近年來慢性疾病在十大死亡原因中就占了七個排名，而慢性病最常見的治療方式就是服用處方藥，為了監控以及簡化領藥程序，醫師會開立連續處方箋 (refillable prescriptions)方便作業，連續處方箋通常為醫生開給慢性疾病 (如高血壓)或是經醫生評估病患可以穩定長期服用相同藥物的病患，由於此類型的病患會在一段時間後回藥局續領藥(refill)，其需求會呈現一定的模式。本研究將病患分成第一次領藥和使用連續處方箋第二次、第三次之後續領藥(refill)的病患，第一次領藥的需求預測由灰色預測和時間序列混合而成預測模型預測，而藥品需求若呈現隨機性或是季節性，也會使用不同的預測方法；對於第二次之後領藥的需求預測，利用醫生開立的連續處方箋的歷史資料，針對病人回藥局領藥的時間以及病人是否會回藥局領藥的機率做分析，將這些病患領藥行為加入預測模型中，得到預測需求量。最後將兩個部分的預測值相加為最終的預測需求。接著使用兩個實際連鎖藥局的資料集，和原始模型比較進行驗證，實驗結果發現，在總共2968種藥品中，有約67%的藥的預測準確度會比原使模型還要準確。另外，本研究試圖從藥品特性、病患行為等方面找尋因子，透過這些因子找出哪種藥品較適用本研究所提出的預測模型，結果發現當準點率高時，大部分的藥品會適用本研究的預測模型。

In recent years, chronic diseases have accounted for seven rankings among the top ten causes of death, and the most common treatment for chronic diseases is using prescription drugs. In order to monitor and simplify the process, the physician will open a refillable prescriptions. Refillable prescription is usually given to patients with chronic diseases (such as hypertension) or doctors believe that patients can take the same drugs for a period of time, and the demand will be a certain pattern. In this paper, patients will be divided into two types, fill the drugs first time and use refillable prescription to refill drugs. The demand of the first type was predicted by gray forecasting method and time series, drugs with different demand pattern like randomness or seasonal will use different forecast method. For the second type of patient, this paper forecast the demand by analyzing patient’s behavior in refillable prescription of historical data. Finally, two parts of the forecast value added to the final forecast demand. Then, using two data sets from the actual pharmacy chain stores, and original model will be compared to proposed model. The experimental results found that a total of 2968 kinds of drugs, about 67% of the drug will be more accurate than the original forecasting model. In addition, this study attempts to find factors from the aspects of drug characteristics, patient behavior, etc. Through these factors to find out which drugs are more suitable for the proposed model. The results found that when the on time rate is high, most of the drugs will be suitable for the proposed model.

目錄
摘要 I
ABSTRACT II
目錄 III
圖目錄 V
表目錄 VI

第一章 緒論 1
1.1 研究背景 1
1.2 研究動機 6
1.3 研究目的 8
1.4 研究架構 9

第二章 文獻探討 10
2.1 藥物順從性 10
2.2 顧客行為之應用 14
2.3 灰色預測法與時間序列法之概念 16
2.3.1灰色預測法 16
2.3.2時間序列法 17

第三章 研究方法與預測系統 20
3.1 病患行為加入需求預測系統主要架構 20
3.2 資料前處理和適用藥品特性 21
3.2.1資料前處理 21
3.2.2藥品特性 22
3.3 單期預測模型 25
3.4 考慮病患行為加入預測模型之單期預測 27
3.4.1機率矩陣 27
3.4.2病患領藥時間狀態 30
3.4.3預測模型步驟 31

第四章 實例驗證與結果分析 36
4.1 資料描述 36
4.2 資料敘述與參數設定 37
4.2.1資料敘述 37
4.2.2參數設定 39
4.3 實例驗證與模型比較 43
4.4 透過藥品特性分類 45

