LawInfo Forum

Medical Outcomes Study: Validation of Items Pertaining to Cerebrovascular Accidents

Lewis O. Amack

Rand Corporation

May 14, 1989


MEDICAL OUTCOMES STUDY

Health Examination Medical History Interview

Section IX (Neurological)

SUBJECT(s): Cerebrovascular accidents

Author: Lewis O. Amack

Date: March 15, 1989

The general style of this documentation is:

A) VARIABLE: Name of scale variable

B) DESCRIPTION: Summary definition of scale variable

C) DERIVATION: Name(s) and exact wording of each item in Medical History Interview used in constructing scale variable

D) PORTION OF STATA PROGRAM USED TO GENERATE VARIABLE

E) CODES: Numerical and verbal description of each of the possible values of the scale variable

F) RESPONSE: Tabulation of the scale variable

G) PROBLEMS: Addressing coding concerns, including:

(1) potential improvements in the STATA program;

(2) strategy for interpretation of missing or wild values;

(3) strategy for interpretation of conflicting evidence.

H) VALIDATION: Analysis of the effectiveness of the scale in measuring health status, including: (1) potential changes in the variable to enhance its prediction of illness (or and (2) the relationship of the scale variable to other scales, particularly key scales used as dependent variables.

I) RELATIONSHIP OF SCALE LEVELS TO DEMOGRAPHICS AND HEALTH:

Omitted

To test the utility of the scale variable as an indicator of health status, the scale is tabulated with four indices of general health. In addition, the scale is regressed against draft several background variables.

NEUROLOGICAL SECTION: OVERVIEW OF SCALES AND MARGINALS

VARIABLE: dxstroke

DESCRIPTION: History of CVA

DERIVATION: ch1cva1 ("Have you ever had a stroke?"), and (if "yes" response to ch1cva1), ch1cva2 ("Do you have any residual effects?"),

[ch1cvay (If yes to ch1cva1 ("In what...year?")].

PORTION OF STATA PROGRAM USED TO GENERATE VARIABLE:

. replace ch1cva1 = 2 if (ch1cva1~=1 & (ch1cva2~=. | (ch1cvay~=. & ch1cvay~=0)))
. replace ch1cva2 =. if (ch1cva2<1 | ch1cva2>2)
. replace ch1cva1 = 2 if (ch1cva2~=. | ch1cvay~=.)
. gen int dxstroke = (ch1cva1==2 | ch1cva2~=. | (ch1cvay~=. & ch1cvay~=0))
. replace dxstroke = 2 if ch1cva2==2
. label variable dxstroke "History of stroke"

CODES: 0: Never had a stroke

1: History of a stroke, but no residual effects

2: History of a stroke with residual effects

RESPONSE:

. tab dxstroke
History of|
     stroke|       Freq.     Percent       Cum.
-------------------------------------------------------
          0    |       2492       96.48        96.48
          1    |         51        1.97        98.45
          2    |         40        1.55       100.00
------------------------------------------------------
      Total    |       2583      100.00

PROBLEMS: The 3 missing values of ch1cva1 probably do not affect validity, since each of them had negative responses to both ch1sync1 and ch1diz1, and therefore probably would have responded "No" to ch1cva1. However, there are 9 troubling cases in which there was an affirmative response to ch1cva1 and no recorded response for ch1cva2. Of these 9, 4 had no response to any of the other neurological items, and consequently there is no facile method by which to predict what their responses might have been for ch1cva2. Another 4 cases each indicated that they had experienced one fainting spell during the past 6 months, as well as dizzy spells, but at a frequency of less than once per month. It is not unreasonable to assume that these individuals, and the individual who indicated that he had dizzy spells "about once a month" but no fainting spells, might have responded "No" to ch1cva2.

Even with complete data on all neurological items, however, it would be difficult to impute a value for ch1cva2 where ch1cva1=2 and ch1cva2=., since ch1sync1 and ch1sync2 do not correlate significantly with ch1cva2.

Furthermore, it is not appropriate to make inferences based upon ch1cvam or ch1cvay, since there may be no residual effects from a cerebrovascular accident which occurred as recently as one day prior to the interview.

