Materialized view of Angus sepsis

[kseverso]

Adapted MIMIC query for eICU for sepsis based on Angus criteria.

[tompollard]

Great, thanks for making a start on this Kristen. Before we merge the code there are a couple of issues that it would be good to resolve.

The main issue is that ICD9 codes in the eICU database include periods and are recorded in a comma separated list, unlike MIMIC. e.g.:

SELECT icd9code 
FROM diagnosis 
LIMIT 5;

    icd9code    
----------------
 414.00, I25.10
 
 491.20, J44.9
 491.20, J44.9
 428.0, I50.9

This means that matching using methods like substring(icd9code,1,5) IN ('99592','78552') will fail. We could consider restructuring the eICU data in a future version, but in the meantime please could you have think about how this issue could be addressed in the code?

[kseverso]

Thanks Tom. As I understand it, the first entry is the ic9code and then the comma separates the icd10code so I just need the first entry in the list. My change was simply to look at substring(icd9code,1,6) IN ('995.92','785.52') which I think will work but please let me know if I've misunderstood.

[alistairewj]

edit: scratch that, see below

[alistairewj]

Tom says he's found the same but I thought differently so we'll raise this at our next meeting with the eICU folks!

[tompollard]

@kseverso, we've not forgotten about this - it's still under discussion!

[eruca]

any progress? @tompollard @kseverso @alistairewj

[alistairewj]

Ultimately we're not sure this is appropriate. Unlike the name implies, the diagnosis table is more of an active problem list rather than billed diagnosis codes. Using the billing codes from Angus et al. wouldn't really be valid.

[eruca]

any suggestion to find the patients with sepsis in eicu dataset ? Thanks a lot!