ablack3 / icdpicr

Injury code classification in R
https://ablack3.github.io/icdpicr/
Other
5 stars 3 forks source link

ICDPICR does not recognize initial encounter ICD10-CM injury codes ending with "B" or "C" #8

Open aldavidson320 opened 5 years ago

aldavidson320 commented 5 years ago

I am using these tables to help identify trauma ICD codes and their body regions/severity level for each ICD10 code and I notice that some tables with the complete 7 character codes do not include codes that have a 7th character designation of "B" for open fractures. Were these missed or do the tables need to be updated?

aldavidson320 commented 5 years ago

There are also codes with a 7th character of "C" also that are initial encounters for open type of fractures; not all initial encounter codes have "A" as their 7th character. Some S72 and S82 codes for example.

ablack3 commented 5 years ago

Thanks very much for pointing these issues out. These codes were missed and we will add them in the next major update to icdpicr. I'll leave the issue open until it is fixed.

nvpmd commented 4 years ago

Any idea when this bug will be fixed? Some of the following codes are still missing: S02.1 – S02.9 S12.0 – S12.6 S22, S32, S42 S49.0 – S49.1 S52, S62, S72, S82, S92 and S99 (also T30-32 codes don't have any letters at all) I think that's about it.

ablack3 commented 4 years ago

Any idea when this bug will be fixed?

Dr. Clark and I are working on an update to ICDPIC but the COVID19 research has taken precedent at the moment. If I knew the correct AIS and body region that these codes should get mapped to then it would be easy to add them. I think that the reason they are not in ICDPIC is because they did not appear in the training data. Fixing this is definitely on the list but we probably won't get to it until after COVID19 because it involves retraining the model. If you see an easy way to add these codes I'm happy to make the change though.

It does look like these codes are included in idcpic but for some reason on are being mapped properly. I will look into it more this weekend.

ablack3 commented 4 years ago

One issue is that ICD-10 codes are assumed to be 7 characters long.

This code works

df <- data.frame(dx1 = "S02.2XXA")
cat_trauma(df, dx_pre = "dx")

While this does not

df <- data.frame(dx1 = "S02.2")
cat_trauma(df, dx_pre = "dx")

The code that would need to change is here: https://github.com/ablack3/icdpicr/blob/03f3007a131cae2c21985d01c9e571d955f057e0/R/cat_trauma.R#L219

However I'm not sure yet if removing this requirement would cause other issues with ICDPIC.

jessyjkn commented 4 years ago

I wanted to add on to the list of icd10 codes that did not have an AIS calculated: T07 T14 (e.g. T1490, T1491, T148) A couple of T2's (e.g. T24211A, T2652XA, etc...) S20 (e.g. S20352A), S30 (e.g. S30814A), S63 (e.g. S63682A), S85 (e.g. S85802A), S90 (e.g. S90912A), S46 (e.g. S46002A)

I believe there a couple more valid codes that were missing but I wasn't able to test every code.

Additionally, the Healthcare Cost and Utilization Project has a list of valid ICD10 trauma codes. They considered a couple more ICD10 codes for trauma. I will provide the text and link here:

Link: https://www.hcup-us.ahrq.gov/db/vars/siddistnote.jsp?var=i10_multinjury

Text:

"Starting in Data Year 2017 ICD-10-CM Diagnosis Codes Used to Identify Injury Discharges

Below is a list of the ICD-10-CM diagnosis codes used to identify injuries starting in data year 2017, only codes indicating an initial encounter (with a 7th character of A, B, C, or missing) are considered:

It should be noted that ICD-10-CM codes in the range of T36-T50 with a 6th character of 5 or 6 are excluded from the criteria to identify injuries. These codes are specific to adverse effects and underdosing of drugs, medicaments and biological substances."

Thanks Adam for your work on calculating severity scores for ICD10. It's fantastic!

jessyjkn commented 2 years ago

Hello Adam, I wanted to follow up. Most of the codes I mentioned above have had an AIS assigned to them. Thanks for updating the R package!

I found a couple more codes that were not assigned AIS scores (e.g. T24619A and T23639A). Were these codes not assigned because they indicated corrosion?