- Email: [email protected]
Surveying the elderly about health, medical care and living conditions. Some issues of response inconsistency
Arch. Gerontol. Geriatr., 11 (1990) 161-173
161
Elsevier AGG 00349
Surveying the elderly about health, medical care and living conditions. Some issues of response inconsistency Mats Thorslund
a
and Bo Warneryd
b
a Department of Social Medicine University Hospital, Uppsala, Sweden and b Statistics Sweden, Stockholm, Sweden
(Received 5 February 1990; revision received 23 July 1990; accepted 25 July 1990)
Summary A sample of elderly persons (75-84 years old) in Sweden was surveyed twice on home conditions, health and medical care. Two data collection methods were used: interviews and a mail survey. It has been shown that it is possible to carry out a mail survey (with low non-response) among the elderly in this group. However, what about other aspects of quality? Is it really possible to use a mail survey instead of the much more expensive and complicated techniques involved in interviewing? The results show that for certain groups of variables there are fairly large differences between how respondents reply in the mail survey and in personal interviews carried out by district nurses. These differences apply both to the degree of inconsistency in answers and to shifts in marginal distributions. Among other things, more ailments are mentioned at interview. The demand for information on the elderly and the differences between the methods in the cost of gathering this information make reliability and validity studies of methods of surveying the elderly increasingly important. Elderly; Survey methodology; Response inconsistency; Living conditions
Introduction T h i s r e p o r t p r e s e n t s t h e results o f a S w e d i s h s t u d y in w h i c h a s a m p l e o f e l d e r l y p e r s o n s ( 7 5 - 8 4 y e a r s o l d ) was s u r v e y e d twice. D a t a w e r e c o l l e c t e d in t w o w a y s , b y i n t e r v i e w a n d b y m a i l survey. O u r a i m is to s h o w h o w t h e t w o m e t h o d s w o r k e d a n d the degree of response inconsistency which arose for different questions. Can a mail s u r v e y b e u s e d i n s t e a d o f the m u c h m o r e e x p e n s i v e a n d c o m p l i c a t e d t e c h n i q u e s i n v o l v e d in t h e m o r e c o m m o n m e t h o d of i n t e r v i e w i n g e l d e r l y p e o p l e in t h e i r h o m e s ?
Correspondence to: Mats Thorslund, Dept. of Social Medicine, University Hospital, S-751 85 Uppsala, Sweden. 0167-4943/90/$03.50 © 1990 Elsevier Science Publishers B.V. (Biomedical Division)
162 Both the number and percentage of elderly in the population have increased in recent years. However, a change in this development is expected. According to forecasts, the percentage of older persons (65 + years) will remain about the same up to the year 2000. But the number of very old will continue to grow. In the year 2000, nearly 800 000 people in the Swedish population will be 75 years or older. This is an increase of 17% since 1988. Over 200000 will be 85 years old or older, an increase of 47% since 1988 (Statistics Sweden, 1989). The same trend can be seen in all industrialized countries. What statistical knowledge do we have of this 8% of the population who are 75 years old or older? Many surveys in Sweden as well as other countries have an upper age limit of 74. Consequently, when results comparing the elderly with other age-groups are presented, they relate to people 65-74 years old, that is to say an age group where most people are still cohabiting, where loneliness is not especially frequent, where the demand for medical care is not too extensive and so on. Furthermore, most surveys of the elderly take the form of personal interviews. Mail surveys are not common. Various reasons have been put forward to explain why elderly people should find it more difficult than others to interpret and respond to survey questions, (Colsher and Wallace, 1989; Collins, 1983; Schmidt, 1975 on the possible effects of age on personal interviews; Herzog and Rodgers (1988) report on telephone interviews with the elderly compared to personal interviews). As for mail surveys, the interceptive function of children, relatives, nurses, etc., is mentioned. There may also, of course, be difficulties with eyesight and other age-related problems. However, those articles we have found which deal with the quality of mail surveys among the elderly, report very positive (Morris and Boutelle, 1985; Victor, 1988; Wilcock, 1979; Vetter and Ford, 1989) or largely positive results (Landry et al., 1988).
Design Since 1972, the Department of Social Medicine, University of Uppsala, has been involved in a series of studies in Tierp, a municipality of about 20 000 inhabitants in central Sweden (Smedby et al., 1987). Studies with focus on the elderly started in 1984. The aims of these studies are to describe the elderly in terms of needs of service and care and the extent to which these needs are met. In the autumn of 1986, interviews and health assessments were carried out on all persons over 84 living at home (about 300). Non-response was 4%. For economic reasons, it was not possible also to interview and assess all 1560 persons in the age-group 75-84. A mail questionnaire was then considered, with a pilot mail survey giving positive results, both in terms of non-response and available indicators of measurement quality, in this case mainly inspection of the returned questionnaires as to completeness of replies, remarks upon questions and so on. It was decided to use this method for the main survey of the total population of 75-84 years of age.
163 Non-response here was also 4%. There are several reasons behind this good postal response. Before the fieldwork started we were able to use the local papers and radio to disseminate positive information about the study (e.g., that it was important for everyone to answer in order for us to get an accurate picture which could serve as the basis for better future planning). The elderly people were thus able to read and hear about the study in the media at the time when they were receiving the introductory letter and the questionnaire. Furthermore, all the staff working in the services for the elderly had been told in advance about the study. Elderly people often turn to their care staff with queries concerning contacts with the 'authorities'. It was thus important that the staff should know what it was about so that they could also help to ensure that the questionnaire was returned. We also feel that the fact that the study concerned all persons over a certain age was of importance. It is often difficult to explain, for example to a person chosen at random, why someone else cannot be used instead. Further, great care was taken in designing the questionnaire. The questions were formulated in such a way that they could be understood without having to refer to any further instructions in the questionnaire. Written instructions were kept to a minimum. A quick flow through the questions was aimed at, with as few references to other questions and as few jumps as possible. It was important that the purpose of each question should be clear to the respondent. A sample of 200 persons was also interviewed and an assessment made of their health. It is on this lesser sample that the results to be discussed here are based. The 200 were drawn at random and were then divided, also randomly, into two halves. One half got the mail survey first and were then interviewed, while the other half went through the reverse process. About three weeks were allowed between the two measurements. The reasons for this design were connected with interest in response inconsistency (as measured in a test-retest-way) and in the possible dependency of the second measurement on the first as related to which method came first. The interview questionnaire was extensive and contained around 140 questions and five health assessment scales. The mail questionnaire comprised six pages. Most of the questions common to both questionnaires were newly constructed and were given a limited pretest in the pilot study. Questions were to be simple and easy to comprehend. It was expected that people would often turn to others for help (spouse, children, home-care personnel) and therefore questions were to be specific and preferably deal with factual conditions, so that the answers would not be influenced by others giving help. One issue, for example, concerned the need for care. This was taken up in questions about whether the respondent could move around, with or without aids, read a newspaper, manage to cook and so on. Roughly grouped, the questions dealt with home conditions, ability to move, symptoms of ill-health, reading and hearing ability, medical care, existence of intimate friends and feelings of health/sickness and contentment with life (Table II gives an overview of items). The questionnaires differed to some extent in layout, e.g., some questions with
164 'yes-no' answers in the interview were only to be answered if the answer was ' y e s ' in the mail survey. Five district nurses were employed as interviewers. Their interviews took place outside their own district so that they could not know their interviewees beforehand. After each interview they made a health assessment.
