▪️ 173,248
mentions on death certificates to 24th Dec'21
▪️ 150,154
within 28 days of positive test reported by 9th Jan'22
▪️ 119,600
age standardised excess deaths from Continuous Mortality Investigation (CMI) to 24th Dec'21
1/
1. Mentions on death cert: filled by physicians who knew the patient w/access to medical history. But includes primary cause (1a) and contributing factor or co-existing. So likely overestimate.
Proportion of 1 or 'due to' covid cause has varied over time 80-90% (ONS).
2/
2. Within 28 days of +ve test: includes incidental in period after test and where C19 may contribute but not 1o CoD (NB every hosp adms tested) = overestimate.
However, when test capacity lacking eg 1st wave spring 2020 this will lead to underestimate. (CMI chart for E&W).
Can this Kaplan-Meier curve apply to countries as a whole?
Thought experiment on survivorship bias in South Africa and question of Omicron 'mildness' (as we wait for better data - UKHSA case control/retrospective matched studies in 2-3 weeks)...
1/
Firstly, we need to recognise that the attack rate in RSA is very very high. Modelling from excess deaths, even if adjusting for a higher low income country IFR would place AR at over 85% and closer to 100%.
278k excess deaths = 0.5% of a demographically young population
2/
Ab serosurveys likewise showed 32-63% positivity in Jan'21 across RSA provinces before Beta wave finished or Delta wave arrived.
Projecting forward for further waves and accounting for Ab wane below limit of detection also places AR >95%.
3rd dose immunogenicity v AZ or PF/BNT 2dose primary course
▪️2883 randomised w/ multiple cntrl groups
▪️7 different vax boosters + some 1/2 doses
▪️1/2 age >70
▪️>84 days from 2nd dose PF and >70d for AZ til boost
On point 1. Glad JCVI exist and appraise vaccine/immunology data.
Booster trial effectiveness is based on 6 month gap (may not be same with shorter gap) and immunogenicity studies show longer the gap between dose2 and 3 stronger the immune response.
The most robust data for vax effectiveness waning is from randomised control trials. Pfizer👇
2doses holds up well- but still complete the course and boost!
This is real world data. Prospective randomisation only reduces confounding we face in retrospective observations.
1/
Sources of bias/confounding in observational data (eg UKHSA case control study below):
▪️ accrued natural immunity in control (>45% 🏴 infected) & only a fraction tested
▪️ difficulty of retrospective matching/case controlling (high clinical risk vaxxed first, not just age)
2/
▪️ break throughs disproportionately sample higher risk (immunocompromised, preexisting conditions, etc)
▪️ lower test seeking behaviour in unvaxxed
▪️ pop highest risk of infection (urban, young, ethic minority, high deprivation) also lower vax uptake