Comparing Scientific Registry of Transplant Recipients posttransplant program-specific outcome ratings at listing with subsequent recipient outcomes after transplant

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.

Textbook Outcome as a Quality Metric in Living and Deceased Donor Kidney Transplantation

BACKGROUND

Quality in kidney transplantation is measured using 1-year patient and graft survival. Because 1-year patient and graft survival exceed 95%, this metric fails to measure a spectrum of quality. Textbook outcomes (TO) are a composite quality metric offering greater depth and resolution. We studied TO after living donor (LD) and deceased donor (DD) kidney transplantation.

STUDY DESIGN

United Network for Organ Sharing data for 69,165 transplant recipients between 2013 and 2017 were analyzed. TO was defined as patient and graft survival of 1 year or greater, 1-year glomerular filtration rate of greater than 40 mL/min, absence of delayed graft function, length of stay of 5 days or less, no readmissions during the first 6 months, and no episodes of rejection during the first year after transplantation. Bivariate analysis identified characteristics associated with TO, and covariates were incorporated into multivariable models. Five-year conditional survival was measured, and center TO rates were corrected for case complexity to allow center-level comparisons.

RESULTS

The national average TO rates were 54.1% and 31.7% for LD and DD transplant recipients. The hazard ratio for death at 5 years for recipients who did not experience TO was 1.92 (95% CI 1.68 to 2.18, p ≤ 0.0001) for LD transplant recipients and 2.08 (95% CI 1.93 to 2.24, p ≤ 0.0001) for DD transplant recipients. Center-level comparisons identify 18% and 24% of centers under-performing in LD and DD transplantation. High rates of TO do not correlate with transplantation center volume.

CONCLUSION

Kidney transplant recipients who experience TO have superior long-term survival. Textbook outcomes add value to the current standards of 1-year patient and graft survival.

A Composite End Point of Graft Status and eGFR at 1 Year to Improve the Scientific Registry of Transplant Recipients' Five-Tier Rating System

BACKGROUND

Performance of kidney transplant programs in the United States is monitored and publicly reported by the Scientific Registry of Transplant Recipients (SRTR). With relatively few allograft failure events per program and increasing homogeneity in program performance, quantifying meaningful differences in program competency based only on 1-year survival rates is challenging.

METHODS

We explored whether the traditional end point of allograft failure at 1 year can be improved by incorporating a measure of allograft function (i.e., eGFR) into a composite end point. We divided SRTR data from 2008 through 2018 into a training and validation set and recreated SRTR tiers, using the traditional and composite end points. The conditional 5-year deceased donor allograft survival and 5-year eGFR were then assessed using each approach.

RESULTS

Compared with the traditional end point, the composite end point of graft failure or eGFR <30 ml/min per 1.73 m² at 1-year post-transplant performed better in stratifying transplant programs based on long-term deceased donor graft survival. For tiers 1 through 5 respectively, the 5-year conditional graft survival was 72.9%, 74.8%, 75.4%, 77.0%, and 79.7% using the traditional end point and 71.1%, 74.4%, 76.9%, 77.0%, and 78.4% with the composite end point. Additionally, with the five-tier system derived from the composite end point, programs in tier 3, tier 4, and tier 5 had significantly higher mean eGFRs at 5 years compared with programs in tier 1. There were no significant eGFR differences among tiers derived from the traditional end point alone.

CONCLUSIONS

This proof-of-concept study suggests that a composite end point incorporating allograft function may improve the post-transplant component of the five-tier system by better differentiating between transplant programs with respect to long-term graft outcomes.

Significant improvements, but consistent disparities in survival for African Americans after liver transplantation

Despite improvements in survival across races in the past 20 years, African Americans have worse liver transplant outcomes after orthotopic liver transplantation (OLT). This study aims at quantifying the change in disparities between African Americans and other races in survival after OLT. We retrospectively analyzed the United Network for Organ Sharing (UNOS) database for patient data for candidates who received a liver transplant between January 1, 2007 and December 31, 2017. Multivariate Cox proportional hazards regression indicated similar decreases in mortality over time for each race with a decrease in mortality for African Americans: 2010-2012 (HR = .930), 2012-2015 (HR = .882), and 2015-2017 (HR = .883) when compared to 2007-2010. Risk of mortality for African Americans compared to Caucasians varied across the 4 eras: 2007-2010 (HR = 1.083), 2010-2012 (HR = 1.090), 2012-2015 (HR = 1.070), and 2015-2017 (HR = 1.125). While African Americans have seen increases in survival in the past decade, a similar increase in survival for other races leaves a significant survival disparity in African Americans.

Variation in Racial Disparities in Liver Transplant Outcomes Across Transplant Centers in the United States

Little is known about the role that transplant centers may play in perpetuating racial disparities after liver transplantation, which are unexplained by patient-level factors. We examined variation in between-center and within-center disparities among 34,114 Black and White liver transplant recipients in the United States from 2010 to 2017 using Scientific Registry of Transplant Recipient (SRTR) data. We used Cox proportional hazards models to calculate transplant center-specific Black-White hazard ratios and hierarchical survival analysis to examine potential effect modification of the race-survival association by transplant center characteristics, including transplant volume, proportion of Black patients, SRTR quality rating, and region. Models were sequentially adjusted for clinical, socioeconomic, and center characteristics. After adjustment, Black patients experienced 1.11 excess deaths after liver transplant per 100 person-years compared with White patients (95% confidence interval [CI], 0.65-1.56), corresponding to a 21% increased mortality risk (95% CI, 1.12-1.31). Although there was substantial variation in this disparity across transplant centers, there was no evidence of effect modification by transplant center volume, proportion of minority patients seen, quality rating, or region. We found significant racial disparities in survival after transplant, with substantial variation in this disparity across transplant centers that was not explained by selected center characteristics. This is the first study to directly evaluate the role transplant centers play in racial disparities in transplant outcomes. Further assessment of the qualitative factors that may drive disparities, such as selection processes and follow-up care, is needed to create effective center-level interventions to address health inequity.

Barriers to High Quality End of Life Care in the Surgical Intensive Care Unit

End of life discussions frequently take place in surgical intensive care units, as a significant number of patients die while admitted to the hospital, and surgery is common during the last month of life. Multiple barriers exist to the initiation of these conversations, including: miscommunication between clinicians and surrogates, a paternalistic approach to surgical patients, and perhaps, conflicts of interest as an unwanted consequence of surgical quality reporting. Goal discordant care refers to the care that is provided to a patient that is incapacitated and that is not concordant to his/her wishes. This is a largely unrecognized medical error with devastating consequences, including inappropriate prolongation of life and non-beneficial therapy utilization. Importantly, hospice and palliative care needs to be recognized as quality care in order to deter the incentives that might persuade clinicians from offering these services.

Posttransplant outcome assessments at listing: Long-term outcomes are more important than short-term outcomes

Posttransplant outcome assessments are publicly reported for patient and regulatory use. However, the currently reported 1-year posttransplant graft survival assessments are commonly criticized for not identifying clinically meaningful differences between programs, and not providing information about longer-term posttransplant outcomes. We investigated the association of different posttransplant outcome assessments available to patients at the time of listing with subsequent posttransplant graft survival. The posttransplant assessments were from period prevalent, rather than incident, cohorts with more timely 1-, 3-, and 5-year follow-up and 6-, 12-, 18-, 24-, and 30-month cohort windows. The association of these assessments at listing with subsequent posttransplant graft survival included candidates listed between July 12, 2011, and December 15, 2015, who subsequently underwent transplant before December 31, 2018. The assessments with 1-year follow-up had uniformly weaker associations than the assessments with 3- and 5-year follow-up. The assessments with 5-year follow-up had the strongest association in kidney and liver transplantation. For kidney, liver, and lung transplantation, assessment windows of at least 18 months typically had the strongest associations with subsequent graft survival. Posttransplant assessments with 5-year follow-up and 18-30-month cohort windows are better than the current posttransplant assessment with 1-year follow-up, particularly at the time of listing.

Comparing Pretransplant and Posttransplant Outcomes When Choosing a Transplant Center: Focus Groups and a Randomized Survey

BACKGROUND

In response to calls for an increased focus on pretransplant outcomes and other patient-centered metrics in public reports of center outcomes, a mixed methods study evaluated how the content and presentation style of new information influences decision-making. The mixed methods design utilized qualitative and quantitative phases where the strengths of one method help address limitations of the other, and multiple methods facilitate comparing results.

METHODS

First, a series of organ-specific focus groups of kidney, liver, heart, and lung patients helped to develop and refine potential displays of center outcomes and understand patient perceptions. A subsequent randomized survey included adult internet users who viewed a single, randomly-selected variation of 6 potential online information displays. Multinomial regression evaluated the effects of graphical presentations of information on decision-making.

RESULTS

One hundred twenty-seven candidates and recipients joined 23 focus groups. Survey responses were analyzed from 975 adults. Qualitative feedback identified patient perceptions of uncertainty in outcome metrics, in particular pretransplant metrics, and suggested a need for clear guidance to interpret the most important metric for organ-specific patient mortality. In the randomized survey, only respondents who viewed a note indicating that transplant rate had the largest impact on survival chose the hospital with the best transplant rate over the hospital with the best posttransplant outcomes (marginal relative risk and 95% confidence interval, 1.161.501.95).

CONCLUSIONS

The presentation of public reports influenced decision-making behavior. The combination of qualitative and quantitative research helped to guide and enhance understanding of the impacts of proposed changes in reported metrics.

Time for reform in transplant program-specific reporting: AST/ASTS transplant metrics taskforce

In accordance with the National Organ Transplant Act and Department of Health and Human Services' Final Rule, the Scientific Registry of Transplant Recipients (SRTR) publicly releases biannual program-specific reports that include analyses of transplant centers' risk-adjusted waitlist mortality, organ acceptance ratios, transplant rates, and graft and patient survival. Since the inception of these center metrics, 1-year posttransplant graft and patient survival have improved, and center variation has decreased, casting uncertainty on their clinical relevance. The SRTR has recently modified center evaluations by ranking centers into 5 tiers rather than 3 tiers in an attempt to discriminate between programs performing within a tight range, further exacerbating this uncertainty. The American Society of Transplantation/American Society of Transplant Surgeons convened an expert taskforce to examine both the utility and unintended consequences of transplant center metrics. Estimates of center variation in outcomes in adjacent tiers are imprecise and fleeting, but can result in consequential changes in clinician and center behavior. The taskforce has concerns that current metrics, based principally on 1-year graft and patient survival, provide minimal if any benefit in informing patient choice and access to transplantation, with the untoward effect of decreased utilization of organs and restriction of research and innovation.