Data and methods of analysis Of the sample of 200, 29 were found during the field work to live in institutions and were thus excluded. Of the remaining 171, 157 (or 92%) were interviewed. Eighty persons were interviewed first and received the mail survey afterwards, while 77 got the mail survey first. Our main presentation here is concerned with the degree of response inconsistency between the two data collection methods. Data are analyzed in (fourfold) tables of the following kind. Measures used are: b+c Gross difference rate, g = ~
• 100
b-c Test for change in marginals = C R = v~ + c (two-sided test, 5% level C R = _+1.96. Correction for continuity for (b + c) between 11 and 20. N o tests for tables where (b + c) < 10) (McNemar, 1957). Measure of inconsistency, M = (b + c ) / ( (a + b ) ( b + d) + (c + d ) ( a + c) ] n n 1 The marginal distribution, of course, set an upper limit for g. A small g may be due to extreme marginals or small variance. Comparisons between different questions can thus only be made to a limited extent. For all questions except those with the most extreme marginal distributions, an alternative measure, M, has been computed. This measure takes the marginals into account. It is related to the measure x, which compares the observed agreement between measurements with that expected by chance, given the marginals. Further-
TABLE I Kind of table used for data analysis Mail surveyresponse Yes No
Interview response Yes a c
No b d
Total a+b c+d
Total
a+ c
b+d
n
165 more, 1-M coincides with the ~-coefficient for the special case of identical marginals. In this case, M is also identical to the index of inconsistency used in the U.S. Current Population Survey. Standard errors for M have been computed using the formula for K (Fleiss, 1981). Some questions had more than two response alternatives. To simplify analysis, a dichotomization was carried out. Some of the response inconsistency has thus been excluded, but this loss is negligible.
Results
Table II shows the degree of inconsistency between interview and mail survey answers. Gross difference rates (g) and, for most o f the questions, the measure M and its S.E. are shown. Results for test of net differences are also given. The greater g is, the greater is also the response inconsistency. Mean g over the 43 questions which were common to both interview and mail survey was 14.6%. The largest g, 28%, was obtained for the question: 'In general, how content are you with your life?' For the question about the standard and situation of the respondent's lavatory, which we would expect to be stable, only one person in the mail survey and three in the interview reported poor standards so the theoretically possible g is thus very small. M varies between 0.12 and 0.92. Although the statistical precision is low, due to the small sample size, we would maintain that the response inconsistency is substantial. (One might recall that it is common to refer to values of 0.50 or more as high where the index of inconsistency is concerned. When this index is used instead of M, the results are very much the same.) It is possible that some of the inconsistencies are the result of possible changes which took place between the first and second measurements. It is also important to remember that the questions dealing with care of the previous 3 months do not refer to the same periods of time. As to marginal distributions, significant shifts were noted for 17 of the 34 questions. The number of shifts is remarkable. We do not maintain that any of the measurement methods is more valid or comes nearer to the true value than the other. Clearly, systematic factors may be at work in both methods. In general, it is hardly surprising that questions about more or less vague, unclearly formulated situations produce relatively great inconsistencies and sometimes even marginal shifts. 'How do you feel?' and 'In general, how content are you with life?' are examples of this type of question. These questions are not, however, representative of those in the survey. On the contrary, as mentioned earlier, considerable effort has been put into formulating questions about concrete, everyday situations in simple language. Experience tells us that these are the sort of questions which should give a good degree of consistency, at least between two measurements carried out under identical conditions. An alternative way of analyzing our material might be to look at the inconsistency between the first and second measurements. When repeated survey studies
10.3 14.0 19.1 13.4 16.6 19.2 12.1 8.9 26.1 19.1
(Q 9) Are you troubled by any of the following: H e a r t / c h e s t discomfort Backpain Stomach pains Constipation Pains in joints Skin problems, itching, leg sores Bronchial problems Pains in shoulders, arms or legs Breathlessness
21.0
(Q 6) Can you get out of bed?
(Q 8) Can you move around by yourself outside?
14.2
(Q 5) Can you get up from a kitchen chair?
6.5
22.7
(Q 4) How do you feel?
(Q 7) C a n you move around by yourself inside?
10.2 3.8 8.9 1.3 13.3 22.9
(Q 2) Is there some aspect of your housing conditions which makes it difficult for you to cope with daily fife? Inside stairs Thresholds C o l d / d r a f t inside Lavatory: poor s t a n d a r d / s i t u a t e d outside your flat Outside stairs Garden maintenance, snow clearing
g a
Degree of inconsistency between interview and mail survey answers (n = 157)
T A B L E II
0.39 0.52 0.79 0.71 0.49 0.47 0.35 0.56 0.46
0.24
0.30
0.66
0.40
0.70
0.72 0.62 0.92 0.71 0.71
M b
0.07 0.08 0.11 0.09 0.07 0.09 0.09 0.07 0.07
0.06
0.09
0.09
0.07
0.08
0.13 0.20 0.12 0.11 0.09
S.E. ( + )
No Yes No Yes Yes Yes No Yes No
No
-
Yes
Yes
Yes
No No No Yes
Significant marginal shift
No
(Q 17) C a n you manage to: Buy food Cook Clean house Visit lavatory Dress and undress
(Q 20) In general, how content are you with your lifc?
a g, gross difference rates. b M, measure.
22.0
(Q 16) D o you have anybody who is close to you and to w h o m you can talk intimately (apart from spouse/cohabitant)?
0.08 0.07 0.07 0.15 0.07
0.48 0.43 0.44 0.39 0.57
28.3
0.09
19.1 15.9 21.7 3.2 3.8
0.64
No
No Yes No
No
No No
3.1 27.4 17.2 5.1 3.8
(Q 13) Have you, during the last 3 months: Been in hospital Visited doctor Visited nurse Got home help Visited chiropractor, naprapath 0.08 0.07 0.08 0.04 0.16
No
0.06
0.31
14.7
(Q 11) Are you bothered by hearing defects? 0.18 0.55 0.48 0.12 0.45
Yes
Yes Yes No Yes Yes Yes
Yes
0.08
0.10 0.08 0.14 0.09 0.08 0.08 0.09 0.08 0.09
0.73
0.53 0.40 0.85 0.50 0.46 0.57 0.67 0.44 0.61
24.2
14.3 13.4 6.4 15.3 15.9 23.6 19.7 12.7 19.1
(Q 10) Can you without difficulty read a daily newspaper (with or without glasses)?
