Machine learning risk estimation and prediction of death in continuing care facilities using administrative data

Area of Science:

• Public Health
• Epidemiology
• Gerontology

Background:

• The coronavirus disease 2019 (COVID-19) pandemic posed significant risks to vulnerable populations, including continuing care residents.
• Identifying mortality predictors in this demographic is crucial for targeted interventions and resource allocation.

Purpose of the Study:

• To identify factors associated with mortality among continuing care residents in Alberta during the COVID-19 pandemic.
• To evaluate the performance of machine learning models and pre-processing methods for mortality risk prediction.

Main Methods:

• Retrospective cohort study of Alberta continuing care residents (March 2020 - March 2021).
• Linked administrative datasets for comprehensive data analysis.
• Utilized univariable and multivariable logistic regression (LR) to identify predictive factors of 60-day all-cause mortality.
• Employed the Youden index for optimal sensitivity-specificity cut-off determination.
• Developed and compared various machine learning models for performance assessment.

Main Results:

• Increased age, male sex, presenting symptoms, previous admissions, and specific comorbidities were significantly associated with higher mortality.
• Machine learning models demonstrated potential for risk prediction, though further validation is required.

Conclusions:

• Several demographic, clinical, and comorbidity factors predict mortality in continuing care residents during the COVID-19 pandemic.
• Machine learning and advanced pre-processing techniques show promise for enhancing mortality risk prediction beyond traditional methods, warranting further investigation.