Here is an example of what we've done comparable to this for each of our datacenters in Service Insights (this one shows link health for ALL, but we also have them broken out by DC):
It's primarily captured by naming conventions on the 'ifAlias/ifDescr' of the interfaces (which we needed to add to the _wildalias_ of the interfaces -- It's frustrating that we can't filter by _instance_ properties in this case, which necessitates us making kinda ridiculously long wildaliases, just so we can dynamically keep these services up to date).
In summary, if we have 6 links out of a given datacenter (2 to each of the other datacenters), then not only can we have severity-based alerting on % of links available, but we can also calculate interface utilization as a function of the _newly reduced_ capacity, and alert on that seperately.
One could _theoretically_ reproduce this in a datasource itself using batchscript data collection -- it was just more convenient for us to do it using Service Insights (which can certainly be a little pricey - I've found a way to only create a 'service' for each datacenter though, and have multiple service datasources aligned with each service).