Plus there is inherent biases in datasets. Folks who have interactions with Medicaid will be more vulnerable by definition.

To quote the standard observability conference line "what gets measured gets managed".