In CMS188v4 (NQF 0147, PN-6) IPP 1 has a curious structure, defining both an Occurrence A and an Occurrence B for an Encounter Inpatient. The two definitions are nearly identical, differing only by "< 120 days" versus "<= 120 days", which itself was presumably an oversight. The two Occurrences are joined by an OR and thus either is sufficient to satisfy the IPP.

From their use in Denominator 1, it appears that these two Occurrences are intended to represent a) inpatient encounters without a preceding ED visit and b) inpatient encounters with a preceding ED visit, though no such distinction can be made from the IPP alone.

In Denominator 1, there are blocks of criteria to check Occurrence A and Occurrence B separately. These blocks too are joined by OR and thus, like the IPP, either is sufficient to satisfy the population.

At a very high level then, the structure of two populations can be viewed as follows:

• IPP:
  • OR: Criteria for Occurrence A
  • OR: Criteria for Occurrence B
• Denominator:
  • OR: Other criteria for Occurrence A
  • OR: Other criteria for Occurrence B

Since disjunctions are used in both populations, it is possible to satisfy each population with separate instances of inpatient encounters, meeting different sets of criteria. For example, the IPP can be satisfied by an encounter that meets the "A" criteria and the Denominator can be satisfied by a different encounter that meets the "B" criteria, despite the fact that that encounter did not meet the IPP's "B" criteria. (It doesn't have to, because the IPP was satisfied by the "A" encounter.) In this situation therefore, the "B" encounter need not fall within the Measurement Period, need not be < 120 days, need not have the requisite diagnoses, and so on. Likewise the "A" encounter, need not meet any requirements of the Denominator since it is satisfied by the "B" encounter.

The two encounters have no relationship to each other (other than not being the same encounter) and yet work together, contrary to the intent, to satisfy different parts of a measure that they would not satisfy independently.

The problem carries forward into the Denominator Exclusions 1 population as well, where even more unintended results are possible. For example, a patient can meet the IPP and Denominator via the "A" criteria, but then would be excluded if the patient ever had a normal CT scan or ever had an inpatient encounter with LOS <= 24 hours because an otherwise unconstrained "B" encounter (that is, without meeting any of the criteria from the IPP or Denominator) could meet any of various Denominator Exclusions criteria.

Similarly, the ED visit is linked only with the "B" encounter, so if the patient met the IPP and Denominator via the "A" route, the Denominator Exclusions criteria would be met if the patient ever had a non-respiratory infection following an ED visit, even though that ED visit is utterly unrelated to the "A" encounter. Other scenarios for unrelated ED visits lead to an exclusion as well.

Finally, Numerator 1 does not reference "Occurrence B" at all so if a patient does manage to get through the IPP and Denominator via the "B" criteria, that encounter has no chance itself of satisfying the Numerator, potentially satisfying it only via an unrelated encounter which, as in previous examples, need not have satisfied the IPP or Denominator criteria.

Population 2 has analogous problems and is further complicated by defects in Denominator 2, described in CQM-1193 and CQM-1196.

This problem is similar to that described in CQM-1155, but the additional logic tied to the Occurrences makes it arguably harder to recognize and resolve.

In summary, using multiple Occurrences across an OR breaks the logic unless further constraints exist to bind the Occurrences to each other (in effect converting the ORs to ANDs) or the criteria are carried along (duplicated) through the populations. The first approach is not applicable here as there is no intent for a relationship between A and B; they are merely intended to represent different, independent scenarios which could satisfy the measure (e.g. with and without a preceding ED visit). The second approach is doable but makes the logic much more complicated. For example, the following high-level structure resolves the problem between the IPP and Denominator:

• IPP:
  • OR: IPP criteria for Occurrence A
  • OR: IPP criteria for Occurrence B
• Denominator:
  • OR:
    • AND: IPP criteria for Occurrence A
    • AND: Other criteria for Occurrence A
  • OR:
    • AND: IPP criteria for Occurrence B
    • AND: Other criteria for Occurrence B

But Denominator Exclusions and subsequent populations must similarly carry forward all of the criteria from all preceding populations, clearly introducing far too much complexity.

Alternatively the problem could be resolved by defining separate populations. The measure currently defines population 1 for patients admitted to ICU and population 2 for patients not admitted to ICU, each with "subpopulations" (the Occurrence A and Occurrence B logic) for preceding ED and no preceding ED. Defining four separate populations (ICU w/ preceding ED, ICU w/o preceding ED, Non-ICU w/ preceding ED, and Non-ICU w/o preceding ED) makes the logic for each population much simpler, though, to be fair, the additional populations introduce complexity of their own.