Swollen legs Vertigo Loss of appetite Problems in urinating Difficulty in sleeping Frequent tiredness Depression Frequent feelings of loneliness Worry, anxiety
168
of the same persons are carried out a dependence may arise between the measurements. One example of the effects which can result from this is that the respondent tries to be consistent on the second occasion and attempts to remember to give the same answers as before. This means that any possible random variation in answering should be reduced on the second occasion. Another possible effect is that the respondent is suddenly reminded on the first occasion of some situation which h e / s h e does not normally think about. On the second occasion, h e / s h e will therefore be able to give a more considered (maybe even more truthful) answer (Thorslund and W~neryd, 1985). However, there is no difference in results if the presentation is made according to first and second measurement, i.e., when variability is concerned. It is another matter with net differences; these would to a large extent vanish if the presentation was made according to the first and second measurement.
Does consistency differ depending on which method is used first? The results one obtains from the second data collection occasion can depend on the situation and the results of the first occasion. The degree of consistency has therefore been studied here with view to seeing whether it differs depending on whether the interview or the mail questionnaire was carried out first. Even though our material is small, we have not found any real difference between the two sample halves in this respect. Table III shows the mean values for the degree of consistency between the two sets of measurements, calculated according to the data collection method used first. The difference is negligible.
T A B L E III Degree of consistency (g) Mean g
n
Entire material
14.6
157
Interview first Mail questionnaire first
14.7 14.2
80 77
Secondary A D L Can cope with all daily tasks oneself Wholly or partly dependent on others Non-response
16.5 13.7 -
98 58 1
Somatic status Completely healthy Diminished health
11.9 17.1
65 92
Received help with filling in questionnaire Did not receive help Non-response
15.1 14.2 -
63 89 5
- , too few comparisons to report; g, gross difference rates.
169 Is consistency greater depending on such factors as state of health? Poor health could be a cause of incongruity in the answers to questions concerning how one copes with everyday life, functional ability, well-being, etc. If failing health meant impairment of judgement, then answers from both the interview and the mail questionnaire would tend to become more random the greater the age of the respondent. Another theoretically possible cause of incongruity between the two rounds of questioning could be a real change in state of health, degree of various types of physical complaint, functional ability and so on in the intervening period. When functioning and health are impaired, state of health and well-being can sometimes fluctuate from day to day ( I ~ et al., 1987). Table III shows the average degree of consistency between the measurement occasions, with regard to different indicators of diminished health and functional ability. The results show relatively small differences. Among the elderly who, in the opinion of the district nurse, could manage all daily tasks independently (secondary ADL), the average degree of inconsistency between the measurements is somewhat greater than among those with impaired functional ability. The opposite is the case for physical health (somatic status): among those considered by the district nurse to be fit and without any real problems, the consistency between the measurements is somewhat better. The mail survey ended by asking the respondent to say whether h e / s h e had been helped to fill in the questionnaire. Such help can indicate that the individual's ability to answer questions of this type is reduced. It can also indicate that one has access to the help of, for example, relatives. This is something which, according to the literature, can influence the old people's appraisal of their own health (Fillenbaum, 1979; Stoller, 1984). However, no difference in consistency between the two measurements occasions could be established between those who had received help and those who had not. It is, though, possible to establish that the answers to this final question correlate with the variables Secondary ADL and Somatic status (Table IV).
TABLE IV Proportion receivinghelp to fill in mail questionnaire SecondaryADL Somatic status All
Proportion receivinghelp Independent Dependent
% 27.5 60.3
No problems Problems
29.2 47.8 40.1
170
What pictures of the elderly do the different methods give? In terms of degree of consistency, the effect of changing the order of the two methods was negligible. However, one question which remains to be answered is whether the significant differences in marginal distribution which could be estabfished for 17 of the 34 questions point in any particular direction. Does one method give results which, when compared with those from the other method, indicate an overestimation or underestimation of, for example, health complaints or problems? An analysis of the types of question which are overestimated or underestimated by the two different methods reveals that all questions which show the respondent as having more complaints according to the interview than according to the mail questionnaire, are sub-questions of the main question: 'Are you troubled by any of the following?' For all 18 sub-questions the interviewers noted down more complaints than the elderly themselves did when filling in the questionnaire. The difference is significant for eleven of the sub-questions. On the other hand, the elderly people themselves noted more often on the questionnaire than at interview that they had problems with looking after the garden/clearing snow, that they didn't feel well, that they could not get up from a kitchen chair or out of bed without difficulty, that they had problems reading the newspaper or that they could not manage to cook.
Discussion
Surveys directed at the elderly will in all probability become increasingly common. Apart from the fact that the numbers of those over 75 are increasing, we are also starting to realise that we need better information about the situation of the elderly for planning purposes (Thorslund, 1986). Screening and casefinding studies aimed at identifying problems at an early stage are going to become more common, a development which has already begun in several countries (Taylor and Buckley, 1987; Rubenstein, 1987). The design and execution of studies directed at the very oldest demand special care (Schmidt, 1975; Hoinville, 1983; R~ et al., 1987). Possibly as a consequence of this, there are markedly few examples of the use of mail questionnaire surveys with the elderly. Mail questionnaires are usually associated with two things - - higher non-response rates and the fact that a great deal of trouble must be put into formulating questions which can be understood and answered correctly without help. In the project of which this method study forms a part, the rate of non-response to the mail questionnaire was 4%, after two followups and a considerable amount of information given before conducting the survey. But what is the quality of the replies to the questionnaire? Can the questionnaire method also be used for the age groups 75-84 instead of the much more expensive and complicated techniques involved in interviewing or other screening and casefinding methods?
