Predicting the Risk Factors Associated With Severe Outcomes Among COVID-19 Patients-Decision Tree Modeling Approach

Mahalakshmi Kumaran; Truong-Minh Pham; Kaiming Wang; Hussain Usman; Colleen M Norris; Judy MacDonald; Gavin Y Oudit; Vineet Saini; Khokan C Sikdar

doi:10.3389/fpubh.2022.838514

Predicting the Risk Factors Associated With Severe Outcomes Among COVID-19 Patients-Decision Tree Modeling Approach

Front Public Health. 2022 May 19:10:838514. doi: 10.3389/fpubh.2022.838514. eCollection 2022.

Authors

Mahalakshmi Kumaran¹, Truong-Minh Pham², Kaiming Wang^{3

4}, Hussain Usman¹, Colleen M Norris^{5

6}, Judy MacDonald^{7

8}, Gavin Y Oudit^{3

4}, Vineet Saini^{9

10}, Khokan C Sikdar^{1

9}

Affiliations

¹ Surveillance and Reporting, Provincial Population and Public Health, Alberta Health Services, Calgary, AB, Canada.
² Surveillance and Reporting, Cancer Research and Analytics, Cancer Care Alberta, Alberta Health Services, Edmonton, AB, Canada.
³ Division of Cardiology, Department of Medicine, University of Alberta, Edmonton, AB, Canada.
⁴ Mazankowski Alberta Heart Institute, Faculty of Medicine and Dentistry, University of Alberta, Edmonton, AB, Canada.
⁵ Faculty of Nursing, University of Alberta, Edmonton, AB, Canada.
⁶ Cardiovascular Health and Stroke Strategic Clinical Network, Alberta Health Services, Edmonton, AB, Canada.
⁷ Department of Community Health Sciences, Cumming School of Medicine, University of Calgary, Calgary, AB, Canada.
⁸ Communicable Disease Control, Provincial Population and Public Health, Alberta Health Services, Calgary, AB, Canada.
⁹ Department of Community Health Sciences and O'Brien Institute for Public Health, Cumming School of Medicine, University of Calgary, Calgary, AB, Canada.
¹⁰ Public Health Evidence and Innovation, Provincial Population and Public Health, Alberta Health Services, Calgary, AB, Canada.

Abstract

Background: The COVID-19 pandemic has seen a large surge in case numbers over several waves, and has critically strained the health care system, with a significant number of cases requiring hospitalization and ICU admission. This study used a decision tree modeling approach to identify the most important predictors of severe outcomes among COVID-19 patients.

Methods: We identified a retrospective population-based cohort (n = 140,182) of adults who tested positive for COVID-19 between 5^th March 2020 and 31^st May 2021. Demographic information, symptoms and co-morbidities were extracted from a communicable disease and outbreak management information system and electronic medical records. Decision tree modeling involving conditional inference tree and random forest models were used to analyze and identify the key factors(s) associated with severe outcomes (hospitalization, ICU admission and death) following COVID-19 infection.

Results: In the study cohort, nearly 6.37% were hospitalized, 1.39% were admitted to ICU and 1.57% died due to COVID-19. Older age (>71Y) and breathing difficulties were the top two factors associated with a poor prognosis, predicting about 50% of severe outcomes in both models. Neurological conditions, diabetes, cardiovascular disease, hypertension, and renal disease were the top five pre-existing conditions that altogether predicted 29% of outcomes. 79% of the cases with poor prognosis were predicted based on the combination of variables. Age stratified models revealed that among younger adults (18-40 Y), obesity was among the top risk factors associated with adverse outcomes.

Conclusion: Decision tree modeling has identified key factors associated with a significant proportion of severe outcomes in COVID-19. Knowledge about these variables will aid in identifying high-risk groups and allocating health care resources.

Keywords: COVID-19; SARS-CoV-2; decision tree modeling; machine learning; outcome.

MeSH terms

Adult
COVID-19* / epidemiology
Decision Trees
Humans
Pandemics
Retrospective Studies
Risk Factors
SARS-CoV-2