Correction to F1156 [B7.Heart Condition] (Final v1.0)

Related Data Product : 

Correction to F1156 [B7. Heart Condition]

1998 HRS Core (Final, Version 1.0)


In the health status respondent-level data file (H98B_R) of the Final Core Data Release Version 1.0, the variable F1156 [B7. HEART CONDITION] has a data error. Specifically, there were a number of AHEAD respondents who reported having heart disease in 1993 or 1995 (and had not disputed the report as of 1995) and were miscoded on the 1998 preload variable for heart disease, F234 (i.e., coded 0 instead of 1 on F234).

The preload variable is what determines which version of the heart disease question (B7) is asked in the current wave. Those coded 1 on the preload variable are simply asked to confirm the prior report of heart disease: "Our records from your last interview in (date) show that you had a heart problem." Unless the R disputes this, the interviewer keys a value of 1 indicating that R still has heart disease. Those coded 0 on the preload variable were instead asked this version of the question: "Since your last interview in (date), has a doctor told you that you have had a heart attack, have coronary heart disease, angina, congestive heart failure, or other heart problems?" If they said no to this question, then they were coded 5 on F1156 and they were not asked any of the heart condition follow-up questions in 1998 (i.e., B7a -- B7q).

To deal with this error, we created a new variable, F1156R, in which the 517 respondents who reported having heart disease in 1993 or 1995 and answered "No" to this question in 1998 were assigned a value of 6 on F1156R. The variables used for this recode are D828 (heart disease status in 1995) and F234 (1998 preload for heart disease). Sample SAS code is provided below. You will want to provide equivalent statements if you are using another statistical software package. 517 cases will change from 5 on F1156 to 6 on F1156R. Frequencies before and after the correction are listed below. Note that this recode will result in a slight overestimate of the number and percent of respondents with heart disease in 1998, because if the 517 respondents had been asked to confirm their prior heart disease, some of them may have disputed it.

If you are interested in using the heart condition follow-up questions in any of your analyses, these are our recommendations: it is unlikely that respondents who had heart disease in the past but no incident report between 1995 and 1998 would respond ‘yes’ to follow-up questions B7d, B7i, B7kb, B7p, and B7q. Thus, respondents with a value of 6 on F1156 can be assigned values of 5 on F1162, F1168, F1171, F1174, and F1175. We do not recommend recoding any of the other follow-up question variables, as it less clear how respondents would have answered those questions.


          Section: B            Level: Respondent         CAI Reference: Q1156
          Type: Numeric         Width: 1                  Decimals: 0

................................................................. ............
Corrected   Uncorrected

   4307        4307     1. YES
    179         179     3. [VOL] DISPUTES PREVIOUS WAVE RECORD
  16368       16885     5. NO
    517                 6. PRELOAD ERROR: Condition reported at prior wave but
                           said no to new event
     13          13     8. DK (don't know); NA (not ascertained)
                        9. RF (refused)
                    Blank. INAP (Inapplicable): [Q456:CS CONTINUE] IS (5);
                           [Q497:CS2b] IS (A) OR [Q542:CS15C2] IS (A)
                           OR [Q518:CS11a] IS (A)


SAS Code for correcting this error:
** Fixing Heart Disease Variable in HRS 98 Core **;

***1995 data with HHID PN and D828;
Data Da95;
  set in95.a95b_r;
  keep hhid pn d828;
Proc Sort out=da95;
          by hhid pn;

***1998 preload data with HHID PN and F234;
Data Da98_pr;
  set in98.h98pr_r;
  keep hhid pn f234;
Proc Sort out=da98_1;
          by hhid pn;

***1998 data with HHID PN and F1156;
Data Da98_b;
  set in98.h98b_r;
  keep hhid pn f1156;
Proc Sort out=da98_2;
          by hhid pn;

***Merge data together using HHID and PN;
***retaining only respondents with data in 1998;
Data All;
  merge da95 da98_1 da98_2 (in=a);
  by hhid pn;
  if a;

  if d828 eq 1 and f234 eq 0 then miscode=1;

  F1156R = .;
    if miscode eq 1 and f1156 eq 5 then F1156R = 6;
    else F1156R = F1156;

***Check frequencies after correction;
Proc Freq data=All;
          tables miscode f1156 f1156r miscode*f1156r f1156*f1156R / nocol                      norow nopercent