急診過度壅塞會導致醫療品質及病人安全疑慮，因此妥善利用急診醫療資源以預防不良事件的發生至關重要。我們期盼可以找到最佳預測工具來準確分析病人狀況是否會惡化。本研究以南台灣某醫學中心急診室為研究標的，收集2017年度1~12月份由急診住院之18,378位病人、共124,334筆電子病歷資料，經去識別化後進行回溯性研究分析，比較急診醫師主觀預測與2018年導入急診之NEWS (National Early Warning Score)客觀預測之病情惡化準確度分析，並導入機器學習，採用Microsoft Azure ML工具來建立模型對上述資料進行預測分析。

NEWS預測之accuracy (90.07%)及precision (76.47%)皆優於急診醫師主觀預測之accuracy (83.04%) 及precision (33.04%)。9種機器學習模型中，Multi-Class Decision Forest預測結果表現最佳，其recall、precision、及accuracy都有98%以上的預測表現。探討病人安全與醫療品質管理面向，為了達到未來機器學習資料分析的即時性及正確性，醫療場所必須要投入資源並導入智慧裝置以滿足生命徵象量測後自動拋轉。

機器學習預測病情惡化優於NEWS及急診醫師主觀預測。Multi-Class Decision Forest是表現最為突出亮眼的機器學習模型，深入探究其模型分析多達了32層之多，是否可能產生過度配適(overfitting)的現象，必須再進一步用其他資料來驗證或實際導入前瞻性研究來釐清機器學習模型的信效度。

Emergency department (ED) overcrowding results in compromise of patient safety and medical quality. Optimal ED resource management to prevent adverse events is crucial. We are looking forward to developing the best tool for predicting patient deterioration. This retrospective study collects 124,334 electronic medical records of 18,378 de-identification patients admitted to an emergency department of a medical center in Southern Taiwan from January to December, 2017. We compare between ED physicians’ subjective predictions and NEWS (National Early Warning Score) predictions for patient deterioration. Microsoft Azure Machine Learning tool was introduced to build models to perform predictive analysis based on the above data in this study.

NEWS prediction results have better accuracy (90.07%) and precision (76.47%), compared to those made by emergency physicians, with accuracy (83.04%) and precision (33.04%). Among 9 machine learning models used in our study, Multi-Class Decision Forest gives best recall, precision, and accuracy results, each higher than 98%. Concerning patient safety and medical quality management, smart medical devices must be introduced for the measurement, recording, and automatic uploading of vital signs, in order to pave the way for the early stage of machine learning in data collection.

The performance of machine learning-based prediction in patient deterioration is superior to those by NEWS and ED physicians. While Multi-Class Decision Forest is the best performing model, the possibility of overfitting should be further investigated since it developed 32 analytic layers in the study. More new data should be collected and follow-up study should be conducted to clarify its prediction reliability and validity.

