採用線上看診之決策樹分類—以台灣為例

Abstract

線上看診透過網路平台使病人可在線上取得醫療看診服務，相較於一般門診的看診方式，線上看診具有更高的效率以及便利性。尤其對於居住在偏鄉地區、行動不便、患有慢性疾病的病人來說，線上看診是更方便的醫療管道。目前已有許多歐美國家開放線上看診等相關醫療服務，而對於尚未開放線上看診的國家，應在開放線上看診或擬定相關政策前，對於開放線上看診所帶來的影響進行全面的預估以及規劃。 本研究提出一以決策樹為基礎的線上看診分類方法分類病人會使用線上看診的情況，我們藉由現有的線上看診研究所提供會使用線上看診的病人特徵，以及病人的看診記錄進行分類。透過決策樹模型得到一組線上看診的分類規則。 本研究以尚未開放線上看診的台灣為例，利用台灣健保資料庫中2012年的資料進行分析，透過線上看診的決策樹分類模型得到一組線上看診的分類規則，了解病人會使用線上看診的情況，並可預估可能使用線上看診的人數比例。 本研究提出的方法利用已開放線上看診國家的研究結果分類看診記錄，並提供一組清楚的線上看診分類規則，可提供尚未開放線上看診的國家進行開放前的評估與計劃。而開放線上看診後，亦可透過此方法更新分類規則，因應病人行為模式的改變，以確保醫療服務的穩定。此方法的概念不只可應用在線上看診的預估，亦可應用在其他新服務或新產品的推行規劃。

E-Visit is the consultation service delivering the health care online, which is more efficient and effective than in-office visit, especially for patients living in rural area, disabled, or with chronic diseases. Some countries have already implemented e-visits. For the countries have not yet implemented e-visits, the governments have to make a comprehensive preparation and predict the effect of implementing e-visits. This study proposes a decision-tree based e-visit classification approach (DTEVCA) to determine clinic visits qualified as e-visits using the clinics’ medical records and patients’ demographic data. This study assumes that health care insurance (i.e., national health insurance) will subsidize the e-visit service cost, in which case it is essential to identify patients who will benefit most from e-visit service. Using a large data set from Taiwan’s National Health Insurance, this study verifies the efficiency and validity of the DTEVCA. The results indicate that this approach can accurately classify clinic in-office visits that could switch to e-visit service. The straightforward rules of this decision tree also give insurance agencies a clear guideline to understand the circumstance of using e-visits and predict the effect of implementing e-visits in Taiwan. The result of this study can help the countries that have not yet implemented e-visits to improve the policy formulation process or academic researches. The DTEVCA can update the classification rules using new data to correct the biases and ensure the stability of the e-visit system. In addition, the concept of this approach is feasible not only for e-visit service but also for other “new services” such as new products or new policies.