Covid prevalence is more than double it's ever been in London, so any random individual attending hospital is more than twice as likely to have covid than ever before.
However hospital attendance is not random, particularly hospital attendance that requires in-patient care, that's much more likely in the elderly, the very young, the vulnerable and births. Rates in these groups are quite different to the average rate, which much reduces the likelihood of in-patients being coincidental - and even if it is incidental it's likely significant complications.
Most of the incidental admissions (broken ankle) would appear in the in-patient data for at most 1 day - same day or next day discharge - so are unlikely to change the stats much at all, and would be reasonably consistent (nights when there's more drinking would add a few)
It is encouraging that in patients with primary diagnosis is going down, but I think the ratio of admissions to in-patients, and average lengths of stay would be more revealing. Sadly not available though.