Prediction of acute kidney injury risk after cardiac surgery: using a hybrid machine learning algorithm.

BACKGROUND: Acute kidney injury (AKI) is a serious complication after cardiac surgery. We derived and internally validated a Machine Learning preoperative model to predict cardiac surgery-associated AKI of any severity and compared its performance with parametric statistical models. METHODS: We conducted a retrospective study of adult patients who underwent major cardiac surgery requiring cardiopulmonary bypass between November 1st, 2009 and March 31st, 2015. AKI was defined according to the KDIGO criteria as stage 1 or greater, within 7 days of surgery. We randomly split the cohort into derivation and validation datasets. We developed three AKI risk models: (1) a hybrid machine learning (ML) algorithm, using Random Forests for variable selection, followed by high performance logistic regression; (2) a traditional logistic regression model and (3) an enhanced logistic regression model with 500 bootstraps, with backward variable selection. For each model, we assigned risk scores to each of the retained covariate and assessed model discrimination (C statistic) and calibration (Hosmer-Lemeshow goodness-of-fit test) in the validation datasets. RESULTS: Of 6522 included patients, 1760 (27.0%) developed AKI. The best performance was achieved by the hybrid ML algorithm to predict AKI of any severity. The ML and enhanced statistical models remained robust after internal validation (C statistic = 0.75; Hosmer-Lemeshow p = 0.804, and AUC = 0.74, Hosmer-Lemeshow p = 0.347, respectively). CONCLUSIONS: We demonstrated that a hybrid ML model provides higher accuracy without sacrificing parsimony, computational efficiency, or interpretability, when compared with parametric statistical models. This score-based model can easily be used at the bedside to identify high-risk patients who may benefit from intensive perioperative monitoring and personalized management strategies.

Derivation and validation of a clinical model to predict death or cardiac hospitalizations while on the cardiac surgery waitlist.

BACKGROUND: Waitlist management is a global challenge. For patients with severe cardiovascular diseases awaiting cardiac surgery, prolonged wait times are associated with unplanned hospitalizations. To facilitate evidence-based resource allocation, we derived and validated a clinical risk model to predict the composite outcome of death and cardiac hospitalization of patients on the waitlist for cardiac surgery. METHODS: We used the CorHealth Ontario Registry and linked ICES health care administrative databases, which have information on all Ontario residents. We included patients 18 years or older who waited at home for coronary artery bypass grafting, valvular or thoracic aorta surgeries between 2008 and 2019. The primary outcome was death or an unplanned cardiac hospitalizaton, defined as nonelective admission for heart failure, myocardial infarction, unstable angina or endocarditis. We randomly divided two-thirds of these patients into derivation and one-third into validation data sets. We derived the model using a multivariable Cox proportional hazard model with backward stepwise variable selection. RESULTS: Among 62 375 patients, 41 729 patients were part of the derivation data set and 20 583 were part of the validation data set. Of the total, 3033 (4.9%) died or had an unplanned cardiac hospitalization while waiting for surgery. The area under the curve of our model at 15, 30, 60 and 89 days was 0.85, 0.82, 0.81 and 0.80, respectively, in the derivation cohort and 0.83, 0.80, 0.78 and 0.78, respctively, in the validation cohort. The model calibrated well at all time points. INTERPRETATION: We derived and validated a clinical risk model that provides accurate prediction of the risk of death and unplanned cardiac hospitalization for patients on the cardiac surgery waitlist. Our model could be used for quality benchmarking and data-driven decision support for managing access to cardiac surgery.

Clinical Impact of Changes in Left Ventricular Function After Aortic Valve Replacement: Analysis From 3112 Patients.

BACKGROUND: Our objectives were to identify correlates of mortality and congestive heart failure after aortic valve replacement (AVR) according to preoperative left ventricular (LV) function and to describe the incidence, time course, and correlates of LV recovery and mass regression postoperatively. METHODS AND RESULTS: A total of 3112 patients with AVR were assessed in a follow-up clinic with echocardiography (median follow-up, 6.0 years). At operation, their mean age was 67.8±13.4 years, one third were female, and 29% had LV dysfunction (ejection fraction <50%). In severe patients with severe aortic stenosis and LV dysfunction, transaortic valve mean pressure gradient <40 mm Hg, longer cardiopulmonary bypass duration, and prosthesis-patient mismatch (indexed effective orifice area ≤0.85 cm(2)/m(2)) were independent correlates of the composite outcome of death or congestive heart failure after AVR. In patients with severe aortic regurgitation and LV dysfunction, older age and higher preoperative LV mass were identified. LV recovery correlated with better survival and freedom from heart failure in patients with aortic stenosis. Maximum LV mass regression took 24 months in patients with aortic stenosis and nearly 5 years with aortic regurgitation; independent correlates included smaller LV end-systolic diameter in patients with aortic stenosis and low New York Heart Association class with aortic regurgitation. CONCLUSIONS: Incomplete LV recovery, prosthesis-patient mismatch, low transaortic valve pressure gradient, and higher LV mass are associated with increased mortality or heart failure after AVR in patients with LV dysfunction. Higher LV end-systolic diameter and symptoms correlate with less LV mass regression, which takes at least 2 years. These findings help surgeons and cardiologists refine the indications, timing, prognostication, and follow-up of patients before and after AVR.

