TL;DR
During Hong Kong’s massive 2022 Omicron wave, excess mortality (13,518 deaths) closely tracked but slightly exceeded laboratory-confirmed COVID-19 deaths (12,228). This high level of concordance suggests that while Hong Kong’s robust testing captured the vast majority of direct COVID-19 fatalities, the sheer scale of the epidemic likely caused additional "indirect" deaths due to healthcare system overload.
What Problem Does This Paper Address?
Counting "confirmed" COVID-19 deaths often underestimates the true toll of a pandemic. Official counts can miss cases due to limited testing capacity, misdiagnosis, or "indirect" effects—such as patients with heart disease or cancer dying because hospitals are too overwhelmed to provide standard care.
Hong Kong presents a unique case study: it successfully maintained a "Zero-COVID" elimination strategy through 2020 and 2021, only to be hit by a massive Omicron surge in early 2022. The authors sought to determine the true mortality burden of this surge by calculating excess mortality—the difference between the number of people who died in 2022 and the number expected to die based on historical trends.
Methods
The researchers employed negative binomial regression analysis to model weekly mortality data from 2010 to 2021. This model allowed them to project a "counterfactual baseline"—essentially, what mortality would have looked like in 2022 if the pandemic hadn't occurred.
Key features of their approach included:
- Seasonality: Using Fourier terms to account for the predictable annual peaks and troughs in death rates (e.g., winter flu seasons).
- Long-term Trends: Accounting for Hong Kong’s aging population and changing health trends over a decade.
- Categorization: Analyzing data by age, sex, and specific causes of death (using ICD-10 codes) to see which populations and health conditions were most impacted.
Datasets
The study utilized two primary high-quality datasets from the Hong Kong government:
- Mortality Data (2010–2022): Individual-level death records from the Census and Statistics Department, covering all-cause and cause-specific deaths.
- COVID-19 Clinical Data: Laboratory-confirmed SARS-CoV-2 death records from the Hospital Authority and the Centre for Health Protection.
- Population Data: Annual mid-year population statistics to calculate mortality risks per 100,000 people.
Key Insights and Ideas
- High Testing Accuracy: Unlike many other regions where excess deaths were 2 to 3 times higher than confirmed COVID-19 deaths, Hong Kong’s numbers were remarkably close. This suggests their massive testing infrastructure (testing ~3% of the population daily at the peak) captured most direct infections.
- The "Healthcare Overload" Gap: The largest discrepancy between excess and confirmed deaths occurred during the peak of the wave (February–April 2022). The authors suggest this "extra" mortality (847 deaths beyond the confirmed count) likely reflects the system's inability to cope with the sudden influx of patients.
- Mortality Displacement: Interestingly, deaths from malignant neoplasms (cancer) actually decreased slightly among older adults in 2022. This suggests a "displacement" effect: individuals who might have died from cancer later in the year may have instead succumbed to COVID-19 earlier, or the virus "competed" with other causes of death.
- Age is the Primary Driver: 90% of all excess deaths occurred in individuals aged 65 and older, with those 80+ bearing the heaviest burden.
Key Results
- Total Excess Deaths: Estimated at 13,518 (95% CI: 13,403–13,633).
- Confirmed COVID-19 Deaths: Recorded at 12,228.
- Mortality Spike: All-cause mortality in 2022 rose to 864 per 100,000 people—a 27% increase over the predicted baseline.
- Cause of Death: 78% of excess deaths were attributed to respiratory causes, while 10% were cardiovascular.
- Demographics: Males showed higher excess mortality (232 per 100,000) compared to females (143 per 100,000).
Limitations and Future Work
- Direct vs. Indirect: The study cannot perfectly disentangle deaths caused directly by the virus from those caused by healthcare disruptions (e.g., a heart attack patient not getting a bed).
- Small Sample Sizes: Accuracy for younger age groups (e.g., children 5–14) was limited by the very low number of weekly deaths in those cohorts.
- Baseline Complexity: The model used 2020–2021 data in its baseline, which were "atypical" years due to social distancing and the absence of influenza. While the authors adjusted for this, it remains a point of statistical complexity.
- Future Directions: The authors suggest future research should focus on developing methods to more clearly define and separate the "indirect" impacts of pandemic-era healthcare strain.
Summary updated