Stream: implementers
Topic: JSON response upon a longitudinal/panel data?
lychenus (Nov 17 2020 at 03:36):
I know that there are a lot of example that is built upon a cross-sectional data like blood pressure.
https://www.hl7.org/FHIR/observation-example-bloodpressure.json.html
But can you built it upon a time-series data also? like can i change time in
{
....
"effectiveDateTime": <...>
....
},
And chain all of those JSON together? It seems like I have seen none of the examples in a longitudinal dataset
and we want to, say, chain every data per 5 seconds (i really do not know if they want data so frequent lol).
And is there anyway to simplify the JSON sent after the first one so that the response sent would be a little bit smaller than it should be? (would the json object be so big and it becomes a few hundred Mb to download?)
ty
Grahame Grieve (Nov 17 2020 at 04:23):
look at the SampledData data type. But that has fairly narrow applicability
lychenus (Nov 17 2020 at 04:38):
all out of blue it seems the correct google term should be IOMT FHIR - high frequency data in internet of medical things.
does anyone have more example on how to construct FHIR on that?
Grahame Grieve (Nov 17 2020 at 04:45):
@Dustin Burson ?
Dustin Burson (Nov 17 2020 at 05:04):
@lychenus what specifically are you looking for? Is it just an example of building Sample Data with Blood Pressure? The code we use is available at https://github.com/microsoft/iomt-fhir.
lychenus (Nov 17 2020 at 05:13):
hello,
we are trying to convert apple watch-like sleep data into FHIR format (what for? honestly i do not know). so that data includes SpO2, heart rate, any likely those everyday numbers you can can name. But it is going to last for a whole night so it is a time series data.
but it seems I have got one example that looks useful we can fork on. thank you
https://www.hl7.org/fhir/observation-example-sample-data.json.html
which is on ECG.
we might continue to work on other data for sleep test (PSG), which includes a way bigger data like EEG, EMG, breathe rate all aggregated. again i do not know where it goes. i was just told to do that.
ty
lychenus (Nov 17 2020 at 10:00):
btw it is weird too.
in valueSampledData.data, why not use array but a string with space in between numbers?
Grahame Grieve (Nov 17 2020 at 10:04):
partly because of how it would look in XML.
lychenus (Nov 17 2020 at 10:05):
i do not understand though, array is used in a lot of examples in json, and honestly that should not really affect the presentation in XML, as long as it is well organized
Dustin Burson (Nov 17 2020 at 22:25):
If you are working with Health Kit data you can look at https://github.com/microsoft/healthkit-on-fhir.
If you are looking to store data in a time series format, SampledData is the supported format in FHIR today. It has some limitations which we called out in the presentation you linked. If there is interest in supporting another data type for time series data from devices I would love to start that discussion :)
But for now we are limited to SampledData or defining your own extension to store the data. I think one challenge you will have with devices there really isn't a start/end for the observation. For the IoMT FHIR connector we handle this by bucketing data according to time (most commonly in one hour bounded chunks) and render the data we receive with in those hour boundaries according to the SampledData type.
Last updated: Apr 12 2022 at 19:14 UTC