171 The age-group 75-84 is of particular interest in this respect, partly because it is quantitatively large, but mainly because it is often within this age-group that screening and casefinding activities are just now being expanded. The results of the present method study demonstrate relatively great inconsistencies in the answers coming from the two different data collection methods. The differences are consistently greater than those one would normally find with re-interview studies in which one and the same data collection technique is used. However, we have found only few examples in the literature of comparable re-interview studies using the same age-group as in this study. Victor (1988), reporting fifty re-interviews in her mail questionnaire study, claims a degree of consistency in answers higher than that found here. We still find it difficult to establish to what extent the differences between measurements in our study may result from the fact that the sample comprises elderly people (i.e., age means loss of quality; discussed in the literature by Herzog and Dielman, 1985; Andrews and Herzog, 1986). The degree of inconsistency does not noticeably seem to vary in line with state of health and functional ability. We have tried to eliminate the effect of the results from the second measurement being dependent on the first by starting half the sample with one method, half with the other. The relatively high proportion of inconsistent answers for several questions - - 15% or more of the respondents reply or are noted to reply differently to the same questions after an interval of a few weeks - - gives rise to analysis problems of various kinds. Even if the inconsistencies seen over whole age-groups were cancelled out, it would mean that a screening and casefinding activity would run the risk of either missing people with real need, since they have failed to note the problem in the questionnaire, or of investing resources in people who do not actually need any help (i.e., lack of sensitivity and specificity). Such inconsistencies also present problems when analyzing associations between various factors and long-term followup of individuals run the risk of being invalidated to a great extent. However, the inconsistencies, in some cases great, are accompanied by significant differences in the marginal distributions. These must be due to procedural differences. The shifts appear independent of order of administration. Furthermore, there seems to be a pattern in the shifts. At interview there is 'over-reporting' of complaints such as aches and pains, constipation, passing water, sleeping, dizziness, etc. compared with when the elderly people fill in the questionnaire. Pihl (1987) in his standard of living study in the Swedish city of Orebro found a similar tendency. On the other hand, reduced functional ability (difficulty in getting up from a kitchen chair or out of bed, problems with reading, doing one's own cooking) is not reported as often at interview as in the questionnaire. One possible explanation is that the attitude a n d / o r behavior of the interviewers (district nurses) influenced the respondents, or that the nurses interpreted the answers differently from when the elderly, more or less independently, were able to answer the questionnaire at their own leisure. The general attitude of district nurses in their 'day-to-day' contact with old persons living at home is positive - - they tend to assume that there is an inherent
172 ability in all elderly people to cope with various situations. Thus, they often concentrate on what is still functioning well for the elderly person, both in terms of health and coping with daily life. In an interview situation, this might well lead to the interviewer wanting the elderly person to answer that h e / s h e can manage various tasks without much difficulty. On the other hand as a district nurse one is also part of the medical establishment and is used to discussing medical complaints at some depth. Similarly, aches and pains, constipation, sleeping, stomach pains, dizziness, etc., are exactly the type of complaints that elderly people are used to discussing with district nurses. This can, in the interview situation, have influenced the respondent to take up certain complaints more then than when filling in the questionnaire. To conclude, the demands for information about the situation of the elderly and their needs of service and care increase with the increase in their number. So far, most surveys of the elderly have taken the form of personal interviews. Our study has shown that it is possible to carry out a mail survey with low non-response among those in the 75-84 year age group. At the same time, the results show that for certain groups of variables there are fairly large differences between how respondents reply in the mail survey and their answers in personal interviews carried out by district nurses. These differences apply both to degree of inconsistency in answers and to shifts in marginal distributions. Among other things, more ailments are mentioned in the interviews. With this study we have not been able to establish which of the two survey methods comes closest to reflecting the ' t r u t h ' regarding health problems, ADL, etc. The demand for information on the elderly and the differences in cost (between the various methods) of gathering this information make reliability and validity studies of methods of surveying the elderly increasingly important.
Acknowledgements This work was supported by the Swedish Commission for Social Research, project G85/46:3 and the Swedish Medical Research Council, project 08163.
References Andrews, F.M. and Herzog, A.R. (1986): The quality of survey data as related to age of respondent. J. Am. Stat. Assoc., 81, 403-410. Colsher, P.L. and Wallace, R.B. (1989): Data quality and age: health and psychobehavioralcorrelates of item nonresponse and inconsistent responses. J. Gerontol., 44, 45-52. Fleiss, J.L. (1981): Statistical Methods for Rates and Proportions. Wiley and Sons, New York. Herzog, A.R. and Dielman, L. (1985): Age differences in response accuracy for factual survey questions. J. Gerontol., 40, 350-357. Herzog, A.R. and Rodgers, W.L. (1988): Interviewingolder adults, mode comparison using data from a face-to-face survey and a telephone survey. Pub. Question. Q., 52, 84-99. Fillenbaum, G. (1979): Social Context and Self Assessment of Health among the Elderly. J. Health Soc. Behav., 20, 45-51.
173 Hoinville, G. (1986): Carrying out surveys among the elderly. Some problems of sampling and interviewing. J. Market Res. Soc., 25, 223-237. Landry, J.A., Smyer, M.A., Tubman, J.G., Lago, D.J., Robert, J. and Simonson, W. (1988): Validation of two methods of data collection of self-reported medicine use among the elderly. Gerontologist, 28, 672-676. McNemar, Q. (1957): Psychological Statistics. Wiley and Sons, New York. Morris, W.W. and Boutelle, S. (1985): Multidimensional functional assessment in two modes. Gerontologist, 25, 638-643. Pihl, A. (1987): Lokala levnadsniv~undersOkningar. En jamfOrelse mellan metoder. F0redrag vid 5: e Nordiska kongressen i allm~inmedicin (In Swedish: Local level of living surveys. A comparison of methods. Paper given at the 5th Nordic Meeting in Family Medicine, Reykjavik). R~, O.C., Hendriksen, C., Kivil~i, S.L. and Thorslund, M. (1987): Intervention studies among elderly people. Scan& J. Primary Health Care, 5, 163-68. Rubenstein, L.Z. (1987): Exposing the iceberg of unrecognised disability. The benefits of functional assessment of the elderly. Arch. Intern. Meal., 147, 419-420. Statistics Sweden. (1986): Sveriges framtida befolkning. Prognos for ~ren 1989-2025. (The future population of Sweden. Projection for the years 1989-2025). Official statistics of Sweden. Stockholm. Schmidt, M.G. (1975): Interviewing the 'old-old'. Gerontologist, 15, 544-547. Smedby, B., CarsjO, K., Haglund, B. and Korpela, M. (1987): A Presentation of the Centre for Primary Care Research in Uppsala. Centre for Primary Care Research, Uppsala. Stoller, E.P. (1984): Self-assessment of health by the elderly: the impact of informal assistance. J. Health Soc. Behav., 25, 260-270. Taylor, R.C. and Buckley, E.G. (Eds.). (1987): Preventive care of the elderly: a review of current developments. Occasional paper 35,-The Royal College of General Practitioners, London. Thorslund, M. (1986): Evaluation research in care of the elderly. Some Swedish experiences. Scan& J. Primary Health Care, 1, 33-38. Thorslund, M. and Warneryd, B. (1985): Testing/assessing question quality - some Swedish experiences. J. Official Stat., 1, 159-178. Vetter, N.J. and Ford, D. (1989): The collection of health data from the elderly in primary care using a postal questionnaire. Arch. Gerontol. Geriatr., 9, 53-58. Victor, C. (1988): Some methodological aspects of using postal questionnaires with the elderly. Arch. Gerontol. Geriatr., 7, 163-172. Wilcock, G.K. (1979): Use of a self-administered questionnaire when screening for health problems in the elderly. Gerontology, 26, 345-349.
161
Elsevier AGG 00349
Surveying the elderly about health, medical care and living conditions. Some issues of response inconsistency Mats Thorslund
a
and Bo Warneryd
b
a Department of Social Medicine University Hospital, Uppsala, Sweden and b Statistics Sweden, Stockholm, Sweden
(Received 5 February 1990; revision received 23 July 1990; accepted 25 July 1990)
Summary A sample of elderly persons (75-84 years old) in Sweden was surveyed twice on home conditions, health and medical care. Two data collection methods were used: interviews and a mail survey. It has been shown that it is possible to carry out a mail survey (with low non-response) among the elderly in this group. However, what about other aspects of quality? Is it really possible to use a mail survey instead of the much more expensive and complicated techniques involved in interviewing? The results show that for certain groups of variables there are fairly large differences between how respondents reply in the mail survey and in personal interviews carried out by district nurses. These differences apply both to the degree of inconsistency in answers and to shifts in marginal distributions. Among other things, more ailments are mentioned at interview. The demand for information on the elderly and the differences between the methods in the cost of gathering this information make reliability and validity studies of methods of surveying the elderly increasingly important. Elderly; Survey methodology; Response inconsistency; Living conditions
Introduction T h i s r e p o r t p r e s e n t s t h e results o f a S w e d i s h s t u d y in w h i c h a s a m p l e o f e l d e r l y p e r s o n s ( 7 5 - 8 4 y e a r s o l d ) was s u r v e y e d twice. D a t a w e r e c o l l e c t e d in t w o w a y s , b y i n t e r v i e w a n d b y m a i l survey. O u r a i m is to s h o w h o w t h e t w o m e t h o d s w o r k e d a n d the degree of response inconsistency which arose for different questions. Can a mail s u r v e y b e u s e d i n s t e a d o f the m u c h m o r e e x p e n s i v e a n d c o m p l i c a t e d t e c h n i q u e s i n v o l v e d in t h e m o r e c o m m o n m e t h o d of i n t e r v i e w i n g e l d e r l y p e o p l e in t h e i r h o m e s ?
Correspondence to: Mats Thorslund, Dept. of Social Medicine, University Hospital, S-751 85 Uppsala, Sweden. 0167-4943/90/$03.50 © 1990 Elsevier Science Publishers B.V. (Biomedical Division)
162 Both the number and percentage of elderly in the population have increased in recent years. However, a change in this development is expected. According to forecasts, the percentage of older persons (65 + years) will remain about the same up to the year 2000. But the number of very old will continue to grow. In the year 2000, nearly 800 000 people in the Swedish population will be 75 years or older. This is an increase of 17% since 1988. Over 200000 will be 85 years old or older, an increase of 47% since 1988 (Statistics Sweden, 1989). The same trend can be seen in all industrialized countries. What statistical knowledge do we have of this 8% of the population who are 75 years old or older? Many surveys in Sweden as well as other countries have an upper age limit of 74. Consequently, when results comparing the elderly with other age-groups are presented, they relate to people 65-74 years old, that is to say an age group where most people are still cohabiting, where loneliness is not especially frequent, where the demand for medical care is not too extensive and so on. Furthermore, most surveys of the elderly take the form of personal interviews. Mail surveys are not common. Various reasons have been put forward to explain why elderly people should find it more difficult than others to interpret and respond to survey questions, (Colsher and Wallace, 1989; Collins, 1983; Schmidt, 1975 on the possible effects of age on personal interviews; Herzog and Rodgers (1988) report on telephone interviews with the elderly compared to personal interviews). As for mail surveys, the interceptive function of children, relatives, nurses, etc., is mentioned. There may also, of course, be difficulties with eyesight and other age-related problems. However, those articles we have found which deal with the quality of mail surveys among the elderly, report very positive (Morris and Boutelle, 1985; Victor, 1988; Wilcock, 1979; Vetter and Ford, 1989) or largely positive results (Landry et al., 1988).
Design Since 1972, the Department of Social Medicine, University of Uppsala, has been involved in a series of studies in Tierp, a municipality of about 20 000 inhabitants in central Sweden (Smedby et al., 1987). Studies with focus on the elderly started in 1984. The aims of these studies are to describe the elderly in terms of needs of service and care and the extent to which these needs are met. In the autumn of 1986, interviews and health assessments were carried out on all persons over 84 living at home (about 300). Non-response was 4%. For economic reasons, it was not possible also to interview and assess all 1560 persons in the age-group 75-84. A mail questionnaire was then considered, with a pilot mail survey giving positive results, both in terms of non-response and available indicators of measurement quality, in this case mainly inspection of the returned questionnaires as to completeness of replies, remarks upon questions and so on. It was decided to use this method for the main survey of the total population of 75-84 years of age.
163 Non-response here was also 4%. There are several reasons behind this good postal response. Before the fieldwork started we were able to use the local papers and radio to disseminate positive information about the study (e.g., that it was important for everyone to answer in order for us to get an accurate picture which could serve as the basis for better future planning). The elderly people were thus able to read and hear about the study in the media at the time when they were receiving the introductory letter and the questionnaire. Furthermore, all the staff working in the services for the elderly had been told in advance about the study. Elderly people often turn to their care staff with queries concerning contacts with the 'authorities'. It was thus important that the staff should know what it was about so that they could also help to ensure that the questionnaire was returned. We also feel that the fact that the study concerned all persons over a certain age was of importance. It is often difficult to explain, for example to a person chosen at random, why someone else cannot be used instead. Further, great care was taken in designing the questionnaire. The questions were formulated in such a way that they could be understood without having to refer to any further instructions in the questionnaire. Written instructions were kept to a minimum. A quick flow through the questions was aimed at, with as few references to other questions and as few jumps as possible. It was important that the purpose of each question should be clear to the respondent. A sample of 200 persons was also interviewed and an assessment made of their health. It is on this lesser sample that the results to be discussed here are based. The 200 were drawn at random and were then divided, also randomly, into two halves. One half got the mail survey first and were then interviewed, while the other half went through the reverse process. About three weeks were allowed between the two measurements. The reasons for this design were connected with interest in response inconsistency (as measured in a test-retest-way) and in the possible dependency of the second measurement on the first as related to which method came first. The interview questionnaire was extensive and contained around 140 questions and five health assessment scales. The mail questionnaire comprised six pages. Most of the questions common to both questionnaires were newly constructed and were given a limited pretest in the pilot study. Questions were to be simple and easy to comprehend. It was expected that people would often turn to others for help (spouse, children, home-care personnel) and therefore questions were to be specific and preferably deal with factual conditions, so that the answers would not be influenced by others giving help. One issue, for example, concerned the need for care. This was taken up in questions about whether the respondent could move around, with or without aids, read a newspaper, manage to cook and so on. Roughly grouped, the questions dealt with home conditions, ability to move, symptoms of ill-health, reading and hearing ability, medical care, existence of intimate friends and feelings of health/sickness and contentment with life (Table II gives an overview of items). The questionnaires differed to some extent in layout, e.g., some questions with
164 'yes-no' answers in the interview were only to be answered if the answer was ' y e s ' in the mail survey. Five district nurses were employed as interviewers. Their interviews took place outside their own district so that they could not know their interviewees beforehand. After each interview they made a health assessment.
