TL;DR
This paper investigates the discrepancy between officially reported COVID-19 deaths and actual excess mortality in Norway and Sweden during the first wave of the pandemic. It finds a good agreement between the two measures, suggesting that mortality displacement (where deaths are merely shifted from one period to another) is not a significant factor in explaining the observed excess deaths.
What Problem Does This Paper Address?
The COVID-19 pandemic has sparked intense debate regarding the effectiveness of various mitigation strategies. A key point of contention, particularly in Scandinavia, is how to accurately measure the true mortality burden of the virus and the resulting years of life lost (YLL). Critics of strict non-pharmaceutical interventions (like those in Norway) have sometimes argued that Sweden's more relaxed approach didn't lead to a significantly higher excess mortality rate compared to previous years, suggesting that reported COVID-19 deaths might be inflated or that deaths are simply displaced from other causes. This paper aims to address this by rigorously estimating excess all-cause mortality and YLL, and specifically investigating the concept of "mortality displacement" to see if it can explain any observed differences.
Methods
The authors developed a method to estimate "expected" mortality by accounting for both the long-term declining trend and the seasonal variations observed in mortality data over the past 20 years. This baseline is then compared to observed weekly all-cause mortality to calculate "excess mortality." For Sweden, they further estimated Years of Life Lost (YLL) by considering the life expectancy of different age groups affected by excess deaths.
A novel aspect of their approach is the investigation of "mortality displacement." They used an auto-regressive model (AR1) to analyze the correlation between excess mortality in one year and the next. This helps determine if a high mortality year is followed by a low mortality year, indicating deaths were "displaced" rather than truly additional.
Datasets
- Weekly mortality data: For Norway and Sweden, covering the last 20 years, obtained from Statistics Sweden (SCB) and Statistics Norway (SSB).
- COVID-19 related deaths: Data sourced from Our World in Data.
- Proportion of deaths by age group and life expectancy by age group: For Sweden, obtained from Statistics Sweden (SCB).
Key Insights and Ideas
- Excess mortality closely tracks reported COVID-19 deaths: The study's core finding is that, for the first epidemic wave in both Norway and Sweden, the estimated excess all-cause mortality aligns well with the officially reported COVID-19 deaths. This challenges the notion that reported figures are significantly inflated or that excess deaths are primarily due to other causes.
- Mortality displacement is minimal: The analysis of auto-correlation functions and the AR1 model suggests that mortality displacement is not a major factor. While there might be a slight negative correlation between years, it's not strong enough to significantly alter the estimation of excess deaths or YLL. This implies that the excess deaths observed are largely attributable to the pandemic itself.
- YLL estimates are substantial: Even after accounting for potential displacement, the estimated Years of Life Lost in Sweden due to COVID-19 are significant, indicating a substantial impact on life expectancy.
Key Results
- Excess all-cause mortality (July 2019 - July 2020):
- Norway: 517 (95% CI: 12, 1074)
- Sweden: 4329 (95% CI: 3331, 5325)
- Reported COVID-19 deaths (July 2019 - July 2020):
- Norway: 255
- Sweden: 5741
- Excess deaths (March 11 - November 11, 2020):
- Sweden: 5517 (95% CI: 4701, 6330)
- Reported COVID-19 deaths (March 11 - November 11, 2020):
- Sweden: 6247
- Estimated YLL in Sweden (without displacement adjustment): 45,850 (95% CI: 13,915, 80,276)
- Estimated YLL in Sweden (with displacement adjustment): 43,073 (95% CI: 12,160, 85,451) - a reduction of about 6%.
Limitations and Future Work
- Focus on the first wave: The study explicitly limits its analysis of excess mortality to the first wave (up to November 2020). The authors note that the second wave saw different transmission patterns (more in younger age groups) and potentially more similar mitigation policies between Norway and Sweden, which could yield different results.
- Mortality displacement model: While the AR1 model is used, the authors acknowledge that the maximum likelihood estimator can be biased for short time series. They use bootstrapping to mitigate this, but it's a point of consideration.
- Future work: The authors suggest that further analysis of the second and subsequent waves would be valuable to understand the evolving impact of the pandemic and mitigation strategies. They also highlight the ongoing debate about the trade-offs between YLL and the costs of strong interventions, implying a need for continued quantitative assessment.
Summary updated