How to apply CMS HCC Hierarchy / trumping logic in SQL? - sql

Question

I've spent a lot of time trying to see if this exists elsewhere, but unfortunately it doesn't. I think I've solved it, but am looking for any advice on how to make this a bit more elegant/streamlined. Hopefully this will help someone else!
CMS publishes a Risk Adjustment model where diseases are grouped into hierarchies. Only the most severe form of the disease is counted toward a patient's Risk Adjustment Score. CMS does publish the model in SAS, but not in SQL. Every other aspect of the model is straightforward apart from applying the hierarchy / trumping logic below.
There are two tables, one containing member/patient IDs and their hierarchical condition categories (HCCs). The other table is the hierarchy table, where only the most severe form of the HCC is meant to be kept:
Members:
memberId
HCC
A
17
A
18
A
19
B
18
B
19
C
19
Hierarchy
HCC
dropHCC
17
18
17
19
18
19
If a member has 17, 18, and 19, only 17 would be kept as a result. If a member only has 19, then 19 would remain. 17 is considered a more severe form of the condition category which includes 18 and 19, but for scoring purposes we'd only want to count 17.
So, applying the Hierarchy to the Members table, the results should be:
memberId
HCC
A
17
B
18
C
19
As mentioned, I've already solved this. I'm wondering if there are any other ways that are more efficient/elegant?
;with members as (
select 123456 as memberID, 17 as hcc
UNION
select 123456 as memberID, 18 as hcc
UNION
select 123456 as memberID, 19 as hcc
UNION
select 2222222 as memberID, 19 as hcc
UNION
select 9999999 as memberID, 18 as hcc
UNION
select 9999999 as memberID, 19 as hcc
)
, Hierarchy as
(
Select 17 as hcc, 18 as dropHCC, 'diabetes1' as hccCategory
UNION
Select 17 as hcc, 19 as dropHCC, 'diabetes1' as hccCategory
UNION
Select 18 as hcc, 19 as dropHCC, 'diabetes2' as hccCategory
)
select m.*--, h2.dropHCC as hccRemovedBy
from members m
left join(
select m.*, r.drophcc
from members m
inner join ( select memberid, m.hcc, h.drophcc
from members m
inner join hierarchy h on h.hcc = m.hcc) r on
r.memberid = m.memberid
and r.dropHCC = m.hcc) h2
on h2.memberID = m.memberID
and h2.hcc = m.hcc
where h2.dropHCC is null --remove this criteria in the event you want to see what dropped

Answer 1

Why can't you just use MIN(HCC)?
If this is simplified data and the real HCC codes are not integers but are truly hierarchical then you might need to look at a recursive CTE. These can be tricky to write, but basically it is a CTE that has two parts, the first part generates a base dataset (the top of the hierarchy) and then UNIONS a second query that references the CTE you are creating, normally with some sort of additive value (Hierarchy depth or concatenated string). You can then match the correct hierarcy record to the rows in your dataset and use them as a filter (e.g. MIN(HierarchyDepth) or LEFT(HierarchyString, LEN(HCC) = HCC)
Depending on the context of where you use it there are recursion limits. By default these are 100, but you can explicity set an option to change that, BUT NOT IF YOU ARE USING IT IN A TABLE FUNCTION!!!!. It is computationally quite expensive so if the data is slow to change you may want to persist it to the database so it can be indexed.