Predicting Pediatric Risk of Death After Hospitalization

描述（由申请人提供）：死亡儿童的前瞻性识别是一项必须完成的重要任务，以提高临终关怀、临终关怀和丧亲（P-EOL-B）服务对死亡儿童及其家庭的有益影响，因为如果要及时采取有效的个人干预措施，这种前瞻性识别是必不可少的。我们假设，使用行政卫生服务利用数据，可以前瞻性地识别相当大比例的儿童谁死于复杂的慢性疾病，并积极主动的识别可以改善提供P-EOL-B照顾这些孩子和他们的家庭。因此，我们的目标是：1.开发一种统计方法，使用行政卫生利用数据将患者分类为低风险、中低风险、中高风险和高风险组，用于随后的1年死亡率。2.发展合作关系，并计划使用来自其他环境的数据进行更大规模的复制研究，以验证能够在当地和不断发展的儿科医疗保健服务系统中重建这种重复算法的技术。该算法将使用逻辑回归和递归分割来开发，该算法将对儿童年龄住院后365天的死亡概率进行建模。我们将利用宾夕法尼亚州卫生保健成本委员会数据库中包含的与生命统计记录相关的医院出院管理数据。有了这个创新的死亡风险预测系统，我们假设P-EOL-B和医院和卫生保健项目中的牧师护理团队将能够识别单个患者或患者类别（例如，脑瘫患者和第一次或第二次因吸入性肺炎住院的患者）适合进行护理质量评估和改善，有针对性的重症监护协调，以尽早将P-EOL-B护理引入这些患者和家庭的综合护理方案中。

DESCRIPTION (provided by applicant): The prospective identification of dying children is an essential task that must be accomplished in order to improve the beneficial impact of palliative, end-of-life, and bereavement (P-EOL-B) services on dying children and their families, since this prospective identification is essential if effective individual-level interventions are to be employed in a timely manner. We hypothesize that using administrative health service utilization data, one can prospectively identify with reasonable accuracy a substantial proportion of children who die from complex chronic conditions, and that proactive identification can improve the delivery of P-EOL-B care to these children and their families. We therefore aim to: 1. Develop a method of prognostication that would use administrative health utilization data to triage patients into low, medium-low, medium-high, and high risk groups for subsequent 1-year mortality. 2. Develop collaborative relationships and plan a larger replication study using data from other contexts, in order to validate techniques that would enable the recreation of this prognostication algorithm in the context of local and evolving systems of pediatric health care services. The algorithm will be developed using logistic regression and recursive partitioning, which will model the probability of dying 365 days after hospitalization among children age. We will utilize administrative hospital discharge data contained in the Pennsylvania Health Care Cost Commission database, linked to vital statistics records. With this innovative mortality risk prediction system in hand, we postulate that P-EOL-B and pastoral care teams in hospitals and health care programs would be able to identify individual patients or classes of patients (eg, those with cerebral palsy and the first or second hospitalization for an aspiration pneumonia) suitable for quality of care assessment and improvement, with targeted intensive care coordination to introduce P-EOL-B care earlier into these patients' and families' comprehensive package of care.