Linna Wang
ABSTRACT
Surgical patients aged 65 and over are facing a 2-10 times higher risk of death after surgery. Early prediction of postoperative mortality is essential, as timely and appropriate treatment can improve survival outcomes. With the development of medical and computer technology, numerous available health-related data can be recorded for research. Among various patient indicators which may affect the accuracy of prediction, it is necessary to find highly relevant and efficient features. The aims of this study were to use machine learning algorithms, specifically Bagging and Boosting Algorithms (e.g. Random Forest, eXtreme Gradient Boosting), to predict the postoperative 30-days mortality in surgical patients aged over 65, and to identify the optimal features using genetic algorithm(GA). This prospective study was developed and validated on the cohort from electronic health records (EHRs) of West China Hospital, Sichuan University, which contained 7467 surgical patients (0.924% mortality rate) who underwent surgery between July 1, 2019 and October 31, 2020. Compared with models like the traditional logistic regression model and the baseline ASA physical status, We found that XGBoost with hyper-parameters had best performance based solely on the automatically obtained features (area under the curve [AUC] of 0.9318, 95% confidence interval [CI] 0.9041 - 0.9594). The AUC of baseline ASA-PS was 0.6787 (95% CI 0.6471 - 0.7103) using XGBoost. When both ASA-PS and the selected features are included as inputs, XGboost achieved the AUC of 0.9345 (95% CI 0.9076 - 0.9613).
- Meara, J.G., (2015). Global Surgery 2030: evidence and solutions for achieving health, welfare, and economic development. The Lancet, 386(9993): p. 569-624.Google Scholar
Cross Ref
- Nepogodiev, D., (2019). Global burden of postoperative death. The Lancet, 93(10170).Google Scholar
- GBD 2016 Causes of Death Collaborators (2017). Global, regional, and national age-sex specific mortality for 264 causes of death, 1980–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 2017; 390: 1151–210.Google Scholar
- Pearse RM, Harrison DA, James P, Watson D, Hinds C, Rhodes A, Grounds RM, Bennett ED (2006). Identification and characterisation of the high-risk surgical population in the United Kingdom. Crit Care. 10(3):R81. doi: 10.1186/cc4928. Epub 2006 Jun 2. PMID: 16749940; PMCID: PMC1550954.Google ScholarCross Ref
- Chiew C J, Liu N, Wong T H, (2020). Utilizing machine learning methods for preoperative prediction of postsurgical mortality and intensive care unit admission[J]. Annals of surgery, 272(6): 1133.Google ScholarCross Ref
- Revenig, L.M., Ogan, K., Guzzo, T.J. (2014). The Use of Frailty as a Surgical Risk Assessment Tool in Elderly Patients. Curr Geri Rep 3, 1-7.Google ScholarCross Ref
- Gilbert T, Neuburger J, Kraindler J, (2018). Development and validation of a Hospital Frailty Risk Score focusing on older people in acute care settings using electronic hospital records: an observational study[J]. The Lancet, 391(10132): 1775-1782.Google ScholarCross Ref
- Preeya K. Mistry, Geoffrey S. Gaunay, David M. Hoenig (2017). Prediction of surgical complications in the elderly: Can we improve outcomes?,Asian Journal of Urology, Volume 4, Issue 1, Pages 44-49,ISSN 2214-3882.Google Scholar
- Aggarwal, G.; Broughton, K.J.; Williams, L.J.; Peden, C.J.; Quiney, N (2020). Early Postoperative Death in Patients Undergoing Emergency High-Risk Surgery: Towards a Better Understanding of Patients for Whom Surgery May not Be Beneficial. J. Clin. Med. 9, 1288.Google ScholarCross Ref
- Naito T, Mitsunaga S, Miura S, (2019). Feasibility of early multimodal interventions for elderly patients with advanced pancreatic and non‐small‐cell lung cancer[J]. Journal of cachexia, sarcopenia and muscle, 2019, 10(1): 73-83.Google Scholar
