運用回診日期調整病人回診預測模型-以糖尿病病人為例

Abstract

糖尿病是一種葡萄糖代謝異常所引起的一種疾病，具備高盛行率、高併發症， 但是是一種能夠治療並且提前預防的一種疾病，其主要特徵會造成血液中的葡萄糖濃度升高。血糖升高後容易具備急性期併發症的可能性，因此對於糖尿病患而言，如何有效控制住血糖對於糖尿病患者來說是達到預防慢性併發症發生的主要方式。 近年來，隨著人口老化、不健康飲食、肥胖與久坐等生活型態，全球的糖尿病患人數急遽上升，估計 2025 年糖尿病患人數將達到 3 億，用藥需求也會隨之增加，因此對於糖尿患用藥需求的預測與分析將更加重要。 隨著大數據的興起，各大產業以及企業也開始重視資料庫行銷的應用，醫院診所對於龐大的就診資訊越來越重視，且在病患人數與就診人口的快速成長下，病人的回診行為以及藥物的存貨控管將日益重要，為了能有效協助醫院診所有效的掌控病人的回診行為，並控管每期的進藥物量，需要更精準的預測模型來掌握病患的就診行為，如此一來將能更有效率的應對用藥需求的不確定性，從中達到存貨成本的降低，並且也可以避免藥物堆積過多造成藥物放置過久而過期、變質等等的風險。 本研究採用 2015-2017 年診所就診之資料庫，針對回診的病人進行研究，期望能建立糖尿病病患之回診行為預測模型，並加入星期幾這個要素來加以調整。本研究所使用的模型將會先以層級貝氏統計分析為基礎，估計參數之後驗機率分配來建立回診間隔的預測模型，再根據回診日期為星期幾來調整上述模型，期望能提供診所更準確的用藥需求預測方式，掌握其病患之回診行為，更精準的去執行藥物存貨管理的策略。

Diabetes is a disease caused by abnormal glucose metabolism. It has a high prevalence and high complications, but it is a disease that can be treated and prevented in advance. Its main characteristics will cause the glucose concentration in the blood to rise. After increasing of blood glucose, the possibility of acute complications would increase. Therefore, for patients with diabetes, how to effectively control blood glucose is the main way to prevent chronic complications. In recent years, with the aging population, unhealthy diet, obesity, and bad lifestyles, the number of people with diabetes has risen sharply in the world. It is estimated that the number of people with diabetes will reach 300 million by 2025, and the demand for medication will increase accordingly. How to exactly forecast the demand of medication will be more and more important. With the development of big data and the rapid growth of the population, lots of enterprises started to apply database marketing. Hospitals and clinics also pay more attention to the patient data, the patient's returning behavior and the inventory of drugs are also becoming more important. In order to assist hospitals and clinics to control the patient's return visits and the amount of drug taken in each period, we need more a more accurate forecast model to understand patient behavior. Therefore, it will be able to decrease the uncertainty of drug demand, decrease the inventory cost, avoid excessive drug accumulation, and drug expiration. This research is based on a medical institution’s database of patient consultation from six municipalities from 2015 to 2017, and we used it to build a model to forecast patient’s return visit, and then adjusted model by the day of the week. This model is based on hierarchical Bayesian statistic. We tried to estimate the parameters of this model and then adjusted the parameters by the day of the week. Through this process, we will be able to build a more accurate forecast model to understand patient’s return visit, so the hospitals and clinics can control their inventory more efficiently.