資訊管理協助醫院空床之圖示分析

Visualization Analysis of Hospital Vacant Beds through Computer Science

傳統上以佔床率最低的病床或護理病房為單位，進行床位的分析與檢討。即便是利用單因子變異數或是多元迴歸分析，也都是以平均佔床率的數字或表格予以呈現。當大量的資訊容易造成使用者迷失之際，醫院新制評鑑又強調PDCA品管循環的觀念與落實。本研究期以視覺化的知覺圖，闡述統計分析的結果，進而利用資訊模組篩檢出「應佔床卻空床」的床位管理與分析。本研究利用電腦資訊系統分析某個案醫院某月883張一般病床，Rasch模式的多相FACETS軟體總製出週內日期別、週別及病房別等因子影響空床機率的等距量尺分析圖，也利用多元量尺法圖示影響空床機率的各獨立變數間之關聯，最後以標準化殘差分析「應佔床卻空床」的篩檢與分析。研究發現週別及週內日期別皆會影響到空床的發生機率。等距量尺分析圖及多元量尺知覺圖可以清楚解析出影響空床的重要因子，標準化殘差分析出最需優先列入檢討的754(2.8％)病床日，此數量遠低於傳統分析的2,303總空床日，因此Z分數空床分析圖可聚焦進行空床的根本原因分析。

Hospital accreditation calls for much attention upon PDCA of Deming cycle. Traditional analysis of bed occupation focuses on locating those beds with the highest vacancy rate. In this study we use Rasch technique to analyze the data of 883 hospital beds in 28 days to quantify the bed vacancy prevalence of each bed and the threshold of each day. The multidimensional scaling analysis plotting perceptual maps help interpret the relations of key factors on bed vacancy. We found that different weeks showed different magnitudes of vacancy prevalence; different dates in a week also showed and influenced magnitudes of threshold. An analysis of the standardized residuals identified 754 most unexpected vacant daily beds (2.8%), which need further investigation with top priority. This number was much smaller than that identified from traditional analysis of 2,303 vacant beds. Hence, it was easier to trace causes of these vacant beds. Through an empirical example, Rasch-specific many facets model and Z-score technique incorporating with multidimensional scaling analysis demonstrated its application on Deming's quality control cycle, which is emphasized by recent hospital accreditation in Taiwan.