Search

Machine learning (ML) algorithms may enhance outcomes prediction and help guide clinical decision making. This study aimed to develop and validate a ML model that predicts postoperative outcomes and costs after cardiac surgery.The Society of Thoracic Surgeons registry data from 4874 patients who underwent cardiac surgery (56% coronary artery bypass grafting, 42% valve surgery, 19% aortic surgery) at our institution were divided into training (80%) and testing (20%) datasets. The Extreme Gradient Boosting decision-tree ML algorithms were trained to predict three outcomes: operative mortality, major morbidity or mortality, and Medicare outlier high hospitalization cost. Algorithm performance was determined using accuracy, F1 score, and area under the precision-recall curve (AUC-PR). The ML algorithms were validated in index surgery cases with The Society of Thoracic Surgeons risk scores for mortality and major morbidities and with logistic regression and were then applied to nonindex cases.The ML algorithms with 25 input parameters predicted operative mortality (accuracy 95%; F1 0.31; AUC-PR 0.21), major morbidity or mortality (accuracy 71%, F1 0.47; AUC-PR 0.47), and high cost (accuracy 84%; F1 0.62; AUC-PR 0.65). Preoperative creatinine, complete blood count, patient height and weight, ventricular function, and liver dysfunction were important predictors for all outcomes. For patients undergoing nonindex cardiac operations, the ML model achieved an AUC-PR of 0.15 (95% CI, 0.05-0.32) for mortality and 0.59 (95% CI, 0.51-0.68) for major morbidity or mortality.The extreme gradient boosting ML algorithms can predict mortality, major morbidity, and high cost after cardiac surgery, including operations without established risk models. These ML algorithms may refine risk prediction after cardiac surgery for a wide range of procedures.

Machine learning may enhance prediction of outcomes after coronary artery bypass grafting (CABG). We sought to develop and validate a dynamic machine learning model to predict CABG outcomes at clinically relevant pre- and postoperative time points.The Society of Thoracic Surgeons (STS) registry data elements from 2086 isolated CABG patients were divided into training and testing datasets and input into Extreme Gradient Boosting decision-tree machine learning algorithms. Two prediction models were developed based on data from preoperative (80 parameters) and postoperative (125 parameters) phases of care. Outcomes included operative mortality, major morbidity or mortality, high cost, and 30-day readmission. Machine learning and STS model performance were assessed using accuracy and the area under the precision-recall curve (AUC-PR).Preoperative machine learning models predicted mortality (accuracy, 98%; AUC-PR = 0.16; F1 = 0.24), major morbidity or mortality (accuracy, 75%; AUC-PR = 0.33; F1 = 0.42), high cost (accuracy, 83%; AUC-PR = 0.51; F1 = 0.52), and 30-day readmission (accuracy, 70%; AUC-PR = 0.47; F1 = 0.49) with high accuracy. Preoperative machine learning models performed similarly to the STS for prediction of mortality (STS AUC-PR = 0.11; P = .409) and outperformed STS for prediction of mortality or major morbidity (STS AUC-PR = 0.28; P.001). Addition of intraoperative parameters further improved machine learning model performance for major morbidity or mortality (AUC-PR = 0.39; P.01) and high cost (AUC-PR = 0.64; P.01), with cross-clamp and bypass times emerging as important additive predictive parameters.Machine learning can predict mortality, major morbidity, high cost, and readmission after isolated CABG. Prediction based on the phase of care allows for dynamic risk assessment through the hospital course, which may benefit quality assessment and clinical decision-making.

