應用專家因素檢定及決策樹分析建構困難插管之風險定性評估

Risk Analysis of Pre-operative Endotracheal Intubation With the Applications of Expert Opinion and Decision Tree

簡梓涵(TzuHan-Chien)

指導教授 ： 王献章

長榮大學/資訊暨工程學院/資訊管理學系(所)/碩士(2010年)

https://doi.org/10.6833/CJCU.2010.00039

摘要

困難插管是在手術麻醉時，麻醉醫師常遇到的問題。導致困難插管的因素有許多，實務上可由術前評估來事先瞭解導致病患困難插管的相關因子。
本研究彙整國內外的相關文獻之氣管插管危險因子共24個，接著與醫院醫師和護理人員共同討論並篩選出可能導致困難插管之危險因子有12個，再運用相關係數比較之後剩9個因子，經由專家訪談後，將危險因子個數濃縮為6個，最後運用決策樹演算法建置困難插管的預測模型。
研究對象為合作對象醫院接受全身麻醉進行插管並使用標準插管工具之51位病患（男性27人，女性24人）。研究對象的基本資料是透過本研究開發的病患術前資料收集系統取得，再用來測試本研究所提之困難插管模型之正確率。
本研究之決策樹模型共有四大類：C5.0決策樹的男女性病患之正確率分別為83%和50%；依兩位專家意見所建立的直覺決策樹之正確率分別為44.44%和25%；因素分析決策樹上，利用因素分群後以分群為節點進行判斷，其正確率在56%和71%；機會補償決策樹則是利用因素分群後的所有分群建構成二大棵決策樹，所有分群在第一棵決策樹中沒通過的，將再給予機會由第二棵決策樹上做判斷，其正確率在63%和75%。
由實驗數據顯示，雖然C5.0決策樹的正確率結果為最高，但因其決策的依據（數據值）不具臨床醫學上的意義，故不適合被採納。本研究所提因素分析決策樹和機會補償決策樹之預測模型，其正確率結果都在56%和71%以上，可被採納建置預測模型，提供較無實際插管經驗的醫師預測插管的困難度，藉此可降低手術時的風險與危險性。在未來研究方面，可朝整合模糊理論至決策樹的建置過程，讓決策之數據判斷更有彈性，以提升預測模型之正確率。

關鍵字

氣管插管 ； 困難插管 ； 危險因子 ； 決策樹 ； 機會補償決策樹

並列摘要

The anesthetists usually encounter, difficult intubation problem, when conducting pre-operational anesthesia. Difficult intubation is affected by many factors, which can be evaluated at pre-operational stapes.
This study first surveyed researches and found 24 risk factors of difficult intubation. After discussed with doctors and nurses, 12 subjective factors remained. Then we used factor analysis and principal component analysis technique to compare the importance of these factors and 3 of them are removed. Finally, expert opinions suggested that the risk factors can be further shrunk into 6 factors. Decision tree technique is then used to set up prediction model for difficult intubation.
The subjects of this study were 51 patients (27 males and 24 females) in a hospital in southern Taiwan; they were full anesthetized with standard intubation tool. The pre-operative data of these patients were collected using a user-friendly interface.
This study had built and tested four types of decision trees which are listed below with their corresponding prediction correct rates: (1) the C5.0 decision tree, the correct rate is 83% for male and 50% for female; (2) the Intuitive decision tree, correct rates are 44.44% and 25%; (3) the factor analysis decision tree, the correct rates are 56% and 71%; (4) the chance redeem decision tree, the correct rates are 63% and 75%.
To sum up, although C5.0 decision tree has the highest correct rate, however since its decision criteria does not contain clinical meaning, it is not suitable for practical application. The decision tree model built based on factor analysis and chance redeem techniques have the correct rate of 56% and 71%, they can be provide as suggestions for inexperienced doctors to reduce the risk and cost during surgery operations. For the future studies, it will be interesting to see how fuzzy theory can be integrated into the model and evaluate the increase of prediction rate.

並列關鍵字

Endotracheal Intubation ； Difficult Intubation ； Risk Factor ； Decision-Tree ； Chance Redeem Tree

參考文獻

[3] 徐芳玲，葉榮懋(2002)，「以主成分分析應用在決策數名目屬性值之二元分割上」，國立成功大學碩士論文。

[5] 徐芳玲，葉榮懋(2002)，「以主成分分析應用在決策數名目屬性值之二元分割上」，國立成功大學碩士論文。

[10] 戴溫然(2004)，「麻醉醫師及手術病患對麻醉醫療資訊供需差異之探討-以北部某醫學中心為例」，台灣大學公共衛生學院醫療機構管理所碩士論文。

[13] A D Paix, J A Williamson, W B Runciman.(2005),“Crisis management during anaesthesia: difficult intubation”, Qual Saf Health Care,14:e5 doi:10.1136/qshc.2002.004135.

[14] Buckley FP, Robinson NB.(1983),“Simonowitz DA, Dellinger EP. Anaesthesia in the morbidly obese: a comparison of anaesthetic and analgesic regimens for upper abdominal surgery. ”, Anaesthesia,38:840–51.