1

I am trying to reproduce the results given in the paper The Artificial Intelligence Clinician learns optimal treatment strategies for sepsis in intensive care.

A MIMIC-3 dataset has been used to generate an AI clinician to suggest a treatment for sepsis patients. The authors shared some patient identifiers under Supplementary Data 1 part at the end.
At the beginning I thought these were subject_id 's of the patients since they fall in the same range but although the authors shared 17083 id 's, only 7934 of them matched a subject_id in MIMIC-3. Moreover, under Supplementary Text and Figures part at the top of Figure 1, the total number of patients without any elimination is given as 61532 which matches the number of rows in ICUSTAYS table.

The problem is, the icustay_id's within the table are within a different range compared to the shared id's but I believe there must still be a connection between two. So if anyone tried to reproduce the results before, could you please tell me what the shared id's correspond to?
I also tried eliminating the patients with the criteria they have given but still I cannot obtain the same numbers.

0

I ran into the same issue. The identifiers represent ICUSTAY_IDs. The problem is that in the database, all the ICUSTAY_IDs have a two at the start to indicate that they're an ICUSTAY_ID (the HADM_IDs have a 1 to identify them etc,) but the AI Clinician paper removed the 2 from their list. Add 200000 to every row in the identifier list and you should find them all in the database.

Hope this helps.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.