Code for categorizing ICD-9 and ICD-10 data

[Sara Glick]

I am analyzing death certificate data that switched from using ICD-9 to ICD-10 codes in 1999, and I want to categorize these data by: a) broad categories (e.g., cancer, heart disease) and b) a few more specific causes of death (e.g., HIV/AIDS). I found the ICD9 program in Stata which is perfect for those data, but it looks like there is not an analagous ICD10 program.

Is anyone familiar with a workaround for the lack of an ICD10 program? Alternatively, is there publicly available Stata code for categorizing ICD9 and ICD10 data? Or even better, code that can merge the two types of ICD data?

[Joe Canner]

Sara,

I'm pretty sure Stata is working on an icd10 command, or at least they were seriously thinking about it as of last summer. With the official changeover (in the US) from ICD9 to ICD10 coming this October, I'm sure the clamor for it will increase. I'm not aware of any other user-written efforts in this area, but I haven''t used ICD10 codes much yet. It would be a massive undertaking, so I'd be surprised if someone went ahead and did it, knowing that it would be superseded by Stata eventually.

Regarding categorization of ICD-9 codes, you can use Clinical Classification Software (CCS), free from the government. It's not actually software per se (or maybe it was at one time), just a couple of spreadsheets and some Stata code. See here: https://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp

For converting between ICD9 and ICD10, try the CDC General Equivalence Mapping (GEM) files: http://www.cdc.gov/nchs/icd/icd10cm.htm

Regards,
Joe

[Dave Airey]

I thought the US government postponed ICD10 (again) until 10/2015? Stata may not do anything about it until version 14. You might want to add your request to another recent thread on desired features of v14, aside from using the links above.

[Joe Canner]

I hadn't heard that, but it looks like you're right. I'm not sure whether Stata can do anything about it or not before version 14, but I would hope that they do it as soon as possible regardless of the US government's plans, because other countries use it and it is used in other contexts (e.g., death certificates).

[Tom Weichle]

Yes, I agree with Joe. Since Stata is used internationally and other countries have implemented ICD-10 long ago, it would be worthwhile for Stata to work on this sooner than later. The delay of the ICD-10 compliance date in the U.S. to not before 10/1/2015 will allow for more training and it would be nice to practice using an ICD-10 command in preparation for the transition.

[Matthew Barclay]

Just to add my voice in support of an ICD10 command; in England ICD10 has been used for many years, and while I generally get by doing things manually without too much time and effort, having a program to do this for me would certainly make things easier, particularly when investigating common comorbidities or causes of death.

[Rebecca Pope (StataCorp)]

Sara asked about an icd10 command in Stata and sparked a lively discussion here on the forum. We've been having lively internal discussions too.

We have wanted to develop an icd10 command like icd9 for a while now*. While it is true that U.S. implementation of ICD-10 in clinical settings has been postponed, this is not the cause of icd10 development delays. We have been held up by bureaucratic licensing issues that we did not encounter when we developed icd9.

The ICD-10 coding system is copyrighted by the World Health Organization (WHO); beginning with ICD-10, they are much more enthusiastic about their copyright. We are now, and have been for some time, in talks with WHO to obtain a license that would allow us to distribute the ICD-10 codes and their descriptions. This is a critical component of developing an icd10 command that functions exactly as the icd9 command that you're familiar with.

At this point, understanding the inner workings of icd9 may help clarify potential losses, so for those of you not familiar with what goes on behind the scenes... icd9 relies on a Stata dataset that is distributed with Stata. The codes in that dataset are the ICD-9-CM codes from the National Center for Health Statistics (NCHS) at the U.S. Centers for Disease Control and Prevention. Each time that NCHS updates the codes, we update the dataset used by icd9 and push that out to you via update. You are thus saved the hassle of checking for updates yourself, ensuring that you have correctly marked which codes are no longer valid, verifying that your colleagues are using the same set of codes to check their data, etc.

With this distributed dataset, icd9 is able to perform what I (speaking here as a former health researcher) think are some of its most useful features. For example, when you submit

icd9 check dx1

icd9 runs a merge in the background and looks for codes in your dataset that do not appear in the master list distributed with Stata. A similar process happens when you use

icd9 generate descr1 = dx1, description

As Matthew pointed out, this is certainly something you could do yourself, but a command that handles the merge for you is much more convenient, not least because you don't have to remember where you saved the file or to consistently name your diagnosis variable to match what is in the using dataset (the complete list of codes and descriptions). The latter I found particularly helpful when working across data gathered from multiple sources and with multiple diagnosis codes that had to be checked in each file. Yes, you can have a series of rename commands, but icd9 spoiled me. (StataCorp employee hat back on now)

Thus far, we have been unable to obtain a license that would enable icd10 to provide the same level of convenience that icd9 does because of WHO's objections to our distribution of the dataset of codes and descriptions. At this point, we are no longer hopeful that we will be able to do so. This means that a future icd10 command cannot simply mimic the icd9 command that exists today. While talks with WHO continue, our development efforts are now focused on alternatives that would provide as much of the functionality of icd9 as possible.

Best,
Rebecca

* This all applies equally to icd9p and a new icd10p.

[Jannik]

It seems very odd that WHO in this way is preventing important public health research.

Which part of WHO is responsible for this decision? Anything that we as users could do, e,g petitions or contact with our national WHO members?

Jannik

[Joe Canner]

Here is a good place to start for understanding WHO's ICD-10 licensing policy: http://www.who.int/about/licensing/classifications/en/

Apparently, there is a non-commercial research license available that might be more flexible than the commercial version that Stata must obtain and abide by. While waiting for Stata to plot their course, perhaps someone clever can figure out how to create a Stata module that utilizes the non-commercial version in a way that mimics the current icd9 functionality.

[Beth Lawson]

Do you still have any issues with converting icd 9 codes to icd 10 codes just look for major sources. If you don't crawl information, I will help you with your issues whatever you want to solve. Latest icd-9 to icd-10 codes 2015 resources and information will help you.