Thursday, May 24, 2012

Dealing with Age in QueryHealth

We discussed some of the challenges around time on the Query Health Technical call today.  We agreed that we needed to look at some specific use cases in order to generate appropriate guidance about how to represent different cases.  One of the cases we needed to look at was age of the patient (which I've extended just a bit to Person for a particular reason).

Age of the Patient with respect to the Measure Period
The most common case in dealing with age is in filtering patients by age for the purpose of computing a measure.  We've already established that the Measure Period defines a default filter for events that are being queried, and that this can be overridden within the various criteria.

A question that results is that when there is a filter on age, does that mean:  Was the patient within this age range at any time during the measure period, or was the patient within this range at the start (or end) of the measure period.

In developing measures, I believe CMS and/or NQF has established the rule that it is the age of the patient at the end of the period (e.g., see NQF Measure 59).

In CCD, a template was constructed to represent Age at Onset which ties the age to the beginning of the time period typically associated with a diagnosis (but could be any observable condition).

The idea described in the Query Health implementation guide is that observations whose effective time overlaps with the measure period are within the scope of the query.

So, start, end, overlap?  Three different ways, and all make sense in some context.  Which should we choose.

Given that the measure period is of a fixed length, you could associated the age criteria at either end without any loss of functionality.  But overlap and begin/end don't work the same way?  Or can they?

Specifying an age criteria that indicates that at some time in the measure period the patient must be at least 18, and must not yet be 75 satisfies the case where the patient must be at least 18 and not older than 75 on December 31st.  But it misses patients who are 75 on January 1st of the year (it catches every other 75-year old).  Changing the observation to <= 75 would catch those, but depending on the precision of the comparison, would let others through.  If the age comparison is done to the precision with which age is specified (years), then it lets through patients who were 75 and some fraction of a year old on January 1st, and will thus be 76 at the end of the measure period.

However, the way that HL7 interprets boundaries is not dependent upon the precision at which the boundary is specified, as Grahame Grieve points out. So, we could address age in during to mean what we want.

If you say in Query Health that you are interested in patients whose age is between 18 and 75 inclusive in the age observation, you will get patients whose age will be at least 18 years, and will be no more than exactly 75 years at the end of the observation period.  So, we can stick with the idea that all acts that overlap with the measure period are within the possible scope of acts to be considered when evaluating the measure, and still deal with age being specified in an "unbound" age criteria, because the default time range is "within the measure period".

Age of a Person with respect to a Specific Event
Which brings us to the next issue.  Often there are cases where you want to suggest a criteria where age is tied to some specific event, such as at diagnosis or condition.  As I mentioned previously, CCD created a template to associate an age with a condition or diagnosis.

Age at onset is simply a special case of age with relationship to a particular event, be it an encounter (e.g., age at admission, age at discharge) or another other event.

One case of special importance that also needs to be considered is that age of another person might also be relevant, especially with respect to family history.  If you have one parent who was diagnosed with diabetes before age 50, your relative risk for the disease greatly increases, and even more so with two parents.  So you might also want to be able to represent age of any other individual as well.

Fortunately, the CCD Age Observation works in both of these cases.  I think want we want to say is that the criteria subject to an age restriction (e.g., an encounter or an observation) is the subject of an age observation criteria element which specifies the age range.  In the case where the age applies to someone other than the patient, the outer observation criteria should indicate who the subject is (e.g., the patient's mother or father), and the inner age observation will be understood via context to apply to the same person.


Post a Comment