1. Brink A, Alsma J, Verdonschot RJCG, Rood PPM, Zietse R, Lingsma HF, Schuit SCE.PLoS One. Predicting mortality in patients with suspected sepsis at the Emergency Department; A retrospective cohort study comparing qSOFA, SIRS and National Early Warning Score. 2019 Jan 25;14(1):e0211133. doi: 10.1371/journal.pone.0211133. eCollection 2019.
2. Nannan Panday RS, Minderhoud TC, Alam N, Nanayakkara PWB. Prognostic value of early warning scores in the emergency department (ED) and acute medical unit (AMU): A narrative review. Eur J Intern Med. 2017 Nov;45:20-31. doi: 10.1016/j.ejim.2017.09.027. Epub 2017 Oct 7. Review.
3. Lim WT, Fang AH, Loo CM, Wong KS, Balakrishnan T. Use of the National Early Warning Score (NEWS) to Identify Acutely Deteriorating Patients with Sepsis in Acute Medical Ward. Ann Acad Med Singapore. 2019 May;48(5):145-149.
4. Zaidi H1, Bader-El-Den M2, McNicholas J3. Using the National Early Warning Score (NEWS/NEWS 2) in different Intensive Care Units (ICUs) to predict the discharge location of patients. BMC Public Health. 2019 Sep 5;19(1):1231. doi: 10.1186/s12889-019-7541-3.
5. Kwon JM, Lee Y, Lee Y, Lee S, Park J. An Algorithm Based on Deep Learning for Predicting In-Hospital Cardiac Arrest. J Am Heart Assoc. 2018 Jun 26;7(13). pii: e008678. doi: 10.1161/JAHA.118.008678.
6. Bente Bilben1, Linda Grandal1 and Signe Søvik1,2. National Early Warning Score (NEWS) as an emergency department predictor of disease severity and 90-day survival in the acutely dyspneic patient – a prospective observational study. Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine (2016) 24:80
7. Faisal M, Richardson D, Scally AJ, Howes R, Beatson K, Speed K, Mohammed MA. Computer-aided National Early Warning Score to predict the risk of sepsis following emergency medical admission to hospital: a model development and external validation study. CMAJ. 2019 Apr 8;191(14):E382-E389. doi: 10.1503/cmaj.181418.
8. Farah Shamout, Tingting Zhu. Machine Learning for the Deterioration of Patients on Hospital Wards. Published 2017
9. Wood C, Chaboyer W, Carr P. How Do Nurses Use Early Warning Scoring Systems to Detect and Act on Patient Deterioration to Ensure Patient Safety? A Scoping Review. Int J Nurs Stud 2019 Jun;94:166-178
10. 天下雜誌https://www.cw.com.tw/article/5066815
11. Zaidi et al. Using the National Early Warning Score(NEWS/NEWS 2) in different Intensive Care Units (ICUs) to predict the discharge location of patients. BMC Public Health (2019) 19:1231
12. Jeongmin Kim et al. Predicting Cardiac Arrest and Respiratory Failure Using Feasible Artificial Intelligence with Simple Trajectories of Patient Data. J Clin Med. 2019, 8(9): 1336. doi: 10.3390/jcm8091336
13. Youngnam Lee et al. Deep Learning in the Medical Domain: Predicting Cardiac Arrest Using Deep Learning. Acute and Critical Care 2018 August 33(3):117-120
14. Smart asset management as a service Deliverable 2.0. Technical Report May 2018. DOI:10.13140/RG.2.2.31027.94244
15. Vistisen, S.T., Johnson, A.E.W. & Scheeren, T.W.L. Predicting vital sign deterioration with artificial intelligence or machine learning. J Clin Monit Comput 33, 949–951 (2019). https://doi.org/10.1007/s10877-019-00343-7
16. Trinkle RM, Flabouris A. Documenting Rapid Response System afferent limb failure and associated patient outcomes. Resuscitation. 2011; 82:810–814.
17. Smith GB, Prytherch DR, Meredith P, Schmidt PE, Featherstone PI. The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and death. Resuscitation. 2013; 84:465–470.
18. Taylor RA, Pare JR, Venkatesh AK, Mowafi H, Melnick ER, Fleischman W, Hall MK. Prediction of in-hospital mortality in emergency department patients with sepsis: a local big data-driven, machine learning approach. Acad Emerg Med. 2016; 23:269–278.
19. Subbe CP, Davies RG, Williams E, Rutherford P, Gemmell L. Effect of introducing the Modified Early Warning score on clinical outcomes, cardio-pulmonary arrests and intensive care utilisation in acute medical admissions. Anaesthesia. 2003; 58:797–802.
20. Suwanpasu S, Sattayasomboon Y. Accuracy of Modified Early Warning Scores for predicting mortality in hospital: a systematic review and meta-analysis. J Intensive Crit Care. 2016; 02:1–11.
