Machine Learning to Predict Cardiac Death Within 1 Hour After Terminal Extubation

Advancing the Scientific Basis for Determining Death in Controlled Organ Donation After Circulatory Determination of Death

In controlled organ donation after circulatory determination of death (cDCDD), accurate and timely death determination is critical, yet knowledge gaps persist. Further research to improve the science of defining and determining death by circulatory criteria is therefore warranted. In a workshop sponsored by the National Heart, Lung, and Blood Institute, experts identified research opportunities pertaining to scientific, conceptual, and ethical understandings of DCDD and associated technologies. This article identifies a research strategy to inform the biomedical definition of death, the criteria for its determination, and circulatory death determination in cDCDD. Highlighting knowledge gaps, we propose that further research is needed to inform the observation period following cessation of circulation in pediatric and neonatal populations, the temporal relationship between the cessation of brain and circulatory function after the withdrawal of life-sustaining measures in all patient populations, and the minimal pulse pressures that sustain brain blood flow, perfusion, activity, and function. Additionally, accurate predictive tools to estimate time to asystole following the withdrawal of treatment and alternative monitoring modalities to establish the cessation of circulatory, brainstem, and brain function are needed. The physiologic and conceptual implications of postmortem interventions that resume circulation in cDCDD donors likewise demand attention to inform organ recovery practices. Finally, because jurisdictionally variable definitions of death and the criteria for its determination may impede collaborative research efforts, further work is required to achieve consensus on the physiologic and conceptual rationale for defining and determining death after circulatory arrest.

The Pediatric Data Science and Analytics Subgroup of the Pediatric Acute Lung Injury and Sepsis Investigators Network: Use of Supervised Machine Learning Applications in Pediatric Critical Care Medicine Research

OBJECTIVE

Perform a scoping review of supervised machine learning in pediatric critical care to identify published applications, methodologies, and implementation frequency to inform best practices for the development, validation, and reporting of predictive models in pediatric critical care.

DESIGN

Scoping review and expert opinion.

SETTING

We queried CINAHL Plus with Full Text (EBSCO), Cochrane Library (Wiley), Embase (Elsevier), Ovid Medline, and PubMed for articles published between 2000 and 2022 related to machine learning concepts and pediatric critical illness. Articles were excluded if the majority of patients were adults or neonates, if unsupervised machine learning was the primary methodology, or if information related to the development, validation, and/or implementation of the model was not reported. Article selection and data extraction were performed using dual review in the Covidence tool, with discrepancies resolved by consensus.

SUBJECTS

Articles reporting on the development, validation, or implementation of supervised machine learning models in the field of pediatric critical care medicine.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Of 5075 identified studies, 141 articles were included. Studies were primarily (57%) performed at a single site. The majority took place in the United States (70%). Most were retrospective observational cohort studies. More than three-quarters of the articles were published between 2018 and 2022. The most common algorithms included logistic regression and random forest. Predicted events were most commonly death, transfer to ICU, and sepsis. Only 14% of articles reported external validation, and only a single model was implemented at publication. Reporting of validation methods, performance assessments, and implementation varied widely. Follow-up with authors suggests that implementation remains uncommon after model publication.

CONCLUSIONS

Publication of supervised machine learning models to address clinical challenges in pediatric critical care medicine has increased dramatically in the last 5 years. While these approaches have the potential to benefit children with critical illness, the literature demonstrates incomplete reporting, absence of external validation, and infrequent clinical implementation.

Pediatric Death After Withdrawal of Life-Sustaining Therapies: A Scoping Review

OBJECTIVES

Evaluate literature on the dying process in children after withdrawal of life sustaining measures (WLSM) in the PICU. We focused on the physiology of dying, prediction of time to death, impact of time to death, and uncertainty of the dying process on families, healthcare workers, and organ donation.

DATA SOURCES

MEDLINE, Embase, Cochrane Central Register of Controlled Trials, PsycINFO, CINAHL, and Web of Science.

STUDY SELECTION

We included studies that discussed the dying process after WLSM in the PICU, with no date or study type restrictions. We excluded studies focused exclusively on adult or neonatal populations, children outside the PICU, or on organ donation or adult/pediatric studies where pediatric data could not be isolated.

DATA EXTRACTION

Inductive qualitative content analysis was performed.

DATA SYNTHESIS

Six thousand two hundred twenty-five studies were screened and 24 included. Results were grouped into four categories: dying process, perspectives of healthcare professionals and family, WLSM and organ donation, and recommendations for future research. Few tools exist to predict time to death after WLSM in children. Most deaths after WLSM occur within 1 hour and during this process, healthcare providers must offer support to families regarding logistics, medications, and expectations. Providers describe the unpredictability of the dying process as emotionally challenging and stressful for family members and staff; however, no reports of families discussing the impact of time to death prediction were found. The unpredictability of death after WLSM makes families less likely to pursue donation. Future research priorities include developing death prediction tools of tools, provider and parental decision-making, and interventions to improve end-of-life care.

