Concordance between COVID-19 mortality statistics derived from clinical audit and death certificates in Östergötland county, Sweden

OBJECTIVES: Despite early notions that correct attribution of deaths caused by SARS-CoV-2 infection is critical to the understanding of the COVID-19 pandemic, three years later the accuracy of COVID-19 deaths counts is still contested. We aimed to compare official death statistics with cause-of-death assessments made in a clinical audit routine by experienced physicians having access to the full medical record. STUDY DESIGN: Health service quality evaluation. METHODS: In Östergötland county (pop. 465,000), Sweden, a clinical audit team assessed from the start of the pandemic the cause of death in individuals having deceased after testing positive for SARS-CoV-2. We estimated the concordance between official data on COVID-19 deaths and data from the clinical audit using correlations (r) between the cause-of-death categories and discrepancies between the absolute numbers of categorised deaths. RESULTS: The concordance between the data sources was poor regarding whether COVID-19 was the underlying or a contributing cause of death. Grouping of the causes increased the correlations to acceptable strength. Also including deaths implicated by a positive SARS-CoV-2 test in the clinical categorisation of COVID-19 deaths reduced the difference in absolute number of deaths; with these modifications the concordance was acceptable before the COVID-19 vaccination program was initiated (r=0.97; Symmetric Mean Absolute Percentage Error (SMAPE)=19%), while a difference in the absolute numbers of deaths remained in the vaccination period (r=0.94; SMAPE=35%). CONCLUSIONS: This study highlights that carefulness is warranted when COVID-19 death statistics are used in health service planning and resonates a need for further research on cause-of-death recording methodologies.