The "output" end refers to the cable itself -- the end where the signal exits. This is not absolutely critical, it will still work if you have it backwards, but ideally the shield should be tied to the output side, because that's the zero reference point for the subsequent circuit.
Note that this style of cabling separates the shielding duty and the signal ground reference into separate conductors, each is then optimized for what it's supposed to do -- the ground line serves as a clean zero-reference for the signal, and the shield acts only as a shield and not as a carrier of the zero reference (even though the shield gets connected to the ground line at one end for unbalanced cabling). Contrast this with the usual coaxial cabling (one signal conductor in the center, wrapped by a shield that serves both as a signal ground carrier and as the shield). The latter of course also works, but to me the 2-conductor with shield method is more ideal.