Checking for misdiagnosis using ICD10

[Bader Bin Adwan]

Hi,

Having a large data, I need to check if any of males misdiagnosed by assigned female-specific ICD10 codes and vice versa. I tried but did not figure it out. I am thinking about local command or macros but not sure if it is the right one or how. Is it okay to enter all males codes as a new variable and similarly for females, because when I did it produced false results? or should the list be in the memory (virtual) without including it in the raw data?

Here is a sample of the data:

* Example generated by -dataex-. For more info, type help dataex
clear
input byte Sex str4 Diag1_code
2 "O342"
1 "R104"
1 "K358"
1 "J931"
2 "O249"
1 "I10" 
2 "N202"
1 "I249"
1 "K649"
2 "O269"
2 "O800"
1 "I501"
2 "M480"
2 "C509"
1 "Q211"
1 "A239"
2 "N133"
2 "N939"
2 "O800"
2 "I64" 
2 "C543"
2 "J988"
2 "O800"
1 "N210"
2 "J459"
1 "A239"
2 "O034"
1 "C169"
2 "O800"
2 "R104"
1 "E111"
1 "J189"
2 "J069"
1 "I739"
1 "K409"
1 "K566"
2 "C509"
1 "U071"
1 "E114"
2 "I214"
2 "O800"
2 "A153"
1 "K358"
2 "O441"
2 "C509"
1 "I219"
2 "N309"
1 "F311"
1 "R572"
2 "K358"
end
label values Sex Sex
label def Sex 1 "Male", modify
label def Sex 2 "Female", modify

Here is a sample of the gender-inked codes (true = 700s codes for females and 120s codes for males):

Female_ICD10_Codes Male_ICD10_Codes
A34 B260
B373 C600
C510 C601
C511 C602
C512 C608
C518 C609
C519 C61
C52 C620
C530 C621
C531 C629
C538 C630
C539 C631
C540 C632
C541 C637
C542 C638
C543 C639
C548 D074
C549 D075
C55 D076
C56 D176


Thanks in advance!

[Andrew Musau]

As you define it, a misdiagnosis is a mismatch of "Diag1_code" and "Sex". A merge on these two variables will identify such mismatches.

* Example generated by -dataex-. For more info, type help dataex
clear
input str4(female_codes male_codes)
"A34"  "B260"
"B373" "C600"
"C510" "C601"
"C511" "C602"
"C512" "C608"
"C518" "C609"
"C519" "C61"
"C52"  "C620"
"C530" "C621"
"C531" "C629"
"C538" "C630"
"C539" "C631"
"C540" "C632"
"C541" "C637"
"C542" "C638"
"C543" "C639"
"C548" "D074"
"C549" "D075"
"C55"  "D076"
"C56"  "D176"
end

g id =_n
reshape long @_codes, i(id) j(which) string
*CREATE MISMATCH HERE
gen Sex= cond(which=="female", 1, 2)
rename _codes Diag1_code
contract Diag1_code Sex, nomiss
tempfile misdiagnosis
save `misdiagnosis'

* Example generated by -dataex-. For more info, type help dataex
clear
input byte Sex str4 Diag1_code
1 "B373"
1 "R104"
2 "K358"
1 "J931"
2 "O249"
1 "I10"
2 "N202"
1 "I249"
1 "K649"
2 "C601"
2 "O800"
2 "I64"
1 "C543"
2 "J988"
2 "B260"
1 "N210"
2 "J459"
2 "C639"
2 "O034"
1 "C169"
end
label values Sex Sex
label def Sex 1 "Male", modify
label def Sex 2 "Female", modify
merge m:1 Diag1_code Sex using `misdiagnosis', keep(master match)
g wanted=_merge==3
drop _*

Res.:

. l, sep(0)

     +----------------------------+
     |    Sex   Diag1_~e   wanted |
     |----------------------------|
  1. | Female       B260        1 |
  2. |   Male       B373        1 |
  3. |   Male       C169        0 |
  4. |   Male       C543        1 |
  5. | Female       C601        1 |
  6. | Female       C639        1 |
  7. |   Male        I10        0 |
  8. |   Male       I249        0 |
  9. | Female        I64        0 |
 10. | Female       J459        0 |
 11. |   Male       J931        0 |
 12. | Female       J988        0 |
 13. | Female       K358        0 |
 14. |   Male       K649        0 |
 15. | Female       N202        0 |
 16. |   Male       N210        0 |
 17. | Female       O034        0 |
 18. | Female       O249        0 |
 19. | Female       O800        0 |
 20. |   Male       R104        0 |
     +----------------------------+

.

