Psychology, Health & Medicine, November 2006; 11(4): 432 – 448
Story-based scales: Development and validation of questionnaires to measure subjective health status and cultural adherence in British Bangladeshis with diabetes
TRISHA GREENHALGH1, MU’MIN CHOWDHURY1, & GARY W. WOOD2 1
University College London Medical School, London, and 2Centre for Lifelong Learning, University of Birmingham
Abstract Questionnaires that measure subjective health status are increasingly used in clinical trials. But scales based on the quantification of subjective traits (‘‘rate your feelings on a scale of 1 to 5’’) and initially developed in western population samples may not be valid for use in minority ethnic groups, even if accurately translated. The measurement of cultural adaptation and assimilation in immigrant groups is important for health research but has well documented methodological challenges. The aim of this study was to develop valid and reliable questionnaires to measure subjective health status and cultural adherence in a minority ethnic group, using the story as the unit of inquiry. The design was a multi-phase study involving (a) narrative interview, (b) vignette construction, (c) questionnaire development, and (d) questionnaire validation in relation to two scales (well-being and cultural adherence) in British Bangladeshis with diabetes. Using data from in-depth narrative interviews (i.e., a non-directive research technique in which the participant is invited to ‘‘tell me the story about your diabetes, starting with when you first noticed anything wrong’’, and the only prompts used are ‘‘tell me more about that’’ or ‘‘what happened next?’’; Greenhalgh, Helman, & Chowdhury, 1998; Muller, 1999), we constructed culturally congruent vignettes to depict different subjective health states and behaviours. We refined these items in focus group interviews and validated the instruments on 98 Bangladeshi participants, randomly sampled from GP diabetes registers in inner London and interviewed by a Bangladeshi anthropologist. We used factor analysis to explore the underlying structure in the responses to questionnaire items, plus Cronbach alpha tests to measure internal consistency of scales. The questionnaires were acceptable and credible to Bangladeshi participants with diabetes. Ninety of 98 participants were able and willing to complete them with interviewer assistance. Following factor analysis, we produced two definitive instruments. The wellbeing scale was a single-factor model with four story-based items (measuring depression, anxiety, physical energy, and social activities), with a Cronbach’s alpha of .92. The cultural adherence scale was a single-factor model with five items (measuring religious restrictions, ethnic practices, and social ties), with a Cronbach’s alpha of .83. In conclusion, this study has produced two important outputs: (a) easy-to-administer, story-based questionnaires that measure well-being and cultural adherence, which are specific to British Bangladeshis with diabetes; (b) a general method for developing story-based instruments to quantify the subjective experience of illness and adherence to cultural norms, which potentially has applications beyond the study population.
Keywords: Cultural adherence, cultural assimilation, health-related quality of life, minority ethnic groups, questionnaires, translation, cross-cultural
Correspondence: Trisha Greenhalgh, Department of Primary Care and Population Sciences, University College London Medical School, Holborn Union Building, Highgate Hill, London N19 5LW, UK. E-mail:
[email protected] ISSN 1354-8506 print/ISSN 1465-3966 online ª 2006 Taylor & Francis DOI: 10.1080/13548500500429379
Story-based scales
433
