How to fix labevents and chartevents charttime in metavision patients on MIMIC-III?

[henrypickler]

Hi!

As it is known (per #784 and #1061) in MIMIC-III, for patients in the metavision system, the LABEVENTS' charttime and CHARTEVENTS' charttime are not compatible (one is 1 hour ahead of the other). As suggested by alistairewj, one way to fix this problem is by adding one hour to the LABEVENTS charttime. However, when looking at a patient that has labevents and has both metavision and carevue ITEMIDs (in this case for Bilirubin) this is what we get

Notice how the itemids 848 (Carevue) and 50885 (Labevents) agree with each other on the charttime column, but 225690 (Metavision) is one hour ahead from both.

This begs the question: should we subtract one hour from CHARTEVENTS' charttime on metavision patients, or is it actually that in this subset of patients both carevue and labevents are behind one hour? In isolation this would not really be a problem (you could add one hour or subtract one hour, it would not make a difference), however I need the accurate time for each measurement since we compare these values with the patient's INTIME and OUTTIME.

This result can be reproduced with the query

SELECT * FROM (
  SELECT HADM_ID, ITEMID, CHARTTIME, VALUENUM FROM `physionet-data.mimiciii_clinical.chartevents` AS chart WHERE chart.ITEMID IN (848, 225690)
  UNION ALL (SELECT HADM_ID, ITEMID, CHARTTIME, VALUENUM FROM `physionet-data.mimiciii_clinical.labevents` as lab WHERE lab.ITEMID = 50885)
) WHERE HADM_ID = 161456 ORDER BY CHARTTIME, ITEMID

I'm creating this topic as a discussion since as stated in #784 this issue will not be fixed on MIMIC-III. However, I'm hoping we can get some insight on this issue so that I can fix it.

Thanks!