Reoperation of left heart valve bioprostheses according to age at implantation.

BACKGROUND: Evidence supporting the use of bioprostheses for heart valve replacement in young adults is accumulating. However, reoperation data, which may help guide clinical decision making in young patients, remains poorly defined in the literature. METHODS AND RESULTS: We examined the need for reoperation in 3975 patients who underwent first-time bioprosthetic aortic valve replacement (AVR) (n=3152) or mitral valve replacement (MVR) (n=823). There were 895 patients below the age of 60 years at bioprosthesis implant (AVR, n=636; MVR, n=259). The median interval to reoperation of contemporary, stented aortic bioprostheses was 7.74 years (95% CI 7.28 to 9.97 years) in patients less than 40 years, and 12.93 years (95% CI 11.10 to 15.76 years) in patients between 40 and 60 years of age. Multivariable risk factors associated with reoperation following bioprosthetic AVR include age (hazard ratio [HR] 0.94 per year, 95% CI 0.91 to 0.96, P<0.001) and concomitant coronary artery bypass grafting (HR 0.34, 95% CI 0.11 to 0.99, P=0.04). The median interval to reoperation of contemporary mitral bioprostheses was 8.11 years (95% CI 5.79 to 16.50 years) in patients less than 40 years, and 10.14 years (95% CI 8.64 to 11.14 years) in patients between 40 and 60 years of age. As for AVR, age (HR 0.96 per year, 95% CI 0.95 to 0.98, P<0.001) and concomitant coronary artery bypass grafting (HR 0.55, 95% CI 0.32 to 0.93, P=0.03) were associated with decreased reoperation risk following bioprosthetic MVR. CONCLUSIONS: These data constitute clinically relevant age-specific prognostic information regarding reoperation in young patients, who may wish to select a bioprosthesis at initial left heart valve replacement.

Aortic Valve Repair Decreases Risks of VRE in AI at 10 Years: A Propensity Score-Matched Analysis.

BACKGROUND: Aortic valve repair (AVr) has emerged as a feasible and effective alternative to AV replacement (AVR) in patients with aortic insufficiency (AI); however, little data exist comparing outcomes. Thus, the objective of this study was to compare early and long-term valve-related complications between AVr and AVR in the treatment of AI. METHODS: This was a single-center, retrospective study of all patients (n = 417) undergoing AVr (n = 264) or AVR (n = 153) for primary AI. Propensity matching using a 1:1 greedy matching algorithm identified 140 patients using 6 covariates (age, sex, left ventricular function, size, presence of aortopathy, and urgency of operation) for comparison. The primary outcome was a composite of all valve-related events (VREs), including endocarditis, myocardial infarction stroke, transient ischemic attack, thromboembolisms, bleeding, and AV reoperation. VREs were defined as per published guidelines. Survival and freedom from VREs were reported using the Kaplan-Meier method. RESULTS: Propensity matching identified 70 well-matched pairs with no major differences in baseline demographics, comorbidities, or AI severity (P = .57). Perioperative outcomes showed no significant differences in VREs (AVR 8 vs AVr 7; P = .78) or mortality (AVR 3 vs AVr 1; P = .62). Event-free survival from the primary outcome at 10 years was significantly better after AVr than after AVR (82% vs 68%; P = .024), with no significant differences in 10-year overall survival between groups (82% vs 72%; P = .29). No significant differences in AI severity (P = .07) or reoperation rate (P = .44) were detected between groups. CONCLUSIONS: This study demonstrated a lower long-term risk of VREs with repair compared with replacement, with low mortality and comparable durability. Further prospective randomized control trials are necessary to formally compare outcomes and determine superiority.

Derivation and validation of a clinical risk score to predict death among patients awaiting cardiac surgery in Ontario, Canada: a population-based study.

