Vaccine coverage and denominators II
What is the vaccine coverage by age groups in England?
The concern was the population denominator in England, affecting interpretations of case rates. Many people covered this issue, including James Ward, Colin Angus, and BBC’s More and Less. Last Sunday, I co-wrote a short column in The Observer with David Spiegelhalter.
What is the statistical issue?
The problem is we do not know how many people there are in England. Since we count the number of vaccinated people, the rest are not vaccinated.
There are major differences between two different sources of population figures. The Office for National Statistics produces mid-year population estimates. These figures take last decade’s census and age the population. Statisticians then add births, remove registered deaths, and apply migration estimates.
The effect of migration is uncertain. In previous years, the main source was the International Passenger Survey. Using a survey to track migration meant high uncertainty in those estimates. Due to the pandemic, that survey halted. In response, the ONS developed a model with administrative data inputs.
The other source in England is the National Immunisation Management System. This is a database, which depends on General Practitioner records. Whilst this is useful for tracking local vaccination programmes, there is a challenge.
Some people register in more than one place. Residents may have secondary homes. Students live in both the university city and their home town. Also, other people may emigrate, leaving their GP registration behind. The data notes for the NHS England weekly vaccination reports state:
Coverage can therefore be viewed as being ‘at least’ the figures calculated using the NIMS denominators.
The NIMS count updates between reports. It is a count of NHS registrations, removing those who died. Those differences yield large disparities in population figures by age: