Predicting earliest safe extubation time in pediatric patients

点击翻译按钮获取中文摘要

PROJECT SUMMARY Determining when to extubate patients in the pediatric intensive care unit (PICU) is a challenge clinicians face each day. Consensus guidelines for pediatric extubation are lacking and, in light of this, most pediatric studies conclude that the decision to extubate relies ultimately on clinician judgment. The resulting variation in care translates to increased morbidity, mortality, and costs that arise from both unnecessary ventilator days from delayed extubation and re-intubation from extubation failure. The long-term goal of this project is to harness the power of artificial intelligence to optimize identification of extubation readiness in the PICU. The objective of this proposal is to create machine learning models using a large electronic health record (EHR) dataset to predict when to extubate patients and to estimate how many ventilator days could be saved if such models were used in practice. Deploying such models in the EHR as a real-time decision support tool could safely shorten extubation times by decreasing variation in care and identifying subsets of patients for earlier, safe extubation. This study will use EHR data from mechanically ventilated PICU patients at the University of California, San Francisco to build models to estimate extubation readiness for PICU patients (Aim 1). The investigators will apply human factor design principles, which aim to increase usefulness of tools and help humans do their jobs with higher reliability, to improve model performance. We will use a novel method, expert-augmented machine learning, to incorporate clinician knowledge directly into our models (Aim 2). The performance of the models will be evaluated with standard metrics, as well as with an estimate of number of ventilator days saved, reflecting the potential health impact (Aim 3). This project will advance extubation practices for critically ill children, yielding a predictive tool ready for prospective testing in the EHR that moves toward delivering high reliability healthcare for patients with respiratory failure. This research will advance NHLBI's mission of using data science to improve treatment of patients with lung diseases. The proposed training, guided by an expert mentorship team, will enrich the applicant's knowledge of and skills in data science, machine learning and prediction, and clinical informatics. The content expertise, research competency, and training in quantitative methods the applicant will receive will prepare her well to improve scientific knowledge and clinical practice in her career as an independent researcher.