BEP010 Medication field

[greydongilmore]

Hi all,

Hoping for some feedback on allowing a medications field in the iEEG json sidecar. Due to the clinical nature of iEEG data, I feel having medication information within the sidecar json would be helpful. Medication can have an impact on the quality of spikes within the iEEG data. The proposed scheme would be:

{
"MedicationsUsed": {
    	"Primidone": {
    		"Dosage": 125
		}
	},
"MedicationUnits": {
    	"Primidone": {
    		"Units": "mg"
		}
	}
}

Thanks! Greydon

[dorahermes]

I think this would be helpful, but this may be a more complex issue. I think this is BIDS general rather than BIDS-iEEG. Here are a couple of issues:

• I am not sure what the frequency of the dosage would be.
• Do the units align with the BIDS units?
• In some cases, we want to know the timing of the medication relative to the measurement. What if a patient had a dose of Ativan a couple of hours ago.

I am not a clinician, we should involve some if we would add this information.

[sappelhoff]

allowing a medications field in the iEEG json sidecar.

This would be backward compatible with the current specification, so it is possible in general terms.

The proposed scheme would be:

Couldn't that be simplified to the following? -->

{
"MedicationsUsed": {
    	"Primidone": {
    		"Dosage": 125,
                "Units": "mg"
		},
        "MyOtherMed": {
                "Dosage": 42,
                "Units": "Kg"
                },
	},
}

Do the units align with the BIDS units?

We generally support all SI units, so "mg" as in the example would certainly be valid, see: appendix V in the spec

I am not sure what the frequency of the dosage would be.

@dorahermes what do you mean by this? --> How frequently the medication is being administered?

In some cases, we want to know the timing of the medication relative to the measurement. What if a patient had a dose of Ativan a couple of hours ago.

perhaps this could still be solved within the JSON file, see this (crude) example:

{
"MedicationsUsed": {
    	"Primidone": {
    		"Dosage": 125,
                "Units": "mg",
                "TimeAdministered": ["1801-05-05T12:04:14"]
		},
        "MyOtherMed": {
                "Dosage": 42,
                "Units": "Kg",
                "TimeAdministered": ["1801-05-05T12:04:14", "1802-05-05T12:04:14", "1803-05-05T12:04:14"]
                },
	},
}

where the "TimeAdministered" field would follow the way we specify dates in BIDS and indicate each timepoint at which at medication was administered.

I am not a clinician, we should involve some if we would add this information.

Me neither :-) ... let me tag some more iEEG people: @bids-standard/raw-electrophys-ieeg

[dorahermes]

@dorahermes what do you mean by this? --> How frequently the medication is being administered?

--> Yes, I meant whether the medication is given with a certain frequency such as 1x or 2x per day or just administered a single time. We work with iEEG patients post-operatively and they have a very complex medication schedule, being on antiepileptic drugs, having surgery with anesthesia, withdrawing some anti-epileptic drugs, readministering some, pain medication etc. I think it is ok to capture a simple case here as long as we also think through the more complex ones such that potential future extensions can be backward compatible (and we are clear about that in the spec).

--> Also, can we specify which word to use for the medication? Brand name or substance name?

[sappelhoff]

I see your concerns ... perhaps this should rather be a small scale "BEP"?

In any case, what is needed are:

1. some real-life examples, going from simple to complex
2. user opinions and a consensus of what's the easiest and best way forward

if a BEP seems too much at this point, I could also imagine that a public, non-formal Google Doc could help with the drafting

@greydongilmore if you are interested in pursuing this, please chime in.

[dorahermes]

+1 on a small scale BEP or non-formal Google Doc to let non-github folks chime in

[robertoostenveld]

At many neuroimaging sites (like the Donders) we also have medication in combination with EEG data (e.g. for ADHD studies) and fMRI data. See for example https://www.eu-aims.eu. I think it would not be smart to represent this linked to a specific modality.

At this moment medication in relation to scans is already presented as example in https://bids-specification.readthedocs.io/en/stable/05-longitudinal-and-multi-site-studies.html. That part of the specification introduces the _sessions.tsv file, which might be used to document medication in other situations as well.

On http://www.fieldtriptoolbox.org/workshop/madrid2019/bids_sedation/ there is an example where sedation is used in an EEG session (sedation levels varied within the session). That data in BIDS format can be explored on ftp://ftp.fieldtriptoolbox.org/pub/fieldtrip/workshop/madrid2019/extra/

[tyarkoni]

As @robertoostenveld notes, there is no current problem representing medication information in a BIDS dataset, because one is free to include any and all relevant variables in scans.tsv, sessions.tsv, or participants.tsv, or for that matter in the phenotype folder as arbitrarily-named .tsv files (with an associated json sidecar).

