiscal years 2002 to 2019. Rates were directly standardized for regional demographics (age, sex, income) and comorbidities (congestive heart failure, diabetes, chronic obstructive pulmonary disease, chronic kidney disease). The extent of regional variation in revascularization and major amputation rates for each time period was quantified by the ratio of 90th over the 10th percentile (PRR). RESULTS: In 2014 to 2019, there were large differences across regions in demographics (rural living [range, 0%-39.4%], lowest neighborhood income quintile [range, 10.1%-25.5%]) and comorbidities (diabetes [range, 14.2%-22.0%], chronic obstructive pulmonary disease [range, 7.8%-17.9%]), and chronic kidney disease [range, 2.1%-4.0%]. Standardized revascularization rates ranged across regions from 52.6 to 132.6/100,000 PY and standardized major amputation rates ranged from 10.0 to 37.7/100,000 PY. The extent of regional variation was large (PRR ≥2.0) for both revascularization and major amputation. From 2002-2004 to 2017-2019, the extent of regional variation increased from moderate to large for revascularization (standardized PRR, 1.87 to 2.04) and major amputation (standardized PRR, 1.94 to 3.07). CONCLUSIONS: Significant regional differences in revascularization and major amputation rates related to PAD remain after standardizing for regional differences in demographics and comorbidities. These differences have not improved over time.

Predicting outcomes following open revascularization for aortoiliac occlusive disease using machine learning.

OBJECTIVE: Open surgical treatment options for aortoiliac occlusive disease carry significant perioperative risks; however, outcome prediction tools remain limited. Using machine learning (ML), we developed automated algorithms that predict 30-day outcomes following open aortoiliac revascularization. METHODS: The National Surgical Quality Improvement Program (NSQIP) targeted vascular database was used to identify patients who underwent open aortoiliac revascularization for atherosclerotic disease between 2011 and 2021. Input features included 38 preoperative demographic/clinical variables. The primary outcome was 30-day major adverse limb event (MALE; composite of untreated loss of patency, major reintervention, or major amputation) or death. The 30-day secondary outcomes were individual components of the primary outcome, major adverse cardiovascular event (MACE; composite of myocardial infarction, stroke, or death), individual components of MACE, wound complication, bleeding, other morbidity, non-home discharge, and unplanned readmission. Our data were split into training (70%) and test (30%) sets. Using 10-fold cross-validation, we trained six ML models using preoperative features. The primary model evaluation metric was area under the receiver operating characteristic curve (AUROC). Model robustness was evaluated with calibration plot and Brier score. Variable importance scores were calculated to determine the top 10 predictive features. Performance was assessed on subgroups based on age, sex, race, ethnicity, symptom status, procedure type, and urgency. RESULTS: Overall, 9649 patients were included. The primary outcome of 30-day MALE or death occurred in 1021 patients (10.6%). Our best performing prediction model for 30-day MALE or death was XGBoost, achieving an AUROC of 0.95 (95% confidence interval [CI], 0.94-0.96). In comparison, logistic regression had an AUROC of 0.79 (95% CI, 0.77-0.81). For 30-day secondary outcomes, XGBoost achieved AUROCs between 0.87 and 0.97 (untreated loss of patency [0.95], major reintervention [0.88], major amputation [0.96], death [0.97], MACE [0.95], myocardial infarction [0.88], stroke [0.93], wound complication [0.94], bleeding [0.87], other morbidity [0.96], non-home discharge [0.90], and unplanned readmission [0.91]). The calibration plot showed good agreement between predicted and observed event probabilities with a Brier score of 0.05. The strongest predictive feature in our algorithm was chronic limb-threatening ischemia. Model performance remained robust on all subgroup analyses of specific demographic/clinical populations. CONCLUSIONS: Our ML models accurately predict 30-day outcomes following open aortoiliac revascularization using preoperative data, performing better than logistic regression. They have potential for important utility in guiding risk-mitigation strategies for patients being considered for open aortoiliac revascularization to improve outcomes.

