Gill Stephen D, Lane Stephen E, Sheridan Michael, Ellis Elizabeth, Smith Darren, Stella Julian
Emergency Department, University Hospital Geelong, Geelong, Victoria, Australia.
Physiotherapy Department, University Hospital Geelong, Geelong, Victoria, Australia.
Emerg Med Australas. 2018 Oct;30(5):641-647. doi: 10.1111/1742-6723.12964. Epub 2018 Mar 23.
Low-acuity 'fast track' patients represent a large portion of Australian EDs' workload and must be managed efficiently to meet the National Emergency Access Target. The current study determined the relative importance and estimated marginal effects of patient and system-related variables in predicting ED fast track patients who stayed longer than 4 h in the ED.
Data for ED presentations between 1 July 2014 and 30 June 2015 were collected from a large regional Australian public hospital. Only 'fast track' patients were included in the analysis. A gradient boosting machine was used to predict which patients would have an ED length of stay greater or less than 4 h. The performance of the final model was tested using a validation data set that was withheld from the initial analysis. A total of 27 variables were analysed.
The model's performance was very good (area under receiver operating characteristic curve 0.89, where 1.0 is perfect prediction). The five most important variables for predicting length of stay were time-dependent and system-related (not patient-related); these were the amount of time taken from when the patient arrived at the ED to: (i) order imaging; (ii) order pathology; (iii) request admission to hospital; (iv) allocate a clinician to care for the patient; and (v) handover a patient between ED clinicians.
We identified the most important variables for predicting length of stay greater than 4 h for fast track patients in our ED. Identifying factors that influence length of stay is a necessary step towards understanding ED patient flow and identifying improvement opportunities.
低急症“快速通道”患者占澳大利亚急诊科工作量的很大一部分，必须进行有效管理以实现国家紧急就诊目标。本研究确定了患者及系统相关变量在预测急诊科快速通道患者留观时间超过4小时方面的相对重要性及估计边际效应。
收集了澳大利亚一家大型地区公立医院2014年7月1日至2015年6月30日期间急诊科就诊的数据。分析仅纳入“快速通道”患者。使用梯度提升机预测哪些患者的急诊科留观时间大于或小于4小时。最终模型的性能使用初始分析时 withheld 的验证数据集进行测试。共分析了27个变量。
该模型的性能非常好（受试者工作特征曲线下面积为0.89，其中1.0为完美预测）。预测留观时间的五个最重要变量与时间和系统相关（而非患者相关）；这些变量是患者到达急诊科后到以下各项所花费的时间：（i）开具影像检查单；（ii）开具病理检查单；（iii）请求住院；（iv）安排临床医生护理患者；（v）在急诊科医生之间交接患者。
我们确定了预测我院急诊科快速通道患者留观时间超过4小时的最重要变量。识别影响留观时间的因素是理解急诊科患者流程并识别改进机会的必要步骤。
