Uncovering Clinical Risk Factors and Predicting Severe COVID-19 Cases Using UK Biobank Data: Machine Learning Approach

JMIR Public Health and Surveillance - Tập 7 Số 9 - Trang e29544
C Y Wong1, Yong Xiang1, Liangying Yin1, Hon‐Cheong So2,3,4,5,6,7,1
1School of Biomedical Sciences, The Chinese University of Hong Kong, Hong Kong, China
2Brain and Mind Institute, The Chinese University of Hong Kong, Hong Kong, China
3CUHK Shenzhen Research Institute, Shenzhen, China
4Department of Psychiatry, The Chinese University of Hong Kong, Hong Kong, China
5Hong Kong Branch of the Chinese Academy of Sciences Center for Excellence in Animal Evolution and Genetics, The Chinese University of Hong Kong, Hong Kong, China
6KIZ-CUHK Joint Laboratory of Bioresources and Molecular Research of Common Diseases, Kunming Institute of Zoology and The Chinese University of Hong Kong, Kunming, China
7Margaret K.L. Cheung Research Centre for Management of Parkinsonism, The Chinese University of Hong Kong, Hong Kong, China

Tóm tắt

Background COVID-19 is a major public health concern. Given the extent of the pandemic, it is urgent to identify risk factors associated with disease severity. More accurate prediction of those at risk of developing severe infections is of high clinical importance. Objective Based on the UK Biobank (UKBB), we aimed to build machine learning models to predict the risk of developing severe or fatal infections, and uncover major risk factors involved. Methods We first restricted the analysis to infected individuals (n=7846), then performed analysis at a population level, considering those with no known infection as controls (ncontrols=465,728). Hospitalization was used as a proxy for severity. A total of 97 clinical variables (collected prior to the COVID-19 outbreak) covering demographic variables, comorbidities, blood measurements (eg, hematological/liver/renal function/metabolic parameters), anthropometric measures, and other risk factors (eg, smoking/drinking) were included as predictors. We also constructed a simplified (lite) prediction model using 27 covariates that can be more easily obtained (demographic and comorbidity data). XGboost (gradient-boosted trees) was used for prediction and predictive performance was assessed by cross-validation. Variable importance was quantified by Shapley values (ShapVal), permutation importance (PermImp), and accuracy gain. Shapley dependency and interaction plots were used to evaluate the pattern of relationships between risk factors and outcomes. Results A total of 2386 severe and 477 fatal cases were identified. For analyses within infected individuals (n=7846), our prediction model achieved area under the receiving-operating characteristic curve (AUC–ROC) of 0.723 (95% CI 0.711-0.736) and 0.814 (95% CI 0.791-0.838) for severe and fatal infections, respectively. The top 5 contributing factors (sorted by ShapVal) for severity were age, number of drugs taken (cnt_tx), cystatin C (reflecting renal function), waist-to-hip ratio (WHR), and Townsend deprivation index (TDI). For mortality, the top features were age, testosterone, cnt_tx, waist circumference (WC), and red cell distribution width. For analyses involving the whole UKBB population, AUCs for severity and fatality were 0.696 (95% CI 0.684-0.708) and 0.825 (95% CI 0.802-0.848), respectively. The same top 5 risk factors were identified for both outcomes, namely, age, cnt_tx, WC, WHR, and TDI. Apart from the above, age, cystatin C, TDI, and cnt_tx were among the top 10 across all 4 analyses. Other diseases top ranked by ShapVal or PermImp were type 2 diabetes mellitus (T2DM), coronary artery disease, atrial fibrillation, and dementia, among others. For the “lite” models, predictive performances were broadly similar, with estimated AUCs of 0.716, 0.818, 0.696, and 0.830, respectively. The top ranked variables were similar to above, including age, cnt_tx, WC, sex (male), and T2DM. Conclusions We identified numerous baseline clinical risk factors for severe/fatal infection by XGboost. For example, age, central obesity, impaired renal function, multiple comorbidities, and cardiometabolic abnormalities may predispose to poorer outcomes. The prediction models may be useful at a population level to identify those susceptible to developing severe/fatal infections, facilitating targeted prevention strategies. A risk-prediction tool is also available online. Further replications in independent cohorts are required to verify our findings.

