運用數據模型預測初始氧合指數與呼吸器使用日數之相關性

Use a Data Model Predicting the Correlation between Initial Oxygenation Index and Days of Ventilator Usage

目的：精準醫療在急性呼吸衰竭個案照護中扮演關鍵角色，而氧合指數是評估肺部氧合狀況與患者能否及早脫離呼吸器的重要指標，對臨床決策及國家健保規範的呼吸照護階段性管理，皆具重要的參考價值。透過回溯性資料，探討急性呼吸衰竭病人在ICU 48小時內的初始特徵中，氧合指數(P/F)與呼吸器使用日數之間的關聯性。方法：運用擅長於醫學預測及診斷的決策樹(Decision Tree)和支持向量機(Support Vector Machine, SVM)等演算法進行實驗。結果：在決策樹C5.0演算法發現，急性呼吸衰竭前的住院日數關係著呼吸器依賴與健保署呼吸照護整合作業；在SVM演算法，實驗揭露BUN、BMI、P/F、APACHE II score、升壓藥物使用、PaCO2等為重要特徵值。結論：本研究驗證氧合指數與呼吸器使用日數具有重要的相關性。然而，急性呼吸衰竭發生前的住院日數及初始供氣模式，亦為影響實驗結果的重要特徵值，是本實驗的重要發現。這些結果在預測ICU急性呼吸衰竭預後，以及健保呼吸照護管理流程，提供了具臨床參考價值的依據。

Objectives. The oxygenation index (P/F ratio) is a crucial metric for assessing pulmonary oxygenation status and guiding clinical decisions regarding ventilator weaning in patients with acute respiratory failure (ARF). According to National Health Insurance Regulations, it provides valuable information for the staged management of respiratory care. This study aimed to retrospectively investigate the relationship between the oxygenation index and the duration of ventilator use by analyzing the initial characteristics of ARF patients within 48 hours of ICU admission.
Methods. This retrospective study analyzed data from ARF patients using decision tree (C5.0) and support vector machine (SVM) algorithms to predict and diagnose factors influencing ventilator dependence.
Results. The decision tree C5.0 algorithm identified the duration of hospitalization before ARF onset as a significant factor associated with ventilator dependence, which was further linked to the National Health Insurance Administration's integrated respiratory care program. The SVM algorithm highlighted critical predictive factors, including BUN, BMI, P/F ratio, APACHE II score, vasopressor use, and PaCO2. The oxygenation index was found to be significantly correlated with the duration of ventilator use. Additionally, the duration of hospitalization prior to ARF onset and the initial ventilation mode were identified as critical variables influencing outcomes.
Conclusion. This study underscores the importance of the oxygenation index in predicting ventilator use duration and identifies additional factors influencing ARF prognosis. These findings offer valuable clinical references for ICU respiratory care management and align with health insurance policy considerations.