Patients without an icustay_id

[lingyiyang]

• [X ] Put an X between the brackets on this line if you have done all of the following:
  • Checked the online documentation: https://mimic.physionet.org/about/mimic/
  • Checked that your issue isn't already addressed: https://github.com/MIT-LCP/mimic-code/issues?utf8=%E2%9C%93&q=

Description

There are 46520 patients admitted from the admissions and patients table, but only 46476 in the icustays table. So there is a difference of 44. I thought that MIMIC data are all patients who have stayed at the ICU, why are some patients not assigned an icustay_id?

I tried to investigate this by looking at the information from the admissions table of

SELECT * FROM admissions WHERE subject_id NOT IN 
(SELECT DISTINCT subject_id FROM icustays)

From the results of this query, we see that 20 of the 44 even have chartdata.

Next I looked at the transfers table

SELECT * FROM transfers WHERE subject_id NOT IN 
(SELECT DISTINCT subject_id FROM icustays)

It seems to me that a few of the individuals were never admitted (i.e. they did not have a row which say 'admit' in admissions table), so were they already at the hospital when the data started in be collected for MIMIC? Subject_id 736 seems to be one such individual, they have a lot of data in chartevents which pre-dates the first entry they have on the transfers table. Their first entry in transfers states that they moved out of MICU into a ward, so this would make sense.

However for most (35 of the 44) individuals they have information on their admissions. I proceeded to look at subject_id 3004 who has chartevents as well as a row saying 'admit' in admissions. They stayed in wards 45 and 49 which did not seem to correspond to a cost centre. I extracted the chartevents information we have on them with

SELECT DISTINCT c.itemid, label 
FROM chartevents c 
INNER JOIN d_items d ON c.itemid = d.itemid WHERE subject_id = 3004

which seems to suggest that they only had laboratory events.

Is my assumption correct that some of these patients were already in ICUs at the start of the recording period?
Were there any other particular reasons these 44 individuals are not assigned an icustay_id? Why were individuals such as these included?

[alistairewj]

This is a result of us having to match the hospital level information (admissions, transfers) with the ICU level information (chartevents) using the patient identifiers documented in the ICU system. Essentially these are typos: patients where the patient hospital identifier was input into the ICU system incorrectly, and so we didn't match them up with a hospital stay appropriately.