I beleive this metric should include everything for the host. We are using native VMware metric "maxTotalLatency", its description says that it represent "maximum latency across all disks used by the host".
At the moment we have a Latency Monitor targeted to Datastores, we made it that way to reduce duplicate alarms from both hosts and the datastore.
In order to make a monitor for a specific performance metric, you'll need to do some SCOM XML. I already posted an example of such a single-metric monitor, for VM metrics and it can be easily changed to the host metric. I'll try to create another example for your use case.
Here is a link to VM metric monitor post
I'll post another update with an axample.