It's an observational study. MOST observational study conclusions end up failing to be reproduced in prospective studies - for reasons you point out.
The end point was 30 day mortality. Readmit rates were unchanged. The latter is typically a better discriminator for 'bad care' - if the patient has to come back to the hospital in a month for the same reason, you can argue that you should have done something different (doesn't always work, but it seems to be a valid generality).
30 day mortality suggests (but does not prove) that a strong possibility is that older doctors where less likely to be aggressive. That can be appropriate or inappropriate care depending on the patient and condition.
All in all, this is *very* preliminary (as the authors duly point out). Unfortunately, it seems to be on the road to the the Medical News de jour and will be plastered about the Internet as the unvarnished truth.