[timr]

Table 3 and Figure 2 make it clear that the hazard ratios for patients with hospitalization and/or encephalopathy are multiples higher for those without hospitalization and/or encephalopathy.

Intracranial haemorrhage is 3-4x more likely in hospitalized patients; ~5x higher in those with encephalopathy. For stroke, 2-3x for both. For first mood disorder, 1.5x/2x more likely.

They don't break down the cross-tabs by age (as they should), but given the patterns here, I would expect to see a strong correlation.

The sicker you are, the sicker you are.

[rcpt]

Yes you are right. I looked into the authors research and found a more recent paper that breaks it down by age

https://www.thelancet.com/journals/lanpsy/article/PIIS2215-0...

Figure 3 has risk ratios sliced by age+sex

[timr]

Yes, that paper, I am familiar with. It is...quite poor (see below). But it does clearly show the age-dependence of the things they're observing.

There's far too much to get into in a single comment, but the TL;DR is that they bury a lot of important information in the supplemental materials [1] that make a strong argument that what they're observing are spurious correlations with a third factor. At the very least, this kind of statistical fishing expedition has a high risk of bias, because the researchers know the outcomes ahead of time.

Consider supplementary tables 10 & 11: these show that a large number of the "psychiatric sequelae" are correlated with the control conditions, and the effect of Covid is not significantly different (even where they are, the authors have clearly gone on a fishing expedition, which should make you skeptical).

Figures 16-21 show that the risk of a patient developing Covid is 1.5x greater if they've had a recent psychological illness. This is on par with the risk ratios discussed in the text of the paper, and indicates that the association is not necessarily causative (i.e. it's not clear if Covid causes psychiatric problems or vice-versa).

Figure 22 shows a particularly interesting series of plots, where the diagnosis rates of the control illnesses (broken bones, etc.) are plotted over time. Without exception, everything drops but Covid. It is almost inevitable that if you look at this dataset, you will find an increase in diagnosis of X after Covid...because Covid patients are being seen at much higher rates!

Overall, my interpretation of the supplementary figures is that there was an intense focus on "Covid patients" in 2020, and all other groups stayed out of the doctor's office. Covid patients were showing up in the clinic, so covid patients were the ones being diagnosed with other illnesses.

[1] https://www.thelancet.com/cms/10.1016/S2215-0366(20)30462-4/...