Từ khóa


Tài liệu tham khảo

10.1056/NEJMoa2001316

10.3760/cma.j.issn.0254-6450.2020.02.003

10.1056/NEJMoa2002032

Johns Hopkins Coronavirus Resource Center2021-06-19https://coronavirus.jhu.edu/map.html

10.1093/cid/ciaa1175

10.1136/bmj.m3339

10.2196/27060

10.1136/bmj.m1328

10.1016/j.imu.2021.100564

10.1016/j.csbj.2021.05.010

10.2196/24018

10.1093/gerona/glaa183

10.1371/journal.pone.0251613

10.1016/j.pcd.2020.05.011

10.11606/s1518-8787.2020054002481

10.3390/ijerph17165974

10.1136/bmjopen-2020-044684

10.1007/s15010-020-01509-1

10.1371/journal.pmed.1001779

Data for COVID-19 research2021-09-08http://biobank.ndph.ox.ac.uk/showcase/exinfo.cgi?src=COVID19

The official UK government website for data and insights on coronavirus (COVID-19)2021-09-08https://coronavirus.data.gov.uk/

10.1136/bmj.n1088

10.15585/mmwr.mm7018e1

10.1056/NEJMoa2101765

10.1016/S0140-6736(21)00947-8

10.1038/s41598-021-84603-0

10.1002/jmv.26890

10.1016/j.bbi.2020.05.059

10.1038/s41586-020-03065-y

10.1016/j.jinf.2020.06.067

COVID-19 test results data2021-09-08http://biobank.ndph.ox.ac.uk/showcase/exinfo.cgi?src=COVID19_tests

10.1038/nmeth.3968

10.1093/bioinformatics/btr597

10.1136/bmjopen-2013-002847

10.1093/aje/kwt312

10.1111/2041-210x.12232

10.1177/0962280213497434

NixonJDusenberryMZhangLJerfelGTranDMeasuring Calibration in Deep LearningCVPR Workshops20192021-09-14https://openaccess.thecvf.com/content_CVPRW_2019/html/Uncertainty_and_Robustness_in_Deep_Visual_Learning/Nixon_Measuring_Calibration_in_Deep_Learning_CVPRW_2019_paper.html

10.1002/sim.6428

10.1038/s41392-020-0159-1

SongHDietheTKullMFlachPDistribution calibration for regression2019Proceedings of the 36th International Conference on Machine LearningJune 10-15, 2019Long Beach, CA5897906

10.1145/1102351.1102430

Jiang, X, 2011, AMIA Jt Summits Transl Sci Proc, 2011, 16

10.1093/eurheartj/ehu207

Niculescu-MizilACaruanaRObtaining Calibrated Probabilities from Boosting200521st Conference on Uncertainty in Artificial Intelligence (UAI)July 26-29, 2005Edinburgh, Scotland41320

LundbergSLeeSA unified approach to interpreting model predictions. Advances in neural information processing systemsProceedings of the 31st International Conference on Neural Information Processing Systems201731st International Conference on Neural Information Processing SystemsDecember 4-9, 2017Long Beach, CARed Hook, NYCurran Associates47684777

10.1038/s42256-019-0138-9

Lundberg, S, 2018, arXiv

10.1093/bioinformatics/btq134

10.1198/jasa.2010.tm09415

10.1111/1467-9868.00293

Boyd, K, 2013, Machine Learning and Knowledge Discovery in Databases, 451

Risk Estimation for Covid-19 - Online Prediction Tool2021-09-08https://labsocuhk.ddns.net:8890/covid19/

10.1038/s41746-021-00456-x

10.2196/26075

10.3389/fmed.2020.583060

10.21037/jmai-20-47

10.1136/bmj.m3731

10.1016/S1473-3099(20)30371-6

10.1038/s41586-020-2521-4

10.1007/s10900-020-00920-x

10.1111/ijcp.13916

10.1002/dmrr.3377

10.1371/journal.pmed.1003321

Understand Your Dataset with XgboostXGBoost R Package20202021-09-08https://xgboost.readthedocs.io/en/latest/R-package/discoverYourData.html

10.1073/pnas.2011086117

10.1111/obr.13128

10.1016/j.dsx.2020.05.020

10.1016/j.numecd.2020.07.031

10.1007/s40620-020-00790-5

10.1093/ndt/gfl073

10.1053/j.ajkd.2013.03.027

10.1111/andr.12836

10.1101/2020.11.16.20232512

10.1016/j.annonc.2020.04.479

10.1159/000510914

10.1002/ajh.25829

10.1038/s41430-020-0652-1

10.1016/s2213-8587(20)30238-2

10.1002/sim.2929

10.1161/circulationaha.106.672402

10.1056/NEJMp068249

10.1093/ije/dyz274

10.1093/aje/kwx246

NielsenDTree Boosting With XGBoostMaster thesis, Norwegian University of Science and Technology20162021-09-08http://pzs.dstu.dp.ua/DataMining/boosting/bibl/Didrik.pdf

Thông tin
Thông tin xuất bản

JMIR Public Health and Surveillance

Tập 7 Số 9

e29544

Thông tin tác giả