A linear mixed model to estimate COVID-19-induced excess mortality

Area of Science:

• Epidemiology
• Biostatistics

Background:

• The COVID-19 pandemic has significantly increased global mortality rates.
• Estimating excess mortality is crucial for understanding the pandemic's true impact, but existing methods for forecasting baseline mortality are often complex and data-intensive.
• Current methods may also be influenced by historical excess mortality, potentially skewing results.

Purpose of the Study:

• To propose and evaluate a novel, user-friendly linear mixed model for estimating excess mortality.
• To address limitations of existing mortality forecasting methods, including complexity, data requirements, and susceptibility to historical biases.
• To improve the accuracy of excess mortality estimation during the COVID-19 pandemic.

Main Methods:

• Development of a linear mixed model designed for ease of application.
• The model utilizes only historical mortality data.
• It incorporates serial correlation and down-weights historical excess mortality influences.
• Model appropriateness was assessed using fit statistics and forecasting accuracy in Belgium and the Netherlands.

Main Results:

• The proposed linear mixed model is easy to apply and requires minimal data.
• It effectively accounts for serial correlation and mitigates the impact of historical excess mortality.
• The model demonstrated improved forecasting accuracy for year-specific mortality compared to the traditional 5-year weekly average method.
• This led to a more precise estimation of excess mortality in Belgium and the Netherlands.

Conclusions:

• The linear mixed model offers a practical and accurate approach to estimating excess mortality.
• It provides a superior alternative to conventional methods for pandemic mortality assessment.
• This method enhances the reliability of epidemiological data during public health crises like COVID-19.