利用成人健檢資料庫預測慢性腎臟病危險因子

Risk Prediction for Early Chronic Kidney Disease: Results from an Adult Health Examination Program

施錦泉(Chi-Chang Chang)

指導教授 ： 楊順發

共同指導教授 ： 張啟昌(Chi-Chang Chang)

中山醫學大學/醫學院/醫學研究所/碩士(2020年)

https://doi.org/10.6834/csmu202000170

摘要

研究目的：慢性腎臟病(CKD)已被公認為全球的公共衛生問題。早期慢性腎臟病不易被發現，若能提早對慢性腎臟病進行預測，不僅讓病人能夠及早發現自己的腎臟問題，也能防止慢性腎臟疾病的惡化。本研究藉由成人健檢資料庫使用機器學習技術，通過使用分類技術和關聯規則來識別各種可分析的危險因子和臨床特徵，進而開發出早期慢性腎臟病的預測模型。
研究方法及資料：所有樣本為成人預防健康檢查數據，收集自32個連鎖診所和三個專門的檢驗所，共計19270有效記錄。根據文獻探討共選擇 11 項獨立變數作為分析慢性腎臟病的候選風險因子，腎絲球濾過率（GFR）為預測變數。使用R語言套件分類技術包括: CART，ELM，C4.5和LDA四種機器學習方法進行分析。績效評估指標包括分類準確性，敏感度，特異度和曲線下面積。
研究結果：在預測早期CKD四種機器學習方法中，C4.5決策樹演算法優於其他三種模型。準確性依序為C4.5(0.820)、CART (0.819)、LDA(0.818)、ELM(0.715)；AUC依序為C4.5(0.788) 、CART (0.779)、LDA(0.773)、ELM(0.692)；敏感度依序為C4.5(0.673)、CART (0.670)、LDA(0.669)、ELM(0.539)；特異度依序為C4.5(0.872)、CART (0.871)、LDA(0.868)、ELM(0.777)。在風險因子的重要性依序為：尿蛋白/尿肌酸酐比值(UPCR)，尿液蛋白(PRO)，尿紅血球(RBC)，空腹血糖(GLU)，三酸甘油脂(TG)，血清總膽固醇(T-CHO)，其中關於年齡和性別扮演重要的關鍵。
結論與建議：慢性腎臟病的及時風險評估和適當的社區初級監測是國家衛生策略的重點。由於慢性腎臟病在早期不易被發現，因此透過早期預測、診斷，進一步的治療防止惡化為末期腎臟病更為重要。本研究提供成人健康檢查數據與所發展的風險預測模型可以支持預測早期慢性腎臟病臨床風險評估模型穩健性的證據。

關鍵字

早期慢性腎臟疾病 ； 機器學習技術 ； 危險因子

並列摘要

Objective：Developing effective risk prediction models is a cost-effective approach to reducing complications of chronic kidney disease (CKD) and mortality rates; however, there is inadequate evidence to support screening for CKD. In this study, four data mining algorithms, including a classification and regression tree, a C4.5 decision tree, a linear discriminant analysis, and an extreme learning machine are used to predict early CKD.
Methods and Materials：The study includes datasets from 19,270 patients, provided by an adult health examination program from 32 chain clinics and three special physical examination centers, between 2015 and 2019. There were 11 independent variables, and the glomerular filtration rate (GFR) was used as the predictive variable. The C4.5 decision tree algorithm outperformed the three comparison models for predicting early CKD based on accuracy, sensitivity, specificity, and area under the curve metrics.
Results：It is, therefore, a promising method for early CKD prediction. The rank of accuracy: C4.5(0.820)、CART (0.819)、LDA(0.818)、ELM(0.715); the rank of AUC: C4.5(0.788), CART(0.779), LDA(0.773), ELM(0.692); the rank of sensitivity: C4.5(0.673), CART (0.670), LDA(0.669), ELM(0.539); the rank of specificity: C4.5(0.872), CART (0.871), LDA(0.868), ELM(0.777). The experimental results showed that UPCR, PRO, RBC, GLU, TG, T-CHO, age, and gender are important risk factors.
Conclusion and Suggestion：CKD care is closely related to primary care level and is recognized as a healthcare priority in national strategy. The proposed risk prediction models can support the important influence of personality and health examination representations in predicting early CKD.

並列關鍵字

early chronic kidney disease ； machine learning ； risk prediction

參考文獻

參考文獻

[1] Nordqvist, C. Chronic kidney disease: causes, symptoms and treatments. IOP Publishing medicalnewstoday,2016 http://www. medicalnewstoday.com/articles/172179.php. (Accessed 14 Jan 2016).

[2] Global, regional, and national burden of chronic kidney disease, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. DOI:https://doi.org/10.1016/S0140-6736(20)30045-3,2020.

[3] Diana Smith. Chronic Kidney Disease: A Global Crisis,2018.

https://www.siemens-healthineers.com/en-be/news/chronic-kidney-disease.html.