Excess mortality during the COVID-19 pandemic

Date of ISAC Approval: 
Lay Summary: 
Tracking of deaths in the population over time (mortality) is an essential part of monitoring and predicting the likely course of the COVID-19 pandemic. The number of deaths attributed to COVID-19 is problematic as likelihood of stating COVID-19 on the death certificate will have changed over time and may have been inconsistently applied across settings. Excess deaths by week from any cause could provide a more objective and comparable way of assessing the scale of the pandemic and formulating lessons to be learned. It is constructed by comparing the deaths over time since the start of the pandemic to the number expected based on the experience of previous non-pandemic years. Crucially this method has as the outcome all-cause mortality data, sidestepping the problems associated with attribution of cause of death to COVID-19. Standalone national mortality data will allow tracking of excess mortality over time, but CPRD uniquely includes timely data on deaths alongside extensive clinical and demographic data, backed up by periodic linkages to national mortality data. This study will exploit this important combination of demographic/clinical and death data to examine how excess mortality during the course of the pandemic differs in individuals with different characteristics (e.g. age, sex, smoking) including pre-existing illnesses (e.g. respiratory disease, heart disease, diabetes). The study will provide key information to inform clinical guidance, resource planning and policies for protecting the most vulnerable.
Technical Summary: 
In recognition of limitations of cause-specific death data, estimates of excess mortality are routinely used to estimate and compare deaths due to seasonal influenza epidemics. These are based on national death registration data which allows for stratification by age, sex and geographic region. We aim to replicate these analyses for the COVID-19 pandemic in the UK with further stratification by COVID-19 effect modifiers that can be measured in primary care data. Time series methods will be used in our primary analyses of observed mortality during the whole period (pre-COVID-19 and during COVID-19) using generalised linear models with a negative binomial error structure. The main outcome is mortality, measured in primary care data using the CPRD derived death date and the main exposure is the pandemic period (1st March 2020 onwards). The model will initially include exposure, age, sex, and terms to capture seasonality and underlying year-on-year trends. Interaction terms will be used to investigate excess mortality during the pandemic according to demographic, lifestyle-related and comorbidity characteristics. Relative and absolute differences in excess mortality will be described. We will compare our primary time series approach to national excess mortality estimates. In a secondary analysis to check conclusions, we will compare weekly mortality with expected mortality based on comparable week-of-year mortality in the years prior to the pandemic (standardised mortality ratio (SMR) approach). We will also explore an approach based on individual-level cohort data, with the pandemic period included as a time-updated variable. In secondary analyses we will use updated linked ONS mortality data when available to repeat the analysis using the ONS death date and HES APC data to improve ascertainment of comorbidities.
Health Outcomes to be Measured: 
All-cause mortality

Krishnan Bhaskaran - Chief Investigator - London School of Hygiene & Tropical Medicine ( LSHTM )
Bianca De Stavola - Collaborator - University College London ( UCL )
Christopher Rentsch - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
David Leon - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Helen Strongman - Corresponding Applicant - London School of Hygiene & Tropical Medicine ( LSHTM )
Ian Douglas - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Liam Smeeth - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )

2011 Rural-Urban (Non-standard) LSOA;HES Admitted;ONS;Patient IMD;Practice Level Carstairs Index