21. DeVita MA, Smith GB, Adam SK, Adams-Pizarro I, Buist M, Bellomo R, BonelloR, Cerchiari E, Farlow B, Goldsmith D, Haskell H, Hillman K, Howell M, HravnakM, Hunt EA, Hvarfner A, Kellett J, Lighthall GK, Lippert A, Lippert FK, MahroofR, Myers JS, Rosen M, Reynolds S, Rotondi A, Rubulotta F, Winters B. “Identifying the hospitalised patient in crisis”—a consensus conference on the afferent limb of rapid response systems. Resuscitation 2010; 81:375–382.
22. Jones DA, DeVita MA, Bellomo R. Rapid-response teams. N Engl J Med. 2011; 365:139–146.
23. Romero-Brufau S, Huddleston JM, Naessens JM, Johnson MG, Hickman J, Morlan BW, Jensen JB, Caples SM, Elmer JL, Schmidt JA, Morgenthaler TI, Santrach PJ. Widely used track and trigger scores: are they ready for automation in practice? Resuscitation. 2014; 85:549–552.
24. Subbe CP, et al. Validation of a modified Early Warning Score in medical admis-sions. QJM 2001; 94:521–6
25. Paterson R, et al. Prediction of in-hospital mortality and length of stay using an early warning scoring system: clinical audit. Clin Med 2006; 6:281–4.
26. Harrison GA, et al. Combinations of early signs of critical illness predict in-hospital death-the SOCCER study (signs of critical conditions and emergency responses). Resuscitation 2006; 71:327–34
27. Duckitt RW, et al. Worthing physiological scoring system: derivation and validation of a physiological early-warning system for medical admis-sions. An observational, population-based single-centre study. Br J Anaesth 2007; 98:769–74
28. Prytherch DR, et al. ViEWS – towards a national early warning score for detecting adult inpatient deterioration. Resuscitation 2010; 81:932–7.
29. Smith GB, et al. The ability of the National Early Warning Score (NEWS) to dis-criminate patients at risk of early cardiac arrest: unanticipated intensive care unit admission, and death. Resuscitation 2013; 84:465–70
30. 讓資料為你產生價值-哈佛商業評論。陳昇瑋解讀〈架構你的資料策略〉。2017年5月號(最佳執行長哪裡不一樣)，2017/4/28
31. Predictive analytics with Microsoft Azure machine learning 2nd edition. Build and deploy actionable solutions in minutes. Roger Barga, Valentine Fontama, Wee Hyong Tok.
32. 李弘毅老師Deep Reinforcement learning (2017 Spring) 【筆記】https://medium.com/change-the-world-with-technology/%E6%9D%8E%E5%AE%8F%E6%AF%85%E8%80%81%E5%B8%AB-deep-reinforcement-learning-2017-spring-%E7%AD%86%E8%A8%98-3784ddb23e0
33. Tommy Huang機器/統計學習：羅吉斯回歸(Logistic regression) https://chih-sheng-huang821.medium.com/%E6%A9%9F%E5%99%A8-%E7%B5%B1%E8%A8%88%E5%AD%B8%E7%BF%92-%E7%BE%85%E5%90%89%E6%96%AF%E5%9B%9E%E6%AD%B8-logistic-regression-aff7a830fb5d
34. Yeh James [資料分析&機器學習] 第3.2講：線性分類-感知器(Perceptron) 介紹 https://medium.com/jameslearningnote/%E8%B3%87%E6%96%99%E5%88%86%E6%9E%90-%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E7%AC%AC3-2%E8%AC%9B-%E7%B7%9A%E6%80%A7%E5%88%86%E9%A1%9E-%E6%84%9F%E7%9F%A5%E5%99%A8-perceptron-%E4%BB%8B%E7%B4%B9-84d8b809f866
35. Tommy Huang機器學習-支撐向量機(support vector machine, SVM)詳細推導 https://chih-sheng-huang821.medium.com/%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E6%94%AF%E6%92%90%E5%90%91%E9%87%8F%E6%A9%9F-support-vector-machine-svm-%E8%A9%B3%E7%B4%B0%E6%8E%A8%E5%B0%8E-c320098a3d2e
36. Yeh James [資料分析&機器學習] 第3.4講：支援向量機(Support Vector Machine)介紹 https://medium.com/jameslearningnote/%E8%B3%87%E6%96%99%E5%88%86%E6%9E%90-%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E7%AC%AC3-4%E8%AC%9B-%E6%94%AF%E6%8F%B4%E5%90%91%E9%87%8F%E6%A9%9F-support-vector-machine-%E4%BB%8B%E7%B4%B9-9c6c6925856b
37. Tommy Huang機器學習：Kernel函數 https://chih-sheng-huang821.medium.com/%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-kernel-%E5%87%BD%E6%95%B8-47c94095171
38. https://ithelp.ithome.com.tw/m/articles/10222840
39. https://www.jamleecute.com/decision-tree-cart-%E6%B1%BA%E7%AD%96%E6%A8%B9/
40. Yeh James [資料分析&機器學習] 第3.5講：決策樹(Decision Tree)以及隨機森林(Random Forest)介紹 https://medium.com/jameslearningnote/%E8%B3%87%E6%96%99%E5%88%86%E6%9E%90-%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E7%AC%AC3-5%E8%AC%9B-%E6%B1%BA%E7%AD%96%E6%A8%B9-decision-tree-%E4%BB%A5%E5%8F%8A%E9%9A%A8%E6%A9%9F%E6%A3%AE%E6%9E%97-random-forest-%E4%BB%8B%E7%B4%B9-7079b0ddfbda
41. ML入門(十七)隨機森林(Random Forest) https://medium.com/chung-yi/ml%E5%85%A5%E9%96%80-%E5%8D%81%E4%B8%83-%E9%9A%A8%E6%A9%9F%E6%A3%AE%E6%9E%97-random-forest-6afc24871857
42. https://ithelp.ithome.com.tw/articles/10191528
43. https://read01.com/zh-tw/xDNNmQ0.html