08-31-2017 09:40 AM - edited 08-31-2017 09:41 AM
Currently writing a program to extract suicide attempts from a 2016 outpatient hospital discharge file. In years past, with ICD-9, an e-code of E950-E959 indicated suicide attempts; easy to extract.
Now with ICD-10, there are ~300 codes that could indicate intentional self-harm/suicide. I have an array written to look for these codes over all of the diagnosis fields; however, I am curious to see if there is a way to consolidate the program without writing out all of these codes. The major problem is that the intentional self harm designation is embedded within a range.
For example: The code for Poisoning by heroin, intentional self-harm is T401X2A. But, there are also codes for accidental (T401X1A), assault (T401X3A) and undetermined (T401X4A) poisonings that I don't need, prohibiting any use of if dx in : ('T401X'), which would extract the three codes I don't need.
Multiply this by 300 or so codes, and that becomes a long program. Especially when the previous one was if ecode in : ('E95')
Any work around would be appreciated!
08-31-2017 01:58 PM
If you have a data set (or even just a list) with the codes then create a custom format that will mark those as "Suicide" and every thing else as "Non-suicide" or whatever term you want.
Proc format library=work;
'T401X2A' = 'Suicide'
other = 'Non-Suicide'
A data set can work to make a cntlin data set with the formatname values for start, label and a little extra code to add the "other".
In a data step you could use thing such as
If (put(code,$suicide.) = 'Suicide' then ...
You may search the internet for SAS format ICD-10 as I have heard of number of projects working on bits of this but whether they have your specific need ...