CONCLUSIONS

The dying process in children is poorly understood and understudied. This knowledge gap leaves families in a vulnerable position and the clinical team without the necessary tools to support patients, families, or themselves. Improving time to death prediction after WLSM may improve care provision and enable identification of potential organ donors.

Palliative Extubation in Pediatric Patients in the Intensive Care Unit and at Home: A Scoping Review

Aim

This scoping review is aimed at systematically mapping the evidence on palliative extubation in the pediatric intensive care unit.

Methods

MEDLINE, EBSCO, and Cochrane databases were searched for articles published between January 2018 and December 2022, in English. Critical appraisal of sources of evidence was done using the Joanna Briggs Institute tools. PRISMA guidelines for scoping reviews were followed.

Results

Six studies were included, with 366 patients, from the USA (n = 4), Brazil (n = 1), and Germany (n = 1). Three were high-quality studies, two were moderate, and one was a low-quality study. Most studies were retrospective analysis; two were narrative approaches; two were evidence-based recommendation and quality improvement project; one study was a prospective intervention.

Conclusion

Symptom control is crucial pre- and postextubation. A checklist (symptom management and family support) and a postdebriefing template improve team communication and staff support postextubation. Critical care transports from the hospital are feasible to provide extubation at home. A framework addressing common planning challenges and resource management is recommended for extubation at home. The provision of pediatric palliative extubation is necessary since futile measures and prolongation of suffering violate the principle of nonmaleficence. Future research on this subject will result in more benefits for patients, parents, and professionals.

Narrowing the gap: expected versus deployment performance

OBJECTIVES

Successful model development requires both an accurate a priori understanding of future performance and high performance on deployment. Optimistic estimations of model performance that are unrealized in real-world clinical settings can contribute to nonuse of predictive models. This study used 2 tasks, predicting ICU mortality and Bi-Level Positive Airway Pressure failure, to quantify: (1) how well internal test performances derived from different methods of partitioning data into development and test sets estimate future deployment performance of Recurrent Neural Network models and (2) the effects of including older data in the training set on models' performance.

MATERIALS AND METHODS

The cohort consisted of patients admitted between 2010 and 2020 to the Pediatric Intensive Care Unit of a large quaternary children's hospital. 2010-2018 data were partitioned into different development and test sets to measure internal test performance. Deployable models were trained on 2010-2018 data and assessed on 2019-2020 data, which was conceptualized to represent a real-world deployment scenario. Optimism, defined as the overestimation of the deployed performance by internal test performance, was measured. Performances of deployable models were also compared with each other to quantify the effect of including older data during training.

RESULTS, DISCUSSION, AND CONCLUSION

Longitudinal partitioning methods, where models are tested on newer data than the development set, yielded the least optimism. Including older years in the training dataset did not degrade deployable model performance. Using all available data for model development fully leveraged longitudinal partitioning by measuring year-to-year performance.

Analgesia and Sedation at Terminal Extubation: A Secondary Analysis From Death One Hour After Terminal Extubation Study Data

OBJECTIVES

To describe the doses of opioids and benzodiazepines administered around the time of terminal extubation (TE) to children who died within 1 hour of TE and to identify their association with the time to death (TTD).

DESIGN

Secondary analysis of data collected for the Death One Hour After Terminal Extubation study.

SETTING

Nine U.S. hospitals.

PATIENTS

Six hundred eighty patients between 0 and 21 years who died within 1 hour after TE (2010-2021).

MEASUREMENTS AND MAIN RESULTS

Medications included total doses of opioids and benzodiazepines 24 hours before and 1 hour after TE. Correlations between drug doses and TTD in minutes were calculated, and multivariable linear regression performed to determine their association with TTD after adjusting for age, sex, last recorded oxygen saturation/F io2 ratio and Glasgow Coma Scale score, inotrope requirement in the last 24 hours, and use of muscle relaxants within 1 hour of TE. Median age of the study population was 2.1 years (interquartile range [IQR], 0.4-11.0 yr). The median TTD was 15 minutes (IQR, 8-23 min). Forty percent patients (278/680) received either opioids or benzodiazepines within 1 hour after TE, with the largest proportion receiving opioids only (23%, 159/680). Among patients who received medications, the median IV morphine equivalent within 1 hour after TE was 0.75 mg/kg/hr (IQR, 0.3-1.8 mg/kg/hr) ( n = 263), and median lorazepam equivalent was 0.22 mg/kg/hr (IQR, 0.11-0.44 mg/kg/hr) ( n = 118). The median morphine equivalent and lorazepam equivalent rates after TE were 7.5-fold and 22-fold greater than the median pre-extubation rates, respectively. No significant direct correlation was observed between either opioid or benzodiazepine doses before or after TE and TTD. After adjusting for confounding variables, regression analysis also failed to show any association between drug dose and TTD.

CONCLUSIONS

Children after TE are often prescribed opioids and benzodiazepines. For patients dying within 1 hour of TE, TTD is not associated with the dose of medication administered as part of comfort care.

Prediction of survival time after terminal extubation: the balance between critical care unit utilization and hospice medicine in the COVID-19 pandemic era

BACKGROUND

We established 1-h and 1-day survival models after terminal extubation to optimize ventilator use and achieve a balance between critical care for COVID-19 and hospice medicine.

