The International Heart Transplant Survival Algorithm (IHTSA): a new model to improve organ sharing and survival

Artificial Intelligence Approaches for Predicting the Risks of Durable Mechanical Circulatory Support Therapy and Cardiac Transplantation

Background: The use of AI-driven technologies in probing big data to generate better risk prediction models has been an ongoing and expanding area of investigation. The AI-driven models may perform better as compared to linear models; however, more investigations are needed in this area to refine their predictability and applicability to the field of durable MCS and cardiac transplantation. Methods: A literature review was carried out using Google Scholar/PubMed from 2000 to 2023. Results: This review defines the knowledge gaps and describes different AI-driven approaches that may be used to further our understanding. Conclusions: The limitations of current models are due to missing data, data imbalances, and the uneven distribution of variables in the datasets from which the models are derived. There is an urgent need for predictive models that can integrate a large number of clinical variables from multicenter data to account for the variability in patient characteristics that influence patient selection, outcomes, and survival for both durable MCS and HT; this may be fulfilled by AI-driven risk prediction models.

Prognosis after heart transplant in patients with hypertrophic and restrictive cardiomyopathy. A nationwide registry analysis

INTRODUCTION AND OBJECTIVES

Posttransplant outcomes among recipients with a diagnosis of hypertrophic cardiomyopathy (HCM) or restrictive cardiomyopathy (RCM) remain controversial.

METHODS

Retrospective analysis of a nationwide registry of first-time recipients undergoing isolated heart transplant between 1984 and 2021. One-year and 5-year mortality in recipients with HCM and RCM were compared with those with dilated cardiomyopathy (DCM).

RESULTS

We included 3703 patients (3112 DCM; 331 HCM; 260 RCM) with a median follow-up of 5.0 [3.1-5.0] years. Compared with DCM, the adjusted 1-year mortality risk was: HCM: HR, 1.38; 95%CI, 1.07-1.78; P=.01, RCM: HR, 1.48; 95%CI, 1.14-1.93; P=.003. The adjusted 5-year mortality risk was: HCM: HR, 1.17; 95%CI, 0.93-1.47; P=.18; RCM: HR, 1.52; 95%CI, 1.22-1.89; P<.001. Over the last 20 years, the RCM group showed significant improvement in 1-year survival (adjusted R2=0.95) and 5-year survival (R2=0.88); the HCM group showed enhanced the 5-year survival (R2=0.59), but the 1-year survival remained stable (R2=0.16).

CONCLUSIONS

Both RCM and HCM were linked to a less favorable early posttransplant prognosis compared with DCM. However, at the 5-year mark, this unfavorable difference was evident only for RCM. Notably, a substantial temporal enhancement in both early and late mortality was observed for RCM, while for HCM, this improvement was mainly evident in late mortality.

Risk factors for 1-year allograft loss in pediatric heart transplant patients using machine learning: An analysis of the pediatric heart transplant society database

BACKGROUND

Pediatric heart transplant patients are at greatest risk of allograft loss in the first year. We assessed whether machine learning could improve 1-year risk assessment using the Pediatric Heart Transplant Society database.

METHODS

Patients transplanted from 2010 to 2019 were included. The primary outcome was 1-year graft loss free survival. We developed a prediction model using cross-validation, by comparing Cox regression, gradient boosting, and random forests. The modeling strategy with the best discrimination and calibration was applied to fit a final prediction model. We used Shapley additive explanation (SHAP) values to perform variable selection and to estimate effect sizes and importance of individual variables when interpreting the final prediction model.

RESULTS

Cumulative incidence of graft loss or mortality was 7.6%. Random forests had favorable discrimination and calibration compared to Cox proportional hazards with a C-statistic (95% confidence interval [CI]) of 0.74 (0.72, 0.76) versus 0.71 (0.69, 0.73), and closer alignment between predicted and observed risk. SHAP values computed using the final prediction model indicated that the diagnosis of congenital heart disease (CHD) increased 1 year predicted risk of graft loss by 1.7 (i.e., from 7.6% to 9.3%), need for mechanical circulatory support increased predicted risk by 2, and single ventricle CHD increased predicted risk by 1.9. These three predictors, respectively, were also estimated to be the most important among the 15 predictors in the final model.

CONCLUSIONS

Risk prediction models used to facilitate patient selection for pediatric heart transplant can be improved without loss of interpretability using machine learning.

Survival analysis for pediatric heart transplant patients using a novel machine learning algorithm: A UNOS analysis

BACKGROUND

Impact of pretransplantation risk factors on mortality in the first year after heart transplantation remains largely unknown. Using machine learning algorithms, we selected clinically relevant identifiers that could predict 1-year mortality after pediatric heart transplantation.

METHODS

Data were obtained from the United Network for Organ Sharing Database for years 2010-2020 for patients 0-17 years receiving their first heart transplant (N = 4150). Features were selected using subject experts and literature review. Scikit-Learn, Scikit-Survival, and Tensorflow were used. A train:test split of 70:30 was used. N-repeated k-fold validation was performed (N = 5, k = 5). Seven models were tested, Hyperparameter tuning performed using Bayesian optimization and the concordance index (C-index) was used for model assessment.

RESULTS

A C-index above 0.6 for test data was considered acceptable for survival analysis models. C-indices obtained were 0.60 (Cox proportional hazards), 0.61 (Cox with elastic net), 0.64 (gradient boosting), 0.64 (support vector machine), 0.68 (random forest), 0.66 (component gradient boosting), and 0.54 (survival trees). Machine learning models show an improvement over the traditional Cox proportional hazards model, with random forest performing the best on the test set. Analysis of the feature importance for the gradient boosted model found that the top 5 features were the most recent serum total bilirubin, the travel distance from the transplant center, the patient body mass index, the deceased donor terminal Serum glutamic pyruvic transaminase/Alanine transaminase (SGPT/ALT), and the donor PCO2.

CONCLUSIONS

Combination of machine learning and expert-based methodology of selecting predictors of survival for pediatric heart transplantation provides a reasonable prediction of 1- and 3-year survival outcomes. SHapley Additive exPlanations can be an effective tool for modeling and visualizing nonlinear interactions.

Fair Survival Time Prediction via Mutual Information Minimization

Survival analysis is a general framework for predicting the time until a specific event occurs, often in the presence of censoring. Although this framework is widely used in practice, few studies to date have considered fairness for time-to-event outcomes, despite recent significant advances in the algorithmic fairness literature more broadly. In this paper, we propose a framework to achieve demographic parity in survival analysis models by minimizing the mutual information between predicted time-to-event and sensitive attributes. We show that our approach effectively minimizes mutual information to encourage statistical independence of time-to-event predictions and sensitive attributes. Furthermore, we propose four types of disparity assessment metrics based on common survival analysis metrics. Through experiments on multiple benchmark datasets, we demonstrate that by minimizing the dependence between the prediction and the sensitive attributes, our method can systematically improve the fairness of survival predictions and is robust to censoring.

Prediction of Outcomes After Heart Transplantation in Pediatric Patients Using National Registry Data: Evaluation of Machine Learning Approaches

BACKGROUND

The prediction of posttransplant health outcomes for pediatric heart transplantation is critical for risk stratification and high-quality posttransplant care.

OBJECTIVE

The purpose of this study was to examine the use of machine learning (ML) models to predict rejection and mortality for pediatric heart transplant recipients.

