o There is an ongoing problem with the diagnosis datatypes.
o All datatypes should be fundamental – meaning that data of a certain datatype should always be of that datatype and should never be able to transition to a different datatype. This should be established as a principle of the QDM. An obvious example is that a medication should never be able to change into an allergy, which of course it cannot. Where a datatype can transition from one type to another, the types should be combined into a single type, and the variable part should become an attribute.
o Where this principle is violated is in the realm of diagnoses. Diagnosis active, Diagnosis inactive, and Diagnosis resolved are independent datatypes. But in actual clinical data, a diagnosis can change states and move from an active state to a resolved state, thus changing datatypes. This should not be allowed.
o The impact of this is the inability of measure authors to specify measures in a meaningful way. Measures refer to “diagnosis, active, starts before time X”, when in fact there are two start times, the time it started, and the time that it became active (these are usually the same). What measure authors really want to say is “the problem was active during time period X”, which they cannot currently do in an unambiguous way.
o The suggested change is that the state of activity of the problem – active, inactive, or resolved, should be an attribute. Furthermore, start datetime and stop datetime should be retired. Instead, there should be a data structure which enables developers to say "was in X status in a given time period".