Using machine learning to predict outcomes following carotid endarterectomy.

OBJECTIVE: Prediction of outcomes following carotid endarterectomy (CEA) remains challenging, with a lack of standardized tools to guide perioperative management. We used machine learning (ML) to develop automated algorithms that predict outcomes following CEA. METHODS: The Vascular Quality Initiative (VQI) database was used to identify patients who underwent CEA between 2003 and 2022. We identified 71 potential predictor variables (features) from the index hospitalization (43 preoperative [demographic/clinical], 21 intraoperative [procedural], and 7 postoperative [in-hospital complications]). The primary outcome was stroke or death at 1 year following CEA. Our data were split into training (70%) and test (30%) sets. Using 10-fold cross-validation, we trained six ML models using preoperative features (Extreme Gradient Boosting [XGBoost], random forest, Naïve Bayes classifier, support vector machine, artificial neural network, and logistic regression). The primary model evaluation metric was area under the receiver operating characteristic curve (AUROC). After selecting the best performing algorithm, additional models were built using intra- and postoperative data. Model robustness was evaluated using calibration plots and Brier scores. Performance was assessed on subgroups based on age, sex, race, ethnicity, insurance status, symptom status, and urgency of surgery. RESULTS: Overall, 166,369 patients underwent CEA during the study period. In total, 7749 patients (4.7%) had the primary outcome of stroke or death at 1 year. Patients with an outcome were older with more comorbidities, had poorer functional status, and demonstrated higher risk anatomic features. They were also more likely to undergo intraoperative surgical re-exploration and have in-hospital complications. Our best performing prediction model at the preoperative stage was XGBoost, achieving an AUROC of 0.90 (95% confidence interval [CI], 0.89-0.91). In comparison, logistic regression had an AUROC of 0.65 (95% CI, 0.63-0.67), and existing tools in the literature demonstrate AUROCs ranging from 0.58 to 0.74. Our XGBoost models maintained excellent performance at the intra- and postoperative stages, with AUROCs of 0.90 (95% CI, 0.89-0.91) and 0.94 (95% CI, 0.93-0.95), respectively. Calibration plots showed good agreement between predicted and observed event probabilities with Brier scores of 0.15 (preoperative), 0.14 (intraoperative), and 0.11 (postoperative). Of the top 10 predictors, eight were preoperative features, including comorbidities, functional status, and previous procedures. Model performance remained robust on all subgroup analyses. CONCLUSIONS: We developed ML models that accurately predict outcomes following CEA. Our algorithms perform better than logistic regression and existing tools, and therefore, have potential for important utility in guiding perioperative risk mitigation strategies to prevent adverse outcomes.

Using machine learning to predict outcomes following open abdominal aortic aneurysm repair.

