Background: during 2020, Italy was one of the first nation hit by SARS-CoV-2, but it was not the hardest-hit country in terms of deaths. In absence of the death certificate, the burden of COVID-19 on mortality is usually calculated from overall deaths or from deaths of patients tested positive for COVID-19. However, these measures do not express the real burden of the disease on the population.
Objectives: identify deaths due to or involving COVID-19 in absence of the death certificates.
Design: deaths for all causes, cause-specific deaths, COVID-19 hospitalization and COVID-19 confirmed cases between 01.01.2020 and 31.12.2021 observed in subjects residing in the territory of the ATS of Milan. Potential deaths due to or involving COVID-19 as those occurring in an optimal time period between the date of death and the date of positive swab and/or COVID-19 hospitalization, were identified. Optimal time period was defined maximizing sensitivity and specificity, comparing potential COVID-19 deaths with 2020 cause-specific mortality as gold standard, stratifying results by time of deaths, age, and number of comorbidities. Then, this method was further validated using a time-series approach to estimate the excess mortality during the COVID-19 outbreak in comparison with the pre-outbreak period 2015-2019. Accuracy of predictions was evaluated with the Root Mean Square Error (RMSE) between observed and predicted values.
Setting and participants: 78,202 deaths for all causes, of which 8,815 due to or involving COVID-19 as classified by the Milan Register of Death Causes for 2020.
Main outcome measures: all-cause mortality, cause-specific mortality.
Results: from the beginning of the epidemic, 30% (23,495) died in the first semester of 2020, 26% (19,988) in the second semester of 2020, 23% (18,189) in the first semester of 2021, and 21% (16,530) in the second semester of 2021. COVID-19 hospitalizations were 13.826 (17%), while confirmed COVID-19 cases were 17,548 (22%). The optimal time intervals capable to identify a potential death due to or involving COVID-19 were 0-61 between the date of death and the date of positive swab and 0-11 between the date of death and the date of COVID-19 hospitalization, with an overall sensitivity of 90%, a specificity of 95%, and a RMSE of 3.6. Comparing the method proposed with the time-series approach, a RMSE in 2021 of 15.8 was found. Results showed different optimal time intervals for 2021 vs 2020 and by years of age and comorbidities.
Conclusions: this study found that deaths due to or involving COVID-19 could be sensitively identified from the date of positive swab and/or COVID-19 hospitalization. This method can be used for public health interventions which provided so far measures in terms of total deaths instead of real numbers of COVID-19 death, in particular those involving the effective reproduction number usually calculated from overall mortality.
【저자키워드】 COVID-19, sensitivity, specificity, excess mortality,