Introduction This work was part of a wider programme of research into health beliefs and behaviour in British Bangladeshis with diabetes, funded by the Wellcome Trust and Diabetes UK. Bangladeshis in Britain have one of the highest levels of deprivation and lowest subjective health status of any ethnic group (Blaxter, 1994). They also have the highest prevalence of adult diabetes (8%), a distinction attributable largely to lifestyle choices such as smoking, diet, and low exercise levels (Greenhalgh, Griffiths, Thomas, Wheeler, & Raleigh, 2001). Our previous qualitative work (Chowdhury, Helman, & Greenhalgh, 2000; Greenhalgh et al., 1998) had suggested two hypotheses which we wished to explore further in future studies: (a) diabetes in British Bangladeshis seemed to be associated with particularly low subjective health status; and (b) traditional Bangledeshi cultural beliefs and practices seemed to account for some health-related lifestyle choices. We therefore sought to develop scales to measure subjective health status (‘‘well being’’) and adherence to traditional ethnic culture (‘‘cultural adherence’’) in British Bangladeshis. Because previous research teams had encountered difficulties using conventional questionnaire formats in people of Bangladeshi origin (Boynton, Wood, & Greenhalgh, 2004), and because we had identified that these individuals readily exchanged information and advice about health through stories (Greenhalgh et al., 1998; Greenhalgh, Collard, & Begum, 2005), we wanted to see whether a story-based questionnaire format would be acceptable and feasible in this group. Story (or vignette)-based questionnaires, which have a long and respectable history in social anthropology (Anderson & Anderson, 1951; Herskovits, 1950), are not common in biomedical research. They have been used in the evaluation of professional practice (Eisemann, Richter, Bauer, Bonelli, & Porzsolt, 1999; Parker, Mahendran, Yeo, Loh, & Jorm, 1999; Richardson, Kitchen, & Livingston, 2003; Schutzwohl et al., 2003), the assessment of lay beliefs about risk behaviour (Love & Thurman, 1991), and very occasionally in the development of quality of life scales (Paterson & Britten, 2000; Salomon, Tandon, & Murray, 2004). Below, we briefly review two relevant bodies of literature: measuring subjective health status in a cross-cultural setting, and measuring cultural adherence. In the Discussion, we return to these themes and review the use of stories in questionnaire research. Measuring subjective health status in a cross-cultural setting Evaluating the subjective experience of illness, and the impact of health interventions, from the patient’s perspective is increasingly important in research and policy making (Scientific Advisory Committee of the Medical Outcomes Trust, 2002). The Medical Research Council warns grant applicants that without a robust measure of health-related quality of life (HRQOL), their trial is unlikely to be funded (see www.mrc.ac.uk/). The National Institute for Clinical Excellence makes extensive use of HRQOL data when judging whether a drug or procedure should be made available on the National Health Service (see www.nice.org.uk/). Researchers’ desire to compare and synthesize the subjective impact of interventions across clinical trials (Garratt, Schmidt, Mackintosh, & Fitzpatrick, 2002) has spawned a growth industry of ‘‘standardized’’ HRQOL scales—both generic (such as the SF-36, the Nottingham Health Profile or the EuroQol; Kind, Dolan, Gudex, & Williams, 1998) and disease-specific (such as the Diabetes Well Being Scale or the Adult Diabetes Dependent Quality of Life scale; Bradley & Speight, 2002). For populations whose language is not English, it is recommended (but, we argue, questionable) practice to translate an established English (or American) instrument and ‘‘validate’’ it by back-translation followed by factor analysis of a pilot dataset (Scientific Advisory Committee of the Medical Outcomes Trust, 2002).
