Stream: conformance
Topic: missing Appointment invariant?
Jens Villadsen (Sep 18 2020 at 11:03):
Shouldn't there be an invariant on Appointments that doesn't allow end to be before start?
Jens Villadsen (Sep 18 2020 at 12:08):
@Grahame Grieve is that a thing for PA WG?
Brian Postlethwaite (Sep 18 2020 at 12:09):
Yes it would be, feel free to log it in jira
Jens Villadsen (Sep 18 2020 at 12:10):
its actually messing up some queries/indexes as I've seen some clients actually violate this physically implicit constraint
Jens Villadsen (Sep 18 2020 at 12:15):
done - see https://jira.hl7.org/browse/FHIR-28578
Grahame Grieve (Sep 18 2020 at 12:16):
I have some vague memory that there's some workflows where the end is before the start transiently, but I can't remember why. But is a resource like that technically invalid? I'm not convinced
Jens Villadsen (Sep 18 2020 at 12:19):
i would say yes - it is technically invalid. One could probably argument that start and end can be equal (in some universe), but i cannot see how an end could be before a start
Lloyd McKenzie (Sep 18 2020 at 12:58):
The precision could be different, so the invariant will need to take that into account
Jens Villadsen (Sep 18 2020 at 19:39):
@Lloyd McKenzie the datatype is an instant ... so I can't see precision as a problem
Jens Villadsen (Sep 18 2020 at 19:43):
@Grahame Grieve image.png is that your case?
Lloyd McKenzie (Sep 18 2020 at 20:06):
Instant seems inappropriate - precision is way too high. @Brian Postlethwaite ?
Jens Villadsen (Sep 18 2020 at 20:07):
its been like that since at least DSTU2
Jens Villadsen (Sep 18 2020 at 21:05):
@Lloyd McKenzie change the datatype and I might as well start handing out the pitchforks and torches ;)
Brian Postlethwaite (Sep 19 2020 at 03:51):
You can have your imprecise appointments Lloyd... The rest of us will know when we are expected to arrive.
Jens Villadsen (Sep 19 2020 at 13:14):
This should do it: Appointment.start.exists() and Appointment.end.exists() implies Appointment.start <= Appointment.end
Grahame Grieve (Sep 19 2020 at 21:37):
@Brian Postlethwaite PA doesn't think that patients record appointments for themselves as just the day?
Jens Villadsen (Sep 19 2020 at 21:39):
that sounds like a presentation thingy
Grahame Grieve (Sep 19 2020 at 21:40):
oh? They have to record a full date/time but it isn't shown to them?
Jens Villadsen (Sep 19 2020 at 21:41):
well as long as the field is of type instant I can't see they can get around such precision
Grahame Grieve (Sep 19 2020 at 21:42):
so I think that use case is about more then just presentation
Jens Villadsen (Sep 19 2020 at 21:43):
that might be true - but that goes beyond the issue reported
Jens Villadsen (Sep 19 2020 at 21:44):
in respect to the current model - I would say there is a missing invariant
Jens Villadsen (Sep 19 2020 at 21:44):
if the current model is subject to be changed ... well thats another case then
Brian Postlethwaite (Sep 19 2020 at 23:22):
You can pencil one in using the requestedperiod which can be as fluffy as you like, or put start and end of the day in the values, and set the minutesduration to be 15 mins, which indicates that there's a slot sometime in that range.
Last updated: Apr 12 2022 at 19:14 UTC