[Bader Bin Adwan]

As you define it, a misdiagnosis is a mismatch of "Diag1_code" and "Sex". A merge on these two variables will identify such mismatches.

* Example generated by -dataex-. For more info, type help dataex
clear
input str4(female_codes male_codes)
"A34" "B260"
"B373" "C600"
"C510" "C601"
"C511" "C602"
"C512" "C608"
"C518" "C609"
"C519" "C61"
"C52" "C620"
"C530" "C621"
"C531" "C629"
"C538" "C630"
"C539" "C631"
"C540" "C632"
"C541" "C637"
"C542" "C638"
"C543" "C639"
"C548" "D074"
"C549" "D075"
"C55" "D076"
"C56" "D176"
end

g id =_n
reshape long @_codes, i(id) j(which) string
*CREATE MISMATCH HERE
gen Sex= cond(which=="female", 1, 2)
rename _codes Diag1_code
contract Diag1_code Sex, nomiss
tempfile misdiagnosis
save `misdiagnosis'

* Example generated by -dataex-. For more info, type help dataex
clear
input byte Sex str4 Diag1_code
1 "B373"
1 "R104"
2 "K358"
1 "J931"
2 "O249"
1 "I10"
2 "N202"
1 "I249"
1 "K649"
2 "C601"
2 "O800"
2 "I64"
1 "C543"
2 "J988"
2 "B260"
1 "N210"
2 "J459"
2 "C639"
2 "O034"
1 "C169"
end
label values Sex Sex
label def Sex 1 "Male", modify
label def Sex 2 "Female", modify
merge m:1 Diag1_code Sex using `misdiagnosis', keep(master match)
g wanted=_merge==3
drop _*

Res.:

. l, sep(0)

+----------------------------+
| Sex Diag1_~e wanted |
|----------------------------|
1. | Female B260 1 |
2. | Male B373 1 |
3. | Male C169 0 |
4. | Male C543 1 |
5. | Female C601 1 |
6. | Female C639 1 |
7. | Male I10 0 |
8. | Male I249 0 |
9. | Female I64 0 |
10. | Female J459 0 |
11. | Male J931 0 |
12. | Female J988 0 |
13. | Female K358 0 |
14. | Male K649 0 |
15. | Female N202 0 |
16. | Male N210 0 |
17. | Female O034 0 |
18. | Female O249 0 |
19. | Female O800 0 |
20. | Male R104 0 |
+----------------------------+

.

Thanks Andrew but it does not work and gave errors such as xij variables found, also wondering if there is another way because I don't think that it is practical to reshape and merge a huge data, the observations is nearly 200,000 ?!

[Bader Bin Adwan]

Just to explain it further, the male and female specific codes were not originally in the data. I have one variable that contain diagnosis codes in general for both females and males and I want to check if male specific code occur in any females obs, and vice versa, want to check if female specific code occur in any males obs? That is why I though to insert two new variables one include female codes and the second include males codes then do verification, but that does not work either!

I appreciate any suggestions to solve this!

[Bader Bin Adwan]

To make it more clear, Diag1_code is a diagnosis ICD10 code for all OBS (females and males altogether), and there is a 127 codes specific to men diseases and 789 codes specific to females diseases (These are not built in the original data but I have them as a list). Sex value is 1== male and 2== female.

What I want is something like this:

Count or browse if Sex==1 & Diag1_code is one of the 789 females codes.

Similarly:

Count or brows if Sex==2 & Diag1_code is one of the 127 males codes ?

[Bader Bin Adwan]

I found the solution and just want to share it for those who may searching for same problem, use command levelsof then foreach.