AR-DRG 9

[genghiskhanofnz]

Hi there, thanks for providing this R package for us.
Do you develop Australian Revision DRG 9, ICD 10.16?
If not, will you add as functions extending to this package?
Cheers.

[wenjie2wang]

Hi,

Do you develop Australian Revision DRG 9, ICD 10.16?

Could you elaborate what specific feature you are looking for?

[genghiskhanofnz]

If i could see this, then it would be much appreciated. AR-DRG 9 / ICD 10.16 combination is currently used in Australia. e.g. DRG T01A infectious and parasitic diseases with GIs was derived from the principal dx of A41.9 sepsis, unspecified + principal procedure of 38677-00 partial thck exc ventricular myocardium. The screenshot is attached.

I want to see a function like drg_map() where i could plug in arguments like principal dx, principal procedure, decimal format yes or no, etc. That will be brilliant for the Australian users.
I hope you will consider it favourably.
Thanks.

[genghiskhanofnz]

The above message means that something like a grouper software. Thanks.

[genghiskhanofnz]

My request is that drg_map(age, gender, discharge, dx, procedures,...). Therefore, i could audit upcoding and downcoding problems. Thanks.

[wenjie2wang]

Thanks for clarification. I am not familiar with the AR-DRG codes. However, I would like to evaluate how much work it will be to add such a feature to the package. Some follow-up questions are

1. Is the expected output of the drg_map(dx, procedures) the corresponding DRG code? In your example, is the output expected to be T01A?
2. Does the mapping also depend on age, gender and discharge, etc.?
3. Where can one find or formulate such a mapping? Is there any reference material? For example, the touch::icd_map() was built based on the mapping rules provided by CMS.gov

[genghiskhanofnz]

1. Yes, T01A
2. A formula is mentioned in an academic paper page 3 or 601 of this reference which can be viewed on the internet. DRG = G(age; sex; discharge; diagnoses; procedures)
   Luo, W., & Gallagher, M. (2010). Unsupervised DRG upcoding detection in healthcare databases. Paper presented at the 2010 IEEE International Conference on Data Mining Workshops.
3. AR-DRG v10 can be downloaded from this website
   https://www.ihpa.gov.au/what-we-do/ar-drg-classification

Please do not hesitate to contact me for any question. Happy to help.
Thanks for your expertise.

[genghiskhanofnz]

The above weblink states that Episodes of admitted acute care are assigned with disease and intervention codes by health information managers or clinical coders. AR-DRGs are then assigned based on these codes and a number of other routinely collected variables including: age, sex, mode of separation (discharge), length of stay, newborn admission weight and hours of mechanical ventilation.

[wenjie2wang]

Thanks for the reference. I thought the mapping is deterministic similar to the determination of comorbidity or icd_map by GEM. However, after reading the paper by Luo & Gallagher, I no longer think so. Could you confirm if my understanding is correct? If so, I do not think I will be able to implement this feature without training data.

[genghiskhanofnz]

Hi there, it is deterministic = yields only one solution that is DRG
I can give you a training dataset.
So, we can work together.
Cheers and thanks.

[wenjie2wang]

I am not clear about why it is deterministic. Is it possible to get different DRG mapped by using different training sets?

[genghiskhanofnz]

Thanks for your q.
I can provide a sufficient record file from that we could subset many datasets to test whether deterministic or not.
Is it OK with you?