Daily admissions and deaths from heart failure were depicted using a 7-day simple moving average (the mean number of daily deaths for that day and the preceding 6 days) from 1 February to 31 May of each year. Poisson regression was used to estimate the incidence rate ratio (IRR) of daily deaths related to heart failure during the pre-COVID-19 and COVID-19 periods, adjusted for time trends by including a cubic spline function of time. We calculated the excess deaths by subtracting the observed total deaths from March 2020 to the average total deaths in 2018 and 2019 in the same period.

Continuous data were presented as medians with interquartile ranges (IQRs), and categorical variables were presented as counts and proportions. We tested differences between the groups using the Chi-squared test for non-parametric data, Student’s T test for normally distributed continuous variables and Wilcoxon rank sum test where continuous data were not normally distributed.

For the NHFA, we used multiple imputation with chained equations to impute data for all variables with missing information. Multivariable logistic regression models were fit to estimate the risk of death for the COVID-19 and pre-COVID-19 periods. In a sensitivity analysis, we repeated the multivariable logistic regression analysis for 30-day mortality amongst patients discharged alive during each period (Supplementary material online, Method section for more details).

Analyses were performed using the Stata/MP 16.1 statistical software (College Station, TX, USA) and R version 4.0.0. All statistical analyses were two-tailed, and an alpha of 5% used throughout.