檢傷時急診滯留時間風險因子

Abstract

研究背景 急診病患在急診的滯留時間與病人就診量、急診病人住院率、醫院佔床率等原因相關。急診滯留時間長短是急診的品質指標，過長的急診待床時間會增加病人住院中不良反應(adverse effect)、病人死亡率和總住院長度。過去研究對於急診滯留時間的分析，一般會區分住院病人和出院病人，去分析急診滯留時間增加的危險因子，但是這樣的研究方式無法套用在新到診，尚不知道是否要住院的病人身上，因此目前缺乏剛到診急診病患的急診滯留時間風險因子分析研究。 研究目的 本研究回溯性分析病患到急診時，不同病患基本資料、檢傷嚴重度和急診室運作結果下，急診滯留小時數目增減的因子為何和相對風險(relative risk, RR)的大小。 研究方法 本研究使用卜瓦松迴歸(Poisson regression model, PRM)和負二項迴歸(negative binomial regression model, NBRM)模型， 運用三種變數選擇的策略，包括向前選擇模型(forward selection model)、後向剃除模型(backward selection model)和群組式的最小絕對緊縮與選擇算子(group least absolute shrinkage and selection operator, gLASSO)的方式，選擇出赤池訊息準則(Akaike information criterion, AIC)最低的變數組合。 研究結果 A醫學中心的急診滯留時間呈現右偏分布，與其他醫院急診滯留時間相同。本研究發現在同樣變數選擇組合下，負二項迴歸的赤池訊息準則皆較卜瓦松迴歸為低，與其他滯留時間相關研究結果相同。病人在急診滯留的時間長短，在病人特質層面部分，年紀越大急診病人滯留小時數目越長；在臨床表現層面部分，檢傷一二級病人、內科、或救護車來診的病人，會相較其他病人有較高的急診滯留小時數目；在組織行政層面部分，一般病房占床率越高、病患之急診等候分鐘數越長，到診時急診病人數量較多，急診病人滯留小時數目越長。 結論與討論 本研究建議未來研究使用負二項迴歸來分析急診滯留時間。本研究的設計為病人層級的急診滯留小時數目研究，其統計方法假設符合急診滯留小時數目的分布。本研究的急診室是一個幾乎不拒絕救護車、無轉院、無未看診就離開病人的急診室，因此本研究找出的危險因子可視為可能影響急診滯留小時數目的原因。

Backgrounds Length of stay (LOS) in the emergency department (ED) is related to the number of in-ED patients amounts, the hospitalization rate of emergency patients, and the hospital bed occupancy rate. LOS in ED is one of the quality indexes of the emergency department. Increased boarding time in ED will increase the adverse effects, the mortality rate, and the total length of hospitalization of patients admitted from the ED. In the past, studies on the analysis of emergency LOS generally distinguished between inpatients and discharged patients to analyze the risk factors for increased LOS, but such research methods cannot be applied to new patients who do not know whether they will be hospitalized or not. Therefore, there is currently a lack of research on the risk factors of emergency stay time among patients who have just arrived in the ED. Purpose This study retrospectively analyzed the factors of the increase and decrease in the number of hours spent in the ED and the relative risks (RR) under different demographical, clinical, and organizational characteristics when patients came to the ED. Methods In this study, Poisson regression (PRM) and negative binomial regression (NBRM) models were used. Three variable selection strategies were used, including the forward selection model, backward selection model, and group least absolute shrinkage and selection operator (gLASSO), to select the variable combination with the lowest Akaike information criterion (AIC). Results ED LOS in medical center A has a right-tailed distribution, which is the same as the ED LOS in other hospitals. This study found that under the same variable selection combination, the AIC of negative binomial regression was lower than that of Poisson regression, which suggests an overdispersion of the ED LOS data. In terms of demographical characteristics, the older the emergency patient is, the longer the number of hours spent in the emergency department. In terms of clinical characteristics, patients with first- and second-level triage, non-traumatic, or ambulance-delivered will have a higher number of hours spent in emergency departments than other patients. In terms of organizational characteristics, the higher the bed occupancy rates, the longer the waiting minutes, the greater in-ED patient amounts when they arrive, and the number of emergency patient stay hours longer. Discussions and Conclusions This study recommends future studies use NBRM to analyze ED LOS. The design of this study is a patient-level study of the number of hours spent in emergency departments, and its statistical method is consistent with the distribution of LOS in ED. The emergency room of this study is an emergency room that almost never refuses ambulances, transfers, and leaves patients without seeing a doctor. Therefore, the influence and directionality of each risk factor mentioned in this study on ED LOS can be considered as a possible cause of prolonged ED LOS.