第五章 結論 47
5.1 結論與建議 47
5.2 未來研究方向 48

參考文獻 50

參考文獻
[1]Balkrishnan, R. (1998). Predictors of medication adherence in the elderly. Clinical therapeutics, 20(4), 764-771.
[2]Benner, J. S., Glynn, R. J., Mogun, H., Neumann, P. J., Weinstein, M. C., & Avorn, J. (2002). Long-term persistence in use of statin therapy in elderly patients. Jama, 288(4), 455-461.
[3]Claxton, A. J., Cramer, J., & Pierce, C. (2001). A systematic review of the associations between dose regimens and medication compliance. Clinical therapeutics, 23(8), 1296-1310.
[4]Desselle, S., Zgarrick, D., & Alston, G. (2016). Pharmacy Management: Essentials for all practice settings: McGraw Hill Professional.
[5]Foxall, G. (2005). Understanding consumer choice: Springer.
[6]Golin, C. E., Liu, H., Hays, R. D., Miller, L. G., Beck, C. K., Ickovics, J., . . . Wenger, N. S. (2002). A prospective study of predictors of adherence to combination antiretroviral medication. Journal of general internal medicine, 17(10), 756-765.
[7]Haynes, R. B., McDonald, H. P., & Garg, A. X. (2002). Helping patients follow prescribed treatment: clinical applications. Jama, 288(22), 2880-2883.
[8]Ho, P. M., Bryson, C. L., & Rumsfeld, J. S. (2009). Medication adherence: its importance in cardiovascular outcomes. Circulation, 119(23), 3028-3035. doi:10.1161/CIRCULATIONAHA.108.768986
[9]Kuester, S. (2012). MKT 301: Strategic marketing & marketing in specific industry contexts. University of Mannheim, 110, 393-404.
[10]LaRosa, J. C. (2000). Poor compliance: the hidden risk factor. Current Atherosclerosis Reports, 2(1), 1-4.
[11]McCaffrey, D., Smith, M., Banahan, B., Frate, D., & Gilbert, F. (1998). A continued look into the financial implications of initial noncompliance in community pharmacies: an unclaimed prescription audit pilot. JOURNAL OF RESEARCH IN PHARMACEUTICAL ECONOMICS, 9, 33-58.
[12]Murphy, D. A., Sarr, M., Durako, S. J., Moscicki, A.-B., Wilson, C. M., & Muenz, L. R. (2003). Barriers to HAART adherence among human immunodeficiency virus–infected adolescents. Archives of pediatrics & adolescent medicine, 157(3), 249-255.
[13]Murray, P. W., Agard, B., & Barajas, M. A. (2015). Forecasting supply chain demand by clustering customers. IFAC-PapersOnLine, 48(3), 1834-1839.
[14]Nikolopoulos, K., Buxton, S., Khammash, M., & Stern, P. (2016). Forecasting branded and generic pharmaceuticals. International Journal of Forecasting, 32(2), 344-357. doi:10.1016/j.ijforecast.2015.08.001
[15]Osterberg, L., & Blaschke, T. (2005). Adherence to medication. New England Journal of Medicine, 353(5), 487-497.
[16]Quilumba, F. L., Lee, W.-J., Huang, H., Wang, D. Y., & Szabados, R. L. (2015). Using smart meter data to improve the accuracy of intraday load forecasting considering customer behavior similarities. IEEE Transactions on Smart Grid, 6(2), 911-918.
[17]Senst, B. L., Achusim, L. E., Genest, R. P., Cosentino, L. A., Ford, C. C., Little, J. A., . . . Bates, D. W. (2001). Practical approach to determining costs and frequency of adverse drug events in a health care network. American Journal of Health-System Pharmacy, 58(12), 1126-1132.
[18]Steiner, J. F., & Prochazka, A. V. (1997). The assessment of refill compliance using pharmacy records: methods, validity, and applications. Journal of clinical epidemiology, 50(1), 105-116.
[19]Wilson, J., Axelsen, K., & Tang, S. (2005). Medicaid prescription drug access restrictions: exploring the effect on patient persistence with hypertension medications. The American journal of managed care, 11, SP27-34.
[20]Wu, C.-H. (2009). Behavior-based spam detection using a hybrid method of rule-based techniques and neural networks. Expert Systems with Applications, 36(3), 4321-4330.
[21]江堉旻. (2015). 應用灰色預測法與時間序列法於連鎖藥局需求預測系統之研究. 國立台灣大學工業工程所碩士論文.
[22]洪新原. (2009). 慢性病連續處方箋之成效評估-以資料探勘技術探討未開立慢性病連續處方之決定因子. 國立中正大學資訊管理研究所碩士論文
[23]劉妙珍. (2014). 影響高血壓慢箋續領之因素探討. 東海大學工業工程與經營資訊學系高階醫務工程與管理碩士在職專班碩士論文