Should datetime be UTC?

[clermontg]

A argument could be made that all times in the database should be UTC to get rid of the time change issue where we go from UTC -5 to -4, and the local datetimes are duplicated in the dataset, while the opposite happens in the spring. What are folks position on this?

[del42]

Big fan of UTC being used for everything in the database. There are no more headaches from those Daylight Saving Time changes, and no worrying about duplicated or missing times. And if we ever need to display it differently, converting from UTC is a breeze. I say we go for it! @clermontg

[Chesterguan]

+1

[clermontg]

So, a workable solution would look like this.

Sites first need to convert EHR datetimes to UTC. For sites that can upload limited data, this is all they need to do.

For sites that do the time shifting themselves, the TS formula has to time an integral number of days, but within with standard time or DST eras. This is because the timing of events such as morning blood draws would not show a difference in local time but be one hour off one day to the next in a patient's data whose encounter bridges the ST/DST divide. This is a script that could be written centrally and deployed at site.

[del42]

@clermontg Is there a way we can indicate timezone in the care site information. We know our sites will get anonymous care site ids in the central db, in order for models to do cross-sites benchmarking.

[clermontg]

@del42 I am not sure this is needed. This is not related to UTC. There is some need to indicate care_site centrally, for a number of reasons. Characterizing heterogeneity across sites and comparing federating learning to pooled learning, etc. But I think this is something we should discuss at the executive

[del42]

@clermontg I thought it help us preserve the time of the event like morning blood draw you mentioned. I agree we need more discussion. Thanks.

[clermontg]

I thought about this some more and I believe you are correct. Keeping track of TZ info for a site would be very good. It would look strange indeed that morning blood draws would happen at a different UTC time across sites. Same thing for change of nursing shifts and definition of day and night. Is this is a Standards decision or a Tools decision?