OBJECTIVE: Prediction of outcomes following open abdominal aortic aneurysm (AAA) repair remains challenging with a lack of widely used tools to guide perioperative management. We developed machine learning (ML) algorithms that predict outcomes following open AAA repair. METHODS: The Vascular Quality Initiative (VQI) database was used to identify patients who underwent elective open AAA repair between 2003 and 2023. Input features included 52 preoperative demographic/clinical variables. All available preoperative variables from VQI were used to maximize predictive performance. The primary outcome was in-hospital major adverse cardiovascular event (MACE; composite of myocardial infarction, stroke, or death). Secondary outcomes were individual components of the primary outcome, other in-hospital complications, and 1-year mortality and any reintervention. We split our data into training (70%) and test (30%) sets. Using 10-fold cross-validation, six ML models were trained using preoperative features (Extreme Gradient Boosting [XGBoost], random forest, Naïve Bayes classifier, support vector machine, artificial neural network, and logistic regression). The primary model evaluation metric was area under the receiver operating characteristic curve (AUROC). Model robustness was evaluated with calibration plot and Brier score. The top 10 predictive features in our final model were determined based on variable importance scores. Performance was assessed on subgroups based on age, sex, race, ethnicity, rurality, median area deprivation index, proximal clamp site, prior aortic surgery, and concomitant procedures. RESULTS: Overall, 12,027 patients were included. The primary outcome of in-hospital MACE occurred in 630 patients (5.2%). Compared with patients without a primary outcome, those who developed in-hospital MACE were older with more comorbidities, demonstrated poorer functional status, had more complex aneurysms, and were more likely to require concomitant procedures. Our best performing prediction model for in-hospital MACE was XGBoost, achieving an AUROC of 0.93 (95% confidence interval, 0.92-0.94). Comparatively, logistic regression had an AUROC of 0.71 (95% confidence interval, 0.70-0.73). For secondary outcomes, XGBoost achieved AUROCs between 0.84 and 0.94. The calibration plot showed good agreement between predicted and observed event probabilities with a Brier score of 0.05. These findings highlight the excellent predictive performance of the XGBoost model. The top three predictive features in our algorithm for in-hospital MACE following open AAA repair were: (1) coronary artery disease; (2) American Society of Anesthesiologists classification; and (3) proximal clamp site. Model performance remained robust on all subgroup analyses. CONCLUSIONS: Open AAA repair outcomes can be accurately predicted using preoperative data with our ML models, which perform better than logistic regression. Our automated algorithms can help guide risk-mitigation strategies for patients being considered for open AAA repair to improve outcomes.

Machine learning to predict outcomes following endovascular abdominal aortic aneurysm repair.

BACKGROUND: Endovascular aneurysm repair (EVAR) for abdominal aortic aneurysm (AAA) carries important perioperative risks; however, there are no widely used outcome prediction tools. The aim of this study was to apply machine learning (ML) to develop automated algorithms that predict 1-year mortality following EVAR. METHODS: The Vascular Quality Initiative database was used to identify patients who underwent elective EVAR for infrarenal AAA between 2003 and 2023. Input features included 47 preoperative demographic/clinical variables. The primary outcome was 1-year all-cause mortality. Data were split into training (70 per cent) and test (30 per cent) sets. Using 10-fold cross-validation, 6 ML models were trained using preoperative features with logistic regression as the baseline comparator. The primary model evaluation metric was area under the receiver operating characteristic curve (AUROC). Model robustness was evaluated with calibration plot and Brier score. RESULTS: Some 63 655 patients were included. One-year mortality occurred in 3122 (4.9 per cent) patients. The best performing prediction model for 1-year mortality was XGBoost, achieving an AUROC (95 per cent c.i.) of 0.96 (0.95-0.97). Comparatively, logistic regression had an AUROC (95 per cent c.i.) of 0.69 (0.68-0.71). The calibration plot showed good agreement between predicted and observed event probabilities with a Brier score of 0.04. The top 3 predictive features in the algorithm were 1) unfit for open AAA repair, 2) functional status, and 3) preoperative dialysis. CONCLUSIONS: In this data set, machine learning was able to predict 1-year mortality following EVAR using preoperative data and outperformed standard logistic regression models.

The influence of diabetes on temporal trends in lower extremity revascularisation and amputation for peripheral artery disease: A population-based repeated cross-sectional analysis.