434
T. Greenhalgh et al.
The translate-and-back-translate approach can demonstrate reliability (i.e., that the questionnaire has been competently translated and produces similar results on repeated use) and ‘‘equivalence’’ (i.e., response patterns of population samples are statistically comparable). But it cannot actually validate the scale (i.e., ensure that it is measuring what we think it is measuring) because it fails to take account of (or even identify) systematic differences in the socio-cultural meaning of translated terms in different groups (Hendrickson, 2003; Bowden & Fox-Rushby, 2003; Herdman, Fox-Rushby, & Badia, 1997; Boynton et al., 2004). Bowden and Fox-Rushby recently evaluated 58 translated HRQOL scales (all of which were alleged to have been ‘‘validated’’; Bowden et al., 2003) using a framework developed by Herdman et al. that has six dimensions: conceptual, item, semantic, operational, measurement and functional (Herdman et al., 1997). They concluded that there is currently ‘‘a misguided pre-occupation with scales rather than the concepts being scaled, and too much reliance on unsubstantiated claims of conceptual equivalence’’. Item equivalence—defined as occurring ‘‘when items measure the same parameters on the latent trait being examined, and when they are equally relevant and accepted in both cultures’’ (Bowden et al., 2003)—is especially relevant to the use of HRQOL instruments in cultures very different from those in which the items were generated. Despite claims by the original authors to the contrary, Bowden et al. found that lack of equivalence was particularly common in studies on non-western population samples. It should be noted in passing that the inherent limitations of ‘‘standardized’’ quality of life scales are not restricted to their use in translation, and in particular that statistical manipulation alone cannot demonstrate the validity of such scales (Bernheim, 1999; Wade, 2003). In relation to our own research, ‘‘validated’’ well-being scales [e.g., the Well-being Scale (Bech, Gudex, & Johansen, 1996) or the Audit of Diabetes Dependent Quality Of Life Scale (Bradley et al., 2002)] did not appear to reflect the types of distress we had picked up in our qualitative study. In other words, ‘‘well-being’’ in our Bangladeshi participants had a different subjective meaning and was differently socially enacted than ‘‘well-being’’ in the groups on which conventional scales had been validated. For example, a common statement made to us by Bangladeshis was ‘‘I have not been out of the house since developing diabetes’’, suggesting that restricted social activity was a key construct that we should try to capture. Measuring cultural adherence Implicit hypotheses about ‘‘culture’’ as the underlying cause of health-related behaviour (such as ‘‘this patient from a minority ethnic group is exhibiting behaviour X; it’s obviously a cultural thing and will diminish as the person becomes more westernized’’) are ubiquitous in clinical practice, especially in relation to diabetes (Greenhalgh et al., 2001). An alternative view is that once socio-economic, education, gender and opportunity differences have been controlled for, ‘‘culture’’ explains few ethnic health differentials, and ‘‘acculturation’’ is merely a proxy for acquiring economic and social capital (Ahmad, 1996; Donovan, 1986; Greenhalgh et al., 2001; Lambert & Sevak, 1996). In order to adjudicate between these hypotheses, a measure of acculturation is needed. In a recent review of acculturation measures in Asian immigrants (most of which have been developed for use in mental health and substance abuse research), Salant and Lauderdale distinguish between acculturation (one cultural group adopting the customs of another), assimilation (integration of a migrant group into the host society through behaviours, social structures and institutions), and ‘‘westernization’’ (adoption of a European or American lifestyle; Salant & Lauderdale, 2003). Scales designed to measure these constructs tend to be developed by researchers whose own ethnicity corresponds to that of the host population, and focus on the extent to which the immigrant has ‘‘picked up’’
Story-based scales
435
host culture. There is often an erroneous assumption that the two cultures have a ‘‘zerosum’’ relationship, that is, the more features of the host culture the individual adopts, the less they will be influenced by traditional beliefs and practices. But superficial indicators of acculturation such as fluency in English or consumption of certain western foods do not necessarily indicate a parallel loss of the ethnic values, beliefs or behaviours that might account (positively or negatively) for health differentials. Our own research began when AMC, an anthropologist who had emigrated from Bangladesh, observed that members of his community could assimilate and ‘‘westernize’’ without losing deeper aspects of their own culture (such as religious food restrictions or arranged marriages). Indeed, many customs (fasting, wearing the hajib) are sometimes adhered to more closely and passionately as the years since immigration pass, and also more by second generation immigrants than their parents (Joppke, Morawska, & Layton-Henry, 2003). In our own empirical work, therefore, we deliberately focussed less on the British culture that had been acquired (‘‘assimilation’’) and more on the extent to which underlying Bangladeshi culture was retained—a construct we termed ‘‘adherence’’. In summary, the aim of this study (which was part of a wider programme of research into the determinants of health outcome in people from minority ethnic groups with diabetes) was to develop valid and reliable questionnaires to measure subjective health status and cultural adherence, using the story as the unit of inquiry.
