資訊管理學報
蔡佳玲;洪新原;袁繼銓;
頁: 139-164
日期: 2011/10
摘要: 中央健康保險局為方便病情穩定之慢性病患者就醫，因而推廣慢性病連續處方箋制度，以提供病患做週期性取藥，藉以降低平均門診次數，並且避免非必要之醫療支出。本研究以健保局資料庫為研究材枓，探討2007年全國醫院層級之門診慢性病案件，符合慢性病連續處方箋開立條件之開立與未開立慢性連續處方箋之記錄，運用C 5.0決策樹演算法之分類功能，將慢性病連續處方箋開立與否之影響因子生成決策樹模型與規則集。本研究以「是否開立慢性病連續處方箋」為分類欄位，來探討醫院權屬別、醫院層級別、醫事機構區域別、醫師性別、醫師年齡、科別、醫師平均每日看診量、患者性別、患者年齡、慢性病疾病範圍等十項因子之區別能力，期能找出慢性病連續處方箋開立與否之影響因素。研究結果顯示決策樹模型整體正確率達79.31%，在開立與未開立慢性病連續處方規則集中，有六條符合開立條件而未開立慢性病連續處方箋之描述規則。分析結果發現：首先，有相關影響性之因子共八項（除了患者年齡與患者性別）。其次，有二十一條符合開立條件而選擇開立慢性病連續處方箋之描述規則，有相關影響性之因子共九項（除了患者性別）。最後，在醫院權屬別、醫院層級別、就醫科別等十項因子中，就醫科別與醫院層級別二項因子，對開立與未開立慢性連續處方箋之規則描述，最具有決定性影響。
關鍵字: 慢性病連續處方箋;影響因子;資料探勘;決策樹;
Abstract: In order to make things easy for patients with stable chronic diseases, Bureau of National Health Insurance (BNHI) popularizes the refilled chronic disease prescriptions (RCDP). Thus, the patients can get the medicine periodically to decrease the average number of times of outpatient services and to avoid unnecessary medical expenses. This study used the database in Bureau of National Health Insurance to investigate the outpatient cases of chronic illness in 2007 which conform to the conditions of refilled chronic disease prescriptions but refilled prescriptions are not prescribed in various levels of hospitals in our county. The C 5.0 decision tree algorithm was taken to generate the decision tree model and rules. ”Prescribing refilled prescriptions or not” is the classification outcome and ten factors including hospital ownership, level of hospital, region of medical institute, gender of doctor, age of doctor, division of medical care, average outpatient service per day, gender of patient, age of patient and chronic illness scope are used to predict ”prescribing or not prescribing refilled prescriptions”.The findings show that the overall accuracy of the Decision Tree Model reaches 79.31%. Six useful rules were found. Eight factors including hospital ownership, level of hospital, region of medical institute, division of medical care, average outpatient service per day, gender of patient, age of patient and chronic illness scope, were also identified as influencing factors. Finally, implications from the findings are also provided.
Keywords: Refilled Chronic Disease Prescriptions;Influencing Factors;Data Mining;Decision Tree;
瀏覽次數: 8830 下載次數: 137
引用 導入Endnote
頁: 139-164
日期: 2011/10
摘要: 中央健康保險局為方便病情穩定之慢性病患者就醫，因而推廣慢性病連續處方箋制度，以提供病患做週期性取藥，藉以降低平均門診次數，並且避免非必要之醫療支出。本研究以健保局資料庫為研究材枓，探討2007年全國醫院層級之門診慢性病案件，符合慢性病連續處方箋開立條件之開立與未開立慢性連續處方箋之記錄，運用C 5.0決策樹演算法之分類功能，將慢性病連續處方箋開立與否之影響因子生成決策樹模型與規則集。本研究以「是否開立慢性病連續處方箋」為分類欄位，來探討醫院權屬別、醫院層級別、醫事機構區域別、醫師性別、醫師年齡、科別、醫師平均每日看診量、患者性別、患者年齡、慢性病疾病範圍等十項因子之區別能力，期能找出慢性病連續處方箋開立與否之影響因素。研究結果顯示決策樹模型整體正確率達79.31%，在開立與未開立慢性病連續處方規則集中，有六條符合開立條件而未開立慢性病連續處方箋之描述規則。分析結果發現：首先，有相關影響性之因子共八項（除了患者年齡與患者性別）。其次，有二十一條符合開立條件而選擇開立慢性病連續處方箋之描述規則，有相關影響性之因子共九項（除了患者性別）。最後，在醫院權屬別、醫院層級別、就醫科別等十項因子中，就醫科別與醫院層級別二項因子，對開立與未開立慢性連續處方箋之規則描述，最具有決定性影響。
關鍵字: 慢性病連續處方箋;影響因子;資料探勘;決策樹;
Applying Data Mining Technique to Identify the Influencing Factors of not Prescribing the Refilled Chronic Disease Prescriptions
Abstract: In order to make things easy for patients with stable chronic diseases, Bureau of National Health Insurance (BNHI) popularizes the refilled chronic disease prescriptions (RCDP). Thus, the patients can get the medicine periodically to decrease the average number of times of outpatient services and to avoid unnecessary medical expenses. This study used the database in Bureau of National Health Insurance to investigate the outpatient cases of chronic illness in 2007 which conform to the conditions of refilled chronic disease prescriptions but refilled prescriptions are not prescribed in various levels of hospitals in our county. The C 5.0 decision tree algorithm was taken to generate the decision tree model and rules. ”Prescribing refilled prescriptions or not” is the classification outcome and ten factors including hospital ownership, level of hospital, region of medical institute, gender of doctor, age of doctor, division of medical care, average outpatient service per day, gender of patient, age of patient and chronic illness scope are used to predict ”prescribing or not prescribing refilled prescriptions”.The findings show that the overall accuracy of the Decision Tree Model reaches 79.31%. Six useful rules were found. Eight factors including hospital ownership, level of hospital, region of medical institute, division of medical care, average outpatient service per day, gender of patient, age of patient and chronic illness scope, were also identified as influencing factors. Finally, implications from the findings are also provided.
Keywords: Refilled Chronic Disease Prescriptions;Influencing Factors;Data Mining;Decision Tree;
瀏覽次數: 8830 下載次數: 137
引用 導入Endnote