Derivation of a Contextually-Appropriate COVID-19 Mortality Scale for Low-Resource Settings

Area of Science:

• Global Health
• Infectious Diseases
• Medical Informatics

Background:

• COVID-19 continues to impact low- and middle-income countries due to delayed vaccinations and underdeveloped healthcare systems.
• Mortality scales are crucial for resource allocation in low-resource settings (LRS) to prioritize care for high-risk patients.
• Existing COVID-19 prognostication tools lack validation in LRS.

Purpose of the Study:

• To develop a pragmatic tool for LRS frontline providers to assess in-hospital COVID-19 mortality risk.
• To utilize easily obtainable demographic and clinical data for the tool.
• To create a contextually relevant mortality index for COVID-19 patients in LRS.

Main Methods:

• Machine learning applied to a retrospective cohort of 467 Sudanese COVID-19 patients.
• Data from two government referral hospitals used for model derivation.
• Assessment of mortality indices using C-statistics.

Main Results:

• Two versions of the AFEM COVID-19 Mortality Scale (AFEM-CMS) were derived.
• The scale uses readily available inputs like age, sex, comorbidities, Glasgow Coma Scale, respiratory rate, and blood pressure.
• Models showed good discrimination, with C-statistics of 0.775 (with pulse oximetry) and 0.719 (without pulse oximetry).

Conclusions:

• The AFEM-CMS offers a practical solution for resource allocation in LRS during the ongoing pandemic.
• The tool aids frontline providers in identifying patients at highest risk of in-hospital mortality.
• Further validation is recommended before wider implementation due to likely narrow generalisability outside of similar LRS.