search

YerevaNN / mimic3-benchmarks

Python suite to construct benchmark machine learning datasets from the MIMIC-III đź’Š clinical database.
https://arxiv.org/abs/1703.07771
MIT License
805 stars 329 forks source link

Question about origin of hcup_ccs_2015_definitions.yaml #115

Closed RJBeetel3 closed 3 years ago

RJBeetel3 commented 3 years ago

I completed a mortality prediction exercise as part of a course a number of years ago. I'm reviewing the project now but can't seem to find the origin of the hcup_ccs_2015_definitions.yaml file that indicates what features to use in the benchmark. Can anyone tell me where these definitions came from?

Thanks!

hrayrhar commented 3 years ago

Hey @RJBeetel3,

If I am not mistaken they come form this page: https://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccsfactsheet.jsp.

The relevant information in our paper is the following paragraph:

In this task we classify 25 conditions that are common in adult ICUs, including 12 critical (and sometimes life-threatening) 
conditions, such as respiratory failure and sepsis; eight chronic conditions that are common comorbidities and risk factors 
in critical care, such as diabetes and metabolic disorders; and five conditions considered “mixed” because they are
recurring or chronic with periodic acute episodes. To identify these conditions, we use the single-level definitions from the 
Health Cost and Utilization (HCUP) Clinical Classification Software (CCS)71. These definitions group ICD-9 billing and 
diagnostic codes into mutually exclusive, largely homogeneous disease categories, reducing some of the noise, 
redundancy, and ambiguity in the original ICD-9 codes. HCUP CCS code groups are used for reporting to state and national 
agencies, so they constitute sensible phenotype labels.

where the reference 71 is Clinical Classifications Software (CCS) for ICD-9-CM fact sheet (Agency for Healthcare Research and Quality, 2012).

RJBeetel3 commented 3 years ago

Hrayr, Thanks very much for your clarification! Best, Rob

On Tue, Jun 1, 2021 at 3:54 PM Hrayr Harutyunyan @.***> wrote:

Hey @RJBeetel3 https://github.com/RJBeetel3,

If I am not mistaken they come form this page: https://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccsfactsheet.jsp.

The relevant information in our paper is the following paragraph:

In this task we classify 25 conditions that are common in adult ICUs, including 12 critical (and sometimes

life-threatening) conditions, such as respiratory failure and sepsis; eight chronic conditions that are common

comorbidities and risk factors in critical care, such as diabetes and metabolic disorders; and fve conditions considered “mixed” because they are recurring or chronic with periodic acute episodes. To identify these conditions, we use the single-level defnitions from the Health Cost and Utilization (HCUP) Clinical Classifcation

Sofware (CCS)71. Tese defnitions group ICD-9 billing and diagnostic codes into mutually exclusive, largely

homogeneous disease categories, reducing some of the noise, redundancy, and ambiguity in the original ICD-9

codes. HCUP CCS code groups are used for reporting to state and national agencies, so they constitute sensible

phenotype labels.

where the reference 71 is Clinical Classifcations Sofware (CCS) for ICD-9-CM fact sheet (Agency for Healthcare Research and Quality, 2012).

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/YerevaNN/mimic3-benchmarks/issues/115#issuecomment-852539172, or unsubscribe https://github.com/notifications/unsubscribe-auth/AE5CV6NHM6KDQLCNJCFBSHLTQVQKHANCNFSM45LC5SFA .