AIM/HYPOTHESIS: To describe the influence of diabetes on temporal changes in rates of lower extremity revascularisation and amputation for peripheral artery disease (PAD) in Ontario, Canada. METHODS: In this population-based repeated cross-sectional study, we calculated annual rates of lower extremity revascularisation (open or endovascular) and amputation (toe, foot or leg) related to PAD among Ontario residents aged ≥40 years between 2002 and 2019. Annual rate ratios (relative to 2002) adjusted for changes in diabetes prevalence alone, as well as fully adjusted for changes in demographics, diabetes and other comorbidities, were estimated using generalized estimating equation models to model population-level effects while accounting for correlation within units of observation. RESULTS: Compared with 2002, the Ontario population in 2019 exhibited a significantly higher prevalence of diabetes (18% vs. 10%). Between 2002 and 2019, the crude rate of revascularisation increased from 75.1 to 90.7/100,000 person-years (unadjusted RR = 1.10, 95% CI = 1.07-1.13). However, after adjustment, there was no longer an increase in the rate of revascularisation (diabetes-adjusted RR = 0.98, 95% CI = 0.96-1.01, fully-adjusted RR = 0.94, 95% CI = 0.91-0.96). The crude rate of amputation decreased from 2002 to 2019 from 49.5 to 45.4/100,000 person-years (unadjusted RR = 0.78, 95% CI = 0.75-0.81), but was more pronounced after adjustment (diabetes-adjusted RR = 0.62, 95% CI = 0.60-0.64; fully-adjusted RR = 0.58, 95% CI = 0.56-0.60). CONCLUSIONS/INTERPRETATION: Diabetes prevalence rates strongly influenced rates of revascularisation and amputation related to PAD. A decrease in amputations related to PAD over time was attenuated by rising diabetes prevalence rates.

The evolving use of robotic surgery: a population-based analysis.

INTRODUCTION: Robotic surgery has integrated into the healthcare system despite limited evidence demonstrating its clinical benefit. Our objectives were (i) to describe secular trends and (ii) patient- and system-level determinants of the receipt of robotic as compared to open or laparoscopic surgery. METHODS: This population-based retrospective cohort study included adult patients who, between 2009 and 2018 in Ontario, Canada, underwent one of four commonly performed robotic procedures: radical prostatectomy, total hysterectomy, thoracic lobectomy, partial nephrectomy. Patients were categorized based on the surgical approach as robotic, open, or laparoscopic for each procedure. Multivariable regression models were used to estimate the temporal trend in robotic surgery use and associations of patient and system characteristics with the surgical approach. RESULTS: The cohort included 24,741 radical prostatectomy, 75,473 total hysterectomy, 18,252 thoracic lobectomy, and 4608 partial nephrectomy patients, of which 6.21% were robotic. After adjusting for patient and system characteristics, the rate of robotic surgery increased by 24% annually (RR 1.24, 95%CI 1.13-1.35): 13% (RR 1.13, 95%CI 1.11-1.16) for robotic radical prostatectomy, 9% (RR 1.09, 95%CI 1.05-1.13) for robotic total hysterectomy, 26% (RR 1.26, 95%CI 1.06-1.50) for thoracic lobectomy and 26% (RR 1.26, 95%CI 1.13-1.40) for partial nephrectomy. Lower comorbidity burden, earlier disease stage (among cancer cases), and early career surgeons with high case volume at a teaching hospital were consistently associated with the receipt of robotic surgery. CONCLUSION: The use of robotic surgery has increased. The study of the real-world clinical outcomes and associated costs is needed before further expanding use among additional providers and hospitals.

A novel Canadian multidisciplinary acute care pathway for people hospitalised with a diabetic foot ulcer.

This manuscript describes the implementation and initial evaluation of a novel Canadian acute care pathway for people with a diabetic foot ulcer (DFU). A multidisciplinary team developed and implemented an acute care pathway for patients with a DFU who presented to the emergency department (ED) and required hospitalisation at a tertiary care hospital in Canada. Processes of care, length of stay (LOS), and hospitalisation costs were considered through retrospective cohort study of all DFU hospitalizations from pathway launch in December 2018 to December 2020. There were 82 DFU-related hospital admissions through the ED of which 55 required invasive intervention: 28 (34.1%) minor amputations, 16 (19.5%) abscess drainage and debridement, 6 (7.3%) lower extremity revascularisations, 5 (6.1%) major amputations. Mean hospital LOS was 8.8 ± 4.9 days. Mean hospitalisation cost was $20 569 (±14 143): $25 901 (±15 965) when surgical intervention was required and $9279 (±7106) when it was not. LOS and hospitalisation costs compared favourably to historical data. An acute care DFU pathway can support the efficient evaluation and management of patients hospitalised with a DFU. A dedicated multidisciplinary DFU care team is a valuable resource for hospitals in Canada.

