MIT-LCP / eicu-code

Code and website related to the eICU Collaborative Research Database
https://eicu-crd.mit.edu
MIT License
307 stars 212 forks source link

Finding start and end time for infused drugs from the eICU infusionDrug table #230

Open lasiadhi opened 1 year ago

lasiadhi commented 1 year ago

In the infusion drug table, we have only one timestamp: infusionOffset (see table schema below). I am wondering how to find the start and end offset time for each infused drug in the table.

image

It is not clear if the given infusionOffset is for the start of the infusion or the end of the infusion. Assuming it is the endoffset time, I am wondering how to calculate the start offset using drugRate, infusionRate, drugAmount, or/and volumeOfFluid, etc. Since a patient can have multiple records of the same drug, it is not clear if it is completed/stopped/in progress before the next drug entry. See the example below for the patientUnitStayID = 242380:

image

Please let me know if you have any idea how to use/read this eICU-infusion drug table to find start and end times. Thank you!

jraffa commented 1 year ago

Hi @lasiadhi,

I think we need to do some more work understanding this. @obadawi may be able to provide insight here.

For most of the studies we have conducted thus far we have focused on presence/absence of drugs, but knowing dosage/duration is of course important too. Working through your example, I would read it this way (keep in mind, I could be wrong!):

  1. Heparin appears to be started at 59min and ends 22 hours later. There are 10 hours @ 1000 units/hr, 4h @ 1100, 8h @ 1300. After which it goes to zero. The total amount given almost equals the initial drugamount (24800 vs 25000). The infusionrates are all some multiple of the drug rates, and the weighted sum of the infusion rates is pretty close to the total volumeoffluid.
  2. Nitroglycerin starts and stops at the same times (plus a 946 entry) as heparin. The first six hours have 3 hours @ 10mcg/min and 3h @ 20mcg/min, going to zero after that for a period of time. The drug amount is probably in mg (50mg), and these rates over the six hours equates with 54mg. So the infusion has 'run out of drug' and it's not surprising that it is zero. It appears to be restarted at a lower rate, which I presume means a new vial, and stops a while later without running out (it still has 28mg left, if I had to guess). The drugrates and infusionrates are proportional, but it's unclear to me how it corresponds to the volumeoffluid (perhaps NS is documented in another table?).

The precision of the measurements and offsets is largely going to depend on the way the data is collected. If this is charted data, then it's not surprising that everything is done at hourly intervals, and the numbers only roughly work out. If the data is automatically collected, then the durations are probably much more accurate. AFAIK there is no way to determine which is the case, other than exploring the data and drawing your own inferences.

Indeed, the data is difficult to work with. I would recommend focusing on a single hospital and a single drug first, and then trying to generalize to other hospitals and other drugs.

Jesse

lasiadhi commented 1 year ago

Thank you very much @jraffa for the explanation and for taking the time to do all the calculations! I really appreciate it.

In the Nitroglycerin example, as you mentioned, the patient got 3 hours @ 10mcg/min and 3h @ 20mcg/min. That means the total drug amount given was 5400 mcg within 6 hours. If I am correct, that is equal to 5.4mg (?) (not 54mg). Then I am not sure how to relate that to the 50mg drug amount.

Anyway, I will go through more examples as you recommended. This is very helpful. Thank you, Jesse!

-- Lasith

lasiadhi commented 1 year ago

Hi @jraffa, @alistairewj , @obadawi and all,

If we revisit the Heparin example above, the infusion starts at 59min and ends at 1379min according to the infusionDrug table. However, if we look at the same patient stay in the medication table (see below), the Heparin drug was started at 19min and ended at 1425min. Which offset time should be considered more reliable for determining the start and end of this drug? Are there any other concerns related to time offsets in the medication table for infused drugs? Thank you!

image

obadawi commented 1 year ago

The medications table will represent the "order" as it was verified by the pharmacist. The pharmacists will generally estimate the start time or round to standard times such as 8am and 8pm. For oral medications, this will be the closest assumption you can make as there is no other record of the actual medication administration time. For a continuous infusion like heparin, the nurse will chart the time the medication is actually started in the flowsheet and this should be considered the most accurate time. The provided example seems very reasonable that the pharmacists may have provided an immediate start time while verifying the order at 19 minutes but it took another 40 minutes for the medication to actually be dispensed and initiated for the patient at 59 minutes. Then the heparin was likely decided to be discontinued at the bedside and stopped at 1379 with the pharmacist shortly after verifying the stop order.

On Wed, Sep 6, 2023 at 10:49 AM Lasith Adhikari @.***> wrote:

Hi @jraffa https://github.com/jraffa, @alistairewj https://github.com/alistairewj , @obadawi https://github.com/obadawi and all,

If we revisit the Heparin example above, the infusion starts at 59min and ends at 1379min according to the infusionDrug table. However, if we look at the same patient stay in the medication table (see below), the Heparin drug was started at 19min and ended at 1425min. Which offset time should be considered more reliable for determining the start and end of this drug? Are there any other concerns related to time offsets in the medication table for infused drugs? Thank you!

[image: image] https://user-images.githubusercontent.com/20522961/266094027-1bb68379-ca27-402b-ac8d-ba65bcb7e9e2.png

— Reply to this email directly, view it on GitHub https://github.com/MIT-LCP/eicu-code/issues/230#issuecomment-1708835919, or unsubscribe https://github.com/notifications/unsubscribe-auth/AD6USVFQADUJLN5CWJQHILDXZCZSXANCNFSM6AAAAAA32FE4VM . You are receiving this because you were mentioned.Message ID: @.***>

lasiadhi commented 12 months ago

Thanks, Omar (@obadawi) for the explanation! This is very helpful to understand the big picture here and how these tables are linked to each other. I believe the last Heparin 5000 UNIT/ML order [not cancelled] is not in the infusionDrug table because it was not continuously administered (?). Thanks again! -- Lasith