search

MIT-LCP / mimic-code

MIMIC Code Repository: Code shared by the research community for the MIMIC family of databases
https://mimic.mit.edu
MIT License
2.61k stars 1.53k forks source link

Rise in vanco level between labs #367

Closed lmockus closed 6 years ago

lmockus commented 6 years ago

There are 500+ instances when vanco levels rise from lab to lab with no vanco administration record in inputevents_mv. I am using the query in https://github.com/MIT-LCP/mimic-code/blob/dedd0faed8273748f0db6034088713d62f226305/notebooks/vancomycin-dosing.ipynb. Some examples are:

SUBJECT_ID, HADM_ID, SOURCE, LABEL, CHARTTIME, ENDTIME, VALUE, VALUEUOM 188, 132401, lab, Vancomycin, 2161-11-04 06:50:00, , 0.0, 0.0, 188, 132401, lab, Vancomycin, 2161-11-05 04:05:00, , 14.8, 14.8, and 99865, 189346, lab, Vancomycin, 2184-12-26 20:31:00, , 18.8, 18.8, 99865, 189346, lab, Vancomycin, 2184-12-28 21:34:00, , 27.0, 27,

I looked into labevents table using the query below and both entries in bold (0ug/mL and 27ug/mL are flagged as abnormal). Does it mean that those entries are considered erroneous and should be ignored or those are good measurements but flagged as abnormally high/low? select le.* from LABEVENTS as le inner join D_LABITEMS di on (le.itemid = di.itemid) inner join VANCO_ADMIN_SUBJ as vas on (le.subject_id=vas.subject_id) where le.SUBJECT_ID=99865 and di.itemid = 51009;

L

alistairewj commented 6 years ago

Abnormal means the lab is considered abnormally high or low. It does not mean it is erroneous.

lmockus commented 6 years ago

Thanks. It implies that some 500 dose administrations are not recorded in inputevents_mv. Is this possible?

L

lmockus commented 6 years ago

Another comment. When fitting the vanco data to one compartmental model I found some strange predictions. By drilling down it became quite obvious that some administrations are missing. By using the query you have developed in https://github.com/MIT-LCP/mimic-code/blob/dedd0faed8273748f0db6034088713d62f226305/notebooks/vancomycin-dosing.ipynb:

SUBJECT_ID, HADM_ID, SOURCE, LABEL, CHARTTIME, ENDTIME, VALUE, VALUEUOM, 40310, 186361, lab, Vancomycin, 2144-08-06 07:42:00, , 9.7, 9.7, , 40310, 186361, poe, Vancomycin, 2144-09-04 00:00:00, 2144-09-28 00:00:00, IV, 1000, , 40310, 186361, poe, Vancomycin, 2144-10-19 00:00:00, 2144-10-24 00:00:00, IV, 1000, , 40310, 186361, lab, Vancomycin, 2144-10-20 18:17:00, , 3.6, 3.6, ,

The level even after 2 months came down from 9.7 to 3.6. There is indication that there was at least one administration based on prescription on 10-19.

Could you suggest any insight how to detect those missing vanco administrations.

L

tompollard commented 6 years ago

It implies that some 500 dose administrations are not recorded in inputevents_mv. Is this possible?

Yes, it is possible that some doses are not captured in the inputevents_mv table. There are several possible explanations , but a key point that seems to have been missed in the examples above is inputevents_mv does not record events outside the ICU.

e.g. Cross referencing the first example with the transfers table (SELECT * FROM transfers WHERE subject_id = 188), the patient was not admitted to the ICU for this particular hospital stay until 2161-12-17.

As he/she was not in the ICU during the period that you are looking at (2161-11-04 - 2161-11-05), we would not expect the medication event to be captured in inputevents_mv. Perhaps the vancomycin was administered during surgery?

It seems to be a similar case for subject_id = 40310. It is possible that the patient received uncharted vancomycin doses between the lab measurements on 2144-08-06 and 2144-10-20 because the patient was not in the ICU for much of this period.

From a quick glance at subject_id = 99865 the reason is less clear, but if you or anyone else chooses to explore in more detail we would welcome additional notes here.

Could you suggest any insight how to detect those missing vanco administrations.

Not currently, although we may supplement MIMIC with non-ICU data in the future.

lmockus commented 6 years ago

Thanks a lot. I guess I understood entry admit in transfer table wrong. It means admission to the hospital, correct?

L

tompollard commented 6 years ago

I guess I understood entry admit in transfer table wrong. It means admission to the hospital, correct?

Yes, that's correct.

chris20881 commented 6 years ago

Sorry, a little late, but this article might be of interest here: "Unexplained increases in serum vancomycin concentration in a morbidly obese patient" http://www.intensivecriticalcarenursing.com/article/S0964-3397(16)30066-0/fulltext Chris

lmockus commented 6 years ago

Chris,

Thanks a lot.

L

On Thu, Feb 1, 2018 at 5:24 PM, Chris Hill notifications@github.com wrote:

Sorry, a little late, but this article might be of interest here: "Unexplained increases in serum vancomycin concentration in a morbidly obese patient" http://www.intensivecriticalcarenursing.com/article/S0964-3397(16) 30066-0/fulltext Chris

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/MIT-LCP/mimic-code/issues/367#issuecomment-362422403, or mute the thread https://github.com/notifications/unsubscribe-auth/Af_3JszGfoZ5nvoo-DNz5nU8qDyNIrLnks5tQjmFgaJpZM4Rc-KK .