Background: MoMo is a mortality monitoring system that guides public health policy in Spain. The COVID-19 pandemic worsened death notification delays, thus biasing downwards the daily (cumulative) excess mortality estimates produced by MoMo. The goal of this study is to find the best model to correct these estimates for the effect of death notification delays.
Methods: The process followed was: 1) estimates for the excess mortality accumulated in Spain since the beginning of the COVID-19 pandemic are published daily by MoMo and gathered in this study for the period 15/04/2020-25/05/2020. 2) the intensity of daily revisions is computed as the ratio of the estimate published each day divided by the estimate published the day before. 3) Adjusted excess mortality estimates result from applying to these ratios five different correcting models (a simple arithmetic mean or a weighted average, as well as linear, quadratic and cubic regressions). 4) The performance of these corrected estimates is compared with the definite values using the root mean square error (RMSE).
Results: The intensity of daily revisions for the cumulative excess of deaths fell to 1 (no revision) as the publication date left behind the date of death. The correcting estimates based on polynomial regressions reduced the error with respect to the definite observed values by 18-25%.
Conclusions: To improve the validity of the daily estimates for the cumulative excess of deaths from MoMo, it is recommended to correct the notification delay of deaths using polynomial regression models estimated with data on previous revisions.