search

h4ste / pronto

Pneumonia Risk predictiON neTwOrk (PRONTO)
MIT License
2 stars 0 forks source link

The ICD9_code in the code is different from the paper. #1

Open Yuz998 opened 4 years ago

Yuz998 commented 4 years ago

Hello, I have a few questions below and I want to ask you for leave.

code in the fetch_cohort_notes.py ： Code

In paper： paper

Problems

  1. The icd9_code in the fetch_cohort_notes.py file starts with 482, but in the paper it is 486. Which one is the main one?

  2. In the material part of the paper, 281,076 patients were obtained in the MIMIC database, but there were only more than 40,000 patients in the MIMIC database. How did these 281,076 patients obtained?

  3. How did the final 1494 patients obtained?

Thank you

h4ste commented 4 years ago

Thank you for the bug report!

  1. I extracted the cohorts in several scripts over many iterations of the project (originally using MIMIC II and then redone with MIMIC-III). I think that is an old script that should probably be removed from the repository. I will investigate and get back to you.

  2. I am not sure where 281,076 comes from, it is definitely not the number of patients; the fact that I said that in the paper is an unfortunate typo on my part. My apologies. I think it may be the number of notes? I will investigate. Regardless, the 1,494 ICU stays and 1,467 patients are the correct and final numbers used for the experiments. I will put a correction on the github page when I figure it out.

  3. As I mentioned before, there were many iterations of this project before the version that was finally published. My recollection is that we (1) identified all patients with an ICD-9 code of 486 (or any subcode 486*); (2) used the patterns in the paper to determine the onset of pneumonia; (3) removed patients whose onset was within the first 48 hours; and (4) removed patients who were coded with pneumonia but for which we were unable to determine the onset.

I will check with the MIMIC folks to see if I can just provide the patient and admission IDs that we used for our cohorts.

Yuz998 commented 4 years ago

Thank you. I look forward to hearing from you again.