Data and methods of analysis Of the sample of 200, 29 were found during the field work to live in institutions and were thus excluded. Of the remaining 171, 157 (or 92%) were interviewed. Eighty persons were interviewed first and received the mail survey afterwards, while 77 got the mail survey first. Our main presentation here is concerned with the degree of response inconsistency between the two data collection methods. Data are analyzed in (fourfold) tables of the following kind. Measures used are: b+c Gross difference rate, g = ~
• 100
b-c Test for change in marginals = C R = v~ + c (two-sided test, 5% level C R = _+1.96. Correction for continuity for (b + c) between 11 and 20. N o tests for tables where (b + c) < 10) (McNemar, 1957). Measure of inconsistency, M = (b + c ) / ( (a + b ) ( b + d) + (c + d ) ( a + c) ] n n 1 The marginal distribution, of course, set an upper limit for g. A small g may be due to extreme marginals or small variance. Comparisons between different questions can thus only be made to a limited extent. For all questions except those with the most extreme marginal distributions, an alternative measure, M, has been computed. This measure takes the marginals into account. It is related to the measure x, which compares the observed agreement between measurements with that expected by chance, given the marginals. Further-
TABLE I Kind of table used for data analysis Mail surveyresponse Yes No
Interview response Yes a c
No b d
Total a+b c+d
Total
a+ c
b+d
n
165 more, 1-M coincides with the ~-coefficient for the special case of identical marginals. In this case, M is also identical to the index of inconsistency used in the U.S. Current Population Survey. Standard errors for M have been computed using the formula for K (Fleiss, 1981). Some questions had more than two response alternatives. To simplify analysis, a dichotomization was carried out. Some of the response inconsistency has thus been excluded, but this loss is negligible.
Results
Table II shows the degree of inconsistency between interview and mail survey answers. Gross difference rates (g) and, for most o f the questions, the measure M and its S.E. are shown. Results for test of net differences are also given. The greater g is, the greater is also the response inconsistency. Mean g over the 43 questions which were common to both interview and mail survey was 14.6%. The largest g, 28%, was obtained for the question: 'In general, how content are you with your life?' For the question about the standard and situation of the respondent's lavatory, which we would expect to be stable, only one person in the mail survey and three in the interview reported poor standards so the theoretically possible g is thus very small. M varies between 0.12 and 0.92. Although the statistical precision is low, due to the small sample size, we would maintain that the response inconsistency is substantial. (One might recall that it is common to refer to values of 0.50 or more as high where the index of inconsistency is concerned. When this index is used instead of M, the results are very much the same.) It is possible that some of the inconsistencies are the result of possible changes which took place between the first and second measurements. It is also important to remember that the questions dealing with care of the previous 3 months do not refer to the same periods of time. As to marginal distributions, significant shifts were noted for 17 of the 34 questions. The number of shifts is remarkable. We do not maintain that any of the measurement methods is more valid or comes nearer to the true value than the other. Clearly, systematic factors may be at work in both methods. In general, it is hardly surprising that questions about more or less vague, unclearly formulated situations produce relatively great inconsistencies and sometimes even marginal shifts. 'How do you feel?' and 'In general, how content are you with life?' are examples of this type of question. These questions are not, however, representative of those in the survey. On the contrary, as mentioned earlier, considerable effort has been put into formulating questions about concrete, everyday situations in simple language. Experience tells us that these are the sort of questions which should give a good degree of consistency, at least between two measurements carried out under identical conditions. An alternative way of analyzing our material might be to look at the inconsistency between the first and second measurements. When repeated survey studies
10.3 14.0 19.1 13.4 16.6 19.2 12.1 8.9 26.1 19.1
(Q 9) Are you troubled by any of the following: H e a r t / c h e s t discomfort Backpain Stomach pains Constipation Pains in joints Skin problems, itching, leg sores Bronchial problems Pains in shoulders, arms or legs Breathlessness
21.0
(Q 6) Can you get out of bed?
(Q 8) Can you move around by yourself outside?
14.2
(Q 5) Can you get up from a kitchen chair?
6.5
22.7
(Q 4) How do you feel?
(Q 7) C a n you move around by yourself inside?
10.2 3.8 8.9 1.3 13.3 22.9
(Q 2) Is there some aspect of your housing conditions which makes it difficult for you to cope with daily fife? Inside stairs Thresholds C o l d / d r a f t inside Lavatory: poor s t a n d a r d / s i t u a t e d outside your flat Outside stairs Garden maintenance, snow clearing
g a
Degree of inconsistency between interview and mail survey answers (n = 157)
T A B L E II
0.39 0.52 0.79 0.71 0.49 0.47 0.35 0.56 0.46
0.24
0.30
0.66
0.40
0.70
0.72 0.62 0.92 0.71 0.71
M b
0.07 0.08 0.11 0.09 0.07 0.09 0.09 0.07 0.07
0.06
0.09
0.09
0.07
0.08
0.13 0.20 0.12 0.11 0.09
S.E. ( + )
No Yes No Yes Yes Yes No Yes No
No
-
Yes
Yes
Yes
No No No Yes
Significant marginal shift
No
(Q 17) C a n you manage to: Buy food Cook Clean house Visit lavatory Dress and undress
(Q 20) In general, how content are you with your lifc?
a g, gross difference rates. b M, measure.
22.0
(Q 16) D o you have anybody who is close to you and to w h o m you can talk intimately (apart from spouse/cohabitant)?
0.08 0.07 0.07 0.15 0.07
0.48 0.43 0.44 0.39 0.57
28.3
0.09
19.1 15.9 21.7 3.2 3.8
0.64
No
No Yes No
No
No No
3.1 27.4 17.2 5.1 3.8
(Q 13) Have you, during the last 3 months: Been in hospital Visited doctor Visited nurse Got home help Visited chiropractor, naprapath 0.08 0.07 0.08 0.04 0.16
No
0.06
0.31
14.7
(Q 11) Are you bothered by hearing defects? 0.18 0.55 0.48 0.12 0.45
Yes
Yes Yes No Yes Yes Yes
Yes
0.08
0.10 0.08 0.14 0.09 0.08 0.08 0.09 0.08 0.09
0.73
0.53 0.40 0.85 0.50 0.46 0.57 0.67 0.44 0.61
24.2
14.3 13.4 6.4 15.3 15.9 23.6 19.7 12.7 19.1
(Q 10) Can you without difficulty read a daily newspaper (with or without glasses)?
Swollen legs Vertigo Loss of appetite Problems in urinating Difficulty in sleeping Frequent tiredness Depression Frequent feelings of loneliness Worry, anxiety
168
of the same persons are carried out a dependence may arise between the measurements. One example of the effects which can result from this is that the respondent tries to be consistent on the second occasion and attempts to remember to give the same answers as before. This means that any possible random variation in answering should be reduced on the second occasion. Another possible effect is that the respondent is suddenly reminded on the first occasion of some situation which h e / s h e does not normally think about. On the second occasion, h e / s h e will therefore be able to give a more considered (maybe even more truthful) answer (Thorslund and W~neryd, 1985). However, there is no difference in results if the presentation is made according to first and second measurement, i.e., when variability is concerned. It is another matter with net differences; these would to a large extent vanish if the presentation was made according to the first and second measurement.
Does consistency differ depending on which method is used first? The results one obtains from the second data collection occasion can depend on the situation and the results of the first occasion. The degree of consistency has therefore been studied here with view to seeing whether it differs depending on whether the interview or the mail questionnaire was carried out first. Even though our material is small, we have not found any real difference between the two sample halves in this respect. Table III shows the mean values for the degree of consistency between the two sets of measurements, calculated according to the data collection method used first. The difference is negligible.
T A B L E III Degree of consistency (g) Mean g
n
Entire material
14.6
157
Interview first Mail questionnaire first
14.7 14.2
80 77
Secondary A D L Can cope with all daily tasks oneself Wholly or partly dependent on others Non-response
16.5 13.7 -
98 58 1
Somatic status Completely healthy Diminished health
11.9 17.1
65 92
Received help with filling in questionnaire Did not receive help Non-response
15.1 14.2 -
63 89 5
- , too few comparisons to report; g, gross difference rates.
169 Is consistency greater depending on such factors as state of health? Poor health could be a cause of incongruity in the answers to questions concerning how one copes with everyday life, functional ability, well-being, etc. If failing health meant impairment of judgement, then answers from both the interview and the mail questionnaire would tend to become more random the greater the age of the respondent. Another theoretically possible cause of incongruity between the two rounds of questioning could be a real change in state of health, degree of various types of physical complaint, functional ability and so on in the intervening period. When functioning and health are impaired, state of health and well-being can sometimes fluctuate from day to day ( I ~ et al., 1987). Table III shows the average degree of consistency between the measurement occasions, with regard to different indicators of diminished health and functional ability. The results show relatively small differences. Among the elderly who, in the opinion of the district nurse, could manage all daily tasks independently (secondary ADL), the average degree of inconsistency between the measurements is somewhat greater than among those with impaired functional ability. The opposite is the case for physical health (somatic status): among those considered by the district nurse to be fit and without any real problems, the consistency between the measurements is somewhat better. The mail survey ended by asking the respondent to say whether h e / s h e had been helped to fill in the questionnaire. Such help can indicate that the individual's ability to answer questions of this type is reduced. It can also indicate that one has access to the help of, for example, relatives. This is something which, according to the literature, can influence the old people's appraisal of their own health (Fillenbaum, 1979; Stoller, 1984). However, no difference in consistency between the two measurements occasions could be established between those who had received help and those who had not. It is, though, possible to establish that the answers to this final question correlate with the variables Secondary ADL and Somatic status (Table IV).
TABLE IV Proportion receivinghelp to fill in mail questionnaire SecondaryADL Somatic status All
Proportion receivinghelp Independent Dependent
% 27.5 60.3
No problems Problems
29.2 47.8 40.1
170
What pictures of the elderly do the different methods give? In terms of degree of consistency, the effect of changing the order of the two methods was negligible. However, one question which remains to be answered is whether the significant differences in marginal distribution which could be estabfished for 17 of the 34 questions point in any particular direction. Does one method give results which, when compared with those from the other method, indicate an overestimation or underestimation of, for example, health complaints or problems? An analysis of the types of question which are overestimated or underestimated by the two different methods reveals that all questions which show the respondent as having more complaints according to the interview than according to the mail questionnaire, are sub-questions of the main question: 'Are you troubled by any of the following?' For all 18 sub-questions the interviewers noted down more complaints than the elderly themselves did when filling in the questionnaire. The difference is significant for eleven of the sub-questions. On the other hand, the elderly people themselves noted more often on the questionnaire than at interview that they had problems with looking after the garden/clearing snow, that they didn't feel well, that they could not get up from a kitchen chair or out of bed without difficulty, that they had problems reading the newspaper or that they could not manage to cook.
Discussion
Surveys directed at the elderly will in all probability become increasingly common. Apart from the fact that the numbers of those over 75 are increasing, we are also starting to realise that we need better information about the situation of the elderly for planning purposes (Thorslund, 1986). Screening and casefinding studies aimed at identifying problems at an early stage are going to become more common, a development which has already begun in several countries (Taylor and Buckley, 1987; Rubenstein, 1987). The design and execution of studies directed at the very oldest demand special care (Schmidt, 1975; Hoinville, 1983; R~ et al., 1987). Possibly as a consequence of this, there are markedly few examples of the use of mail questionnaire surveys with the elderly. Mail questionnaires are usually associated with two things - - higher non-response rates and the fact that a great deal of trouble must be put into formulating questions which can be understood and answered correctly without help. In the project of which this method study forms a part, the rate of non-response to the mail questionnaire was 4%, after two followups and a considerable amount of information given before conducting the survey. But what is the quality of the replies to the questionnaire? Can the questionnaire method also be used for the age groups 75-84 instead of the much more expensive and complicated techniques involved in interviewing or other screening and casefinding methods?
