以卜瓦松混合模型探討醫院用藥情形：以臺北市立聯合醫院為例

Abstract

如今已有大量的資料存在於資料庫中，各行各業皆可善加利用這些數據、找出具重要意義的部份並加以分析，以期促成優良的決策，不論是對自身的了解或是未來的方向都能有相當的助益。而在醫院中，因為藥品具時效性，若醫院一次購入相較於使用量更多的藥物，除了當期成本增加外，期限內無法處置妥當可能還會造成諸多醫院內部問題；反之，若是備藥不及使用量的情況下，不僅影響病患的治療，也會波及醫院的信譽。 本研究以醫院作為藥物供給者的角度，探討給予病患的藥物使用量是否會有固定的模式，針對其中變數和藥物的支出數量進行建模。考量到醫院的異質性，故採用階層貝氏模型，結合兩個成分的混合模型，定義出其機率較小的部份為非規律的用藥行為、機率較大的部分為規律的用藥行為，希望能藉此判斷出藥品的使用情形。若當月使用量為不規律，又比起規律的使用量偏少，則可能會造成後續囤貨的趨勢產生；反之，若使用量偏多，則可能造成備藥不足。 筆者認為使用量對於購買量及庫存量間有一定關聯，因此將藥品的購入數量和結存數量應用至此模型中，以進一步觀察三者間的變化，若是醫院能考量這些藥品紀錄背後的行為模式，可擬定出對醫院整體更合適、有效率的用藥策略。

Nowadays, a large amount of data stores in the database. All companies can make good use of the data to promote the good decision-making. If the hospital purchases more drugs than the ones used, it may increase the current cost and make much waste because the drugs are time-sensitive. If the medicine is out of stock, it will not only affect the treatment of the patient, but also affect the reputation of the hospital. This study takes hospitals as drug suppliers to explore whether there is a fixed pattern of medication consumption. Considering the heterogeneity of the hospital, we use the hierarchical Bayesian model and combine two components to a mixture model. This can be used to determine the behavior of the medicine consumption. If the usage of a month is irregular and less than the regular part, it may cause the trend of stockpiling. On the contrary, if the irregular usage is too large, it may cause insufficient preparation. The author believes that there is a certain correlation between the amount of purchase and the amount of stock. Therefore, the purchase quantity and balance of the medicines are applied to this model to observe the changes between the three. If the hospital can consider these behavior patterns, it may draw up more appropriate and efficient medication strategy for the hospital.