BACKGROUND: Surgical delay may result in unintended harm to patients needing cardiac surgery, who are at risk for death if their condition is left untreated. Our objective was to derive and internally validate a clinical risk score to predict death among patients awaiting major cardiac surgery. METHODS: We used the CorHealth Ontario Registry and linked ICES health administrative databases with information on all Ontario residents to identify patients aged 18 years or more who were referred for isolated coronary artery bypass grafting (CABG), valvular procedures, combined CABG-valvular procedures or thoracic aorta procedures between Oct. 1, 2008, and Sept. 30, 2019. We used a hybrid modelling approach with the random forest method for initial variable selection, followed by backward stepwise logistic regression modelling for clinical interpretability and parsimony. We internally validated the logistic regression model, termed the CardiOttawa Waitlist Mortality Score, using 200 bootstraps. RESULTS: Of the 112 266 patients referred for cardiac surgery, 269 (0.2%) died while awaiting surgery (118/72 366 [0.2%] isolated CABG, 81/24 461 [0.3%] valvular procedures, 63/12 046 [0.5%] combined CABG-valvular procedures and 7/3393 [0.2%] thoracic aorta procedures). Age, sex, surgery type, left main stenosis, Canadian Cardiovascular Society classification, left ventricular ejection fraction, heart failure, atrial fibrillation, dialysis, psychosis and operative priority were predictors of waitlist mortality. The model discriminated (C-statistic 0.76 [optimism-corrected 0.73]). It calibrated well in the overall cohort (Hosmer-Lemeshow p = 0.2) and across surgery types. INTERPRETATION: The CardiOttawa Waitlist Mortality Score is a simple clinical risk model that predicts the likelihood of death while awaiting cardiac surgery. It has the potential to provide data-driven decision support for managing access to cardiac care and preserve system capacity during the COVID-19 pandemic, the recovery period and beyond.

Long-term clinical and hemodynamic performance of the Hancock II versus the Perimount aortic bioprostheses.

BACKGROUND: The Medtronic Hancock II and the Carpentier-Edwards Perimount are among the world's most commonly used aortic bioprostheses. However, a direct comparison of their clinical performance is lacking. To minimize biases inherent to between-center comparisons, we examined these prostheses within a large, contemporary, single-center cohort. METHODS AND RESULTS: Between 1990 and 2007, 1659 patients (mean age, 73.1±9.3 years) underwent aortic valve replacement with either the Hancock II (N=1021) or the Perimount (N=638). Patients were prospectively followed-up with serial clinic visits and echocardiograms for up to 16 years (mean, 5.0±3.3 years). There was no significant difference in aortic root size preoperatively (P=0.7). Aortic root enlargement was more commonly performed with the Perimount (P<0.001), and the manufacturer valve size of the implanted prosthesis was larger with the Hancock II (P<0.001). Postoperatively, peak and mean transprosthesis gradients were higher for the Hancock II (32.7±0.7 and 16.0±0.3 mm Hg, respectively) than for the Perimount (24.9±0.7 and 13.4±0.4 mm Hg, respectively; P<0.001). However, no difference in left ventricular mass regression was observed at late follow-up (P=0.9). Unadjusted 10-year survival was 59.4%±2.4% for the Hancock II and 70.2%±3.8% for the Perimount (P=0.07). Multivariable predictors of survival did not include prosthesis type (P=0.2). CONCLUSIONS: For the same manufacturer valve size, the Perimount is larger, which may warrant enlarging the aortic root more often, and it is associated with better hemodynamics than the Hancock II. These differences do not impact survival or left ventricular mass regression, and the long-term clinical performances of the Hancock II and Perimount bioprostheses are equivalent.

Aspirin plus clopidogrel versus aspirin alone after coronary artery bypass grafting: the clopidogrel after surgery for coronary artery disease (CASCADE) Trial.

BACKGROUND: clopidogrel inhibits intimal hyperplasia in animal studies and therefore may reduce saphenous vein graft (SVG) intimal hyperplasia after coronary artery bypass grafting. The Clopidogrel After Surgery for Coronary Artery DiseasE (CASCADE) study was undertaken to evaluate whether the addition of clopidogrel to aspirin inhibits SVG disease after coronary artery bypass grafting, as assessed at 1 year by intravascular ultrasound. METHODS AND RESULTS: in this double-blind phase II trial, 113 patients undergoing coronary artery bypass grafting with SVGs were randomized to receive aspirin 162 mg plus clopidogrel 75 mg daily or aspirin 162 mg plus placebo daily for 1 year. The primary outcome was SVG intimal hyperplasia (mean intimal area) as determined by intravascular ultrasound at 1 year. Secondary outcomes were graft patency, major adverse cardiovascular events, and major bleeding. One-year intravascular ultrasound and coronary angiography were performed in 92 patients (81.4%). At 1 year, SVG intimal area did not differ significantly between the 2 groups (4.1 ± 2.0 versus 4.5 ± 2.1 mm(2), aspirin-clopidogrel versus aspirin-placebo, P=0.44). Overall 1-year graft patency was 95.2% in the aspirin-clopidogrel group compared with 95.5% in the aspirin-placebo group (P=0.90), and SVG patency was 94.3% in the aspirin-clopidogrel group versus 93.2% in the aspirin-placebo group (P=0.69). Freedom from major adverse cardiovascular events at 1 year was 92.9 ± 3.4% in the aspirin-clopidogrel group and 91.1 ± 3.8% in the aspirin-placebo group (P=0.76). The incidence of major bleeding at 1 year was similar for the 2 groups (1.8% versus 0%, aspirin-clopidogrel versus aspirin-placebo, P=0.50). CONCLUSIONS: compared with aspirin monotherapy, the combination of aspirin plus clopidogrel did not significantly reduce the process of SVG intimal hyperplasia 1 year after coronary artery bypass grafting.

