Using the example above if the takeoff minima is 200ft and 2km vis, but the only instrument approach is the RNAV which has a minima of 500ft and cloud is OVC003, I wouldn't takeoff until the cloud will allow an instrument approach at the departure aerodrome, or another local aerodrome.
Sounds great in theory. But let's look at say YPDN, a common place for a piston twin to operate. And let's take a pretty descent SEROC of 250fpm. The lowest minima is the ILS (290ft), great we only have to wait for the ceiling to raise 90ft. But the approach starts at 3000ft, so that would take you 12 mins to climb, which 12 mins ~100kts means you need 20nm to climb in. Not too bad as the approach starts 8.5nm away and if you took off rwy 29 as well you can add some track miles pretty easily on the way out there but it will take you 7 minutes to get to the MSA. Although that doesnt take into account loss of performance during turns. Now obviously you can use the MVA and while you are there get ATC to vector you onto the ILS at a closer location, but it's still pretty fu
ed.
I'd prefer the DGA, stay within 3nm (the closest step) and only need to climb to 700ft, that will take you 3 mins, which is 5nm of tracking. Probably add 50ft on to that to provide you a bit of buffer and 2.4km vis so you can do the approach. So now we are looking at 750ft/2.4km vis.
BUT to do that, you need to think about how you will manoeuvre to stay within the 3nm boundary and the sector. So it's a bit more complicated than "whats the approach MDA"? And well worth a serious think about in the flight planning stage.