Questions about ED_diagnoses and MIMIC_IV_diagnoses_icd

[wyyyzw]

So far, the Note_Table is not publicly available. Thus, we cannot accquire free-text clinical notes for CTPA and echocardiography in MIMIC IV. Then, we find some ICD-9 and ICD-10 code diagnoses of pulmonary embolism (e.g. 41519 and I2699) for the primary diagnosis. Is the diagnosis of pulmonary embolism based on gold standards such as CTPA or other angiography?

[SichengH]

In retrospective studies, people are using ICD codes to identify patient cohorts. The codes themselves don't provide any information on diagnoses.

I would suggest searching the information that you want in other tables such as "chartevents" as well as "procedures_icd" and "procedureevents"

[Taichi-Pink]

Hi @SichengH, would you like to explain detailedly for me why ICD codes cannot be used as diagnoses in MIMIC III? According to https://www.aappublications.org/news/2019/02/06/coding020619, ICD codes could be used to document diagnoses.
Thanks for your help!

[wenhailiu]

Hi @SichengH, would you like to explain detailedly for me why ICD codes cannot be used as diagnoses in MIMIC III? According to https://www.aappublications.org/news/2019/02/06/coding020619, ICD codes could be used to document diagnoses. Thanks for your help!

Hi @SichengH, I'd like to add one more question about the diagnosis information. On the documentation page, they mentioned they did not recommend using ICD codes to stratify patients. Does it mean we can only trust the diagnosis from the free text summary, instead of the ICD codes?

[SichengH]

@Taichi-Pink @wenhailiu ICD codes are billing codes, by definition, are made by the billing department based on information during the hospital stay, more like a summarization. And procedures performed during the actual diagnoses are documented somewhere else.

However, in contrast, the diagnosis column in the admission table is like a first impression, the problem is it is not comprehensive and not using oncology terms.

The original question was asking can ICD-9 and ICD-10 codes presume certain diagnostic procedures. To my best knowledge, the answer is ideally yes and not definitely in practice. However, would you get a decent patient cohort by using ICD codes in a case-control study, sometimes. One counter-example being for sepsis, ICD code for sepsis and sepsis shock include a small portion of patients, therefore we sometimes use criteria such as Angus or Sepsis-III.

My answer above was trying to convey the message that ICD billing codes and the diagnoses procedures are not matching perfectly. If you find it misleading, I apologized. Please let me know if there is any further question about the data.