J Chron Dis 1972. Vol. 25, pp. 433-440. Pergamon Press. Printed in Great Britain
COMPARISON OF PERSONAL INTERVIEW AND POSTAL INQUIRY METHODS FOR ASSESSING PREVALENCE OF ANGINA AND POSSIBLE INFARCTION TRULSZEINER-HENRIKSEN The Norwegian Council on Cardiovascular Diseases, The Cancer Registry of Norway, The Norwegian Radium Hospital, Oslo, Norway (Received 20 September 1971;
in
final form 27 March 1972)
INTRODUCTION
THE QUESTIONNAIRE developed at the London School of Hygiene and Tropical Medicine and later recommended by the World Health Organization for use in field studies of cardiovascular disease [ 11, has been extensively used in various populations [2]. While originally developed for personal interviews, this questionnaire has also been employed for postal inquiries [3-51. The postal inquiry method is of course much cheaper than personal interviewing and is without interviewer error. A Finnish-Norwegian lung cancer study [6, 71 offered an opportunity to evaluate the repeatability at interview of the cardiac pain questionnaire [8], and to compare the interview symptom results with those of a similar postal inquiry. The last project, confined to a postal inquiry of the chest pain questions in a sub-sample of the 4092 men interviewed, was launched in April 1965, 23-3 yr after the original interviews. The objective was to compare the postal inquiry method with the personal interview method as a means of assessing the prevalence of angina and possible infarction, and to estimate the validity of the symptom findings by mortality follow-up in subsequent years. A secondary object was to examine the non-response in a postal inquiry among persons with and without cardiac pain symptoms according to the interview findings. MATERIAL
AND
METHODS
Sample
It was decided that a stratified sampling method with a larger sampling fraction of chest pain positives and questionables than of negatives should be employed. The interview material was limited to 237 chest pain positives and 226 questionables when previously re-interviewed were excluded; all of them were included in the sample. The chest pain negatives were sampled with a fraction corresponding to 0.13. With the purpose of obtaining a uniform geographical distribution the chest pain negatives were selected in a number of approximately twice the positives within each of the 12 counties where interviewing had taken place. Within each county the negatives were drawn at random. The geographical part of the analysis is omitted here. 433
TRULY ZEINER-HENRIKSEN
434
The criteria for chest pain positive were Yes to the first chest pain question and the ‘when you hurry’ question, or Yes to the severe chest pain question; the criteria for chest pain questionable were Yes to the first chest pain question only, see the postal inquiry questionnaire in Appendix 1 and the relevant chest pain questions at the interview in Appendix 2. Chest pain negative covered the remaining questionnaires. The year of birth distribution in the postal sample was similar to the total interviewed with one-third of the men born in each of the periods 1898-1902, 1903-1907, and 1908-1917 [6]. The median age was 57 at interview and 60 at the postal inquiry. Lists of names of the 943 men in the sample were first mailed to each local Population Register for checking. A total of 30 deaths were reported. Questionnaire, covering letter and return envelope were sent to the remaining persons in the sample. A further 13 deaths were reported and 9 persons could not be traced. One reminder was sent after an interval of 3 months. Symptom questions
The questions in the postal questionnaire (Appendix 1) were similar to the chest pain questions at interview (Appendix 2), except for the following modifications: In the postal questionnaire the questions on location of pain were replaced by a figure for marking the location of pain, and a question on when the pain was first experienced was included with the purpose of identifying new cases developed after the interview. The postal questionnaire covered chest pain questions only, while the interview included questions on many other items, such as cough, phlegm, previous illnesses and detailed smoking history [6]. The criteria for angina pectoris (AP) were the occurrence of pain or discomfort in the chest when walking uphill or hurrying, to which the respondent reacts by stopping or slowing down, and which goes away in 10 min or less if one stands still, when the site of the pain includes either the sternum or the left arm. The criteria for possible infarction (PI) were one or more episodes of severe pain across the front of the chest lasting for half an hour or more. ‘Non-speczjics’ were those not AP or PI positive with Yes to the first chest pain question, while other symptom negatives had a No answer to this question. RESULTS
Symptom prevalence
The response is shown in Table 1. With subtraction of the known deaths from the total sample a response per cent of 86 was obtained. Included in the 127 non-respondTABLE 1. NUMBER OF MENIN THE SAMPLE,NUMBERRESPONDED M MAIL QUESTIONNAIRE, AND RESPONSE PER CENT AMONG MEN ALIVE, BY CHESTPAIN WSITIVES, QUEsTIONABLE& AND NEGATZVES AT INTERVIEW Interview chest pain status
Sample
Positive
2;
Questionable Negative Total
226 480 943
Responded
b 178 189 406 773
Not responded Other Not alive
2c4 8 11 43
d 35 29 63 127
Rate not alive
Rate of response
(c/a)
(b/a-c)
per cent 10.1 3.5 2.3 4.6
per cent 83.6 86.7 86.6 85.9
Comparison
435
of Methods for Assessing Prevalence of Angina and Possible Infarction
ents are 9 men who were not found. Based upon men alive the response was slightly lower in the positive than in the questionable and negative chest pain groups. The number of deaths since the interviewing was highest among the chest pain positives. Questions not filled in or incompletely answered were of course more frequent in the postal inquiry than in the interview material; as many as 12 per cent of the postal respondents had missing answers to the first or the last chest pain question. The distribution of the respondents by their symptom status at interview and postal inquiry is shown in Table 2. Excluded are 67 men who reported that their first attack of pain occurred after the interview. Included in Table 2 are figures for the total number interviewed in Norway and the ‘inflation factor’ (total number interviewed divided by the number completing postal inquiry) by symptom status at interview. The inflation factor was under 3, except for the ‘other negative’ group where it was 9.4. TABLE 2. AND
POSTAL
NUMBER OF RESPONDENTS,EXCLUDING 67 ‘NEW CASES’, BY SYMPTOMSTATUSAT INTERVIEW INQUIRY.
THE
TOTAL
INTERVIEWED
POPULATION
STATUS
AND
‘INFLATION
FACTOR’
BY
SYMPTOM
INCLUDED.
PI : Possible infarction
AP: Angina pectoris Interview
Postal inquiry
PI+AP PI only AP only PI/AP negative Non-specific Other negative Incomplete Total (n) Total interviewed (N) Inflation factor (N/n)
PI A+p
PI only
AP only
23 14 3
15 18 5
14 20
2 2 44
8 3 2 51
85 1.93
PI/AP negative Other Nonneg. spec.
Incomplete
6 24 12
-
2;
56
54 62 22 180
37 365
-! 10
101 352 61 706
119
97
340
3423
28
4092
2.33
1.73
1.89
9
8 5
8 17
9.38
1
Total
3 5
2.80
54 78 60
5.80
It is seen from Table 2 that the number of PI increased from interview (95) to the postal inquiry (132), while the number of AP without PI (AP only) remained about the same (56 and 60). However, the sampling method employed implies an underrepresentation of the chest pain negatives, and accordingly also the proportion converting from negatives to symptom positives, in the study population. In Table 3 the estimates of postal inquiry characteristics of the total interviewed population are calculated by inflating the results from Table 2 to the symptom distribution of the population at interview, and reduced to 1000 total. The following points from Table 3 are noted: (1) Of the total 5.9 per cent were symptom positives (AP or PI) and 74.8 per cent negatives at both interview and postal inquiry, while 1.4 per cent converted from positive to negative and 7.6 per cent the other way from interview to postal inquiry. Incompletely filled-in questionnaires were 0.7 per cent at interview and 9.6 per cent at postal inquiry.
TRULYZEINER-HENRMSEN
436 TABLE3.
ESTIMATES OF POSTALINQUIRYPREVALENCE RATES(PER 1000) IN THETOTAL
INTERVIBWED
POPULATIONBYSYMPTOMSTATUS
PI: Possible infarction
AP: Angina pectoris Interview
Postal inquiry
PI+AP PI only AP only PI/AP negative Non-specific Other negative Incomplete Total
PI
PI/AP negative NonOther spec. neg.
i&
PI only
AP only
10.8 6.6 1.4
8.5 10.3 2.9
3.8 ;:z
2.8 11.1 5.5
4.6 1.7
3.4 2.1 23.7
24.9 28.6 10.2 83.1
0”:;
2;6
2;::
Incomplete
Total
18.3 39.0
0.7 2.1
26.6 52.2 59.4
55.0 639.5 84.8 836.6
3.4 0.7
92.2 673.5 96.1 1000.0
;9
(2) The conversion to symptom negative status among interview positives was 19 per cent; lower among positives to AP and PI than among positives to PI only or AP only.
(3)
Men with AP only status at interview had a conversion to PI positives at postal inquiry of 41 per cent. Among the interview non-specifics 23 per cent converted to symptom positives, against 7 per cent of the other symptom negatives.
(4)
Questionnaires incompletely filled-in at the postal inquiry were more frequent among the interview negatives (10 per cent) than among the symptom positives (1.5 per cent).
(5) The symptom prevalence rates increased from interview to postal inquiry: AP and PI from 2.1 to 2.7 per 100 (ratio 1.3), PI only from 2.9 to 5.2 (ratio 1.8),
AP only from 2.4 to 5.9 (ratio 2.5), PI total from 5.0 to 7.9 (ratio 1.6), and AP total from 4.4 to 8.6 (ratio 1.9). (6)
The ‘Repeatability’, defined as positives at both interview and postal inquiry in percentage of positives at interview and/or postal inquiry, was 39 for PI total, 23 for AP total, and 11 for AP only.
Mortality follow-up
A 5-yr mortality follow-up was reported for the total interview material [7]. In the postal inquiry material 43 deaths were registered from April 1965 to the end of 1967. Four of these deaths were among the ‘new cases’. The remaining number of deaths, with deaths from coronary heart disease (CHD) separately identified, is shown by symptom status at interview and postal inquiry in Table 4. A small decrease from interview to postal inquiry in per cent dying was found in both the symptom positive and negative group, as should be expected from the larger proportion of symptom positives at postal inquiry. The per cent died was highest for PI (13.7 at interview, 12.9 at postal inquiry) with the AP only group lower (8.9 at interview, 8.3 at postal inquiry), and the symptom negatives definitely on a lower level (3.8 at interview, 3.3 at postal inquiry). The 30 PI or AP symptom positives at interview who were negatives at the postal inquiry yielded no deaths (3.6 expected from interview mortality among symptom
Comparison
of Methods for Assessing Prevalence of Angina and Possible Infarction
437
NIJMBEROFDEATHS IN A POLLOW-up OF RESPONDENTS (‘NEW CASES'EXCLUDED) MAY 1965 TO DECEMBER1967 BY SYMPTOM STATUS AT INTERVIEW AND POSTAL INQUIRY. DEATHSFROMCORONARY HEARTDISEASE(CHD) SPECIFIED PI : Possible infarction AP: Angina pectoris TABLET.
Interview
PI+AP
PI only
AP only
Neither
Total
Postal inquiry
Number followed up
Deaths All causes
c PI+AP PI only AP onlv \ Neither
23 14 3 4
c PIfAP PI only AP only Neither
15 18 5 13
-
-
1 AP PI+AP Neither only PI only
;:13 9
- :1
-1 1
r PI+AP PI only AP only Neither
I 32 32 484 706
2 1 1
CHD
-
2 1 -
6 3
1 3 17 39
5 1
5 16
positives). The 71 postal inquiry positives among interview negatives yielded four deaths (2.7 expected), none of them with CHD as main cause of death. There were 16 deaths from CHD in all, a number which is esteemed to be too small for informative analyses. DISCUSSION
There is considerable experience with the cardiac pain questionnaire through a standard interview [I, 2, 8, 91. It has been shown that this method correlates well with electrocardiographic evidence of ischemia and high risk of subsequent death from myocardial infarction [2]. During the last decennium the same questionnaire has been extensively used by a self-administered postal method [4]. Direct comparison of the interview and postal methods in the use of the cardiac pain questionnaire has, as far as we can see, not been published. The present study demonstrates that for Norwegian middle-aged males the two methods are yielding much variety in distinction between symptom positives and negatives and are not interchangeable. The postal inquiry method produced higher prevalence rates, more so for angina than possible infarction, notwithstanding that many interview positives converted to negatives, especially for angina without infarction. This was expressed by a low repeatability, lower for angina than for possible infarction, and lowest for angina without infarction. It should be noted that all the respondents had been interviewed on the same questions; even if this was almost 3 yr previously it may have influenced the answering. A similar group not interviewed might have yielded even higher postal inquiry symptom prevalence. The intervention of personal interviewer will, as a rule, involve a higher level of criticism to positive symptom answers. Decisive in this connection is how far the interviewer is active or leading.
438
TRULSZEINER-HENIUKSEN
The conversion of positive to negative symptom status found in the Oslo reinterview [8] was higher than in the present material. A great deal of flux among angina symptom positives was reported by Rose [9]. The results of the present material should therefore not be interpreted without due regard to the limitation of symptom repeatability also present with the personal interview method. A 2-yr mortality follow-up of British and Norwegian migrants to U.S., who had answered the chest pain questions by a postal inquiry showed that the postal method was able to separate groups with quite different probabilities of dying [5]. The followup of the present material yielded similar experiences for middle aged Norwegian men, with almost as high discriminative effect on mortality among symptom positives and negatives by postal inquiry as by personal interviewing. But until further follow-up results are available the material does not permit a validation of how the excess risks are concentrated in symptom-related causes of death or an evaluation of the ability to measure differences in risk among various sub-groups by the two methods. One of the points of concern with the postal inquiry method is a high percentage of unanswered and incompletely filled-in questionnaires. For the evaluation of the symptom status of the non-respondents it is noted that in the present material the non-responding had a slightly higher representation among the chest pain negatives than the positives evaluated from the interview status. The incompletely filled-in questionnaires were definitely more frequent among the interview negatives than the symptom positives. SUMMARY
The cardiac pain questionnaire administered by a postal inquiry was answered by 773 Norwegian men in a stratisfied sample of the interviewed on the same questions almost 3 yr previously. Median age was 57 at interview and 60 at the postal inquiry. Response rate was 86 per cent. Prevalence of angina (AP) and possible infarction (PI) was assessed from the symptom questions. The estimates of postal inquiry symptom prevalence of the interviewed population presented higher rates by the postal inquiry than from the interview: AP increased from 4.4 to 8.6, PI from 5.0 to 7.9, and AP only from 2.4 to 5.9 per 100. The conversion to symptom positives at postal inquiry was higher among interview non-specifics than other negatives. As much as 41 per cent of men with AP only status at interview converted to PI at postal inquiry. The conversion of interview symptom positives to negatives at postal inquiry was 19 per cent; lower among AP and PI than among PI only or AP only. It is concluded that though both methods seem to have a high ability to identify groups liable to excess mortality, the postal inquiry method is yielding higher prevalence rates of angina and possible infarction than the personal interview method in a Norwegian male population. REFERENCES 1.
2. 3.
Rose GA: The diagnosis of ischaemic heart pain and intermittent claudication in field surveys. Bull WHO 27: 645-658, 1962 Rose GA: Chest pain questionnaire. Milbank Mem Fund Quart 43 (2): 32-39, 1965 Cederldf, Jonsson E, Lundman T: On the validity of mailed questionnaires in diagnosing angina pectoris and bronchitis. Arch Env Health 13: 738-742, 1966
Comparison 4.
5. 6. 7. 8. 9.
of Methods for Assessing Prevalence of Angina and Possible Infarction
439
Reid DD, Cornfield J, Markush RE et al: Studies of disease among migrants and native populations in Great Britain, Norway, and the United States. III. Prevalence of cardiorespiratory symptoms among migrants and native-born in the United States. Nat Cancer Inst Monogr 19: 321-346, 1966 Kreuger DE, Rogot E, Blackwelder WC, Reid DD: The predictive value of a postal questionnaire on cardio-respiratory symptoms. J Chron Dis 23: 411-421, 1970 Pedersen E, Saxen E, Hakama M et al: Lung cancer in Finland and Norway. Acta Path Stand (Suppl 199) 74 pp, 1969 Zeiner-Hemiksen T: Cardiovascular disease symptoms in Norway. A study of prevalence and a mortality follow-up. J Cbron Dis 24: 553-567, 1971 Zeiner-Henriksen T: The repeatability at interview of symptoms of angina and possible infarction. J Cbron Dis 25: 407-414, 1972 Rose GA: Variability of angina. Brit J Prev Med 22: 12-15, 1968
APPENDIX Name:
,...,
1
,......,.,........ ..,.,_,,..,. .._...._.._...................... ..,.._., .._ ..__............ .,,.
.._....._..........................................
Please answer the following questions and return the scheme in the enclosed envelope. Cross the appropriate q for answer. 1.
Have you ever had pain or discomfort
in your chest? 0 Yes
2.
Do you get this when you walk uphill or hurry?
0 Yes
0 No 17 No
If Yes : 3.
Do you get this when you walk at an ordinary pace on the level?
0 Yes
[7 No
If No to questions 1-2 continue from question 9. 4.
What do you do when you get it while walking?
5.
What happens to it if you stand still?
Relieved ............................................................................ ...... n Not relieved ....................................................................... U
6.
How soon is the pain relieved?
10 min or less ................................................................ U More than 10 min.. ............................. ...........................Cl
7.
Please record on the drawing with X the locations where you feel the pain (discomfort).
8.
When did you for the first time feel such pain (discomfort)?
Stop or slow down .................................................... n Carry on ................................ ................................................. 0 (Record ‘Stop or slow down’ if you carry on after taking nitroglycerine or similar medicine.)
Month ..,,........,.. ,....,....,..Year . .._......_............
Have you ever had a severe pain across the front of your chest lasting for half an hour or more? 0 Yes 0 No If ‘Yes’: 10. When did you have such pain for the first time? Month.. Year. ..__............
9.
11. How many such attacks have you had? ., .,,,.,.,,,. .,,,,,...,......,.,,,,..........,.........,,,,...,.,,...,,,.., ,,,.,.,..,..,........ .._..........
TRULSZEINER-HENRIKSEN
APPENDIX 2 The Finnish Norwegian Lung Cancer Study Interview-questionnaire Chest pain questions: Section A: Chest pain on effort 1. Have you ever had any pain or discomfort in your chest? 2. Do you get it when you walk uphill or hurry on the level? 3. Do you get it when you walk at ordinary pace on the level? 4. What do you do when you get it while you are walking? Stop or slow down-Carry on. 5. If you stand still, what happens to it? Relieved-Not relieved. 6. How soon relieved? 10 min or less-More than 10 min. 7. Will you show where you feel the pain (discomfort)? Sternum (upper or middle); Sternum (lower); Left anterior chest; Left arm; Other places. Section B: Possible infarction 10. Have you ever had a severe pain across the front of your chest lasting for half an hour or more? 11. How many of these attacks have you had? Date First attack: Latest attack:
Duration of pain