I applaud your wanting to learn instead of blindly copying the reference design. I guess I'm the 2% of engineers Kortuk refers to. I may look at a reference design, but I'm not just going to follow it. I try to understand what the important points are and why they did what they did, then make sure to incorporate those if I deem them applicable to my design. Datasheets are often written in isolation, and any real design has other tradeoffs and issues that must be considered. I'd say that 98% of good engineers don't just copy whatever the datasheet examples are. Of course you have to read the datasheet carefully and make sure you understand what they are saying about particular needs of the part.
So to answer your question. The characteristics of audio are that it is low frequency but high signal to noise ratio. That means you don't have to worry about transmission line effects and the like. However, you do have to think about every little place noise can come in and try to prevent it in as many ways as possible. Separate filters on the power to any IC is a good idea, something like a small ferrite in series and a cap to ground right at the chip. This is for anything that doesn't handle the final output power. That needs a low impedance connection to the power supply.
Capacitive coupling onto signal traces has to be considered. This can be handled in routing, and sometimes you route extra ground traces around a signal path just for this reason. Keeping signal nets low impedance helps, but that's not always possible. Keep the sensitive traces away from traces with large voltage swings, like the final power output. Keep the power supply away from signal traces as best you can. Eventually the power supply needs to power the circuitry, but make sure it is well filtered before it gets there to that it it not a source of capacitive noise. In some cases you have to consider inductive coupling between traces, but that's usually not as big a deal as capacitive coupling, especially if you keep the high current final output traces away from the sensitive input traces.
Another source of noise is external coupling from the power line frequency or radio stations. Keeping out power line noise is one of the few places a shield can actually be a good idea. Putting the circuit in a metal box that is tied to the signal ground in one place is a good place to start. Simple R-C low pass filters well above the audio frequency but still well below radio helps keep down radio pickup. For example, one R-C pole in the 50-100kHz region won't effect audio, but will attenuate even AM radio.
There are a lot more details, and there are probably whole books written on this, but this should give you a place to start. A good way to learn is to try these things, then play around with them and see how they effect the output.