METHODS

Various ML models were used to predict rejection and mortality at 1, 3, and 5 years after transplantation in pediatric heart transplant recipients using United Network for Organ Sharing data from 1987 to 2019. The variables used for predicting posttransplant outcomes included donor and recipient as well as medical and social factors. We evaluated 7 ML models-extreme gradient boosting (XGBoost), logistic regression, support vector machine, random forest (RF), stochastic gradient descent, multilayer perceptron, and adaptive boosting (AdaBoost)-as well as a deep learning model with 2 hidden layers with 100 neurons and a rectified linear unit (ReLU) activation function followed by batch normalization for each and a classification head with a softmax activation function. We used 10-fold cross-validation to evaluate model performance. Shapley additive explanations (SHAP) values were calculated to estimate the importance of each variable for prediction.

RESULTS

RF and AdaBoost models were the best-performing algorithms for different prediction windows across outcomes. RF outperformed other ML algorithms in predicting 5 of the 6 outcomes (area under the receiver operating characteristic curve [AUROC] 0.664 and 0.706 for 1-year and 3-year rejection, respectively, and AUROC 0.697, 0.758, and 0.763 for 1-year, 3-year, and 5-year mortality, respectively). AdaBoost achieved the best performance for prediction of 5-year rejection (AUROC 0.705).

CONCLUSIONS

This study demonstrates the comparative utility of ML approaches for modeling posttransplant health outcomes using registry data. ML approaches can identify unique risk factors and their complex relationship with outcomes, thereby identifying patients considered to be at risk and informing the transplant community about the potential of these innovative approaches to improve pediatric care after heart transplantation. Future studies are required to translate the information derived from prediction models to optimize counseling, clinical care, and decision-making within pediatric organ transplant centers.

Artificial intelligence, big data and heart transplantation: Actualities

BACKGROUND

As diagnostic and prognostic models developed by traditional statistics perform poorly in real-world, artificial intelligence (AI) and Big Data (BD) may improve the supply chain of heart transplantation (HTx), allocation opportunities, correct treatments, and finally optimize HTx outcome. We explored available studies, and discussed opportunities and limits of medical application of AI to the field of HTx.

METHOD

A systematic overview of studies published up to December 31st, 2022, in English on peer-revied journals, have been identified through PUBMED-MEDLINE-WEB of Science, referring to HTx, AI, BD. Studies were grouped in 4 domains based on main studies' objectives and results: etiology, diagnosis, prognosis, treatment. A systematic attempt was made to evaluate studies by the Prediction model Risk Of Bias ASsessment Tool (PROBAST) and the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD).

RESULTS

Among the 27 publications selected, none used AI applied to BD. Of the selected studies, 4 fell in the domain of etiology, 6 in the domain of diagnosis, 3 in the domain of treatment, and 17 in that of prognosis, as AI was most frequently used for algorithmic prediction and discrimination of survival, but in retrospective cohorts and registries. AI-based algorithms appeared superior to probabilistic functions to predict patterns, but external validation was rarely employed. Indeed, based on PROBAST, selected studies showed, to some extent, significant risk of bias (especially in the domain of predictors and analysis). In addition, as example of applicability in the real-world, a free-use prediction algorithm developed through AI failed to predict 1-year mortality post-HTx in cases from our center.

CONCLUSIONS

While AI-based prognostic and diagnostic functions performed better than those developed by traditional statistics, risk of bias, lack of external validation, and relatively poor applicability, may affect AI-based tools. More unbiased research with high quality BD meant for AI, transparency and external validations, are needed to have medical AI as a systematic aid to clinical decision making in HTx.

A machine learning model for prediction of 30-day primary graft failure after heart transplantation

Background

Primary graft failure (PGF) remains the most common cause of short-term mortality after heart transplantation. The main objective was to develop and validate a risk model for prediction of short-term mortality due to PGF after heart transplantation using the ISHLT Heart Transplant Registry.

Methods

We developed a non-linear artificial neural networks (ANN) model to evaluate the association between recipient-donor variables and post-transplant PGF. Patients in the ISHLT registry were randomly divided into derivation and an independent internal validation cohort. The primary endpoint was PGF defined as death within 30 days due to Graft failure or Cardiovascular causes or retransplant within 30 days for causes other than rejection.

Results

Among 64,964 adult recipients transplanted between 1994 and 2013, mean age was 51 years and 22% were female. The incidence of PGF up to 30 days was 3.7%. The ANN model selected 33 of 77 risk variables as relevant for PGF prediction. The C-index in the test cohort was 0.70 (95% CI: 0.68-0.71). The risk variables which most influenced the PGF were underlying HF diagnosis, ischemia time and sex, while renal function had a lower influence.

Conclusion

An ANN model to predict primary graft dysfunction was derived and independently validated. The good discrimination of the ANN model likely results from its flexibility to model potentially non-linear relationships and interactions. Whether this model with improved discrimination can assist in clinical decisions at the time of transplant should be tested.

Artificial intelligence guidance of advanced heart failure therapies: A systematic scoping review

Introduction

Artificial intelligence can recognize complex patterns in large datasets. It is a promising technology to advance heart failure practice, as many decisions rely on expert opinions in the absence of high-quality data-driven evidence.

Methods

We searched Embase, Web of Science, and PubMed databases for articles containing "artificial intelligence," "machine learning," or "deep learning" and any of the phrases "heart transplantation," "ventricular assist device," or "cardiogenic shock" from inception until August 2022. We only included original research addressing post heart transplantation (HTx) or mechanical circulatory support (MCS) clinical care. Review and data extraction were performed in accordance with PRISMA-Scr guidelines.

Results

Of 584 unique publications detected, 31 met the inclusion criteria. The majority focused on outcome prediction post HTx (n = 13) and post durable MCS (n = 7), as well as post HTx and MCS management (n = 7, n = 3, respectively). One study addressed temporary mechanical circulatory support. Most studies advocated for rapid integration of AI into clinical practice, acknowledging potential improvements in management guidance and reliability of outcomes prediction. There was a notable paucity of external data validation and integration of multiple data modalities.

Conclusion

Our review showed mounting innovation in AI application in management of MCS and HTx, with the largest evidence showing improved mortality outcome prediction.

Donor heart selection: Evidence-based guidelines for providers

The proposed donor heart selection guidelines provide evidence-based and expert-consensus recommendations for the selection of donor hearts following brain death. These recommendations were compiled by an international panel of experts based on an extensive literature review.

Enhanced survival prediction using explainable artificial intelligence in heart transplantation

The most limiting factor in heart transplantation is the lack of donor organs. With enhanced prediction of outcome, it may be possible to increase the life-years from the organs that become available. Applications of machine learning to tabular data, typical of clinical decision support, pose the practical question of interpretation, which has technical and potential ethical implications. In particular, there is an issue of principle about the predictability of complex data and whether this is inherent in the data or strongly dependent on the choice of machine learning model, leading to the so-called accuracy-interpretability trade-off. We model 1-year mortality in heart transplantation data with a self-explaining neural network, which is benchmarked against a deep learning model on the same development data, in an external validation study with two data sets: (1) UNOS transplants in 2017-2018 (n = 4750) for which the self-explaining and deep learning models are comparable in their AUROC 0.628 [0.602,0.654] cf. 0.635 [0.609,0.662] and (2) Scandinavian transplants during 1997-2018 (n = 2293), showing good calibration with AUROCs of 0.626 [0.588,0.665] and 0.634 [0.570, 0.698], respectively, with and without missing data (n = 982). This shows that for tabular data, predictive models can be transparent and capture important nonlinearities, retaining full predictive performance.

