Relational data is organised in tuples. Tuples have n dimensions and can easily describe functional mappings from m-dimensional vectors to scalar or k-dimensional values with a single relation (table). (n = m + k)
If what you need/suggest is for presentation layer only then you can do what Eric Hogue suggests, but if this is somewhere mid process or at the start of another process then Fabrik's suggestion is the proper way to go.
GROUP_CONCAT makes data non relational and hence much harder to use as basis for subsequent processing with SQL. So, the purpose of this command is to make output pertty if your client layer can not do it for you. Do not try to model your core relations according to presentation issues.