The question is really just whether a standardized representation offers the community meaningful benefits given the associated costs in terms of increased spec complexity and additional tooling requirements. The primary benefit I can see is in facilitating standardized analyses that automatically take into account medication information. This seems to me like a rather daunting proposal in this case, inasmuch as we probably don't want BIDS to have to include a complete ontology of drugs, units, indications, etc. It's hard to imagine more than a fraction of researchers taking the time to learn the relevant extension, let alone encoding their medication information in that format, let alone constructing analysis pipelines that look for that information in a predictable place. So I guess my own inclination would be to say that this is outside the scope of the spec, and people should feel free to encode such information in whatever way the spec currently affords.

An intermediate solution might be to settle on a soft recommendation (though not a requirement!) that medication information be stored in phenotype/medications.tsv, with each subject a row, and each column a medication (which could be a string like "Primidone 125mg bid"). If it fits comfortably in a single paragraph in the core spec, it might be worth the increase in complexity.

[greydongilmore]

Hi all, My apologies for the delay. I would be more than happy to start a google doc to hash out some of these ideas. I could see there is a growing need for a medication standard as more clinical data is being collected. For instance, our group has two ongoing studies where we are collected 24hr iEEG from epilepsy patients in our unit. Knowing the timing of medication dosages would be helpful during analysis.

I like the temporary solution by @tyarkoni of placing a medications.tsv within the phenotype directory. Could this be a .json instead? I like the look and feel of the proposed structure from @sappelhoff

{
"MedicationsUsed": {
    	"Primidone": {
    		"Dosage": 125,
                "Units": "mg",
                "TimeAdministered": ["1801-05-05T12:04:14"]
		},
        "MyOtherMed": {
                "Dosage": 42,
                "Units": "Kg",
                "TimeAdministered": ["1801-05-05T12:04:14", "1802-05-05T12:04:14", "1803-05-05T12:04:14"]
                },
	},
}

I will put together a draft google doc and provide the link shortly.

Greydon

[CPernet]

+1 for general and not specific +1 @tyarkoni solution: given the flexibility of additional fields, what is the advantage of hard recommendations on medication description

[robertoostenveld]

Hi Greydon

I don’t get enthousiastic from the thought of it being a JSON file, since that can in principle be arbitrarily complex and hence difficult to standardize, unless something like https://www.jsonschemavalidator.net https://www.jsonschemavalidator.net/ were used. A TSV is (due to its constraints) easier to standardize and maintain human and machine readability. In this case it could also be documented as

MedicationsUsed Dosage Units TimeAdministered Primidone 125 mg 1801-05-05T12:04:14 MyOtherMed 42 kg 1801-05-05T12:04:14 MyOtherMed 42 kg 1802-05-05T12:04:14 MyOtherMed 42 kg 1803-05-05T12:04:14 BandAid n/a n/a 1803-05-05T14:23:10

Regarding the phenotype directory, according to https://bids-specification.readthedocs.io/en/stable/03-modality-agnostic-files.html#participants-file https://bids-specification.readthedocs.io/en/stable/03-modality-agnostic-files.html#participants-file I understand that as a directory with one TSV file per tool, and "Each participant needs to be described by one and only one row”. It does not lend itself to repeated administration of medication to a single participant. I would consider it more appropriate to have a medication.tsv file under the sub-XXX directory.

Regarding the (relative) timing of administration versus recording/scanning of brain data: the optional scans.tsv file (see https://bids-specification.readthedocs.io/en/stable/03-modality-agnostic-files.html#scans-file https://bids-specification.readthedocs.io/en/stable/03-modality-agnostic-files.html#scans-file) contains the acquisition start of each recording. Please document in your proposal e.g. whether TimeAdministered would make scans.tsv required, and how to deal with (partial) deidentification of the data by changing/shifting the dates. I think it is largely obvious how it should be done, but nevertheless it would be valuable to be explicit about it.

best Robert

On 17 Sep 2019, at 00:53, Greydon Gilmore [email protected] wrote:

Hi all, My apologies for the delay. I would be more than happy to start a google doc to hash out some of these ideas. I could see there is a growing need for a medication standard as more clinical data is being collected. For instance, our group has two ongoing studies where we are collected 24hr iEEG from epilepsy patients in our unit. Knowing the timing of medication dosages would be helpful during analysis.

I like the temporary solution by @tyarkoni https://github.com/tyarkoni of placing a medications.tsv within the phenotype directory. Could this be a .json instead? I like the look and feel of the proposed structure from @sappelhoff https://github.com/sappelhoff { "MedicationsUsed": { "Primidone": { "Dosage": 125, "Units": "mg", "TimeAdministered": ["1801-05-05T12:04:14"] }, "MyOtherMed": { "Dosage": 42, "Units": "Kg", "TimeAdministered": ["1801-05-05T12:04:14", "1802-05-05T12:04:14", "1803-05-05T12:04:14"] }, }, } I will put together a draft google doc and provide the link shortly.

Greydon

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/bids-standard/bids-specification/issues/319?email_source=notifications&email_token=AAG3PY4RXX2LEWWMRP5PNSTQKAE5PA5CNFSM4IQGSM3KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD62YA5I#issuecomment-531988597, or mute the thread https://github.com/notifications/unsubscribe-auth/AAG3PYZJOT4TRYEKOS4Q6IDQKAE5PANCNFSM4IQGSM3A.