- Han J, Kamber M, Pei J (2011). Data mining concepts and techniques third edition[J]. The Morgan Kaufmann Series in Data Management Systems, 5(4): 83-124.Google Scholar
- Witten I H, Frank E (2002). Data mining: practical machine learning tools and techniques with Java implementations[J]. Acm Sigmod Record, 31(1): 76-77.Google ScholarDigital Library
- Jordan M I, Mitchell T M. Machine learning: Trends, perspectives, and prospects[J]. Science, 2015, 349(6245): 255-260.Google Scholar
- Babatunde O H, Armstrong L, Leng J, (2014). A genetic algorithm-based feature selection[J].Google Scholar
- Etzioni D A, Liu J H, Maggard M A, (2003). The aging population and its impact on the surgery workforce[J]. Annals of surgery, 238(2): 170.Google ScholarCross Ref
- The International Surgical Outcomes Study group, Global patient outcomes after elective surgery: prospective cohort study in 27 low-, middle- and high-income countries, BJA: British Journal of Anaesthesia, Volume 117, Issue 5, November 2016, Pages 601–609, https://doi.org/10.1093/bja/aew316.Google Scholar
- Glance L G, Lustik S J, Hannan E L, (2012). The Surgical Mortality Probability Model: derivation and validation of a simple risk prediction rule for noncardiac surgery[J]. Annals of surgery, 255(4): 696-702.Google ScholarCross Ref
- Le Manach Y, Collins G, Rodseth R, (2016). Preoperative score to predict postoperative mortality (POSPOM). Anesthesiology, 124: 570e9.Google ScholarCross Ref
- Protopapa K L, Simpson J C, Smith N C E, (2014). Development and validation of the surgical outcome risk tool (SORT)[J]. The British journal of surgery, 101(13): 1774.Google Scholar
- Charlson M, Szatrowski T P, Peterson J, (1994). Validation of a combined comorbidity index[J]. Journal of clinical epidemiology, 47(11): 1245-1251.Google ScholarCross Ref
- Bilimoria K Y, Liu Y, Paruch J L, (2013). Development and evaluation of the universal ACS NSQIP surgical risk calculator: a decision aid and informed consent tool for patients and surgeons[J]. Journal of the American College of Surgeons, 217(5): 833-842. e3.Google ScholarCross Ref
- Cohen M E, Bilimoria K Y, Ko C Y, (2009). Effect of subjective preoperative variables on risk-adjusted assessment of hospital morbidity and mortality[J]. Annals of surgery, 249(4): 682-689.Google ScholarCross Ref
- Hill B L, Brown R, Gabel E, (2019). An automated machine learning-based model predicts postoperative mortality using readily-extractable preoperative electronic health record data[J]. British journal of anaesthesia, 123(6): 877-886.Google Scholar
- Pirracchio R, Petersen M L, Carone M, (2015). Mortality prediction in intensive care units with the Super ICU Learner Algorithm (SICULA): a population-based study[J]. The Lancet Respiratory Medicine, 3(1): 42-52.Google ScholarCross Ref
- Guyon I, Elisseeff A (2003). An introduction to variable and feature selection[J]. Journal of machine learning research, 3(Mar): 1157-1182.Google ScholarDigital Library
- Mao Y, Yang Y (2019). A wrapper feature subset selection method based on randomized search and multilayer structure[J]. BioMed research international, 2019.Google Scholar
- WHO. International Statistical Classification of Diseases and Related Health Problems-10th Revision[M]. 1993.Google Scholar
- Haldun Akoglu (2018). User's guide to correlation coefficients,Turkish Journal of Emergency Medicine,Volume 18, Issue 3, Pages 91-93, ISSN 2452-2473.Google Scholar
- Mazumder R, Hastie T, Tibshirani R (2010). Spectral regularization algorithms for learning large incomplete matrices[J]. The Journal of Machine Learning Research, 11: 2287-2322.Google ScholarDigital Library
- Chawla N V, Bowyer K W, Hall L O, (2002). SMOTE: synthetic minority over-sampling technique[J]. Journal of artificial intelligence research, 16: 321-357.Google ScholarCross Ref
- Lemaître G, Nogueira F, Aridas C K (2017). Imbalanced-learn: A python toolbox to tackle the curse of imbalanced datasets in machine learning[J]. The Journal of Machine Learning Research, 18(1): 559-563.Google ScholarDigital Library
- B. Xue, M. Zhang, W. N. Browne and X. Yao (2016). A Survey on Evolutionary Computation Approaches to Feature Selection," in IEEE Transactions on Evolutionary Computation, vol. 20, no. 4, pp. 606-626, doi: 10.1109/TEVC.2015.2504420.Google ScholarDigital Library
- Chandrashekar G, Sahin F (2014). A survey on feature selection methods[J]. Computers & Electrical Engineering, 40(1): 16-28.Google ScholarDigital Library
- Golberg D E (1989). Genetic algorithms in search, optimization, and machine learning[J]. Addion wesley, 1989(102): 36.Google Scholar
- Yang J, Honavar V (1998). Feature subset selection using a genetic algorithm[M]//Feature extraction, construction and selection. Springer, Boston, MA, 117-136.Google Scholar
- Anbarasi M, Anupriya E, Iyengar N (2010). Enhanced prediction of heart disease with feature subset selection using genetic algorithm[J]. International Journal of Engineering Science and Technology, 2(10): 5370-5376.Google Scholar
- Quinlan J R (1986). Induction of decision trees[J]. Machine learning, 1(1): 81-106.Google Scholar
- Fortin F A, De Rainville F M, Gardner M A G, (2012). DEAP: Evolutionary algorithms made easy[J]. The Journal of Machine Learning Research, 13(1): 2171-2175.Google ScholarDigital Library
- Guyon I, Weston J, Barnhill S, (2002). Gene selection for cancer classification using support vector machines[J]. Machine learning, 2002, 46(1): 389-422.Google Scholar
- Breiman L (2001). Random forests[J]. Machine learning, 2001, 45(1): 5-32.Google Scholar
- Chen T, Guestrin C (2016). Xgboost: A scalable tree boosting system[C]//Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, 785-794.Google ScholarDigital Library
- Hastie T, Tibshirani R, Friedman J (2009). The elements of statistical learning: data mining, inference, and prediction[M]. Springer Science & Business Media.Google ScholarCross Ref
- Pedregosa F, Varoquaux G, Gramfort A, (2011). Scikit-learn: Machine learning in Python[J]. the Journal of machine Learning research, 12: 2825-2830.Google Scholar
Index Terms
- Machine Learning-based Prediction of Postoperative 30-days Mortality
Recommendations
Machine Learning-based Prediction of ICU Patient Mortality at Time of Admission
ICISDM '18: Proceedings of the 2nd International Conference on Information System and Data MiningIn this paper we describe the application of machine learning techniques to the prediction of hospital Intensive Care Unit (ICU) patient mortality. A large dataset of over 58,000 ICU admissions from the MIMIC III database was used in the development, ...
The Impact of Functional Dependence and Related Surgical Complications on Postoperative Mortality
AbstractPurposeFunctional dependency is a known determinant of surgical risk. To enhance our understanding of the relationship between dependency and adverse surgical outcomes, we studied how postoperative mortality following a surgical complication was ...
Mortality Prediction of ICU patients using Machine Leaning: A survey
ICCDA '17: Proceedings of the International Conference on Compute and Data AnalysisRecently health care researchers are working a lot on outcome prediction on Intensive Care Unit (ICU) and trauma. Outcome prediction in intensive care is a difficult process. Accurate synthesis of quality data and application of prior experience to the ...
Comments