以非線性時間序列方法分析急診類流感就診人次與急診等待住院人數之相關性

Abstract

背景 每年流行性感冒病毒在全球造成至少500萬人次的重症及50萬人死亡，長期以來為人類史中的公共健康安全威脅，每當進入流行季節時，大量類流感患者湧入急診室中，恐造成急診壅塞現象，此外，大量滯留於急診室的類流感病患，恐交叉感染進而引發群聚疫情，嚴重影響急診救護品質。 急診壅塞是當前重要的衛生議題，然而迄今對於急診壅塞尚無統一的定義，且其成因復雜，近期研究指出過多的急診等待住院人數是主因之一，流感疫情與急診壅塞間的關聯性為何尚不清楚。 目的 本研究旨在分析每日類流感就診人次數與每日急診等待住院人數是否存在相關。 材料與方法 研究樣本為18家全國重度及急救責任醫院，2016-2017年之急診室就診資料，其係屬計數時間序列資料，包括：每日類流感就診人次、每日類流感就診人次、每日急診總人次與每日急診等待住院人數，資料分別來自於衛生福利部之疾病流行早期即時監測系統（Real-time Outbreak and Disease Surveillance, RODS) 與全國重度級急救責任醫院公開急診即時資訊。 研究採用描述性、相關性分析來探索每日類流感就診人次、每日類流感就診人次與急診等待住院人數的關係，並利用自迴歸條件均值模型配適不同期間，包括：2016年、2017年以及流感季節與非流感季節，來釐清流感疫情對於每日急診等待住院人數的影響為何。 結果 描述性統計分析結果顯示整體而言當平均每日急診就診總人次增加，平均每日類流感就診人次、平均每日非類流感就診人次與平均每日急診等待住院人數也會遞增，於相關性分析中，類流感就診人次、非類流感就診人次與急診等待住院人數之互相關函數及斯皮曼等級相關係數，隨著滯後天數的增加相關性也跟著增加，多數醫院當滯後天數為1日、2日時相關性為最高且為正相關。 以自迴歸條件均值模型分析，發現所有醫院之一日前急診等待住院人數在不同的研究期間下與平均每日急診等待住院人數呈現正相關且均達到統計上顯著意義，部分醫院之二日前急診等待住院人數亦顯著，但與平均每日急診等待住院人數呈現負相關。當以流感與非流感季節所分析同期類流感就診人次、同期非類流感就診人次、一日前類流感就診人次與二日前類流感就診人次對於急診等待住院人數的影響，高中就診人次的醫院間的結果存在異質性，但於低就診人次的醫院中，則可以見到在2016年流感季期間類流感就診人次對於急診等待住院人數則有顯著的影響。另外，在多數高中就診人次的醫院中前述變項對於急診等待住院人數的影響，在不同年度、季節影響均相同。 結論 在低就診人次醫院中，每日急診等待住院人數在2016流感季有顯著的上升，但在中高急診就診人次的醫院間，前述影響則不明顯。另外，發現到在大多數的醫院裡，部分急診就診病患可能常需於急診室中留觀1-2日以上，因此，醫療院所需建立相關應變、分流機制與感控措施以降低於流感季節時當有過多類流感病患滯留於急診室時，發生流感群聚事件、院內感染的可能性。

Background The influenza has been a global health security threat for long-lasting in human history. It causes over 5 million critical cases and 500 thousand deaths annually worldwide. If a large number of patients with influenza-like illness symptoms remain in emergency department (ED), it may cause cluster infection among patients and seriously affects the quality of emergency medical care. During influenza season, massive influenza-like illness visits surge into ED as well. The relationship between influenza-like illnesses and crowding in emergency department is still unclear. The emergency crowding is an important public health issue. However, it has no uniform definition and its formation is quite complex. Many studies have indicated that one primary cause of crowding is boarding in emergency department. The relationship between influenza-like illnesses and crowding in emergency department is still unclear. Aims This study aims to examine if the daily numbers of influenza-like illnesses in ED are associated with the daily numbers of boarded patients. Material and Methods The time-series count data including the daily numbers of influenza-like illness visits（ILI）, noninfluenza-like illness visits（non-ILI）as well as the daily numbers of visits in the ED , which are retrospectively collected from Real-time Outbreak and Disease Surveillance（RODS）of Taiwan CDC and daily numbers of boarded patients are from open and real time information of 18 intensive emergency care responsibility hospitals between 2016 and 2017. Univariate and bivariate descriptive statistics and correlation analyses are used to explore the relationship between the daily numbers of ILI and non-ILI in ED as well as the daily numbers of boarded patients in ED. The autoregressive conditional mean models are used to fit the time-series count data in different study periods including 2016, 2017, influenza seasons and non-influenza seasons to understand how the influenza epidemic affects the daily numbers of boarded patients in ED. Result Descriptive statistics show that the daily numbers of boarded patients, ILI, non-ILI and in ED tend to increase as the average daily total numbers of patient visits in the emergency department increase. The results of correlation analysis indicate that the strength of correlations between the daily numbers of boarded patients and the daily numbers of influenza-like illness cases as well as the daily numbers of noninfluenza-like illnesses increases as the lags in day decrease. In particular, such correlations of 1-day lag and 2-day lag are highest with statistical significance in most hospitals. For fitting the autoregressive conditional mean models, the results show that the number of boarded patient one day ago is significantly associated with increased average daily numbers of boarded patients during any types of study period in all hospitals, and in some hospitals, the number of boarded patients two days ago is also significantly with reduced association between average daily numbers of boarded patients. In addition, the impacts of the daily numbers of ILI at the same day, one day ago as well as two day ago, and the non-ILI at the same day to the daily numbers of boarded patients during flu and non-flu seasons are quite heterogeneious among hospitals with more than 300 and 200-300 ED visits per day. Nevertheless, in hospitals with less than 200 daily ED visits, the daily number of ILI visits significantly affect the average daily number of boarded patients during 2016 flu season. Furthermore, in most of hospitals with over 200 daily ED visits, their impacts on the boarded patients are the same during flu and non-flu seasons. Conclusion For hospitals with less than 200 daily ED visits, the daily number of boarded patients tended to increase during the 2016 influenza season, but such increase was not apparent for hospitals with over 200 daily ED visits. In most hospitals, it is common that some patients may stay in the emergency room for more than one or two days. Therefore, during the influenza season, it is necessary to formulate contingency plans, diversion mechanisms and infection control in all the hospitals to prevent any possible epidemics as massive patients with influenza-like symptoms remain in ED.