Management and In-hospital Mortality of 2203 Patients With a Traumatic Intimal Tear of the Thoracic Aorta.

OBJECTIVE: Our goal was to describe contemporary management and inhospital mortality associated with blunt thoracic aortic intimal tears (IT) within the American College of Surgeons Trauma Quality Improvement Program. SUMMARY BACKGROUND DATA: The evidence basis for nonoperative expectant management of traumatic iT of the thoracic aorta remains weak. METHODS: All adult patients with a thoracic aortic IT following blunt trauma were captured from Level I and II North American Centers enrolled in Trauma Quality Improvement Program from 2010 to 2017. For each patient, we extracted demographics, injury characteristics, the timing and approach of thoracic aortic repair and in-hospital mortality. Mortality attributable to IT was calculated by comparing IT patients to a propensity-score matched control cohort of severely injured blunt trauma patients without aortic injury. RESULTS: There were 2203 IT patients across 315 facilities. Injury most often resulted from motor vehicle collision (75%). A total of 758 patients (34%) underwent operative management, with 93% (N = 708) of repairs performed via an endovascular approach. Median time to surgery was 11 hours (IQR 4- 40). The frequency of operative management was higher in patients without traumatic brain injury (TBI) (35%, N = 674) compared to those with TBI (29%, N = 84) (P = 0.024). Compared to severely injured blunt trauma patients without aortic injury, ITwas not associated with additional in-hospital mortality (10.7% for IT vs 11.7% for no IT, absolute risk difference: -1.0%, 95% CI: -2.9% to 0.8%). CONCLUSIONS: The majority of blunt thoracic IT are managed nonoperatively and IT does not confer additional in-hospital mortality risk. Future studies should focus on the risk of injury progression.

A Population-based Analysis of the COVID-19 Generated Surgical Backlog and Associated Emergency Department Presentations for Inguinal Hernias and Gallstone Disease.

OBJECTIVE: To evaluate the downstream effects of the COVID-19 generated surgical backlog. BACKGROUND: Delayed elective surgeries may result in emergency department (ED) presentations and the need for urgent interventions. METHODS: Population-based repeated cross-sectional study utilizing administrative data. We quantified rates of elective cholecystectomy and inguinal hernia repair and rates of ED presentations, urgent interventions, and outcomes during the first and second waves of COVID-19 (March 1, 2020- February 28, 2021) as compared to a 3-year pre-COVID-19 period (January 1, 2017-February 29, 2020) in Ontario, Canada. Poisson generalized estimating equation models were used to predict expected rates during COVID-19 based on the pre-COVID-19 period. The ratio of observed (actual events) to expected rates was generated for surgical procedures (SRRs) and ED visits (ED-RRs). RESULTS: We identified 74,709 elective cholecystectomies and 60,038 elective inguinal hernia repairs. During the COVID-19 period, elective inguinal hernia repairs decreased by 21% (SRR 0.791; 0.760-0.824) whereas elective cholecystectomies decreased by 23% (SRR 0.773; 0.732-0.816). ED visits for inguinal hernia decreased by 17% (ED-RR 0.829; 0.786 - 0.874) whereas ED visits for gallstones decreased by 8% (ED-RR 0.922; 0.878 - 0.967). A higher population rate of urgent cholecystectomy was observed, particularly after the first wave (SRR 1.076; 1.000-1.158). No difference was seen in inguinal hernias. CONCLUSIONS: An over 20% reduction in elective surgeries and an increase in urgent cholecystectomies was observed during the COVID-19 period suggesting a rebound effect secondary to the surgical backlog. The COVID-19 generated surgical backlog will have a heterogeneous downstream effect with significant implications for surgical recovery planning.