The current and future role of artificial intelligence in optimizing donor organ utilization and recipient outcomes in heart transplantation

Heart failure (HF) is a leading cause of morbidity and mortality in the United States. While medical management and mechanical circulatory support have undergone significant advancement in recent years, orthotopic heart transplantation (OHT) remains the most definitive therapy for refractory HF. OHT has seen steady improvement in patient survival and quality of life (QoL) since its inception, with one-year mortality now under 8%. However, a significant number of HF patients are unable to receive OHT due to scarcity of donor hearts. The United Network for Organ Sharing has recently revised its organ allocation criteria in an effort to provide more equitable access to OHT. Despite these changes, there are many potential donor hearts that are inevitably rejected. Arbitrary regulations from the centers for Medicare and Medicaid services and fear of repercussions if one-year mortality falls below established values has led to a current state of excessive risk aversion for which organs are accepted for OHT. Furthermore, non-standardized utilization of extended criteria donors and donation after circulatory death, exacerbate the organ shortage. Data-driven systems can improve donor-recipient matching, better predict patient QoL post-OHT, and decrease needless organ waste through more uniform application of acceptance criteria. Thus, we propose a data-driven future for OHT and a move to patient-centric and holistic transplantation care processes.

Utility of the IMPACT score for predicting heart transplant mortality. Analysis on a contemporary cohort of the Spanish Heart Transplant Registry

INTRODUCTION AND OBJECTIVES

The Index for Mortality Prediction After Cardiac Transplantation (IMPACT) score was derived and validated as a predictor of mortality after heart transplantation (HT). The primary objective of this work is to externally validate the IMPACT score in a contemporary Spanish cohort.

METHODS

Spanish Heart Transplant Registry data were used to identify adult (>16 years) HT patients between January 2000 and December 2015. Retransplantation, multiorgan transplantation and patients in whom at least one of the variables required to calculate the IMPACT score was missing were excluded from the analysis (N = 2,810).

RESULTS

Median value of the IMPACT score was 5 points (IQR: 3, 8). Overall 1-year survival rate was 79.1%. Kaplan-Meier 1-year survival rates by IMPACT score categories (0-2, 3-5, 6-9, 10-14, ≥ 15) were 84.4%, 81.5%, 79.3%, 77.3% and 58.5% respectively (Log-Rank test: p<0.001). Performance analysis showed a good calibration (Hosmer-Lemeshow chi-square for one year was 7.56; p = 0.47) and poor discrimination ability (AUC-ROC 0.59) of the IMPACT score as a predictive model.

CONCLUSIONS

In a contemporary Spanish cohort, the IMPACT score failed to accurately predict the risk of death after HT. This article is protected by copyright. All rights reserved.

Machine learning and artificial intelligence in cardiac transplantation: A systematic review

Background

This review aims to systematically evaluate the currently available evidence investigating the use of artificial intelligence (AI) and machine learning (ML) in the field of cardiac transplantation. Furthermore, based on the challenges identified we aim to provide a series of recommendations and a knowledge base for future research in the field of ML and heart transplantation.

Methods

A systematic database search was conducted of original articles that explored the use of ML and/or AI in heart transplantation in EMBASE, MEDLINE, Cochrane database, and Google Scholar, from inception to November 2021.

Results

Our search yielded 237 articles, of which 13 studies were included in this review, featuring 463 850 patients. Three main areas of application were identified: (1) ML for predictive modeling of heart transplantation mortality outcomes; (2) ML in graft failure outcomes; (3) ML to aid imaging in heart transplantation. The results of the included studies suggest that AI and ML are more accurate in predicting graft failure and mortality than traditional scoring systems and conventional regression analysis. Major predictors of graft failure and mortality identified in ML models were: length of hospital stay, immunosuppressive regimen, recipient's age, congenital heart disease, and organ ischemia time. Other potential benefits include analyzing initial lab investigations and imaging, assisting a patient with medication adherence, and creating positive behavioral changes to minimize further cardiovascular risk.

Conclusion

ML demonstrated promising applications for improving heart transplantation outcomes and patient‐centered care, nevertheless, there remain important limitations relating to implementing AI into everyday surgical practices.

Validation of existing risk scores for mortality prediction after a heart transplant in a Chinese population

OBJECTIVES

The objectives of this study were to validate 3 existing heart transplant risk scores with a single-centre cohort in China and evaluate the efficacy of the 3 systems in predicting mortality.

METHODS

We retrospectively studied 428 patients from a single centre who underwent heart transplants from January 2015 to December 2019. All patients were scored using the Index for Mortality Prediction After Cardiac Transplantation (IMPACT) and the United Network for Organ Sharing (UNOS) and risk stratification scores (RSSs). We assessed the efficacy of the risk scores by comparing the observed and the predicted 1-year mortality. Binary logistic regression was used to evaluate the predictive accuracy of the 3 risk scores. Model discrimination was assessed by measuring the area under the receiver operating curves. Kaplan-Meier survival analyses were performed after the patients were divided into different risk groups.

RESULTS

Based on our cohort, the observed mortality was 6.54%, whereas the predicted mortality of the IMPACT and UNOS scores and the RSSs was 10.59%, 10.74% and 12.89%, respectively. Logistic regression analysis showed that the IMPACT [odds ratio (OR), 1.25; 95% confidence interval (CI), 1.15-1.36; P < 0.001], UNOS (OR, 1.68; 95% CI, 1.37-2.07; P < 0.001) and risk stratification (OR, 1.61; 95% CI, 1.30-2.00; P < 0.001) scores were predictive of 1-year mortality. The discriminative power was numerically higher for the IMPACT score [area under the curve (AUC) of 0.691)] than for the UNOS score (AUC 0.685) and the RSS (AUC 0.648).

CONCLUSIONS

We validated the IMPACT and UNOS scores and the RSSs as predictors of 1-year mortality after a heart transplant, but all 3 risk scores had unsatisfactory discriminative powers that overestimated the observed mortality for the Chinese cohort.

Machine learning methods for perioperative anesthetic management in cardiac surgery patients: a scoping review

Background

Machine learning (ML) is developing fast with promising prospects within medicine and already has several applications in perioperative care. We conducted a scoping review to examine the extent and potential limitations of ML implementation in perioperative anesthetic care, specifically in cardiac surgery patients.

Methods

We mapped the current literature by searching three databases: MEDLINE (Ovid), EMBASE (Ovid), and Cochrane Library. Articles were eligible if they reported on perioperative ML use in the field of cardiac surgery with relevance to anesthetic practices. Data on the applicability of ML and comparability to conventional statistical methods were extracted.

Results

Forty-six articles on ML relevant to the work of the anesthesiologist in cardiac surgery were identified. Three main categories emerged: (I) event and risk prediction, (II) hemodynamic monitoring, and (III) automation of echocardiography. Prediction models based on ML tend to behave similarly to conventional statistical methods. Using dynamic hemodynamic or ultrasound data in ML models, however, shifts the potential to promising results.

Conclusions

ML in cardiac surgery is increasingly used in perioperative anesthetic management. The majority is used for prediction purposes similar to conventional clinical scores. Remarkable ML model performances are achieved when using real-time dynamic parameters. However, beneficial clinical outcomes of ML integration have yet to be determined. Nonetheless, the first steps introducing ML in perioperative anesthetic care for cardiac surgery have been taken.

Determining the Level of Importance of Variables in Predicting Kidney Transplant Survival Based on a Novel Ranking Method

BACKGROUND

Kidney transplantation is the best alternative treatment for end-stage renal disease. To optimal use of donated kidneys, graft predicted survival can be used as a factor to allocate kidneys. The performance of prediction techniques is highly dependent on the correct selection of predictors. Hence, the main objective of this research is to propose a novel method for ranking the effective variables for predicting the kidney transplant survival.