Disability-free survival after major cardiac surgery: a population-based retrospective cohort study.

BACKGROUND: Cardiovascular research has traditionally been dedicated to "tombstone" outcomes, with little attention dedicated to the patient's perspective. We evaluated disability-free survival as a patient-defined outcome after cardiac surgery. METHODS: We conducted a retrospective cohort study of patients aged 40 years and older who underwent coronary artery bypass grafting (CABG) or single or multiple valve (aortic, mitral, tricuspid) surgery in Ontario between Oct. 1, 2008, and Dec. 31, 2016. The primary outcome was disability (a composite of stroke, 3 or more nonelective hospital admissions and admission to a long-term care facility) within 1 year after surgery. We assessed the procedure-specific risk of disability using cumulative incidence functions, and the relative effect of covariates on the subdistribution hazard using Fine and Gray models. RESULTS: The study included 72 824 patients. The 1-year incidence of disability and death was 2431 (4.6%) and 1839 (3.5%) for CABG, 677 (6.5%) and 539 (5.2%) for single valve, 118 (9.0%) and 140 (10.7%) for multiple valve, 718 (9.0%) and 730 (9.2%) for CABG and single valve, and 87 (13.1%) and 94 (14.1%) for CABG and multiple valve surgery, respectively. With CABG as the reference group, the adjusted hazard ratios for disability were 1.34 (95% confidence interval [CI] 1.21-1.48) after single valve, 1.43 (95% CI 1.18-1.75) after multiple valve, 1.38 (95% CI 1.26-1.51) after CABG and single valve, and 1.78 (95% CI 1.43-2.23) after CABG and multiple valve surgery. Combined CABG and multiple valve surgery, heart failure, creatinine 180 µmol/L or greater, alcohol use disorder, dementia and depression were independent risk factors for disability. INTERPRETATION: The cumulative incidence of disability was lowest after CABG and highest after combined CABG and multiple valve surgery. Our findings point to a need for models that predict personalized disability risk to enable better patient-centred care.

Impact of surgeon and anaesthesiologist sex on patient outcomes after cardiac surgery: a population-based study.

BACKGROUND: Effective teamwork between anaesthesiologists and surgeons is essential for optimising patient safety in the cardiac operating room. While many factors may influence the relationship between these two physicians, the role of sex and gender have yet to be investigated. OBJECTIVES: We sought to determine the association between cardiac physician team sex discordance and patient outcomes. DESIGN: We performed a population-based, retrospective cohort study. PARTICIPANTS AND SETTING: Adult patients who underwent coronary artery bypass grafting (CABG) and/or aortic, mitral or tricuspid valve surgery between 2008 and 2018 in Ontario, Canada. PRIMARY AND SECONDARY OUTCOME MEASURES: The primary outcome was all-cause 30-day mortality. Secondary outcomes included major adverse cardiovascular events at 30 days and hospital and intensive care unit lengths of stay (LOS). Mixed effects logistic regression was used for categorical outcomes and Poisson regression for continuous outcomes. RESULTS: 79 862 patients underwent cardiac surgery by 98 surgeons (11.2% female) and 279 anaesthesiologists (23.3% female); 19 893 (24.9%) were treated by sex-discordant physician teams. Physician sex discordance was not associated with overall patient mortality or LOS; however, patients who underwent isolated CABG experienced longer hospital LOS when treated by an all-male physician team as compared with an all-female team (adjusted OR=1.07; p=0.049). When examining the impact of individual physician sex, the length of hospital stay was longer when isolated CABG procedures were attended by a male surgeon (OR=1.10; p=0.004) or anaesthesiologist (OR=1.02; p=0.01). CONCLUSIONS: Patient mortality and length of stay after cardiac surgery may vary by sex concordance of the attending surgeon-anaesthesiologist team. Further research is needed to examine the underlying mechanisms of these observed relationships.