在台灣平均每年約有3萬多人死於癌症,其中約80%癌症末期病患未能在安寧緩和醫療的照顧下安然走完人生。有鑑於此,2005年行政院衛生署國民健康局開始推動安寧緩和醫療照護團隊與原診療團隊之安寧共同照護模式,收案條件是病人有身心靈上需求,需要安寧共同照護小組介入。癌症末期病患接受安寧共同照護服務益處有:提昇病人及家屬生活品質、提昇醫護團隊滿意度及減少醫療費用支出。因此,本研究利用傳統統計工具分析影響存活天數之預後因子有:性別、身體功能狀態、便秘、呼吸困難、意識不清、虛弱疲倦、腹水、水腫、暈眩、吞嚥困難、胺基丙胺酸轉移酶、尿素氮等症狀。並運用資料探勘中的決策樹、類神經網路、邏輯斯迴歸進行分析,研究結果發現,決策樹建立的規則較邏輯斯迴歸優,而類神經網路較決策樹優,就實務上而言,決策樹的樹狀圖型較能解釋及判讀。本研究之結果期望提供醫療人員另一種預測癌症末期病患之存活天數之參考規則。
In Taiwan, an annual average of about 30,000 people dies of cancer, of which approximately 80% of terminally ill cancer patients never receive palliative care for the rest of their days. Regarding to this, in 2005 the Bureau of Health Promotion, Department of Health began promoting palliative hospice care model which involves palliative care team and original clinic team to cooperate. By receiving palliative care service, terminally ill cancer patients obtain the following benefits: enhanced quality of life of patients and their families, improved satisfaction of the health care team, and reduced medical expenses. In this study, we first analyzed the data with statistical tools and identified the following prognostic factors of survival days: Gender, Physical functional status, Constipation, Dyspnoea, Conscious disturbance, Weakness, Ascites, Edema, Dizziness, Dysphagic, Alanine aminotransferase, and Blood urea nitrogen. Then, we developed the survival prediction models with a statistical tool, logistic regression analysis, and two data mining techniques, namely decision tree and neural networks, respectively, and compare the performance of the three models in terms of prediction accuracy. We found that the decision tree model performed better than the logistic regression model did. The neural networks model performed better than the decision tree model did, however, in practice, the rules established by the decision tree model were easier to comprehend. This study provides medical personnel a useful reference for predicting survival days of terminally ill cancer patients.