台灣人口老化，加上醫療技術發展迅速，使得有復健需求的民眾日益增多，復健治療為長期治療，因此累積醫療費用愈來愈高。而全民健保為使醫療供給機構提高成本意識、控制醫療費用，已全面實施總額支付制度，年度預先協商各部門總額預算。本研究藉由機器學習預測、分類復健部病人的花費，以探討在總額支付制度下的成本控管並提升醫療品質。本研究將蒐集台灣某區域醫院復健部病患的人口統計學特性、就醫情形、疾病診斷碼與醫療費用等歷史資料，藉由機器學習的決策樹，採監督式學習，將資料的一部分作為訓練資料建立數學模型，其餘資料作為模型驗證，找出醫療費用分組較高的疾病分類，並預測病人醫療費用。經由機器學習之決策樹技術，能更了解復健部各類疾病的醫療費用情形，並能成功預測復健病人的醫療費用，以提供醫院管理人員預估病人費用消耗情況，對於費用較高的疾病分類分析並解釋可能原因，促使醫療機構改善醫療照護模式，加強成本控制與品質提升，並將機器學習方法應用在復健醫療管理，預測醫療趨勢，作為復健部未來發展策略與目標設定的依據。

Rehabilitation needs are increasing rapidly due to the aging population and the advance of medical technology. Since rehabilitation is a long term, on-going process, it leads to high medical expenses. Therefore, the National Health Insurance in Taiwan has implemented the global budget payment system to manage medical costs, to enhance efficiency and to improve the quality of healthcare. To propose machine learning algorithms to identify patients who need more medical expenditure and to predict potential medical expenses in the rehabilitation department, data of the patients in the rehabilitation department of a regional hospital in Taiwan will be collected from the electronic medical records, starting from 2016 till 2019, which will include several major categories, such as demographics, chronic conditions, major diagnoses, and medical spending variables. We use supervised learning to build a mathematical model and split data into a training set (outpatient observations: 4901 observations, inpatient observations: 8230 observations) for model construction and a test set (outpatient observations: 1930 observations, inpatient observations: 981 observations) for out-of-sample evaluation. Through this study, we can discover the categories of diseases which cost more to treat in rehabilitation patients, and make a precise prediction of the medical cost of the rehabilitation patients.

1.前言…………………………………………………………………1
2.文獻回顧………………………………………………………3
3.研究方法………………………………………………………5
4.實證結果………………………………………………………8
5.結論………………………………………………………………22
參考文獻…………………………………………………………26
附錄……………………………………………………………………28

1.Jesus, T. S., Landry, M. D., & Hoenig, H. (2019). Global need for physical rehabilitation: systematic analysis from the Global Burden of Disease Study 2017. International Journal of Environmental Research and Public Health, 16(6), 980-999.
2.Kan, H. J., Kharrazi, H., Chang, H.-Y., Bodycombe, D., Lemke, K., & Weiner, J. P. (2019). Exploring the use of machine learning for risk adjustment: A comparison of standard and penalized linear regression models in predicting health care costs in older adults. PLoS One, 14(3), e0213258.
3.Relles, D., Ridgeway, G., & Carter, G. (2002). Data mining and the implementation of a prospective payment system for inpatient rehabilitation. Health Services and Outcomes Research Methodology, 3(3-4), 247-266.
4.Rose, S. (2016). A machine learning framework for plan payment risk adjustment. Health Services Research, 51(6), 2358-2374.
5.Rose, S. (2018). Robust machine learning variable importance analyses of medical conditions for health care spending. Health Services Research, 53(5), 3836-3854.
6.Shrestha, A., Bergquist, S., Montz, E., & Rose, S. (2018). Mental health risk adjustment with clinical categories and machine learning. Health Services Research, 53, 3189-3206.
7.Yan J, Linn K.A., Powers B.W., et al. (2019). Applying machine learning algorithms to segment high-cost patient populations. Journal of General Internal Medicine, 34(2), 211-217.
8.李玉春, 黃昱瞳, 黃光華, 葉玲玲, 與陳珮青 (2014)，「全民健保支付制度改革之回顧與展望」《台灣醫學》，18(1)，53-66。
9.周歆凱, 蘇喜, 黃興進, 蔡明足, 與翁林仲 (2006)，「 運用決策樹技術探討急診病患醫療費用之消耗」，《台灣公共衛生雜誌》，25(6)，430-439。
10.謝如蘭, 連倚南, 謝霖芬等人 (1996)，「國人接受復健醫療之疾病分類研究：以北部, 東部八家醫院為例」，《中華民國復健醫學會雜誌》，24(1)，35-40。
11.蘇宗柏, 陳思遠, 王亭貴, 王顏和, 與連倚南 (2010)，「復健醫療服務之疾病分類研究：國內某醫學中心近期經驗」，《台灣復健醫學雜誌》，38(4)，229-236。