Info about keep modules

[aboccali]

Hi,

I created a Keep module to select only Alzheimer’s patients with age > 60 years and MMSE < 20. I managed to get the expected population, but I have a question:
Looking at the data, I noticed that each patient has at least one observation with MMSE < 20, but not necessary all (specifically, there are up to 3 MMSE observations in the dementia state machine).
Is it always the most recent MMSE observation that satisfies the condition (< 20), or is it just any one of the available observations?

Another question:
I tried to do the same with hypertension, adding a filter to select patients with certain blood pressure thresholds (SBP > 100 & DBP > 30). However, in this case the Keep module doesn’t work — it doesn’t generate any patients and often throws this error:

ava.lang.NullPointerException: Required observation [system=LOINC, code=8480-6, display=Systolic Blood Pressure, version=null, valueSet=null] is null.

Does this happen because blood pressure data, unlike MMSE, are not created directly in the state machine but instead come from the vital-signs module? Or is there another reason?

Thanks in advance!

[jawalonoski]

The observation checked should always be the latest.

BP may not work because it is a "multi-observation", so SBP and DBP are sub-Observations or components within:

"system": "LOINC",
"code": "85354-9",
"display": "Blood pressure panel with all children optional"

I haven't checked the code enough to zero-in on the issue though.

You might try to use the "Vital Sign" conditional type instead, with the Strings as Systolic Blood Pressure and Diastolic Blood Pressure. You can find the specific String sequences in VitalSign.java.

However, the issue with using the "Vital Sign" conditional type is that the patient with Hypertension may be on medication that is keeping their BP under control (in which case the BP will be normal, and your logic will only pick up those with uncontrolled BP).

So, you're better off just checking to see if they have a hypertension diagnosis, using the "Active Condition" condition type.

"system": "SNOMED-CT",
"code": "59621000",
"display": "Essential hypertension (disorder)"

[aboccali]

Thanks,

there’s another thing I don’t fully understand:
when analyzing the hypertension.json state machine, I see that the encounters where blood pressure is recorded are 3. Therefore, I would expect that in the observations.csv, each patient would have at most 3 blood pressure measurements -- similar to what happens in the Alzheimer module, where each subject has up to 4 MMSE measurements, consistent with the dementia.json state machine -- instead I found patients with more than 10 BP measurements.

Does this difference in behavior occur because blood pressure is a generic vital sign that can also be measured in other pathologies' state machines (unlike MMSE, which is specific to Alzheimer’s), or is there another reason for this?

Thanks.