Doesn't severe illness usually precede death?
It is reasonable to suspect that the population of people sampled by this study is not actually representative of the whole patient population.
There are a number of ways that the way they found patients who'd died could mislead:
Patients dying at, or beyond their 'normal' time is not newsworthy, and will not be reported as widely.
Carers for older patients may be less engaged online.
Carers for older patients may have died of old age before being contacted.
...
'directionally lower' - 'directional difference' is a statistical test to see if a lower value is real, or might have happened due to factors that are not understood or compensated for.
The study is suggestive that there may be a lower age of death - and it proves that - but only for the patients selected - who may have been selected accidentally in part because they died early.