. tab ch1cva1, missing
Ever had a|
       stroke   |      Freq.     Percent        Cum.
---------------------------------------------------------
          1     |       2489       96.36         96.36
          2     |         91        3.52         99.88
          .     |          3        0.12        100.00
----------------|------------------------------------------
      Total     |       2583      100.00
. tab ch1cva2, missing
Stroke--residual|
effects               |      Freq.     Percent        Cum.
-----------------------------------------------------------------
          1              |          42          1.63           1.63
          2              |          40          1.55           3.17
          .              |        2501         96.83         100.00
----------------------------------------------------------------
      Total           |       2583     100.00
. tab ch1sync1 ch1diz1 if (ch1cva1==. & ch1cva2==.), missing
     Past 6     | Past 6 mos--had dizzy spells
   mos--had     |
   fainting     |
      spell     |         1      Total
--------------------------------------
       1        |         3 |         3
--------------------------------------
       Total    |         3 |         3
. tab ch1sync2 ch1diz2 if (ch1cva1==2 & ch1cva2==.), missing
     Past 6      | Past 6 mos--freq. of dizziness
 mos--freq.      |
of fainting      |         1          4          .      Total
---------------------------------------------------------
         1       |         4          0          0 |          4
         .       |         0          1          4 |          5
---------------------------------------------------------
      Total      |         4          1          4 |          9
. tab ch1sync1 ch1diz1 if (ch1cva1==2 & ch1cva2==. & ch1sync2==. & ch1diz2==.),
>  missing
     Past 6    | Past 6 mos--had dizzy spells
   mos--had    |
   fainting    |
      spell    |         1          .      Total
------------------------------------------------
         .     |         1          3 |         4
------------------------------------------------
      Total    |         1          3 |         4
. tab ch1sync1 ch1diz1 if (ch1cva1==2 & ch1cva2==. & ch1diz2==4), missing
     Past 6    | Past 6 mos--had dizzy spells
   mos--had    |
   fainting    |
      spell    |         2      Total
-------------------------------------
         1     |         1 |         1
-------------------------------------
      Total    |         1 |         1
VALIDATION:
. tab dxstroke, summ(sxsync)
  History of|      Summary of Fainting spells
      stroke  |    Mean          Std. Dev.        Freq.
----------------------------------------------------------------
          0      |   .03210273       .22794685        2492
          1      |   .11764706       .38194548          51
          2      |   .075            .34990841          40
----------------------------------------------------------------
      Total      |   .03445606       .23445754        2583
. tab dxstroke, summ(sxdizzy)
  History of|       Summary of Dizzy spells
      stroke |     Mean      Std. Dev.      Freq.
-------------------------------------------------------------
          0      |    .4394061   1.0681045      2492
          1      |   1.1764706   1.669801         51
          2      |    .85        1.3310879        40
------------------------------------------------------------
      Total      |    .46031746  1.0928403      2583

See Appendix (Correlation Matrix) for further validation data.

VARIABLE: sxsync

DESCRIPTION: Frequency of fainting spells during the past 6 months

DERIVATION: ch1sync1 ("In the past 6 months, have you had fainting spells?"), and (if "Yes" to ch1sync1), ch1sync2 ("How often did you have fainting spells in the past 6 months: Once (=1), Two to four times (=2), or Five times or more (=3)?").

PORTION OF STATA PROGRAM USED TO GENERATE VARIABLE:

. replace ch1sync1 = 2 if (ch1sync1~=1 & ch1sync2~=.)
. replace ch1sync2 =. if (ch1sync2<1 | ch1sync2>3)
. replace ch1sync1 = 2 if ch1sync2~=.
. replace ch1sync1 = 1 if (ch1sync1~=2 & ch1sync1~=.)
. gen int sxsync = ch1sync1==2
. replace sxsync = ch1sync2 if ch1sync2>=1 & ch1sync2<=3
. label variable sxsync "Fainting spells"
CODES:  0:  No fainting spells during the preceding 6 months
        1:  One fainting spell during the past 6 months
        2:  Two to four fainting spells during the past 6 months
        3:  Five or more fainting spells during the past 6 months
RESPONSE:
. tab sxsync
    Fainting|
      spells  |      Freq.      Percent          Cum.
-----------------------------------------------------------
       0      |       2519       97.52           97.52
       1      |         42        1.63           99.15
       2      |         19        0.74           99.88
       3      |          3        0.12          100.00
----------------------------------------------------------
      Total   |       2583      100.00

PROBLEMS: There are 10 cases in which ch1sync1 has a missing value. Of these, 5 have missing values for all other neurological variables, and are the most difficult to assess, although the number of cases is small enough that an imputation that sxsync = 0 will probably not affect measurement validity substantially. Another 2 cases indicated having dizzy spells less than once per month, but left ch1sync1 blank. Two cases had a stroke with no residual effects, and neither responded "Yes" to ch1diz1. The only troubling case is a respondent with residual effects from a stroke who indicated that he was not suffering from dizzy spells. That case, as well as the previous 4 cited cases, can all reasonably be imputed as sxsync = 0.

