Recoding Multiple Variables and give them consecutive values

[Bader Bin Adwan]

Hello,

I am dealing with ICD10 database for nearly 250K observations, what I would like to do in essence is to classified all the diagnoses into the 21 ICD10 chapters. I have created variables ICD10_Chapter1 - ICD10_Chapter21 and each of these variables has a 0 and 1 values. What I wand to do now is to recode all of them ICD_Chapter* but instead to have a 0 and 1 in the new variable, I want to recode them as follow:
ICD10_Chapter1 = 1
ICD10_Chapter2 = 2
ICD10_Chapter3 = 3
ICD10_Chapter4 = 4
ICD10_Chapter5 = 5
ICD10_Chapter21 = 21

to give you an idea of what I did, I used icd10 gen to code each chapter separately, for example, chapter 11 as this: icd10 gen ICD10_Chapters11 = PrincDiag, range(K00/K93)

I included the main diagnosis codes if someone has better and efficient approach.

* Example generated by -dataex-. For more info, type help dataex
clear
input str4 PrincDiag byte(ICD10_Chapters ICD10_Chapters2 ICD10_Chapters3 ICD10_Chapters4 ICD10_Chapters5 ICD10_Chapters6 ICD10_Chapters7 ICD10_Chapters8 ICD10_Chapters9 ICD10_Chapters10)
"O342" 0 0 0 0 0 0 0 0 0 0
"R104" 0 0 0 0 0 0 0 0 0 0
"K358" 0 0 0 0 0 0 0 0 0 0
"J931" 0 0 0 0 0 0 0 0 0 1
"O249" 0 0 0 0 0 0 0 0 0 0
"I10"  0 0 0 0 0 0 0 0 1 0
"N202" 0 0 0 0 0 0 0 0 0 0
"I249" 0 0 0 0 0 0 0 0 1 0
"K649" 0 0 0 0 0 0 0 0 0 0
"O269" 0 0 0 0 0 0 0 0 0 0
"O800" 0 0 0 0 0 0 0 0 0 0
"I501" 0 0 0 0 0 0 0 0 1 0
"M480" 0 0 0 0 0 0 0 0 0 0
"C509" 0 1 0 0 0 0 0 0 0 0
"Q211" 0 0 0 0 0 0 0 0 0 0
"A239" 1 0 0 0 0 0 0 0 0 0
"N133" 0 0 0 0 0 0 0 0 0 0
"N939" 0 0 0 0 0 0 0 0 0 0
"O800" 0 0 0 0 0 0 0 0 0 0
"I64"  0 0 0 0 0 0 0 0 1 0
"C543" 0 1 0 0 0 0 0 0 0 0
"J988" 0 0 0 0 0 0 0 0 0 1
"O800" 0 0 0 0 0 0 0 0 0 0
"N210" 0 0 0 0 0 0 0 0 0 0
"J459" 0 0 0 0 0 0 0 0 0 1
"A239" 1 0 0 0 0 0 0 0 0 0
"O034" 0 0 0 0 0 0 0 0 0 0
"C169" 0 1 0 0 0 0 0 0 0 0
"O800" 0 0 0 0 0 0 0 0 0 0
"R104" 0 0 0 0 0 0 0 0 0 0
"E111" 0 0 0 1 0 0 0 0 0 0
"J189" 0 0 0 0 0 0 0 0 0 1
"J069" 0 0 0 0 0 0 0 0 0 1
"I739" 0 0 0 0 0 0 0 0 1 0
"K409" 0 0 0 0 0 0 0 0 0 0
"K566" 0 0 0 0 0 0 0 0 0 0
"C509" 0 1 0 0 0 0 0 0 0 0
"U071" 0 0 0 0 0 0 0 0 0 0
"E114" 0 0 0 1 0 0 0 0 0 0
"I214" 0 0 0 0 0 0 0 0 1 0
"O800" 0 0 0 0 0 0 0 0 0 0
"A153" 1 0 0 0 0 0 0 0 0 0
"K358" 0 0 0 0 0 0 0 0 0 0
"O441" 0 0 0 0 0 0 0 0 0 0
"C509" 0 1 0 0 0 0 0 0 0 0
"I219" 0 0 0 0 0 0 0 0 1 0
"N309" 0 0 0 0 0 0 0 0 0 0
"F311" 0 0 0 0 1 0 0 0 0 0
"R572" 0 0 0 0 0 0 0 0 0 0
"K358" 0 0 0 0 0 0 0 0 0 0
end

Thank you

[Nick Cox]

In your data example you don't have any of the variables ICD10_Chapter1 - ICD10_Chapter21

The names on display are ICD10_Chapters ICD10_Chapters2 through to ICD10_Chapters10

Guessing from a mix of what you claim to have and what you show, this might help

rename ICD10_Chapters ICD10_Chapters1 

forval j = 1/21 { 
     replace ICD10_Chapters`j'  = IC10_Chapters`j' * `j' 
}

but make sure you have saved your data before you do this, so that you can back up if this messes up your data.

[Bader Bin Adwan]

Thanks Nick, yea I tried to include all the 21 variables but Stata said too many variables specified that is why I limited it to 10 as an example.

Is there a way to generate a new variable instead of replacing the existing ones because I don't want to mess up with the original data as you suggest.
Also, when I run your command above after renaming the first variable, I got an error message "IC10_Chapters1 not found" although it exist?

[Bader Bin Adwan]

I think I have not explain it well. In each variable of ICD10_Chapters(1-21) there are 0 and 1 values, so

ICD10_Chapters1 has for example 4500=1 and the rest =0
ICD10_Chapters2 has for example 6320=1 and the rest =0
ICD10_Chapters3 has for example 846=1 and the rest =0
ICD10_Chapters4 has for example 1238=1 and the rest =0
ICD10_Chapters5 has for example 794=1 and the rest =0
etc..

I want to generate a new variable and recode all of these variables and give them a consecutive values according to their chapter number, so as above example:

ICD10_ALL_Chapter = 1 if ICD10_Chapters1==1

replace ICD10_ALL_Chapter = 2 if ICD10_Chapters2 ==1

replace ICD10_ALL_Chapter = 3 if ICD10_Chapters3 ==1

replace ICD10_ALL_Chapter = 4 if ICD10_Chapters4 ==1

[Nick Cox]

The error message

IC10_Chapters1 not found 

is the result of a typo on my part. Sorry about that. The code naturally should have been

  
 replace ICD10_Chapters`j'  = ICD10_Chapters`j' * `j'

The request in #4 is different and asks for a single summary variable. But you can use the advice already given in #2 -- use a loop, not a statement for every such variable.

For example,

gen ICD10_ALL_Chapter = 0

forval j = 1/21 {
      replace ICD10_ALL_Chapter = `j'  if ICD10_Chapters`j' ==1
}

However, this summary is of limited use if any patient is 1 on two or more variables. It then becomes the code of the last such variable.

[Bader Bin Adwan]

Works well. It would not because it is impossible to have two or more different primary diagnosis code for one observation. Thank Nick!