Stratification risk computation of primary graft gysfunction in heart transplant recipients on a smartphone application

Abstract Introduction There are no standardized guidelines for helping graft allocation in heart transplant (HTx) and accepting transplants from marginal donors and donors after circulatory arrest make decision making even more complex and time consuming than ever. The organ allocation process should include a risk analysis for primary graft dysfunction (PGD) that represents the most dreadful complication of HTx. There are 3 score systems validated to predict PGD but none has clear superior predictive capability compared to the others and the complexity of score computation potentially discourage their clinical use. A score integrating all the variables presented in a user-friendly format should improve efficacy and safeness of the organ allocation process. Methods Severe PGD defined as need for temporary mechanical circulatory support (ECMO) in the first 24 hours following HTx was considered the end-point for risk stratification computation. As a derivation cohort we used meta-analytic approach for identifying variables that are independently associated to severe PGD. The overall prevalence of each variable was determined to define the absolute risk increase value in presence of each factor. We then created a scoring system applying multivariate logistic regression model. As a validation cohort we studied the performance of the score in a retospective series of HTx performed at our institution from 2013 to 2022. We used C statistic equivalent to the area under a receiver-operating characteristic curve for dichotomous outcomes. Values above 0.7 were considered acceptable. We eventually developed an application for smartphone for risk computation. Results On 64 papers screened, 5 studies were considered eligible for the derivation cohort definition. Cumulative number of patients considered was 10’175 with an incidence of severe PGD of 7.1% (CI 5.3 - 10.4) and a 30-day mortality of 41%. 13 variables were identified (table 1). The ROC curve predicted the use of ECMO with the specificity and the sensitivity of 0,81 and 0,92 respectively. We have included in the validation cohort 128 out of 150 transplanted patients, excluding pediatric and double organs transplants. The mean donor age was 45 +/- 12,4 years, 72 (54%) were men. The mean recipient age was 57 +/- 9,3; 92 were men; 32 recipients (25%) needed ECMO support within 24 hours post-transplant; 75% of the donors in the ECMO group were > 40 y.o.; 30 days mortality was 3.1% and 6,2% in the non-ECMO and ECMO group respectively. We validated the model on our cohort defying a binary outcome (ECMO yes/no). Conclusion Clinicians fill donor, recipient and surgical parameters in user-friendly smartphone application identifying patients that are most likely to need ECMO support post HTx. The decision-making process during organ allocation evolves from what was once little more than personal experience and intuition to a reliable evidence-based method based on personalized risk prediction.Table 1