171 The age-group 75-84 is of particular interest in this respect, partly because it is quantitatively large, but mainly because it is often within this age-group that screening and casefinding activities are just now being expanded. The results of the present method study demonstrate relatively great inconsistencies in the answers coming from the two different data collection methods. The differences are consistently greater than those one would normally find with re-interview studies in which one and the same data collection technique is used. However, we have found only few examples in the literature of comparable re-interview studies using the same age-group as in this study. Victor (1988), reporting fifty re-interviews in her mail questionnaire study, claims a degree of consistency in answers higher than that found here. We still find it difficult to establish to what extent the differences between measurements in our study may result from the fact that the sample comprises elderly people (i.e., age means loss of quality; discussed in the literature by Herzog and Dielman, 1985; Andrews and Herzog, 1986). The degree of inconsistency does not noticeably seem to vary in line with state of health and functional ability. We have tried to eliminate the effect of the results from the second measurement being dependent on the first by starting half the sample with one method, half with the other. The relatively high proportion of inconsistent answers for several questions - - 15% or more of the respondents reply or are noted to reply differently to the same questions after an interval of a few weeks - - gives rise to analysis problems of various kinds. Even if the inconsistencies seen over whole age-groups were cancelled out, it would mean that a screening and casefinding activity would run the risk of either missing people with real need, since they have failed to note the problem in the questionnaire, or of investing resources in people who do not actually need any help (i.e., lack of sensitivity and specificity). Such inconsistencies also present problems when analyzing associations between various factors and long-term followup of individuals run the risk of being invalidated to a great extent. However, the inconsistencies, in some cases great, are accompanied by significant differences in the marginal distributions. These must be due to procedural differences. The shifts appear independent of order of administration. Furthermore, there seems to be a pattern in the shifts. At interview there is 'over-reporting' of complaints such as aches and pains, constipation, passing water, sleeping, dizziness, etc. compared with when the elderly people fill in the questionnaire. Pihl (1987) in his standard of living study in the Swedish city of Orebro found a similar tendency. On the other hand, reduced functional ability (difficulty in getting up from a kitchen chair or out of bed, problems with reading, doing one's own cooking) is not reported as often at interview as in the questionnaire. One possible explanation is that the attitude a n d / o r behavior of the interviewers (district nurses) influenced the respondents, or that the nurses interpreted the answers differently from when the elderly, more or less independently, were able to answer the questionnaire at their own leisure. The general attitude of district nurses in their 'day-to-day' contact with old persons living at home is positive - - they tend to assume that there is an inherent
172 ability in all elderly people to cope with various situations. Thus, they often concentrate on what is still functioning well for the elderly person, both in terms of health and coping with daily life. In an interview situation, this might well lead to the interviewer wanting the elderly person to answer that h e / s h e can manage various tasks without much difficulty. On the other hand as a district nurse one is also part of the medical establishment and is used to discussing medical complaints at some depth. Similarly, aches and pains, constipation, sleeping, stomach pains, dizziness, etc., are exactly the type of complaints that elderly people are used to discussing with district nurses. This can, in the interview situation, have influenced the respondent to take up certain complaints more then than when filling in the questionnaire. To conclude, the demands for information about the situation of the elderly and their needs of service and care increase with the increase in their number. So far, most surveys of the elderly have taken the form of personal interviews. Our study has shown that it is possible to carry out a mail survey with low non-response among those in the 75-84 year age group. At the same time, the results show that for certain groups of variables there are fairly large differences between how respondents reply in the mail survey and their answers in personal interviews carried out by district nurses. These differences apply both to degree of inconsistency in answers and to shifts in marginal distributions. Among other things, more ailments are mentioned in the interviews. With this study we have not been able to establish which of the two survey methods comes closest to reflecting the ' t r u t h ' regarding health problems, ADL, etc. The demand for information on the elderly and the differences in cost (between the various methods) of gathering this information make reliability and validity studies of methods of surveying the elderly increasingly important.
Acknowledgements This work was supported by the Swedish Commission for Social Research, project G85/46:3 and the Swedish Medical Research Council, project 08163.
References Andrews, F.M. and Herzog, A.R. (1986): The quality of survey data as related to age of respondent. J. Am. Stat. Assoc., 81, 403-410. Colsher, P.L. and Wallace, R.B. (1989): Data quality and age: health and psychobehavioralcorrelates of item nonresponse and inconsistent responses. J. Gerontol., 44, 45-52. Fleiss, J.L. (1981): Statistical Methods for Rates and Proportions. Wiley and Sons, New York. Herzog, A.R. and Dielman, L. (1985): Age differences in response accuracy for factual survey questions. J. Gerontol., 40, 350-357. Herzog, A.R. and Rodgers, W.L. (1988): Interviewingolder adults, mode comparison using data from a face-to-face survey and a telephone survey. Pub. Question. Q., 52, 84-99. Fillenbaum, G. (1979): Social Context and Self Assessment of Health among the Elderly. J. Health Soc. Behav., 20, 45-51.
173 Hoinville, G. (1986): Carrying out surveys among the elderly. Some problems of sampling and interviewing. J. Market Res. Soc., 25, 223-237. Landry, J.A., Smyer, M.A., Tubman, J.G., Lago, D.J., Robert, J. and Simonson, W. (1988): Validation of two methods of data collection of self-reported medicine use among the elderly. Gerontologist, 28, 672-676. McNemar, Q. (1957): Psychological Statistics. Wiley and Sons, New York. Morris, W.W. and Boutelle, S. (1985): Multidimensional functional assessment in two modes. Gerontologist, 25, 638-643. Pihl, A. (1987): Lokala levnadsniv~undersOkningar. En jamfOrelse mellan metoder. F0redrag vid 5: e Nordiska kongressen i allm~inmedicin (In Swedish: Local level of living surveys. A comparison of methods. Paper given at the 5th Nordic Meeting in Family Medicine, Reykjavik). R~, O.C., Hendriksen, C., Kivil~i, S.L. and Thorslund, M. (1987): Intervention studies among elderly people. Scan& J. Primary Health Care, 5, 163-68. Rubenstein, L.Z. (1987): Exposing the iceberg of unrecognised disability. The benefits of functional assessment of the elderly. Arch. Intern. Meal., 147, 419-420. Statistics Sweden. (1986): Sveriges framtida befolkning. Prognos for ~ren 1989-2025. (The future population of Sweden. Projection for the years 1989-2025). Official statistics of Sweden. Stockholm. Schmidt, M.G. (1975): Interviewing the 'old-old'. Gerontologist, 15, 544-547. Smedby, B., CarsjO, K., Haglund, B. and Korpela, M. (1987): A Presentation of the Centre for Primary Care Research in Uppsala. Centre for Primary Care Research, Uppsala. Stoller, E.P. (1984): Self-assessment of health by the elderly: the impact of informal assistance. J. Health Soc. Behav., 25, 260-270. Taylor, R.C. and Buckley, E.G. (Eds.). (1987): Preventive care of the elderly: a review of current developments. Occasional paper 35,-The Royal College of General Practitioners, London. Thorslund, M. (1986): Evaluation research in care of the elderly. Some Swedish experiences. Scan& J. Primary Health Care, 1, 33-38. Thorslund, M. and Warneryd, B. (1985): Testing/assessing question quality - some Swedish experiences. J. Official Stat., 1, 159-178. Vetter, N.J. and Ford, D. (1989): The collection of health data from the elderly in primary care using a postal questionnaire. Arch. Gerontol. Geriatr., 9, 53-58. Victor, C. (1988): Some methodological aspects of using postal questionnaires with the elderly. Arch. Gerontol. Geriatr., 7, 163-172. Wilcock, G.K. (1979): Use of a self-administered questionnaire when screening for health problems in the elderly. Gerontology, 26, 345-349.