An important question has been posted by
@Dusty911.
As this is confusing people, let me try to answer it.
Basically hazard ratios are always calculated by comparing two population samples. One is the reference (or denominator if you are mathematically inclined). Thus you always have to carefully look at what the reference population is when interpreting a hazard ratio.
Let me illustrate this with a few examples from the two Valabhji papers,
https://www.england.nhs.uk/publicat...es-and-covid-19-related-mortality-in-england/. In these papers hazard is the risk of in hospital deaths with Covid-19. In the figure on page 21 of the whole population study (paper with 24 pages) it says
female 1.
male 1.94
In this example the reference sample are females. For males the number 1.94 means that the hazard ratio of males is about twice that of females. The hazard ratio for females is of course 1, because you would compare females with themselves.
In the same figure it says
No diabetes 1.
Type 1 3.50
Type 2 2.03
Here the reference sample are people who do not have diabetes. And as discussed in this thread, the overall risk for T1s is 3.5 times that of people without diabetes and for T2s it is a factor 2.
Now let's go to the other paper which studies the population of people with diabetes, see figure on page 20 (paper with 22 pages) where it says
age < 40 0.22
age 40 to 49 0.27
age 50 to 59 0.54
age 60 to 69 1.00
age 70 to 79 1.92
age > 80 4.36
Here the reference population is people with diabetes T2 between ages of 60 and 69. We find that for people with T2 and below 50 the hazard ratio is about a factor of four lower and for people with T2 over 80 the risk is about four times higher than that for those T2s in the reference population (60 to 69). This illustrates how age is the main risk factor. After 50 it doubles every 10 years of age. On page 19 you can find the corresponding results for people with T1. Many of these have large bars through the numbers as there is limited data and the statistical error is large. However, the overall picture is the same. The average risk ratio of a T1 below 50 is at least a factor of fiver lower than a T1 in the reference sample of 60 to 69. In fact, the paper does not report any T1 deaths below 50, so I am wondering myself how these numbers have been calculated. To me they should be upper limits of risk ratios, but the conclusions do not depend on this.
In summary when reading a number always look for the reference population where the hazard risk ratio is set to 1. And make sure you understand what the population is. Is it the full population or only people with diabetes or with T1 or T2?
I hope that helps understanding these numbers. And my usual warning at the end. These are averages, each person is an individual and our risk could vary substantially from these depending on their natural variations which is usually much larger than hazard ratios close to 1 and in particular if there are other factors.