METHODS

Five classification models were used to classify kidney recipients in long- and short-term survival classes. Synthetic minority oversampling and random undersampling were used to overcome the imbalanced class problem. In dealing with missing values, 2 approaches were used (eliminating and imputing them). All variables were categorized into 4 levels. The ranking was evaluated using the sensitivity analysis approach.

RESULTS

Thirty-four of the 41 variables were identified as important variables, of which, 5 variables were categorized in very important level ("Recipient creatinine at discharge," "Recipient dialysis time," "Donor history of diabetes," "Donor kidney biopsy," and "Donor cause of death"), 17 variables in important level, and 12 variables in the low important level.

CONCLUSIONS

In this study, we identify new variables that have not been addressed in any of the previous studies (eg, AGE_DIF and MATCH_GEN). On the other hand, in kidney allocation systems, 2 main criteria are considered: equity and utility. One of the utility subcriteria is the graft survival. Our study findings can be used in the design of systems to predict the graft survival.

30-Year Trends in Graft Survival After Heart Transplant: Modeled Analyses of a Transplant Registry

BACKGROUND

Heart failure is an epidemic in the United States, and transplantation remains the most definitive therapy. We describe multi-decade trends in post-transplant graft survival, adjusted for concurrent changes in the population, over the 30 years antecedent to the most recent heart allocation policy change.

METHODS

Scientific Registry of Transplant Recipients data were used to identify all primary adult heart recipients 1989 through 2017. We described temporal changes in population characteristics (recipient/donor demographics and comorbidities, pretransplant interventions, clinical transplant measures, and providers). The primary outcome was graft survival, defined as freedom from all-cause death and graft failure, within 6 months post-transplant. Modified Poisson logistic regression estimated relative changes in risk of outcomes compared to 1989, with and without adjustment for changing population characteristics. We identified risk factors, quantified by associated risk ratios.

RESULTS

Among 56,488 primary adult heart recipients, we observed 5,529 (9.8%) all-cause deaths and 1,933 (3.4%) graft failure events within 6 months post-transplant. Prevalence of known recipient risk factors increased over time. Unadjusted modeling demonstrated a significant 30-year improvement in graft survival, averaging 2.6% (95%CI:2.4-2.9%) per year (p-for-trend<0.001). After adjusting for population changes, the 30-year trend remained significant and graft survival improved on average 3.0% (95%CI:2.6-3.3%) yearly. Regression modeling identified multiple predictors of graft survival. Modeling two additional outcomes of 6-month mortality and 6-month graft failure produced like results.

CONCLUSIONS

Short-term graft survival after heart transplantation has improved significantly leading up to the 2018 heart allocation policy change, despite concurrent increase in prevalence of higher risk population characteristics.

Using machine learning to improve survival prediction after heart transplantation

BACKGROUND

This study investigates the use of modern machine learning (ML) techniques to improve prediction of survival after orthotopic heart transplantation (OHT).

METHODS

Retrospective study of adult patients undergoing primary, isolated OHT between 2000 and 2019 as identified in the United Network for Organ Sharing (UNOS) registry. The primary outcome was 1-year post-transplant survival. Patients were randomly divided into training (80%) and validation (20%) sets. Dimensionality reduction and data re-sampling were employed during training. Multiple machine learning algorithms were combined into a final ensemble ML model. The discriminatory capability was assessed using the area under receiver-operating-characteristic curve (AUROC), net reclassification index (NRI), and decision curve analysis (DCA).

RESULTS

A total of 33,657 OHT patients were evaluated. One-year mortality was 11% (n = 3738). In the validation cohort, the AUROC of singular logistic regression was 0.649 (95% CI, 0.628-0.670) compared to 0.691 (95% CI, 0.671-0.711) with random forest, 0.691 (95% CI, 0.671-0.712) with deep neural network, and 0.653 (95% CI, 0.632-0.674) with Adaboost. A final ensemble ML model was created that demonstrated the greatest improvement in AUROC: 0.764 (95% CI, 0.745-0.782) (p < .001). The ensemble ML model improved predictive performance by 72.9% ±3.8% (p < .001) as assessed by NRI compared to logistic regression. DCA showed the final ensemble method improved risk prediction across the entire spectrum of predicted risk as compared to all other models (p < .001).

CONCLUSIONS

Modern ML techniques can improve risk prediction in OHT compared to traditional approaches. This may have important implications in patient selection, programmatic evaluation, allocation policy, and patient counseling and prognostication.

Machine learning in medicine: It has arrived, let's embrace it

Machine learning and artificial intelligence (AI) have arrived in medicine and the healthcare community is experiencing significant growth in their adoption across numerous patient care settings. There are countless applications for machine learning and AI in medicine ranging from patient outcome prediction, to clinical decision support, to predicting future patient therapeutic setpoints. This commentary discusses a recent application leveraging machine learning to predict one-year patient survival following orthotopic heart transplantation. This modeling approach has significant implications in terms of improving clinical decision-making, patient counseling, and ultimately organ allocation and has been shown to significantly outperform pre-existing algorithms. This commentary also discusses how adoption and advancement of this modeling approach in the future can provide increased personalization of patient care. The continued expansion of information systems and growth of electronic patient data sources in health care will continue to pave the way for increased use and adoption of data science in medicine. Personalized medicine has been a long-standing goal of the healthcare community and with machine learning and AI now being continually incorporated into clinical settings and practice, this technology is well on the pathway to make a considerable impact to greatly improve patient care in the near future.

Mechanical ventilation at the time of heart transplantation and associations with clinical outcomes

AIMS

The impact of mechanical ventilation (MV) at the time of heart transplantation is not well understood. In addition, MV was recently removed as a criterion from the new US heart transplantation allocation system. We sought to assess for the association between MV at transplantation and 1-year mortality.

METHODS AND RESULTS

We utilized the United Network for Organ Sharing database and included all adult, single organ heart transplantations from 1990 to 2019. We utilized multivariable logistic regression adjusting for demographics, comorbidities, and markers of clinical acuity. We identified 60 980 patients who underwent heart transplantation, 2.4% (n = 1431) of which required MV at transplantation. Ventilated patients were more likely to require temporary mechanical support, previous dialysis, and had a shorter median waitlist time (21 vs. 95 days, P < 0.001). At 1 year, the mortality was 33.7% (n = 484) for ventilated patients and 11.7% (n = 6967) for those not ventilated at the time of transplantation (log-rank P < 0.001). After multivariable adjustment, patients requiring MV continued to have a substantially higher 90-day [odds ratio (OR) 3.20, 95% confidence interval (CI): 2.79-3.66, P < 0.001] and 1-year mortality (OR 2.67, 95% CI: 2.36-3.03, P < 0.001). For those that survived to 90 days, the adjusted mortality at 1 year continued to be higher (OR 1.48, 95% CI: 1.16-1.89, P = 0.002).

CONCLUSION

We found a strong association between the presence of MV at heart transplantation and 90-day and 1-year mortality. Future studies are needed to identify which patients requiring MV have reasonable outcomes, and which are associated with substantially poorer outcomes.

The Future Role of Machine Learning in Clinical Transplantation

The use of artificial intelligence and machine learning (ML) has revolutionized our daily lives and will soon be instrumental in healthcare delivery. The rise of ML is due to multiple factors: increasing access to massive datasets, exponential increases in processing power, and key algorithmic developments that allow ML models to tackle increasingly challenging questions. Progressively more transplantation research is exploring the potential utility of ML models throughout the patient journey, although this has not yet widely transitioned into the clinical domain. In this review, we explore common approaches used in ML in solid organ clinical transplantation and consider opportunities for ML to help clinicians and patients. We discuss ways in which ML can aid leverage of large complex datasets, generate cutting-edge prediction models, perform clinical image analysis, discover novel markers in molecular data, and fuse datasets to generate novel insights in modern transplantation practice. We focus on key areas in transplantation in which ML is driving progress, explore the future potential roles of ML, and discuss the challenges and limitations of these powerful tools.