. tab ch1sync1, missing
      Past 6   |
    mos--had   |
    fainting   |
       spell   |      Freq.    Percent       Cum.
----------------------------------------------------------
          1    |      2509       97.14        97.14
          2    |        64        2.48        99.61
          .    |        10        0.39       100.00
----------------------------------------------------------
      Total    |       2583    100.00
. tab ch1sync2, missing
      Past 6   |
  mos--freq.   |
 of fainting   |    Freq.    Percent     Cum.
------------+---------------------------------------------
          1    |      41        1.59          1.59
          2    |      19        0.74          2.32
          3    |       3        0.12          2.44
          .    |    2520       97.56        100.00
---------------------------------------------------------
      Total    |    2583      100.00
. tab ch1diz2 ch1cva2 if (ch1sync1==. & ch1sync2==.), missing
     Past 6   | Stroke--residual effects
 mos--freq.   |
         of   |
  dizziness   |         1          2          .      Total
---------------------------------------------------------
         5    |         0          0          2 |         2
         .    |         2          1          5 |         8
---------------------------------------------------------
      Total   |         2          1          7 |       10
. tab ch1diz1 ch1diz2 if (ch1sync1==. & ch1cva2==1), missing
     Past 6    | Past 6 mos--freq. of dizziness
   mos--had    |
      dizzy    |
     spells    |         .      Total
-------------------------------------
         1     |         1 |         1
         .     |         1 |         1
-------------------------------------
      Total    |         2 |         2
. tab ch1diz2 ch1cva1 if (ch1sync1==2 & ch1sync2==.), missing
     Past 6   | Ever had a stroke
 mos--freq.   |
         of   |
  dizziness   |         1      Total
-------------------------------------
         2    |         1 |        1
-------------------------------------
      Total   |         1 |        1
. tab ch1diz1 ch1diz2 if (ch1sync1==. & ch1cva2==2), missing
     Past 6    | Past 6 mos--freq. of dizziness
   mos--had    |
      dizzy    |
     spells    |         .      Total
-----------------------------------------------
         1     |         1 |       1
------------------------------------------------
      Total    |         1 |       1
VALIDATION:
. tab sxsync, summ(sxdizzy)
    Fainting    |       Summary of Dizzy spells
  spells    |       Mean          Std. Dev.      Freq.
--------------------------------------------------------------
          0     |     .42477173      1.0445772       2519
          1     |    1.5238095       1.864143          42
          2     |    2.4210526       1.6437014         19
          3     |    3               1                  3
--------------------------------------------------------------
      Total     |     .46031746      1.0928403       2583
. tab sxsync, summ(dxstroke)
    Fainting|     Summary of History of stroke
      spells  |     Mean       Std. Dev.       Freq.
--------------------------------------------------------------
          0       |   .04843192   .27620147        2519
          1       |   .14285714   .41739194          42
          2       |   .15789474   .50145986          19
          3       |           0           0           3
-------------------------------------------------------------
      Total       |   .05071622   .28132982        2583

See Appendix (Correlation Matrix) for further validation data.

VARIABLE: sxdizzy

DESCRIPTION: Frequency of dizzy spells during the past 6 months

DERIVATION: ch1diz1 ("In the past 6 months, have you had dizzy spells?"), and (if "Yes" to ch1diz1), ch1diz2 ("How often did you have dizzy spells: Every day (=1), About three times a week (=2), About once a week (=3), About once a month (=4), or Less than once a month (=5)?").

PORTION OF STATA PROGRAM USED TO GENERATE VARIABLE:

. replace ch1diz1 = 2 if (ch1diz1~=1 & ch1diz2~=.)
. replace ch1diz2 =. if (ch1diz2<1 | ch1diz2>5)
. replace ch1diz1 = 2 if ch1diz2~=.
. replace ch1diz1 = 1 if (ch1diz1~=2 & ch1diz1~=.)
. gen int sxdizzy = ch1diz1==2
. replace sxdizzy=6-ch1diz2 if ch1diz2>=1 & ch1diz2<=5
. label variable sxdizzy "Dizzy spells"
CODES:  0:  No fainting spells during the past 6 months
        1:  Fainting spells less than once per month
        2:  Fainting spells about once per month
        3:  Fainting spells about once per week
        4:  Fainting spells about three times a week
        5:  Fainting spells every day
RESPONSE:
. tab sxdizzy
Dizzy spells  |      Freq.     Percent        Cum.
---------------------------------------------------------
     0        |       2079       80.49        80.49
     1        |        169        6.54        87.03
     2        |        131        5.07        92.10
     3        |         98        3.79        95.90
     4        |         66        2.56        98.45
     5        |         40        1.55       100.00
----------------------------------------------------------
     Total    |       2583      100.00

PROBLEMS: There are only 4 missing values for ch1diz1. Yet the coding of this data set poses some difficult problems, since all four of these cases were coded as having a history of a stroke. Of these 4, 3 have missing values for ch1cva2. The other case indicated that there were no residual effects (i.e., ch1cva2 = 1). Apparently, coding ch1cva1=2 for wild values of ch1cva2 was inappropriate, since 4 such cases had faulty or insufficient evidence throughout the neurological section.