Safety-net hospitals provide essential services to vulnerable patients with complex medical and socioeconomic circumstances. We hypothesized that matched patients at safety-net hospitals and non-safety-net hospitals would have comparable outcomes, costs, and readmission rates after isolated surgical aortic valve replacement (AVR) or mitral valve replacement (MVR).The National Readmissions Database was queried to identify patients who underwent isolated AVR (n = 109 744) or MVR (n = 31 475) from 2016 to 2018. Safety-net burden was defined as the percentage of patients who were uninsured or insured with Medicaid, with hospitals in the top quartile designated as safety-net hospitals. After propensity score matching, outcomes for AVR and MVR at safety-net hospitals vs non-safety-net hospitals were compared.Overall, 17 925 AVRs (16%) and 5516 MVRs (18%) were performed at safety-net hospitals, and these patients had higher comorbidity rates, had lower socioeconomic status, and more frequently required urgent surgery. Observed inhospital mortality was similar between safety-net hospitals and non-safety-net hospitals (AVR 2.2% vs 2.1%, P = .4; MVR 4.8% vs 4.3%, P = .1). After matching, rates of inhospital mortality, major morbidity, and readmission were similar; however, safety-net hospitals had longer length of stay after AVR (7 vs 6 days, P = .001) and higher total cost after AVR ($49 015 vs $42 473, P.001) and MVR ($59 253 vs $52 392, P.001).Isolated surgical AVR and MVR are both performed at safety-net hospitals with outcomes comparable to those at non-safety-net hospitals, supporting efforts to expand access to these procedures for underserved populations. Investment in care coordination resources to reduce length of stay and curtail cost at safety-net hospitals is warranted.

This study evaluated outcomes and risk factors for surgical aortic valve replacement (SAVR) for aortic insufficiency (AI) in a national cohort. We analyzed the incidence, outcomes, and risk factors for SAVR for AI in the Society of Thoracic Surgeons Adult Cardiac Surgery Database.The national database was queried for patients with moderate or greater AI undergoing isolated SAVR between July 2011 and December 2018. Patients with moderate or greater aortic stenosis, acute dissection, active endocarditis, concomitant procedures, or emergent operation were excluded. AI was staged using guideline criteria based on symptoms and ventricular remodeling. Operative mortality and morbidity were compared between stages, and risk factors for operative death were identified.A total of 12,564 patients underwent isolated SAVR for AI from 2011 to 2018. Patients were most frequently AI stage D (7019 [57.5%]), compared with B (1405 [11.2%]), C1 (1128 [9.0%]), or C2 (1325 [10.5%]). Operative mortality was 1.1% overall, and increased between stage C1, C2, and D (0.4% vs 0.7% vs 1.6%, respectively, P.01), along with major morbidity (5.1% vs 7.5% vs 9.9%, respectively; P.01). Mortality was higher in patients with severe ventricular dilation and an ejection fraction of less than 0.30 (2.7% vs 1.0%, P.01). Risk factors for death were symptomatic AI, decreased ejection fraction, age, weight, body surface area, and dialysis.Operative mortality and morbidity for isolated SAVR for AI is very low in a national cohort, providing a benchmark for future transcatheter approaches. Operative risk increases with advanced ventricular remodeling. SAVR before development of ventricular remodeling may be appropriate in patients with severe AI.

OBJECTIVES: Safety-net hospitals (SNHs) provide essential services to predominantly underserved patients regardless of their ability to pay. We hypothesized that patients who underwent coronary artery bypass grafting (CABG) would have inferior observed outcomes at SNHs compared with non-SNHs but that matched cohorts would have comparable outcomes. METHODS: We queried the Nationwide Readmissions Database for patients who underwent isolated CABG from 2016 to 2018. We ranked hospitals by the percentage of all admissions in which the patient was uninsured or insured with Medicaid; hospitals in the top quartile were designated as SNHs. We used propensity-score matching to mitigate the effect of confounding factors and compare outcomes between SNHs and non-SNHs. RESULTS: A total of 525,179 patients underwent CABG, including 96,133 (18.3%) at SNHs, who had a greater burden of baseline comorbidities (median Elixhauser score 8 vs 7; P = .04) and more frequently required urgent surgery (57.1% vs 52.8%; P

Choosing between a bioprosthetic and a mechanical mitral valve is an important decision for both patients and surgeons. We compared patient outcomes and readmission rates after bioprosthetic mitral valve replacement (Bio-MVR) vs mechanical mitral valve replacement (Mech-MVR).The Nationwide Readmissions Database was queried to identify 31 474 patients who underwent isolated MVR (22 998 Bio-MVR, 8476 Mech-MVR) between January 1, 2016, and December 31, 2018. Propensity score matching by age, sex, elective status, and comorbidities was used to compare outcomes between matched cohorts by prosthesis type. Freedom from readmission within the first calendar year was estimated by Kaplan-Meier analysis and compared between matched cohorts.Bio-MVR patients were older (median age, 69 vs 57 years; P.001) and had more comorbidities (median Elixhauser score, 14 vs 11; P.001) compared with Mech-MVR patients. After propensity score matching (n = 15 549), Bio-MVR patients had similar operative mortality (3.5% vs 3.4%; P = .97) and costs ($50 958 vs $49 782; P = .16) but shorter lengths of stay (8 vs 9 days; P.001) and fewer 30-day (16.0% vs 18.1%; P = .04) and 90-day (23.8% vs 26.8%; P = .01) readmissions compared with Mech-MVR patients. The difference in readmissions persisted at 1 year (P = .045). Readmission for bleeding or coagulopathy complications was less common with Bio-MVR (5.7% vs 10.1%; P .001).Readmission was more common after Mech-MVR than after Bio-MVR. Identifying and closely observing patients at high risk for bleeding complications may bridge the readmissions gap between Bio-MVR and Mech-MVR.

Acute kidney injury after cardiac surgery increases morbidity and mortality. Diagnosis relies on oliguria or increased serum creatinine, which develop 48 to 72 hours after injury. We hypothesized machine learning incorporating preoperative, operative, and intensive care unit data could dynamically predict acute kidney injury before conventional identification.Cardiac surgery patients at a tertiary hospital (2008-2019) were identified using electronic medical records in the Medical Information Mart for Intensive Care IV database. Preoperative and intraoperative parameters included demographics, Charlson Comorbidity subcategories, and operative details. Intensive care unit data included hemodynamics, medications, fluid intake/output, and laboratory results. Kidney Disease: Improving Global Outcomes creatinine criteria were used for acute kidney injury diagnosis. An ensemble machine learning model was trained for hourly predictions of future acute kidney injury within 48 hours. Performance was evaluated by area under the receiver operating characteristic curve and balanced accuracy.Within the cohort (n = 4267), there were approximately 7 million data points. Median baseline creatinine was 1.0 g/dL (interquartile range, 0.8-1.2), with 17% (735/4267) of patients having chronic kidney disease. Postoperative stage 1 acute kidney injury occurred in 50% (2129/4267), stage 2 occurred in 8% (324/4267), and stage 3 occurred in 4% (183/4267). For hourly prediction of any acute kidney injury over the next 48 hours, area under the receiver operating characteristic curve was 0.82, and balanced accuracy was 75%. For hourly prediction of stage 2 or greater acute kidney injury over the next 48 hours, area under the receiver operating characteristic curve was 0.95 and balanced accuracy was 86%. The model predicted acute kidney injury before clinical detection in 89% of cases.Ensemble machine learning models using electronic medical records data can dynamically predict acute kidney injury risk after cardiac surgery. Continuous postoperative risk assessment could facilitate interventions to limit or prevent renal injury.

Bariatric surgery helps many morbidly obese patients lose substantial weight. However, few data exist on its long-term safety and effectiveness in patients who also have continuous-flow left ventricular assist devices and in whom heart transplantation is contemplated. We retrospectively identified patients at our institution who had undergone ventricular assist device implantation and subsequent laparoscopic sleeve gastrectomy from June 2015 through September 2017, and we evaluated their baseline demographic data, preoperative characteristics, and postoperative outcomes. Four patients (3 men), ranging in age from 32 to 44 years and in body mass index from 40 to 57, underwent sleeve gastrectomy from 858 to 1,849 days after left ventricular assist device implantation to treat nonischemic cardiomyopathy. All had multiple comorbidities. At a median follow-up duration of 42 months (range, 24–47 mo), median body mass index decreased to 31.9 (range, 28.3–44.3) at maximal weight loss, with a median percentage of excess body mass index lost of 72.5% (range, 38.7%–87.4%). After achieving target weight, one patient was listed for heart transplantation, another awaited listing, one was kept on destination therapy because of positive drug screens, and one regained weight and remained ineligible. On long-term follow-up, laparoscopic sleeve gastrectomy appears to be safe and feasible for morbidly obese patients with ventricular assist devices who must lose weight for transplantation consideration. Additional studies are warranted to evaluate this weight-loss strategy after transplantation and immunosuppression.

We examined readmissions and resource use during the first postoperative year in patients who underwent thoracic endovascular aortic repair or open surgical repair of Stanford type B aortic dissection.The Nationwide Readmissions Database (2016-2018) was queried for patients with type B aortic dissection who underwent thoracic endovascular aortic repair or open surgical repair. The primary outcome was readmission during the first postoperative year. Secondary outcomes included 30-day and 90-day readmission rates, in-hospital mortality, length of stay, and cost. A Cox proportional hazards model was used to determine risk factors for readmission.During the study period, type B aortic dissection repair was performed in 6456 patients, of whom 3517 (54.5%) underwent thoracic endovascular aortic repair and 2939 (45.5%) underwent open surgical repair. Patients undergoing thoracic endovascular aortic repair were older (63 vs 59 years;Approximately one-third of all patients with type B aortic dissection were readmitted within 90 days after aortic intervention. Surprisingly, readmission during the first postoperative year was similar in the open surgical repair and thoracic endovascular aortic repair cohorts, despite marked differences in preoperative patient characteristics and interventions.

Studies have noted racial/ethnic disparities in coronary artery disease intervention strategies. We investigated trends and outcomes of coronary artery disease treatment choice (coronary artery bypass grafting or percutaneous coronary intervention) stratified by race/ethnicity.We queried the National Inpatient Sample for patients who underwent isolated coronary artery bypass grafting or percutaneous coronary intervention (2002-2017). Outcomes were stratified by race/ethnicity (White, African American, Hispanic, Asian). Multivariable logistic regression evaluated associations between race/ethnicity and receiving coronary artery bypass grafting versus percutaneous coronary intervention, in-hospital mortality, and costs.Over the 15-year period, 2,426,917 isolated coronary artery bypass grafting surgeries and 7,184,515 percutaneous coronary interventions were performed. Compared with White patients, African American patients were younger (62 [interquartile range, 53-70] vs 66 [interquartile range, 57-75] years), were more likely to have Medicaid insurance (12.2% vs 4.4%), and had more comorbidities (Charlson-Deyo index, 1.9 ± 1.6 vs 1.7 ± 1.6) (all P .01). After adjustment for patient comorbidities, presence of acute myocardial infarction, insurance status, and geography, African Americans were the least likely of all racial/ethnic groups to undergo coronary artery bypass grafting (odds ratio, 0.76; P .01), a consistent trend throughout the study. African American patients had higher risk-adjusted mortality after coronary artery bypass grafting (odds ratio, 1.09; P .01). Race/ethnicity was not associated with increased mortality after percutaneous coronary intervention. African American patients had higher hospitalization costs for coronary artery bypass grafting (+$5816; P .01) and percutaneous coronary intervention (+$856; P .01) after controlling for confounders.In this contemporary national analysis, risk-adjusted frequency of coronary artery bypass grafting versus percutaneous coronary intervention for coronary artery disease differed by race/ethnicity. African American patients had lower odds of undergoing coronary artery bypass grafting and worse outcomes. Reasons for these differences merit further investigation to identify opportunities to reduce potential disparities.

Cultural elements may affect the establishment of patient care best practices such as transfusion rates.

Functional status and health-related quality of life (HRQoL) are important outcomes, particularly among older patients. However, data on such patient-centered outcomes after cardiac surgery are limited. We evaluated the incidence and predictors of decline in functional status and HRQoL among older patients hospitalized for acute myocardial infarction (AMI). Participants were age 75 years or older hospitalized for AMI at 94 US sites. We examined decline in functional status (defined as decline in 1 or more activities of daily living, ADLs), as well as mental (MCS) and physical component scales (PCS) of the SF-12 to assess HRQoL (5-point decline or greater in each scale) between 1 month prior to the hospitalization and 6 months after. Multivariable model compared the risk of decline after coronary artery bypass graft (CABG), percutaneous coronary intervention (PCI) and medical management. Among 3041 patients (1708 PCI, 362 CABG, and 971 medical management), 1525 (50.2%) experienced decline in 1 or more domain: 633 (20.8%) declined in ADLs, 786 (25.9%) declined in the MCS, and 1078 (35.5%) declined in the PCS. The unadjusted incidence of ADL decline was the lowest among patients who underwent CABG (n = 50, 13.8%) compared with PCI (n = 271, 15.9%) or medical management (n = 312, 32.1%). Patients who underwent CABG and PCI had lower adjusted risk of decline in functional and HRQoL compared with those who received medical therapy. The risks after CABG and PCI were not significantly different. Over half of older patients significantly declined in function or HRQoL after AMI. Compared with medical management, risk of decline was lower in those who underwent revascularization.

Background Laboratory studies demonstrate glucose-insulin-potassium (GIK) as a potent cardioprotective intervention, but clinical trials have yielded mixed results, likely because of varying formulas and timing of GIK treatment and different clinical settings. This study sought to evaluate the effects of modified GIK regimen given perioperatively with an insulin-glucose ratio of 1:3 in patients undergoing cardiopulmonary bypass surgery. Methods and Results In this prospective, randomized, double-blinded trial with 930 patients referred for cardiac surgery with cardiopulmonary bypass, GIK (200 g/L glucose, 66.7 U/L insulin, and 80 mmol/L KCl) or placebo treatment was administered intravenously at 1 mL/kg per hour 10 minutes before anesthesia and continuously for 12.5 hours. The primary outcome was the incidence of in-hospital major adverse cardiac events including all-cause death, low cardiac output syndrome, acute myocardial infarction, cardiac arrest with successful resuscitation, congestive heart failure, and arrhythmia. GIK therapy reduced the incidence of major adverse cardiac events and enhanced cardiac function recovery without increasing perioperative blood glucose compared with the control group. Mechanistically, this treatment resulted in increased glucose uptake and less lactate excretion calculated by the differences between arterial and coronary sinus, and increased phosphorylation of insulin receptor substrate-1 and protein kinase B in the hearts of GIK-treated patients. Systemic blood lactate was also reduced in GIK-treated patients during cardiopulmonary bypass surgery. Conclusions A modified GIK regimen administered perioperatively reduces the incidence of in-hospital major adverse cardiac events in patients undergoing cardiopulmonary bypass surgery. These benefits are likely a result of enhanced systemic tissue perfusion and improved myocardial metabolism via activation of insulin signaling by GIK. Clinical Trial Registration URL: clinicaltrials.gov. Identifier: NCT01516138.

Leveraging intrinsic cardiomyocyte repair mechanisms may allow for restoration of myocardial function in heart failure, but extensive further preclinical validation of these strategies is needed.

Observational studies suggest a survival advantage with bilateral single internal thoracic artery (BITA) versus single internal thoracic artery grafting for coronary surgery, whereas this conclusion is not supported by randomized trials. We hypothesized that this inconsistency is attributed to unmeasured confounders intrinsic to observational studies. To test our hypothesis, we performed a meta-analysis of the observational literature comparing BITA and single internal thoracic artery, deriving incident rate ratio for mortality at end of follow-up and at 1 year. We postulated that BITA would not affect 1-year survival based on the natural history of coronary artery bypass occlusion, so that a difference between groups at 1 year could not be attributed to the intervention.We searched MEDLINE and Pubmed to identify all observational studies comparing the outcome of BITA versus single internal thoracic artery. One-year and long-term mortality for BITA and single internal thoracic artery were compared in the propensity-score-matched (PSM) series, that is, the form of observational evidence less prone to confounders. Thirty-eight observational studies (174 205 total patients) were selected for final comparison. In the 12 propensity-score-matched series (34 019 patients), the mortality reduction for BITA was similar at 1 year and at the end of follow-up (incident rate ratio, 0.70; 95% confidence interval, 0.60-0.82 versus 0.77; 95% confidence interval, 0.70-0.85;Unmeasured confounders, rather than biological superiority, may explain the survival advantage of BITA in observational series.