醫療大數據-憂鬱症用藥指標建置及預測台灣菸品銷量

Abstract

目前全球約有3億人罹患憂鬱症，症狀嚴重時會影響患者的家庭生活、工作機能和社交等面向，間接造成生產力降低，最糟糕的情況是會導致自殺，而直接的影響是造成健保費用支出壓力。放眼台灣至少有20萬人罹患重鬱症，以及其他的憂鬱症相關疾病，尤其憂鬱症與抽菸同時出現的問題更是受到關注。 在本研究我們從就診用藥資料庫撈取憂鬱症疾病，利用就診時間、藥量變化區間建立活躍性指標 (CAI)，再分別與台灣健康福利捐和網路聲量進行迴歸分析，找出彼此之間的預測關聯性。資料集合有10,926筆就診用藥資料、80位病患被納入此CAI模型。我們發現CAI模型有60%至70%的解釋力可以預期未來期間的健康捐，作為菸品銷量預測以及60%解釋力預期當期網路聲量。 我們希望本研究作為初探藥品資料庫之應用，以其未來研究可以有更多相關研究在醫療資料庫，估計人類健康狀態。

Currently, there are approximately 300 million people worldwide suffering from depression. When the symptoms are severe, they affect patients’ family life, work performance, social interactions, and other aspects. Depression indirectly reduces patients’ productivity, and in the worst case, it may cause suicide. On the other hand, depression makes a direct impact on increasing health insurance costs. The reason why this study pays attention to depression is that there are at least 200,000 people suffering from major depression, as well as other depression-related diseases in Taiwan. In particular, the issues resulting from both depression and smoking are more concerned. In this study, we extract depression data from medical visits database to establish an activity index (CAI) based on the visit times and the dose change interval. After that, this study conducts regression analysis with Tobacco Health Welfare Surcharge and social media share of voice to find out the relationship between the two factors. The dataset contains 10,926 drug consumption records and 80 patients are included in this CAI model. The result shows that the explanatory power of this CAI model is 60% to 70% for predicting Tobacco Health Welfare Surcharge and the sales of cigarettes; the explanatory power of this CAI model is 60% for predicting the current social media share of voice. We hope that this study may shed some light on using drug database, and look forward to having more relevant researches in the medical database to capture human health status.