. tab ch1diz1, missing
      Past 6|
    mos--had|
dizzy spells|      Freq.     Percent         Cum.
-----------------------------------------------------------
   1        |       2075       80.33         80.33
   2        |        504       19.51         99.85
   .        |          4        0.15        100.00
-----------------------------------------------------------
  Total     |       2583      100.00
. tab ch1diz2, missing
   ast 6    |
 mos--freq. |
of dizziness|    --  Freq.     Percent        Cum.
------------------------------------------------------------
    1 ------|         40  ---      1.55     ---   1.55
    2 ------|         66  ---      2.56   ---     4.10
    3 ------|         98    ---    3.79   ---     7.90
    4 ------|        131   --     5.07   --    12.97
    5 ------|        160    --    6.19   --    19.16
    . ------|       2088       80.84    -  100.00
------------------------------------------------------------
      Total     |       2583     100.00
. list ch1sync1 ch1sync2 ch1diz1 ch1diz2 ch1cva1 ch1cva2 if (ch1cva2<1 | ch1cva
> 2>2) & ch1cva2~=.
         ch1sync1  ch1sync2   ch1diz1   ch1diz2   ch1cva1   ch1cva2
  599.        .                .                 .               .                .                4
1390.        .                .                 5              .                9               3
1909.        .                .                 .               .                .                4
2484.        .                .                 .               .                .                3
. list ch1sync1 ch1sync2 ch1diz1 ch1diz2 ch1cva1 ch1cva2 if ch1cva1~=1 & ch1cva
> 2==.
         ch1sync1  ch1sync2   ch1diz1   ch1diz2   ch1cva1   ch1cva2
    48.        1               .                1               .                .                .
  224.        1               .                1               .                .                .
1202.        .               1                .               1                .                .
1682.        .               1                .               1                .                .
1948.        .               1                .               1                .                .
2273.        1               .                1               .                .                .
2538.        .               1                .               1                .                .
VALIDATION:
. tab sxdizzy, summ(sxsync)
                |Summary of Fainting spells
Dizzy spells|   Mean         Std. Dev.      Freq.
------------------------------------------------------------
              0 |   .01298701   .12534716       2079
              1 |   .07100592   .33760297        169
              2 |   .07633588   .36412848        131
              3 |   .18367347   .58056032         98
              4 |   .1969697    .58755548         66
              5 |   .225        .53048116         40
-----------------------------------------------------------
          Total |   .03445606   .23445754       2583
. tab sxdizzy, summ(dxstroke)
                     |Summary of History of stroke
Dizzy spells|        Mean        Std. Dev.       Freq.
-----------------------------------------------------------------
          0        |   .03896104   .24997031         2079
          1        |   .05325444   .27299823          169
          2        |   .16030534   .49329027          131
          3        |   .10204082   .39297428           98
          4        |   .04545455   .27354191           66
          5        |   .175        .44649633           40
----------------------------------------------------------------
      Total        |   .05071622   .28132982         2583

See Appendix (Correlation Matrix) for further validation data.

APPENDIX:  CORRELATION MATRIX OF NEUROLOGICAL VARIABLES
. corr sx* dx* ch1sy* ch1diz* ch1cva1 ch1cva2
(obs=2583)
        | sxsync  sxdizzy  dxstroke  ch1sync1  ch1sync2  ch1diz1  ch1diz2
------------------------------------------------------------------------------------------------------------
sxsync  |  1.0000
sxdizzy |  0.2147  1.0000
dxstroke|  0.0440  0.0853   1.0000
ch1sync1|  0.8568  0.1944  -0.0142   1.0000
ch1sync2|  0.9965  0.2100   0.0444   0.8500    1.0000
ch1diz1 |  0.1853  0.8515   0.0712   0.1822    0.1826    1.0000
ch1diz2 |  0.1375  0.6103   0.0731   0.1109    0.1365    0.9158    1.0000
ch1cva1 |  0.0525  0.1010   0.9279  -0.0226    0.0530    0.0774    0.0823
ch1cva2 |  0.0227  0.0641   0.9778  -0.0130    0.0231    0.0718    0.0730
        |  ch1cva1  ch1cva2
----------------------------------------
ch1cva1|   1.0000
ch1cva2|   0.8818   1.0000

The algorithm used to generate the correlation matrix above was as follows:

replace ch1sync1 = 0 if (ch1sync1==. & ch1sync2==.)
replace ch1sync2 = 0 if ch1sync2==.
replace ch1diz1 = 0 if (ch1diz1==. & ch1diz2==.)
replace ch1diz2 = 0 if ch1diz2==.
replace ch1cva1 = 0 if (ch1cva1==. & ch1cva2==. & (ch1cvay==. | ch1cvay==0))
replace ch1cva2 = 0 if ch1cva2==.
corr sx* dx* ch1sy* ch1diz* ch1cva1 ch1cva2
LawInfo Forum