I can think of two reasons why I'd prefer to not do that:

1 - It requires another computer to gateway the feed from the camera. More stuff to setup. More stuff to go wrong.

2 - A transcode pass is from H264 to NDI is not going to help the image quality from the camera. OK, probably won't hurt it much, but it seems unnecessary.

I wonder if there's another way to access the audio stream?

The surveillance cameras that I use are from Grandstream. They're a little unusual in that they support SIP connectivity. That makes them a lot like an IP phone or a video conference end-point.

Using such cameras you can get the video stream via RTSP as has been described. You can also get the audio stream via a SIP "phone call" that can be added into vMix separately, if required.

In truth, you can get V+A via a SIP call, but that would need to be gatewayed into vMix in some fashion, so it's not that useful.

I do like that I can "call" one of the cameras from the Polycom phone on my desk, see the stream and hear who is at the gate.