Death One Hour After Terminal Extubation (DONATE) 2.0 Study - PROJECT SUMMARY/ABSTRACT Children have the highest mortality rate on the organ transplant waiting list compared to any other age group, including elderly adults. Organ donation after circulatory death (DCD) has increased the number of available organs, but the process is challenging. In DCD, organ procurement necessitates death within a short period of time after terminal extubation, typically one hour, to avoid ischemic damage to the donor organ. If a child does not die within the organ recovery period, they are considered ineligible to donate. While DCD has the potential to save the lives of children who would otherwise die due to lack of transplantable organs, unsuccessful DCD attempts are highly resource intensive and can place additional emotional strain on a grieving family. To balance these competing issues, there is a critical need to accurately predict which patients will die within the organ recovery period. Our team trained and tested the high-performing Death One Hour After Terminal Extubation (DONATE) machine learning model using multi-site data to predict whether a patient would die within one hour after terminal extubation. Using this model, we designed a prototype of a clinical decision support (CDS) tool to aid clinicians in their assessment of DCD candidacy, which we aim to test and eventually use in clinical practice. Unfortunately, even successful models and CDS tools often fail translation into practice because this is attempted without sufficient attention to real-world use, clinician input, and clinical context. Implementation science frameworks must be used to rigorously study the users and environment prior to deployment of any new model or tool. The proposed research has two specific aims: (1) to validate the DONATE model’s performance at different time points prior to terminal extubation so that the recommended time range for clinical use of the model can be defined, and (2) to define the facilitators of and barriers to implementation of the DONATE model in the pediatric intensive care unit (PICU) and to test the usability of the CDS tool prototype. To accomplish Aim 1, we will perform a retrospective cohort study of patients terminally extubated in 5 PICUs from 2022- 2023. We will assess the model’s performance when it is used at different time points prior to terminal extubation. Metrics will include area under the receiver operator characteristic curve, sensitivity, false positive rate, positive predictive value, and number needed to alert. For Aim 2, we will perform semi-structured interviews with PICU clinicians at two sites and test the usability of the CDS tool prototype with simulated patient data. We will perform coded analysis to establish themes based on the Consolidated Framework for Implementation Research. Participants will answer validated survey questions on the user experience. We will make user-informed adaptations to the CDS tool prototype. After completion of these aims, we will have generated a user-informed CDS tool prototype that is ready for development and EHR integration, as well as candidate implementation strategies to support a Type 1 multi-site hybrid effectiveness-implementation trial. Click here to enter text.