METHODS

Data were obtained from patients with end-of-life status at terminal extubation from 2015 to 2020. The associations between APACHE II scores and parameters with survival time were analyzed. Parameters with a p-value ≤ 0.2 in univariate analysis were included in multivariate models. Cox proportional hazards regression analysis was used for the multivariate analysis of survival time at 1 h and 1 day.

RESULTS

Of the 140 enrolled patients, 76 (54.3%) died within 1 h and 35 (25%) survived beyond 24 h. No spontaneous breathing trial (SBT) within the past 24 h, minute ventilation (MV) ≥ 12 L/min, and APACHE II score ≥ 25 were associated with shorter survival in the 1 h regression model. Lower MV, SpO2 ≥ 96% and SBT were related to longer survival in the 1-day model. Hospice medications did not influence survival time.

CONCLUSION

An APACHE II score of ≥ 25 at 1 h and SpO2 ≥ 96% at 1 day were strong predictors of disposition of patients to intensivists. These factors can help to objectively tailor pathways for post-extubation transition and rapidly allocate intensive care unit resources without sacrificing the quality of palliative care in the era of COVID-19. Trial registration They study was retrospectively registered. IRB No.: 202101929B0.

Paediatric death after withdrawal of life-sustaining therapies: a scoping review protocol

INTRODUCTION

The physiology of dying after withdrawal of life-sustaining measures (WLSM) is not well described in children. This lack of knowledge makes predicting the duration of the dying process difficult. For families, not knowing this process's duration interferes with planning of rituals related to dying, travel for distant relatives and emotional strain during the wait for death. Time-to-death also impacts end-of-life care and determines whether a child will be eligible for donation after circulatory determination of death. This scoping review will summarise the current literature about what is known about the dying process in children after WLSM in paediatric intensive care units (PICUs).

METHODS AND ANALYSIS

This review will use Joanna Briggs Institute methodology for scoping reviews. Databases searched will include Ovid MEDLINE, Ovid Embase, Cochrane Central Register of Controlled Trials via EBM Reviews Ovid, Ovid PsycINFO, CINAHL and Web of Science. Literature reporting on the physiology of dying process after WLSM, or tools that predict time of death in children after WLSM among children aged 0-18 years in PICUs worldwide will be considered. Literature describing the impact of prediction or timing of death after WLSM on families, healthcare workers and the organ donation process will also be included. Quantitative and qualitative studies will be evaluated. Two independent reviewers will screen references by title and abstract, and then by full text, and complete data extraction and analysis.

ETHICS AND DISSEMINATION

The review uses published data and does not require ethics review. Review results will be published in a peer-reviewed scientific journal.

Predicting Time to Death After Withdrawal of Life-Sustaining Treatment in Children

Accurately predicting time to death after withdrawal of life-sustaining treatment is valuable for family counseling and for identifying candidates for organ donation after cardiac death. This topic has been well studied in adults, but literature is scant in pediatrics. The purpose of this report is to assess the performance and clinical utility of the available tools for predicting time to death after treatment withdrawal in children.

Continuous Prediction of Mortality in the PICU: A Recurrent Neural Network Model in a Single-Center Dataset

OBJECTIVES

Develop, as a proof of concept, a recurrent neural network model using electronic medical records data capable of continuously assessing an individual child's risk of mortality throughout their ICU stay as a proxy measure of severity of illness.

DESIGN

Retrospective cohort study.

SETTING

PICU in a tertiary care academic children's hospital.

PATIENTS/SUBJECTS

Twelve thousand five hundred sixteen episodes (9,070 children) admitted to the PICU between January 2010 and February 2019, partitioned into training (50%), validation (25%), and test (25%) sets.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

On 2,475 test set episodes lasting greater than or equal to 24 hours in the PICU, the area under the receiver operating characteristic curve of the recurrent neural network's 12th hour predictions was 0.94 (CI, 0.93-0.95), higher than those of Pediatric Index of Mortality 2 (0.88; CI, [0.85-0.91]; p < 0.02), Pediatric Risk of Mortality III (12th hr) (0.89; CI, [0.86-0.92]; p < 0.05), and Pediatric Logistic Organ Dysfunction day 1 (0.85; [0.81-0.89]; p < 0.002). The recurrent neural network's discrimination increased with more acquired data and smaller lead time, achieving a 0.99 area under the receiver operating characteristic curve 24 hours prior to discharge. Despite not having diagnostic information, the recurrent neural network performed well across different primary diagnostic categories, generally achieving higher area under the receiver operating characteristic curve for these groups than the other three scores. On 692 test set episodes lasting greater than or equal to 5 days in the PICU, the recurrent neural network area under the receiver operating characteristic curves significantly outperformed their daily Pediatric Logistic Organ Dysfunction counterparts (p < 0.005).

CONCLUSIONS

The recurrent neural network model can process hundreds of input variables contained in a patient's electronic medical record and integrate them dynamically as measurements become available. Its high discrimination suggests the recurrent neural network's potential to provide an accurate, continuous, and real-time assessment of a child in the ICU.