I'm no Cloud advocate, but losing your systems because of cabling damage is not specifically a "cloud" issue. Large organizations (Banks, TfL, Argos et. al.) who run their own DCs have always been in this position since bit barns are typically remotely sited, not squirreled away in office basements.
Your Doctors clinic would be most secure (in this sense) with paper records and failing that, everything on a laptop (because laptops have inbuilt UPS). If they had an on prem server then they would still be vulnerable to a backhoe hitting the power instead of the fiber and would likely take longer to recover from a server failure since they probably don't have on site technicians and hot spares which the remote facility probably does.
If you're worried about all your employees idling because their computer systems won't work then by all means host all servers locally, after you've installed full building UPS and auxiliary generators, hired or trained an IT DR team, purchased relevant hot spares and installed a couple of microwave relays to keep your external connectivity operational if the fiber is down.
That describes my house by the way. In the last three years the UPS has cut in four times and the router switched to 4G cellular backup (two different networks) seven times. The remote stuff I've moved around, but in over a year I've detected no Linode or AWS downtime. If the cloud component goes down, it won't be critical since everything is mirrored from the local NAS, but that is only relevant because the (demonstrably more fragile) local services are secured failure conditions.