Methods Study design The study design is summarized in Figure 1. Ethical approval was obtained from Tower Hamlets NHS Local Research Ethics Committee. AMC (a Bangladeshi postdoctoral anthropologist who is fluent in English, standard Bengali and Sylheti) led the fieldwork. Phase 1: Narrative interviews Phase 1 of the work—a set of 40 in-depth qualitative interviews with adult Bangladeshis with type 2 diabetes, recruited from GP practices in East London—was described in detail in two previous papers (Chowdhury et al., 2000; Greenhalgh et al., 1998). In this phase, we used a narrative interview format to obtain 40 in-depth accounts of the experience of diabetes in British Bangladeshis, and established that storytelling was an effective way to prompt focus group discussion about health-related topics (Chowdhury et al., 2000; Greenhalgh et al., 1998). We subsequently developed group-based storytelling as an educational intervention (Greenhalgh et al., 2005). In phase 1, we also explored a new technique called the ‘‘structured vignette’’, in which we told a standardized story about a fictitious person with diabetes, presented as the same age and gender as the person being interviewed, to seek participants’ views on sensitive subjects such as impotence (Greenhalgh et al., 1998). The structured vignette potentially allowed the interviewee to project anxieties, fears and hopes onto a fictitious character by explaining how this individual ‘‘would feel’’ about particular issues (Anderson et al., 1951; Herskovits, 1950). Phase 2: Developing draft vignettes In phase 2, we extended and formalized the structured vignette approach by drawing on a technique that Winter has called fictional-critical writing (Winter, 1986). Using our phase 1
436
T. Greenhalgh et al.
Figure 1. Summary of study design.
narrative data and the NUDISTTM software, we first conducted a thematic content analysis on all sections of text relating to either subjective health status (for the well-being scale) or adherence to Bangladeshi culture (for the cultural adherence scale). The technique is identical to that used in the analysis of a text by thematic content analysis (Green & Thorogood, 2004). The thematic analysis was undertaken separately by TG and AMC on translated transcripts, and differences resolved by discussion, returning to the original Sylheti tape-recording where necessary. Having identified the key themes, we then wrote short vignettes in which each theme was represented. In other words, we systematically extracted the key elements of real stories told by people with diabetes and then fed these into new stories about comparable fictional characters. Phase 3: Refining the vignettes to form questionnaire items In phase 3 (Figure 1), we piloted and refined these draft vignettes in three focus groups of eight participants (one male, one female, and one mixed), held in local community centres and conducted by AMC with TG observing and taking field notes. Our sample comprised 24 Bangladeshi adults with diabetes (12 of each gender) recruited by ‘‘snowballing’’ from our phase 1 participants, that is, asking two participants in the original study to nominate a friend, and continuing until we had eight participants for each focus group. We did not collect demographic data on these individuals other than ensuring an equal gender balance and confirming that they had diabetes. To illustrate how the focus groups were used to refine the vignettes, the story about the man who sometimes takes his children to eat at McDonald’s (Box 2) was suggested by participants effectively to identify individuals who occasionally ‘‘lapse’’ into eating non-halal meat. It held greater credibility with all three focus groups than our original story about a
Story-based scales
437
man who ‘‘occasionally’’ consumed non-halal meat when he went out to dinner—a scenario that was deemed implausible by participants. Phase 4: Validating the questionnaires For phase 4, we sought a representative sample of Bangladeshi adults with diabetes. We planned to recruit these from diabetes registers in seven general practices in Tower Hamlets, an area of inner London with a large Bangladeshi population. The seven practices were chosen because of the 15 practices in Tower Hamlets at the time of the study, these were known to serve a predominantly ethnic Bangladeshi population. We randomly sampled (either using a search function for computerized registers or manually for card index registers) one patient in six, and confirmed with the GP or practice nurse that (a) the patient was