As a civil engineer (albeit not a wastewater one), I was thinking about this issue too. On one hand, flushing unused medicine does add to the downstream contamination. But on the other hand, even if it were used it (or its metabolites, which might be equally bad) would still end up being flushed anyway. And even then, medicines aren't the only (or even the largest) problem: there's also pesticides, detergents, etc. to deal with. this study (which, it should be noted, measured streams as opposed to treatment plant outflow, which means some pollution sources were untreated runoff) has this to say:
The most frequently detected chemicals (found in more than half of the streams) were coprostanol (fecal steroid), cholesterol (plant and animal steroid), N-N-diethyltoluamide (insect repellent), caffeine (stimulant), triclosan (antimicrobial disinfectant), tri (2-chloroethyl) phosphate (fire retardant), and 4-nonylphenol (nonionic detergent metabolite). Steroids, nonprescription drugs, and insect repellent were the chemical groups most frequently detected. Detergent metabolites, steroids, and plasticizers generally were measured at the highest concentrations.
IMO, the only complete solution would be to change the treatment criteria on its head: currently, we only even consider treating for substances that are proven to be harmful (i.e. default-allow with a blacklist, to use a computer firewall analogy). Instead, we should switch to a default-deny policy and use better wastewater treatment techniques to remove all non-H2O chemicals from the water. The trouble is, it would be a lot more expensive.