The infection fatality rate (IFR) is not always lower than the confirmed case fatality ratio (cCFR or just) you initially estimate for a new disease. Initial estimates can be really wrong in both directions. To see why, note that there's two things you're measuring wrong early on:
• Some people you already know have the disease will die, but haven't died yet.
• Some people have the disease but you don't know they do because it's not causing severe symptoms.
The IFR:
'Number of ever die' / 'Number who ever have the disease'
Or alternatively:
('Already died' + 'Yet to die') / ('Known to have the illness' + 'Not known to have the illness')
Initially you don't know how severe/virulent the disease is in general.
A severe disease means more people who are already known to be sick aren't going to make it. On top of that it means there are few unidentified cases, because the symptoms will usually be very visible.
So higher severity has a powerful impact on the ratio because it pushes up both the numerator, by increasing 'Yet to die', while simultaneously lowering the denominators by decreasing 'Not known the have the illness'.
H1N1 turned out to rarely be severe so lots of most people already diagnosed with it recovered. And lots of people had it who weren't sick enough to report their symptoms or get tested. As a result the IFR ended up being way lower than it initially looked to be.
By contrast, SARS turned out to be severe. Many people already in hospitals later died, and we weren't missing many mild cases. So our final estimate of the IFR was ~10%, higher than our initial CFR (1-10%).
Ebola would have exhibited the same thing early on — almost all cases are counted, and depending on how fast the pandemic was spreading the fatality rate be around ~25% early on, and rise to ~50% over time.
So the CFR to IFR relationship is non-linear with respect to the actual natural severity of the new disease. (See header image)
(A further non-linearity is that if a disease is severe enough it can overwhelm the healthcare system trying to deal with it, raising its fatality rate further still.)