diabetic and (b) they were able to give informed consent and (c) they were not involved in any other research studies. We recruited by letter followed up by phone call. We approached 101 potential participants for testing the pilot questionnaire. Our target was 100 participants, in order to yield a sufficient case to variable ratio for factor analysis (Tabachnick & Fidell, 1996). Participants were interviewed in their own homes by AMC, in the Sylheti dialect except for two people who preferred to speak in English. Each item in the pilot questionnaire consisted of three story characters, each with a common Bangladeshi name and representing different levels of subjective feelings and associated behaviours in a particular domain. For the well-being scale, the stories were presented in two versions—one male and one female— and administered to participants of the corresponding gender. For the cultural adherence scale, only one version was used since the story items generally described the behaviour and values of a family rather than an individual. For each item, AMC read out the three story options to the participant, and asked ‘‘which of these people/families do you think you are most like?’’. They were also offered an intermediate option (‘‘between Mr A and Mr B’’), allowing five possible responses for each item (Box 1). Box 1. Example of story-based questionnaire item in well-being scale (anxiety dimension). Mrs Miah has had diabetes for some years. Today, she feels most nervous and anxious. She often feels afraid for no reason at all. She gets upset and panicky, and often feels she is falling apart. She is never calm, and is unable to sit still. She has trouble getting to sleep and is restless during the night. Mrs Hussein has had diabetes for some years. On some days she feels nervous and anxious, but on other days she feels fairly calm. Occasionally she feels afraid or panicky. She sometimes finds she can sit still and relax, although she does become restless at times. She sleeps well on some days but on other days she has some problems getting to sleep. Mrs Rahman has had diabetes for some years. She very rarely feels nervous or anxious, and she does not feel unduly afraid of anything. She is not a panicker. She is easily able to sit still and relax, and she has no trouble at all sleeping. Which of these people is most like you? ¤ ¤ ¤ ¤ ¤
Mrs Miah Someone between Mrs Miah and Mrs Hussein Mrs Hussein Someone between Mrs Hussein and Mrs Rahman Mrs Rahman
438
T. Greenhalgh et al.
Analysing the questionnaire data Using SPSS for Windows, we performed factor analyses to explore the underlying structure in the response to each story-based item, with a view to reducing the number of items in each scale. Initial analysis of the statistics showed that the data were suitable for factor analysis (Bartlett Test of Sphericity was 221.40 (df ¼ 21), p 5 .0005). The Kaiser – Meyer – Olkin Measure of Sampling Adequacy was .78 (approaching Kaiser’s ‘‘meritorious’’ range; Kaiser, 1974). Firstly, we performed a principal components analysis to examining the communalities of test items, that is, with a review to screening out items that had little in common with other items. The rationale for exclusion was a communality value of 5.5. Secondly, after excluding one item with a low communality we performed a principal components analysis and a varimax rotation. We used the rule that factors should have an eigenvalue of greater than 1.0 for extraction. For both analyses, all small loadings (5.3) were suppressed for ease of interpretation. We also performed Cronbach’s alpha tests as confirmatory measures of internal consistency. Finally, we undertook hypothesis-driven regression analyses to investigate the correlation between key demographic variables (age, gender, education level), cultural adherence, and well-being.
Main findings Response rates All seven GP practices approached agreed to participate in the study, but two did not have accessible diabetes registers so were excluded. Of those who were still living at the contact address, 98 of 101 individuals approached agreed to be interviewed (97% response rate), and 90 produced usable data (two were unable to understand the instructions and six became tired or distracted very early in the interview). The demographic details of these participants are given in Table I.
Table I. Demographic details of participants. Mean agea Gender Marital status Country of birth Years in UKa Education Time since diagnosis of diabetesa Type of treatment Body mass index Waist – hip ratio a
+SD.
52 (+11) years 47 female, 49 male 86 married, 9 widowed, 1 separated 90 Bangladesh, 6 UK 25 (+11) years Pre-secondary 55, secondary plus 41 7 (+5) years Diet only 17, tablets 76, insulin 3 25.8 (+3.3) kg/m2 0.94 (+0.05)
Story-based scales
439
Well-being scale Initial analysis of the statistics showed that the data were suitable for factor analysis (Bartlett Test of Sphericity was 262.09 (df ¼ 6) and was significant (p 5 .0005)). The Kaiser – Meyer – Olkin Measure of Sampling Adequacy was .81, which is in Kaiser’s ‘‘meritorious’’ range (Kaiser, 1974). Principal components analysis yielded a one-factor solution accounting for 79.9% of the total variance between respondents. A follow-up Cronbach’s alpha test (.92) revealed ‘‘excellent’’ internal consistency, that is, all items appeared to be measuring aspects of the same construct. An example of an item on the wellbeing scale is shown in Box 1; the full scale (four items) is reproduced in Appendix 1. The scale comprises two dimensions of negative well-being (‘‘depression’’ and ‘‘anxiety’’), as well as physical energy and social activity. The definitive scale scores 1 through 5 on each of these dimensions, giving an overall well-being scale ranging from 4 to 20. More than two in five (42%) of our respondents scored below 12 (the mid-point) on this scale, confirming our previous impressions that low well-being is common in ethnic Bangladeshis with diabetes in this deprived area of London.
Cultural adherence scale Exploratory factor analysis on the definitive dataset revealed a two-factor model that accounted for 66.0% of the variance in scores. Factor one (accounting for 49.6% of the variance) comprised items 1 – 5 (religious food restrictions, ethnic food choices, food shopping, upbringing of children, and marriage and social ties), each scoring from 1 to 5 and making a possible score range of 5 to 25. A follow-up Cronbach’s alpha test showed ‘‘good’’ internal consistency (.83). Factor two (accounting for 16.4% of the variance) comprised items 6 and 7 (beliefs about cause and treatment of illness), making a possible score range of 2 to 10. A followup Cronbach’s alpha test showed ‘‘acceptable’’ internal consistency (.72). We decided to omit the second factor given its lower internal consistency and the proportion of variance accounted for. Together with the omission of the item with a low communality (preference for Bangladeshi doctor) we reduce the overall number of items from eight to five. The final scale (factor one only) has a possible score range of 5 (very strict adherence to Bangladeshi culture) to 25 (no adherence). An example of an item from this scale is shown in Box 2, and the scale is reproduced in full in Appendix 2. A regression analysis confirmed our decision to drop the second cultural adherence factor. We used the criterion variable of ‘‘well-being score’’ with predictor variables cultural adherence factor one, cultural adherence factor two, level of education (pre-secondary vs. post-secondary education), gender, age, years in UK, and years since diagnosis. The resultant model yielded an R-value of .59, explaining 34.2% of the variance in well-being (R2). The only two significant predictors of well-being were educational level (p 5 .01) and cultural adherence factor one ( p 5 .01). Our findings highlighted the important distinction between cultural adherence (following particular customs of one’s own social group) and cultural assimilation (adopting aspects of
440
T. Greenhalgh et al.
Box 2. Example of full text of story-based questionnaire item in cultural adherence scale (ethnic food choices). At Mrs Miah’s home, the main meals are always rice and curry. Her daughter wants to make green salad but Mrs Miah says green salad does not go well with Bangladeshi food. For guests and on special occasions she always cooks Bangladeshi dishes. In Mrs Rahman’s home rice and curry meals are generally eaten. But at weekends, she makes fish and chips, spaghetti, or pizza, which her sons fancy. She and her husband like to have green salad with these English foods. For guests and on special occasions, Mrs Rahman prepares cutlets and roast beef alongside Bangladeshi dishes. Mr Karim likes good food. Mrs Karim prepares all sorts of food from the cookery books. She makes very good grilled chicken and fish kebab. She is proud of her vegetable soup. Which of these people is most like you? ¤ ¤ ¤ ¤ ¤
Mrs Miah Someone between Mrs Miah and Mrs Rahman Mrs Rahman Someone between Mrs Rahman and Mrs Karim Mrs Karim
Our findings highlighted the important distinction between cultural adherence (following particular customs of one’s own social group) and cultural assimilation (adopting aspects of the lifestyle of the indigenous population). For example, we demonstrated only a moderate negative correlation (Pearson coefficient ¼ .49, p 5 .0005, N ¼ 87) between religious adherence (measured by consumption of halal meat) and assimilation into British food habits (e.g., eating salad and pizza, or shopping in British supermarkets). In other words, these superficially ‘‘westernized’’ food habits explained less than 25% of the variance in how strictly these participants followed the halal restriction.
Discussion This study has developed and validated two easy-to-administer, story-based scales that were highly acceptable to participants and which measured well-being and cultural adherence in British Bangladeshis with diabetes. Our method potentially has much wider applications for developing story-based instruments to tap into the subjective experience of illness in nonwestern populations—and perhaps in other groups for whom closed-item questionnaires prove difficult. Jerome Bruner, an early protagonist for the centrality of meaning in research into human behaviour, has criticized his fellow psychologists for building a discipline based on a stimulus-response paradigm, in which meaning is ‘‘black-boxed’’ out of the focus of enquiry (Bruner, 1990). Because the meaning of experience is socially constructed and culturally shared, ‘‘psychometric equivalence’’ of HRQOL scales across cultural groups cannot be demonstrated simply by confirming that similar proportions of people tick the different boxes when the instrument is translated. The premise on which we based our present work was that when measuring subjective health status in culturally disparate groups, an entirely new set of items must be developed, based on a detailed qualitative investigation of the
Story-based scales
441
subjective meaning of the symptoms and experiences being studied within the target population. It as been argued that the story is the most efficient tool yet discovered for conveying the complexity of human experience (Charon, 2001). Stories are image-rich; they create engagement by drawing the reader or listener into the plight of the character; they are especially powerful for depicting suffering and sorrow; and they are sense-making devices that give form and coherence to disparate perceptions and feelings (Greenhalgh & Hurwitz, 1999). Common to many of the previous studies on vignette-based scales (see Background) is an attempt to convey, in a standardized form, a complex social phenomenon that is assumed to influence an action or action-related decision. In other words, vignettebased questionnaires appear to have emerged mainly in research that asks, ‘‘What would you do if . . . ?’’. Interestingly, our early research had suggested that items based entirely on feelings held little meaning for many Bangladeshi respondents, but that feelings could be brought alive by vignettes depicting how a person who feels like this would do. The production of entirely new scales involved a trade-off: it is impossible to compare our findings directly with other published studies, but we believe that these findings have greater validity in the population we studied than would have been achieved through an ‘‘off-thepeg’’ scale (Bowden et al., 2003). Whilst we fully agree with authors who call for the use of standardized instruments as far as possible in researching HRQOL (Garratt et al., 2002; Scientific Advisory Committee of the Medical Outcomes Trust, 2002), we are concerned that ‘‘standardization’’ for the purposes of cross-cultural research should not be achieved at the expense of the validity of the instrument in any particular cultural group. However, whilst the content of any instrument is not transferable, we believe the method for developing story-based scales can be transferred to many different contexts and cultural groups—and may well prove universal in its applications. In summary, this preliminary study challenges conventional approaches to the measurement of subjective health states across cultural and linguistic barriers, and raises some epistemological and methodological questions about the use of stories in questionnaire-based research. We are undertaking further analyses to link the findings from this survey with health-related beliefs and behaviours in the same sample. We invite others to explore further the techniques introduced here, with a view to identifying the particular situations in which stories can add value to questionnaire research. Acknowledgements This study was funded by the Wellcome Trust. We are grateful to the GP practices and study participants for their time and co-operation. We also thank two anonymous referees for helpful comments on previous drafts of this paper. References Ahmad, W. I. U. (1996). The trouble with culture. In D. Kelleher & S. Hillier (Eds.), Researching cultural differences in health (pp. 190 – 219). London: Routledge. Anderson, H. H., & Anderson, G. L. (1951). An introduction to projective techniques and other devices for understanding human behaviour. Englewood Cliffs: Prentice Hall. