Predicting Length of Stay of Coronary Artery Bypass Grafting Patients Using Machine Learning.

BACKGROUND: There is a growing need to identify which bits of information are most valuable for healthcare providers. The aim of this study was to search for the highest impact variables in predicting postsurgery length of stay (LOS) for patients who undergo coronary artery bypass grafting (CABG). MATERIALS AND METHODS: Using a single institution's Society of Thoracic Surgeons (STS) Registry data, 2121 patients with elective or urgent, isolated CABG were analyzed across 116 variables. Two machine learning techniques of random forest and artificial neural networks (ANNs) were used to search for the highest impact variables in predicting LOS, and results were compared against multiple linear regression. Out-of-sample validation of the models was performed on 105 patients. RESULTS: Of the 10 highest impact variables identified in predicting LOS, four of the most impactful variables were duration intubated, last preoperative creatinine, age, and number of intraoperative packed red blood cell transfusions. The best performing model was an ANN using the ten highest impact variables (testing sample mean absolute error (MAE) = 1.685 d, R2 = 0.232), which performed consistently in the out-of-sample validation (MAE = 1.612 d, R2 = 0.150). CONCLUSION: Using machine learning, this study identified several novel predictors of postsurgery LOS and reinforced certain known risk factors. Out of the entire STS database, only a few variables carry most of the predictive value for LOS in this population. With this knowledge, a simpler linear regression model has been shared and could be used elsewhere after further validation.

Preparing for Bundled Payments: Impact of Complications Post-Coronary Artery Bypass Grafting on Costs.

BACKGROUND: Bundled payments for coronary artery bypass grafting (CABG) provide a single reimbursement for care provided from admission through 90 days post-discharge. We aim to explore the impact of complications on total institutional costs, as well as the drivers of high costs for index hospitalization. METHODS: We linked clinical and internal cost data for patients undergoing CABG from 2014 to 2017 at a single institution. We compared unadjusted average variable direct costs, reporting excess cost from an uncomplicated baseline. We stratified by The Society of Thoracic Surgeons preoperative risk and quality outcome measures as well as value-based outcomes (readmission, post-acute care utilization). We performed multivariable linear regression to evaluate drivers of high costs, adjusting for preoperative and intraoperative characteristics and postoperative complications. RESULTS: We reviewed 1789 patients undergoing CABG with an average of 2.7 vessels (SD 0.89). A significant proportion of patients were diabetic (51.2%) and obese (mean body mass index 30.6, SD 6.1). Factors associated with increased adjusted costs were preoperative renal failure (P = .001), diabetes (P = .001) and body mass index (P = .05), and postoperative stroke (P < .001), prolonged ventilation (P < .001), rebleeding requiring reoperation (P < .001) and renal failure (P < .001) with varying magnitude. Preoperative ejection fraction and insurance status were not associated with increased adjusted costs. CONCLUSIONS: Preoperative characteristics had less of an impact on costs post-CABG than postoperative complications. Postoperative complications vary in their impact on internal costs, with reoperation, stroke, and renal failure having the greatest impact. In preparation for bundled payments, hospitals should focus on understanding and preventing drivers of high cost.

Mean platelet volume and cardiac-surgery-associated acute kidney injury: a retrospective study.

PURPOSE: Increased mean platelet volume (MPV) may indicate platelet activation, platelet aggregation, and a resulting prothrombotic state. Such changes in the postoperative period have been associated with organ injury and adverse outcomes. We hypothesized that changes in MPV after cardiac surgery are associated with both a higher risk of acute kidney injury (AKI) and mortality. METHODS: In this retrospective study, we evaluated consecutive patients undergoing adult cardiac surgery patients between 12 December 2011 and 5 June 2018. The change in MPV was derived by calculating the difference between the baseline MPV before surgery and the average postoperative MPV just prior to the occurrence of AKI. We defined postoperative AKI according to Kidney Disease: Improving Global Outcomes Clinical Practice Guideline for Acute Kidney Injury as either a ≥ 50% increase in serum creatinine in the first ten postoperative days, or an increase of ≥ 0.3 mg·dL-1 during any 48-hr window across the ten-day postoperative period. Multivariable logistic regression analysis was used to examine the association between MPV change and postoperative AKI and mortality. RESULTS: Of the 4,204 patients studied, 1,373 (32.7%) developed postoperative AKI, including 83 (2.0%) and 38 (0.9%) who developed stages II and III AKI, respectively. Compared with patients who had an increase in median postoperative MPV of 0.2 femtolitre (fL), those with an increase of 0.8 fL had an 80% increase in the odds of developing AKI (adjusted odds ratio [aOR], 1.80; 95% confidence interval [CI],1.36 to 2.38; P < 0.001) and were almost twice as likely to progress to a higher severity AKI (aOR, 1.66; 95% CI, 1.28 to 2.16; P < 0.001). Change in MPV was not associated with mortality (aOR,1.32; 95% CI, 0.92 to 1.89; P = 0.14). CONCLUSION: Increased MPV change in the postoperative period was associated with both increased risk and severity of AKI, but not mortality.

RéSUMé: OBJECTIF: Un volume plaquettaire moyen (VPM) augmenté peut être indicatif d'une activation plaquettaire, d'une agrégation plaquettaire, et de l'état prothrombotique qui en résulte. De tels changements en période postopératoire ont été associés à des lésions aux organes et à des devenirs défavorables. Nous avons émis l'hypothèse que des changements du VPM après une chirurgie cardiaque seraient associés à un risque plus élevé d'insuffisance rénale aiguë et de mortalité. MéTHODE: Dans cette étude rétrospective, nous avons évalué des patients adultes consécutifs subissant une chirurgie cardiaque entre le 12 décembre 2011 et le 5 juin 2018. Le changement de VPM a été dérivé en calculant la différence entre le VPM de base avant la chirurgie et le VPM postopératoire moyen juste avant la survenue de l'IRA. Nous avons défini une IRA postopératoire sur la base des Directives Kidney Disease: Improving Global Outcomes Clinical Practice Guideline for Acute Kidney Injury (Les maladies rénales: Guide d'exercice clinique pour améliorer les devenirs globaux pour l'insuffisance rénale aiguë) en tant qu'une augmentation ≥ 50 % de la créatine sérique au cours des dix premiers jours postopératoires, ou une augmentation de ≥ 0,3 mg·dL−1 pendant toute fenêtre de 48 h au cours des dix premiers jours postopératoires. Une analyse multivariée de régression logistique a été utilisée pour examiner l'association entre le changement de VPM et l'IRA postopératoire et la mortalité. RéSULTATS: Parmi les 4204 patients à l'étude, 1373 (32,7 %) ont souffert d'IRA postopératoire, y compris 83 (2,0 %) et 38 (0,9 %) qui ont développé des IRA de stade II et III, respectivement. Par rapport aux patients ayant manifesté une augmentation du VPM postopératoire médian de 0,2 femtolitre (fL), ceux affichant une augmentation de 0,8 fL ont démontré une augmentation de 80 % de la probabilité d'IRA (rapport de cotes ajusté [RCA], 1,80; intervalle de confiance [IC] 95 %, 1,36 à 2,38; P < 0,001) et couraient un risque pratiquement deux fois plus élevé de voir leur IRA progresser à un stade plus grave (RCA, 1,66; IC 95 %, 1,28 à 2,16; P < 0,001). Les changements de VPM n'étaient pas associés à la mortalité (RCA, 1,32; IC 95 %, 0,92 à 1,89; P = 0,14). CONCLUSION: Une augmentation accrue du VPM en période postopératoire a été associée à un risque et une gravité accrus d'IRA, mais pas à la mortalité.

Predictors of Stroke After Minimally Invasive Mitral Valve Surgery Without the Cross-Clamp.

Minimally invasive mitral valve surgery (mini-MVS) with hypothermic fibrillatory arrest has been associated with an increased risk of stroke. We aim to investigate the incidence, predictors, and outcomes of stroke in a large cohort of patient who underwent clampless mini-MVS. Between January 2008 and June 2017, we performed 1247 mini-MVSs. The clinical, operative, and postoperative outcomes were analyzed. Univariable and multivariable regression analyses were used to identify predictors of postoperative stroke. The median follow-up was 5.2 years (interquartile range 2.6-7.5). The etiology of mitral valve (MV) disease was degenerative (60.4%, n = 753), functional (12.8%, n = 160), rheumatic (8.7%, n = 109), endocarditis (3.1%, n = 39), and reoperative MV surgery (14.9%, n = 186). The overall incidence of postoperative neurologic event was 2.5% (n = 31/1247). Univariable predictors of stroke were a higher Society of Thoracic Surgeons mortality risk (6.0 ± 11.8% vs 3.3 ± 5.2%, P < 0.001), advanced age, (69.6 ± 12.1 years vs 63.0 ± 13.6 years, P = 0.002), female gender (71.0% vs 46.3%, P = 0.007), and a history of a cerebrovascular accident (22.6% vs 8.7%, P = 0.008). Stroke patients had a higher 30-day mortality (22.6% vs 1.6%, P < 0.001) and a higher risk for long-term mortality (hazard ratio = 5.56, 95% confidence interval [CI] 3.2-9.6, P < 0.001). Advanced age (odds ratio [OR] 2.1; 95% CI 1.1-4.0; P = 0.02), female gender (OR 2.3; 95% CI 0.9-5.2; P = 0.05), and history of cerebrovascular accident (OR 3.1; 95% CI 0.98-10.1; P = 0.05) remained as independent predictors of stroke in the multivariable analysis. Our decade-long experience indicates that clampless mini-MVS is associated with a low incidence of postoperative stroke, and that the predictors of stroke are not specific to this approach.