Why I am in a higher priority group for a vaccine than younger people with chronic health conditions

The Joint Committee for Vaccines and Immunisation (JCVI) has released its priority list for getting a Covid vaccine, which is as follows:

1. residents in a care home for older adults and their carers
2. all those 80 years of age and over and frontline health and social care workers
3. all those 75 years of age and over
4. all those 70 years of age and over and clinically extremely vulnerable individuals
5. all those 65 years of age and over
6. all individuals aged 16 years to 64 years with underlying health conditions which put them at higher risk of serious disease and mortality
7. all those 60 years of age and over
8. all those 55 years of age and over
9. all those 50 years of age and over

I’m 67 and healthy. So why am I in a higher priority (group 5) than younger people with underlying health conditions (group 6)?

JCVI provide an explanation about why they have chosen to prioritise people in terms of vulnerability, meaning at higher risk of catching and dying from Covid. JCVI provide links to studies on which they base their analysis, and are clear that the risk from Covid is dominated by age, as is the normal risk of dying.

Risk from Covid is a combination of the risk of catching it, and the risk of serious consequences of infected. Considering the second component, being infected by the virus in the first wave led on average to a lethal risk roughly matching the annual risk of dying of something else: as Figure 1 shows, this was suggested back in March, but has been confirmed by more recent analysis based on over 200,000 deaths.

Figure 1 Age-specific mortality rates following infection estimated in March and October 2020, compared with actuarial proportion of people dying each year of any cause

To see how other risk factors compare with age, perhaps it is best to look at the ‘Covid age’ calculator. This expresses an individualised risk of dying if infected in terms of the age of a healthy white man with the same risk, with the analysis based on the Open Safely study that JCVI refer to.

I am a healthy white man, so my ‘Covid age’ is 67, and the calculator says that If infection occurs, the probability that it will be fatal is expected to lie between 4.7 per 1000 and 19 per 1000. A table provides a point estimate of 9.4 per 1000, or around 1% risk of dying if I get infected. (This is slightly lower than the 1.4% average annual risk of death for 67-year-old men, due to me being healthier than average.)

Someones ‘Covid age’ is decreased by 5 years if they are female, and increased for being non-white, for example, being Asian increases it by 5 years.

Clinical conditions would then further increase someone’s Covid age. For example, a 40 year-old white man with Type 2 diabetes, who would be considered clinically vulnerable, would have a Covid age of 40+21 = 61, which is less than me.

The amount that clinical conditions increase your Covid age can depend on your chronological age: for example, a 50-year-old man with Type 2 diabetes would have a Covid age of 50+18 = 68, which is greater than my Covid age and so would give them a marginally higher risk than me of dying if infected. But their Covid age would only be 63 if they were female.

So, the take-home message is that a younger person could have a lower vulnerability than me, even if they had a chronic clinical condition.

My point is not that I deserve to be a higher priority, and I certainly hope that all those in group 6 will get the vaccine at the same time as group 5. However the fact that age is such a dominant factor means that I appear to be at higher risk of death than most younger people, even if they have chronic health conditions.

It’s important to note that clinically extremely vulnerable younger adults will be at higher priority — this includes conditions such as solid organ transplants, cystic fibrosis or Down’s syndrome. JCVI state that “Considering data from the first wave in the UK, the overall risk of mortality for clinically extremely vulnerable younger adults is estimated to be roughly the same as the risk to persons aged 70 to 74 years.” And so they have been placed in group 4, comprising “all those 70 years of age and over and clinically extremely vulnerable individuals”.

But shouldn’t we take into account the number of years of life lost?

This analysis is all based on risk of dying if you get it, regardless of whether someone is 8 or 80. But what if we looked at potential loss of life expectancy IF you catch it?

JCVI have thought of this, and report that “Mathematical modelling indicates that the optimal strategy for minimising future deaths or quality adjusted life year (QALY) losses is to offer vaccination to older age groups first.”

This modelling is available here, and is detailed and rather complex, but we can do some back-of-envelope calculations to get a feel for what is going on.

The UK life tables provide, for each year of age, the average risk of dying each year q, and the average life expectancy LE. We have seen from Figure 1 that, for each age, the average chance of dying, if infected, is roughly q. The life expectancy if infected is therefore q . 0 + (1-q) . LE , assuming that death if infected leads to immediate death, and life expectancy is unchanged if someone survives the infection. The loss in life expectancy due to infection is, under these assumptions, LE - (1-q) LE = q LE, which is easily calculated from the life tables.

Figure 2. Average loss in life expectancy after infection with coronavirus.

Figure 2 shows, for women and men, that the average loss of life expectancy increases steadily with age. Importantly, this means that the ranking of vulnerability is unaffected by considering length of life rather than just risk of immediate death.

The loss in life-expectancy runs between 2 and 8 months for those between 60 and 90 years-old. For 67-year-old men like me getting Covid, their loss of life expectancy is on average around 3 months. Although this is one of those hopelessly unrepresentative averages, like the average number of testicles in the population.

(The fact that the loss in life expectancy q LE is increasing with age follows from the fact that q increases exponentially (ie multiplicatively ) with age, while LE decreases linearly, and exponential wins!)

To get the total expected loss of life-years from, say, 1,000 infections, we need to know the age-distribution of infections. Weekly Public Health England Surveillance reports provide the number of positive tests by age-groups and sex, and so assuming that distribution for those testing positive holds for all infections, we can estimate that the average loss of life expectancy per infection is around 1.2 months, or 0.1 years, or 5 to 6 weeks. This corresponds to the risk faced by someone in their early 50s, which is the average age of infection.

1,000 new infections would therefore, on average, lead to the loss around 100 years of life.

Using the data shown in Figure 1, the recent estimate of the average infection fatality rate is 1.1%, and this matches the expected proportion of deaths using my analysis above. So at an average 1.1% infection fatality rate and 1,000 infections, we would expect around 11 deaths with an average loss of 9 years of life. This closely matches an estimate for the US using a much more complex methodology, but is slightly less than some UK estimates of the years of life lost per COVID-19 death, which were 14 for men and 12 for women. Clearly people who die from Covid were not on death’s door.

Currently (early December), there are perhaps around 30,000 new infections a day in the UK, which would mean around 330 deaths and 3,000 years of life lost.

This is a very rough analysis: the risk of dying if infected is now probably lower than the first wave, due to improved care and possibly lower viral loads due to social distancing, which would reduce the impact of being infected, while there may be long-term harms to survivors, which would tend to increase the impact.

We could go on to work out the average gain from giving someone the vaccine, given different risks of catching the virus, and so assess the cost-effectiveness in terms of individual benefit. Any analysis becomes very value-laden at this point. But the benefit of vaccines goes well beyond the individual, in terms of inducing herd immunity and so benefiting even those who do not take the vaccine, for whatever reason. JCVI may have modelled all of this, and doubtless have concluded that the vaccine is well-within the criteria for a cost-effective intervention.

In conclusion, prioritisation of this eagerly-awaited vaccine is bound to be controversial, but JCVI appear to have taken a reasonable risk-based approach. And their ranking is unaffected by considering length of life.

Added December 7th: my comparator has been changed to someone with diabetes — previously I used someone with a transplant, but they would be on the clinically extremely vulnerable group and therefore would be above me in group 4.