Development and validation of specific post-transplant risk scores according to the circulatory support status at transplant: A UNOS cohort analysis

BACKGROUND

The clinical use of post-transplant risk scores is limited by their poor statistical performance. We hypothesized that developing specific prognostic models for each type of circulatory support at transplant may improve risk stratification.

METHODS

We analyzed the UNOS database including contemporary, first, non-combined heart transplantations (2013-2018). The endpoint was death or retransplantation during the first year post-transplant. Three different circulatory support statuses at transplant were considered: no support, durable mechanical support and temporary support (inotropes, temporary mechanical support). We generated 1,000 bootstrap samples that we randomly split into derivation and test sets. In each sample, we derived an overall model and 3 specific models (1 for each type of circulatory support) using Cox regressions, and compared, in the test set, their statistical performance for each type of circulatory support.

RESULTS

A total of 13,729 patients were included; 1,220 patients (8.9%) met the composite endpoint. Circulatory support status at transplant was associated with important differences in baseline characteristics and distinct prognosis (p = 0.01), interacted significantly with important predictive variables included in the overall model, and had a major impact on post-transplant predictive models (type of variables included and their corresponding hazard ratios). However, specific models suffered from poor discriminative performance and significantly improved risk stratification (discrimination, reclassification indices, calibration) compared to overall models in a very limited proportion of bootstrap samples (<15%). These results were consistent across several sensitivity analyzes.

CONCLUSION

Circulatory support status at transplant reflected different disease states that influenced predictive models. However, developing specific models for each circulatory support status did not significantly improve risk stratification.

State-of-the-art machine learning algorithms for the prediction of outcomes after contemporary heart transplantation: Results from the UNOS database

PURPOSE

We sought to develop and validate machine learning (ML) models to increase the predictive accuracy of mortality after heart transplantation (HT).

METHODS AND RESULTS

We included adult HT recipients from the United Network for Organ Sharing (UNOS) database between 2010 and 2018 using solely pre-transplant variables. The study cohort comprised 18 625 patients (53 ± 13 years, 73% males) and was randomly split into a derivation and a validation cohort with a 3:1 ratio. At 1-year after HT, there were 2334 (12.5%) deaths. Out of a total of 134 pre-transplant variables, 39 were selected as highly predictive of 1-year mortality via feature selection algorithm and were used to train five ML models. AUC for the prediction of 1-year survival was .689, .642, .649, .637, .526 for the Adaboost, Logistic Regression, Decision Tree, Support Vector Machine, and K-nearest neighbor models, respectively, whereas the Index for Mortality Prediction after Cardiac Transplantation (IMPACT) score had an AUC of .569. Local interpretable model-agnostic explanations (LIME) analysis was used in the best performing model to identify the relative impact of key predictors. ML models for 3- and 5-year survival as well as acute rejection were also developed in a secondary analysis and yielded AUCs of .629, .609, and .610 using 27, 31, and 91 selected variables respectively.

CONCLUSION

Machine learning models showed good predictive accuracy of outcomes after heart transplantation.

Statistical performance of 16 posttransplant risk scores in a contemporary cohort of heart transplant recipients

Accurate risk stratification of early heart transplant failure is required to avoid futile transplants and rationalize donor selection. We aimed to evaluate the statistical performance of existing risk scores on a contemporary cohort of heart transplant recipients. After an exhaustive search, we identified 16 relevant risk scores. From the UNOS database, we selected all first noncombined adult heart transplants performed between 2014 and 2017 for validation. The primary endpoint was death or retransplant during the first year posttransplant. For all scores, we analyzed their association with outcomes, sensitivity, specificity, likelihood ratios, and discrimination (concordance index and overlap of individual scores). The cohort included 9396 patients. All scores were significantly associated with the primary outcome (P < .001 for all scores). Their likelihood ratios, both negative and positive, were poor. The discriminative performance of all scores was limited, with concordance index ranging from 0.544 to 0.646 (median 0.594) and an important overlap of individual scores between patients with or without the primary endpoint. Subgroup analyses revealed important variation in discrimination according to donor age, recipient age, and the type of assist device used at transplant. Our findings raise concerns about the use of currently available scores in the clinical field.

Machine learning-based prediction of health outcomes in pediatric organ transplantation recipients

OBJECTIVES

Prediction of post-transplant health outcomes and identification of key factors remain important issues for pediatric transplant teams and researchers. Outcomes research has generally relied on general linear modeling or similar techniques offering limited predictive validity. Thus far, data-driven modeling and machine learning (ML) approaches have had limited application and success in pediatric transplant outcomes research. The purpose of the current study was to examine ML models predicting post-transplant hospitalization in a sample of pediatric kidney, liver, and heart transplant recipients from a large solid organ transplant program.

MATERIALS AND METHODS

Various logistic regression, naive Bayes, support vector machine, and deep learning (DL) methods were used to predict 1-, 3-, and 5-year post-transplant hospitalization using patient and administrative data from a large pediatric organ transplant center.

RESULTS

DL models generally outperformed traditional ML models across organtypes and prediction windows with area under the receiver operating characteristic curve values ranging from 0.750 to 0.851. Shapley additive explanations (SHAP) were used to increase the interpretability of DL model results. Various medical, patient, and social variables were identified as salient predictors across organ types.

DISCUSSION

Results demonstrate the utility of DL modeling for health outcome prediction with pediatric patients, and its use represents an important development in the prediction of post-transplant outcomes in pediatric transplantation compared to prior research.

CONCLUSION

Results point to DL models as potentially useful tools in decision-support systems assisting physicians and transplant teams in identifying patients at a greater risk for poor post-transplant outcomes.

Impact of donor smoking history on post heart transplant outcomes: A propensity-matched analysis of ISHLT registry

PURPOSE

Smoking is a major public health issue, and its effect on cardiovascular outcomes is well established. This study evaluates the impact of donor smoking on heart transplant (HT) outcomes.

METHODS

HT recipients between January 1, 2005, and December 31, 2016, with known donor smoking status were queried from the International Society of Heart and Lung Transplantation (ISHLT) registry. The primary outcome was all-cause mortality, and secondary endpoints were graft failure, acute rejection, and cardiac allograft vasculopathy. We utilized propensity-score matching to identify cohorts of recipients with and without a history of donor smoking. Hazard ratios for post-transplant outcomes for the matched sample were estimated from separate Cox proportional hazard models.

RESULTS

Of 26 390 patients in the cohort, 18.9% had history of donor smoking. Donors with history of smoking were older, predominantly male and had higher incidence of diabetes, hypertension, cocaine use, and "high-risk" status. In propensity-matched analysis, recipients with a history of donor smoking had increased risk of death (HR 1.11, 95% CI 1.03-1.20) and higher risk of graft failure (HR 1.11, 95% CI 1.03-1.20).

CONCLUSION

Donor smoking was associated with increased mortality and higher incidence of graft failure following HT. Consideration of donor smoking history is warranted while evaluating donor hearts.

Twenty-year survival following orthotopic heart transplantation in the United States

BACKGROUND

This study evaluated 20-year survival after adult orthotopic heart transplantation (OHT).

METHODS

The United Network of Organ Sharing Registry database was queried to study adult OHT recipients between 1987 and 1998 with over 20-year posttransplant follow-up. The primary and secondary outcomes were 20-year survival and cause of death after OHT, respectively. Multivariable logistic regression was used to identify significant independent predictors of long-term survival, and long-term survival was compared among cohorts stratified by number of predictors using Kaplan Meier survival analysis.

RESULTS

20,658 patients undergoing OHT were included, with a median follow-up of 9.0 (IQR, 3.2-15.4) years. Kaplan-Meier estimates of 10-, 15-, and 20-year survival were 50.2%, 30.1%, and 17.2%, respectively. Median survival was 10.1 (IQR, 3.9-16.9) years. Increasing recipient age (>65 years), increasing donor age (>40 years), increasing recipient body mass index (>30), black race, ischemic cardiomyopathy, and longer cold ischemic time (>4 h) were adversely associated with a 20-year survival. Of these 6 negative predictors, presence of 0 risk factors had the greatest 10-year (59.7%) and 20-year survival (26.2%), with decreasing survival with additional negative predictors. The most common cause of death in 20-year survivors was renal, liver, and/or multisystem organ failure whereas graft failure more greatly impacted earlier mortality.

CONCLUSIONS

This study identifies six negative preoperative predictors of 20-year survival with 20-year survival rates exceeding 25% in the absence of these factors. These data highlight the potential for very long-term survival after OHT in patients with end-stage heart failure and may be useful for patient selection and prognostication.

Outcomes of mechanical support for cardiogenic shock associated with late cardiac allograft failure

BACKGROUND

Late graft failure (LGF) is an unresolved issue after orthotopic heart transplant (OHT). In this study, we report characteristics and outcomes of severe LGF requiring mechanical circulatory support (MCS).

METHODS

All patients undergoing OHT from 2000 to 2018 at our center were reviewed. Patients re-admitted to the hospital for late graft failure (>3 months after initial discharge) and developing cardiogenic shock requiring MCS were identified. Outcomes and mortality were evaluated.

RESULTS

Twenty-six patients were identified. Median age was 37.3 years (interquartile range: 28.2-47.6) and 69% were male. Median time from initial transplant to MCS was 2.9 years. Etiology of graft failure was rejection in 19 patients (73%), transplant coronary artery disease (tCAD) in 3 (12%), with mixed tCAD or rejection in 4 (15%).

Artificial intelligence and automation in valvular heart diseases

Artificial intelligence (AI) is gradually changing every aspect of social life, and healthcare is no exception. The clinical procedures that were supposed to, and could previously only be handled by human experts can now be carried out by machines in a more accurate and efficient way. The coming era of big data and the advent of supercomputers provides great opportunities to the development of AI technology for the enhancement of diagnosis and clinical decision-making. This review provides an introduction to AI and highlights its applications in the clinical flow of diagnosing and treating valvular heart diseases (VHDs). More specifically, this review first introduces some key concepts and subareas in AI. Secondly, it discusses the application of AI in heart sound auscultation and medical image analysis for assistance in diagnosing VHDs. Thirdly, it introduces using AI algorithms to identify risk factors and predict mortality of cardiac surgery. This review also describes the state-of-the-art autonomous surgical robots and their roles in cardiac surgery and intervention.

A nonrandomized open-label phase 2 trial of nonischemic heart preservation for human heart transplantation

Pre-clinical heart transplantation studies have shown that ex vivo non-ischemic heart preservation (NIHP) can be safely used for 24 h. Here we perform a prospective, open-label, non-randomized phase II study comparing NIHP to static cold preservation (SCS), the current standard for adult heart transplantation. All adult recipients on waiting lists for heart transplantation were included in the study, unless they met any exclusion criteria. The same standard acceptance criteria for donor hearts were used in both study arms. NIHP was scheduled in advance based on availability of device and trained team members. The primary endpoint was a composite of survival free of severe primary graft dysfunction, free of ECMO use within 7 days, and free of acute cellular rejection ≥2R within 180 days. Secondary endpoints were I/R-tissue injury, immediate graft function, and adverse events. Of the 31 eligible patients, six were assigned to NIHP and 25 to SCS. The median preservation time was 223 min (IQR, 202–263) for NIHP and 194 min (IQR, 164–223) for SCS. Over the first six months, all of the patients assigned to NIHP achieved event-free survival, compared with 18 of those assigned to SCS (Kaplan-Meier estimate of event free survival 72.0% [95% CI 50.0–86.0%]). CK-MB assessed 6 ± 2 h after ending perfusion was 76 (IQR, 50–101) ng/mL for NIHP compared with 138 (IQR, 72–198) ng/mL for SCS. Four deaths within six months after transplantation and three cardiac-related adverse events were reported in the SCS group compared with no deaths or cardiac-related adverse events in the NIHP group. This first-in-human study shows the feasibility and safety of NIHP for clinical use in heart transplantation. ClinicalTrial.gov, number NCT03150147

Ischemia and reperfusion damage contribute to early graft dysfunction and recipient’s death. Here the authors show the feasibility and safety of a non-ischemic heart preservation method for heart transplantation in a non-randomized trial.

Heart Transplantation: An In-Depth Survival Analysis

OBJECTIVES

This study aims to understand the complex factors affecting heart transplant survival and to determine the importance of possible sex-specific risk factors.

BACKGROUND

Heart transplant allocation is primarily focused on preventing waitlist mortality. To prevent organ wastage, future allocation must balance risk of waitlist mortality with post-transplantation mortality. However, more information regarding risk factors after heart transplantation is needed.

METHODS

We included all adults (30,606) in the Scientific Registry of Transplant Recipients database who underwent isolated heart transplantation from January 1, 2004, to July 1, 2018. Mortality (8,278 deaths) was verified with the complete Social Security Death Index with a median follow-up of 3.9 years. Temporal decomposition was used to identify phases of survival and phase-specific risk factors. The random survival forests method was used to determine importance of mortality risk factors and their interactions.

RESULTS

We identified 3 phases of mortality risk: early post-transplantation, constant, and late. Sex was not a significant risk factor. There were several interactions predicting early mortality such as pretransplantation mechanical ventilation with presence of end-organ function (bilirubin, renal function) and interactions predicting later mortality such as diabetes and older age (donor and recipient). More complex interactions predicting early-, mid-, and late-mortality existed and were identified with machine learning (i.e., elevated bilirubin, mechanical ventilation, and dialysis).

CONCLUSIONS

Post-heart transplant mortality risk is complex and dynamic, changing with time and events. Sex is not an important mortality risk factor. To prevent organ wastage, end-organ dysfunction should be resolved before transplantation as much as possible.

Review of interactions between high-risk pediatric heart transplant recipients and marginal donors including utilization of risk score models

BACKGROUND

Donor organ acceptance practices vary among pediatric heart transplant professionals. We sought to understand what is known about the interactions between the "high-risk" recipient and the "marginal donor," and how donor risk scores can impact this discussion.

METHODS

A systematic review of published literature on pediatric HTx was undertaken with the assistance of a medical librarian. Two authors independently assessed search results, and papers were reviewed for inclusion.

RESULTS

We found that there are a large number of individual factors, and clusters of factors, that have been used to label individual recipients "high-risk" and individual donors "marginal." The terms "high-risk recipient" and "marginal donor" have been used broadly in the literature making it virtually impossible to make comparisons between publications. In general, the data support that patients who could be easily agreed to be "sicker recipients" are at more risk compared to those who are clearly "healthier," albeit still "sick enough" to need transplantation. Given this variability in the literature, we were unable to define how being a "high-risk" recipient interplays with accepting a "marginal donor." Existing risk scores are described, but none were felt to adequately predict outcomes from factors available at the time of offer acceptance.

CONCLUSIONS

We could not determine what makes a donor "marginal," a recipient "high-risk," or how these factors interplay within the specific recipient-donor pair to determine outcomes. Until there are better risk scores predicting outcomes at the time of organ acceptance, programs should continue to evaluate each organ and recipient individually.

Risk prediction models for survival after heart transplantation: A systematic review

Risk prediction scores have been developed to predict survival following heart transplantation (HT). Our objective was to systematically review the model characteristics and performance for all available scores that predict survival after HT. Ovid Medline and Epub Ahead of Print and In-Process & Other Non-Indexed Citations, Ovid Embase, Cochrane Database of Systematic Reviews, and Cochrane Central Register of Controlled Clinical Trials were searched to December 2018. Eligible articles reported a score to predict mortality following HT. Of the 5392 studies screened, 21 studies were included that derived and/or validated 16 scores. Seven (44%) scores were validated in external cohorts and 8 (50%) assessed model performance. Overall model discrimination ranged from poor to moderate (C-statistic/area under the receiver operating characteristics 0.54-0.77). The IMPACT score was the most widely validated, was well calibrated in two large registries, and was best at discriminating 3-month survival (C-statistic 0.76). Most scores did not perform particularly well in any cohort in which they were assessed. This review shows that there are insufficient data to recommend the use of one model over the others for prediction of post-HT outcomes.

Risk Indices in Deceased-donor Organ Allocation for Transplantation: Review From an Australian Perspective

Over the last decade, organ donation and transplantation rates have increased in Australia and worldwide. Donor and recipient characteristics for most organ types have generally broadened, resulting in the need to consider more complex data in transplant decision-making. As a result of some of these pressures, the Australian software used for donor and recipient data management is currently being updated. Because of the in-built capacity for improved data management, organ allocation processes will have the opportunity to be significantly reviewed, in particular the possible use of risk indices (RIs) to guide organ allocation and transplantation decisions. We aimed to review RIs used in organ allocation policies worldwide and to compare their use to current Australian protocols. Significant donor, recipient, and transplant variables in the indices were summarized. We conclude that Australia has the opportunity to incorporate greater use of RIs in its allocation policies and in transplant decision-making processes. However, while RIs can assist with organ allocation and help guide prognosis, they often have significant limitations which need to be properly appreciated when deciding how to best use them to guide clinical decisions.

Improving risk prediction in heart failure using machine learning

BACKGROUND

Predicting mortality is important in patients with heart failure (HF). However, current strategies for predicting risk are only modestly successful, likely because they are derived from statistical analysis methods that fail to capture prognostic information in large data sets containing multi-dimensional interactions.

METHODS AND RESULTS

We used a machine learning algorithm to capture correlations between patient characteristics and mortality. A model was built by training a boosted decision tree algorithm to relate a subset of the patient data with a very high or very low mortality risk in a cohort of 5822 hospitalized and ambulatory patients with HF. From this model we derived a risk score that accurately discriminated between low and high-risk of death by identifying eight variables (diastolic blood pressure, creatinine, blood urea nitrogen, haemoglobin, white blood cell count, platelets, albumin, and red blood cell distribution width). This risk score had an area under the curve (AUC) of 0.88 and was predictive across the full spectrum of risk. External validation in two separate HF populations gave AUCs of 0.84 and 0.81, which were superior to those obtained with two available risk scores in these same populations.

CONCLUSIONS

Using machine learning and readily available variables, we generated and validated a mortality risk score in patients with HF that was more accurate than other risk scores to which it was compared. These results support the use of this machine learning approach for the evaluation of patients with HF and in other settings where predicting risk has been challenging.

Predicting ideal outcome after pediatric liver transplantation: An exploratory study using machine learning analyses to leverage Studies of Pediatric Liver Transplantation Data

Machine learning analyses allow for the consideration of numerous variables in order to accommodate complex relationships that would not otherwise be apparent in traditional statistical methods to better classify patient risk. The SPLIT registry data were analyzed to determine whether baseline demographic factors and clinical/biochemical factors in the first-year post-transplant could predict ideal outcome at 3 years (IO-3) after LT. Participants who received their first, isolated LT between 2002 and 2006 and had follow-up data 3 years post-LT were included. IO-3 was defined as alive at 3 years, normal ALT (<50) or GGT (<50), normal GFR, no non-liver transplants, no cytopenias, and no PTLD. Heat map analysis and RFA were used to characterize the impact of baseline and 1-year factors on IO-3. 887/1482 SPLIT participants met inclusion criteria; 334 had IO-3. Demographic, biochemical, and clinical variables did not elucidate a visual signal on heat map analysis. RFA identified non-white race (vs white race), increased length of operation, vascular and biliary complications within 30 days, and duct-to-duct biliary anastomosis to be negatively associated with IO-3. UNOS regions 2 and 5 were also identified as important factors. RFA had an accuracy rate of 0.71 (95% CI: 0.68-0.74), PPV = 0.83, and NPV = 0.70. RFA identified participant variables that predicted IO-3. These findings may allow for better risk stratification and personalization of care following pediatric liver transplantation.

Using machine learning to estimate survival curves for patients receiving an increased risk for disease transmission heart, liver, or lung versus waiting for a standard organ

Introduction

Over 19% of deceased organ donors are labeled increased risk for disease transmission (IRD) for viral blood‐borne disease transmission. Many potential organ recipients need to decide between accepting an IRD organ offer and waiting for a non–IRD organ.

Methods

Using machine learning and simulation, we built transplant and waitlist survival models and compared the survival for patients accepting IRD organ offers or waiting for non–IRD organs for the heart, liver, and lung. The simulation consisted of generating 20 000 different scenarios of a recipient either receiving an IRD organ or waiting and receiving a non–IRD organ.

Results

In the simulations, the 5‐year survival probabilities of heart, liver, and lung recipients who accepted IRD organ offers increased on average by 10.2%, 12.7%, and 7.2%, respectively, compared with receiving a non–IRD organ after average wait times (190, 228, and 223 days, respectively). When the estimated waitlist time was at least 5 days for the liver, and 1 day for the heart and lung, 50% or more of the simulations resulted in a higher chance of 5‐year survival when the patient received an IRD organ versus when the patient remained on the waitlist. We also developed a simple equation to estimate the benefits, in terms of 5‐year survival probabilities, of receiving an IRD organ versus waiting for a non–IRD organ, for a particular set of recipient/donor characteristics.

Conclusion

For all three organs, the majority of patients are predicted to have higher 5‐year survival accepting an IRD organ offer compared with waiting for a non–IRD organ.

Prediction of mortality following pediatric heart transplant using machine learning algorithms

BACKGROUND

Optimizing transplant candidates' priority for donor organs depends on the accurate assessment of post-transplant outcomes. Due to the complexity of transplantation and the wide range of possible serious complications, recipient outcomes are difficult to predict accurately using conventional multivariable regression. Therefore, we evaluated the utility of 3 ML algorithms for predicting mortality after pediatric HTx.

METHODS

We identified patients <18 years of age receiving HTx in 2006-2015 in the UNOS Registry database. Mortality within 1, 3, or 5 years was predicted using classification and regression trees, RFs, and ANN. Each model was trained using cross-validation, then validated in a separate testing set. Model performance was primarily evaluated by the area under the receiver operating characteristic (AUC) curve.

RESULTS

The training set included 2802 patients, whereas 700 were included in the testing set. RF achieved the best fit to the training data with AUCs of 0.74, 0.68, and 0.64 for 1-, 3-, and 5-year mortality, respectively, and performed best in the testing data, with AUCs of 0.72, 0.61, and 0.60, respectively. Nevertheless, sensitivity was poor across models (training: 0.22-0.58; testing: 0.07-0.49).

DISCUSSION

ML algorithms demonstrated fair predictive utility in both training and testing data, but the sensitivity of these algorithms was generally poor. With the registry missing data on many determinants of long-term survival, the ability of ML methods to predict mortality after pediatric HTx may be fundamentally limited.

Optimization of heart allocation: The transplant risk score

The new French heart allocation system is designed to minimize waitlist mortality and extend the donor pool without a detrimental effect on posttransplant survival. This study was designed to construct a 1-year posttransplant graft-loss risk score incorporating recipient and donor characteristics. The study included all adult first single-organ recipients transplanted between 2010 and 2014 (N = 1776). This population was randomly divided in a 2:1 ratio into derivation and validation cohorts. The association of variables with 1-year graft loss was determined with a mixed Cox model with center as random effect. The predictors were used to generate a transplant-risk score (TRS). Donor-recipient matching was assessed using 2 separate recipient- and donor-risk scores. Factors associated with 1-year graft loss were recipient age >50 years, valvular cardiomyopathy and congenital heart disease, previous cardiac surgery, diabetes, mechanical ventilation, glomerular filtration rate and bilirubin, donor age >55 years, and donor sex: female. The C-index of the final model was 0.70. Correlation between observed and predicted graft loss rate was excellent for the overall cohort (r = 0.90). Hearts from high-risk donors transplanted to low-risk recipients had similar survival as those from low-risk donors. The TRS provides an accurate prediction of 1-year graft-loss risk and allows optimal donor-recipient matching.

Simulating the Outcome of Heart Allocation Policies Using Deep Neural Networks

We created a system to simulate the heart allocation process in a transplant queue, using a discrete event model and a neural network algorithm, which we named the Lund Deep Learning Transplant Algorithm (LuDeLTA). LuDeLTA is utilized to predict the survival of the patients both in the queue and after transplant. We tried four different allocation policies: wait time, clinical rules and allocating the patients using either LuDeLTA or The International Heart Transplant Survival Algorithm (IHTSA) model. Both IHTSA and LuDeLTA were used to evaluate the results. The predicted mean survival for allocating according to wait time was about 4,300 days, clinical rules 4,300 days and using neural networks 4,700 days.

Predicting the outcome for patients in a heart transplantation queue using deep learning

Heart transplantations have made it possible to extend the median survival time to 12 years for patients with end-stage heart diseases. This operation is unfortunately limited by the availability of donor organs and patients have to wait on average about 200 days in a waiting list before being operated. This waiting time varies considerably across the patients. In this paper, we studied the outcome for patients entering a transplantation waiting list using deep learning techniques. We implemented a model in the form of two-layer neural networks and we predicted the outcome as still waiting, transplanted or dead in the waiting list, at three different time points: 180 days, 365 days, and 730 days. As data source, we used the United Network for Organ Sharing (UNOS) registry, where we extracted adult patients (>17 years) from January 2000 to December 2011. We trained our model using the Keras framework, and we report F1 macro scores of respectively 0.674, 0.680, and 0.680 compared to a baseline of 0.271. We also applied a backward elimination procedure, using our neural network, to extract the 10 most significant parameters predicting the patient status for the three different time points.

Improving prediction of heart transplantation outcome using deep learning techniques

The primary objective of this study is to compare the accuracy of two risk models, International Heart Transplantation Survival Algorithm (IHTSA), developed using deep learning technique, and Index for Mortality Prediction After Cardiac Transplantation (IMPACT), to predict survival after heart transplantation. Data from adult heart transplanted patients between January 1997 to December 2011 were collected from the UNOS registry. The study included 27,860 heart transplantations, corresponding to 27,705 patients. The study cohorts were divided into patients transplanted before 2009 (derivation cohort) and from 2009 (test cohort). The receiver operating characteristic (ROC) values, for the validation cohort, computed for one-year mortality, were 0.654 (95% CI: 0.629-0.679) for IHTSA and 0.608 (0.583-0.634) for the IMPACT model. The discrimination reached a C-index for long-term survival of 0.627 (0.608-0.646) for IHTSA, compared with 0.584 (0.564-0.605) for the IMPACT model. These figures correspond to an error reduction of 12% for ROC and 10% for C-index by using deep learning technique. The predicted one-year mortality rates for were 12% and 22% for IHTSA and IMPACT, respectively, versus an actual mortality rate of 10%. The IHTSA model showed superior discriminatory power to predict one-year mortality and survival over time after heart transplantation compared to the IMPACT model.

Role of Depression and Social Isolation at Time of Waitlisting for Survival 8 Years After Heart Transplantation

Background

We evaluated depression and social isolation assessed at time of waitlisting as predictors of survival in heart transplant (HTx) recipients.

Methods and Results

Between 2005 and 2006, 318 adult HTx candidates were enrolled in the Waiting for a New Heart Study, and 164 received transplantation. Patients were followed until February 2013. Psychosocial characteristics were assessed by questionnaires. Eurotransplant provided medical data at waitlisting, transplantation dates, and donor characteristics; hospitals reported medical data at HTx and date of death after HTx. During a median follow‐up of 70 months (<1–93 months post‐HTx), 56 (38%) of 148 transplanted patients with complete data died. Depression scores were unrelated to social isolation, and neither correlated with disease severity. Higher depression scores increased the risk of dying (hazard ratio=1.07, 95% confidence interval, 1.01, 1.15, P=0.032), which was moderated by social isolation scores (significant interaction term; hazard ratio = 0.985, 95% confidence interval, 0.973, 0.998; P=0.022). These findings were maintained in multivariate models controlling for covariates (P values 0.020–0.039). Actuarial 1‐year/5‐year survival was best for patients with low depression who were not socially isolated at waitlisting (86% after 1 year, 79% after 5 years). Survival of those who were either depressed, or socially isolated or both, was lower, especially 5 years posttransplant (56%, 60%, and 62%, respectively).

Conclusions

Low depression in conjunction with social integration at time of waitlisting is related to enhanced chances for survival after HTx. Both factors should be considered for inclusion in standardized assessments and interventions for HTx candidates.

Impact of hyperlipidemia on alloimmunity

PURPOSE OF REVIEW

Hyperlipidemia is a comorbidity affecting a significant number of transplant patients despite treatment with cholesterol lowering drugs. Recently, it has been shown that hyperlipidemia can significantly alter T-cell responses to cardiac allografts in mice, and graft rejection is accelerated in dyslipidemic mice. Here, we review recent advances in our understanding of hyperlipidemia in graft rejection.

RECENT FINDINGS

Hyperlipidemic mice have significant increases in serum levels of proinflammatory cytokines, and neutralization of interleukin 17 (IL-17) slows graft rejection, suggesting that IL-17 production by Th17 cells was necessary but not sufficient for rejection. Hyperlipidemia also causes an increase in alloreactive T-cell responses prior to antigen exposure. Analysis of peripheral tolerance mechanisms indicated that this was at least in part due to alterations in FoxP3 T cells that led to reduced Treg function and the expansion of FoxP3 CD4 T cells expressing low levels of CD25. Functionally, alterations in Treg function prevented the ability to induce operational tolerance to fully allogeneic heart transplants through costimulatory-molecule blockade, a strategy that requires Tregs.

SUMMARY

These findings highlight the importance of considering the contribution of inflammatory comorbidities to cardiac allograft rejection, and point to the potential importance of managing hyperlipidemia in the transplant population.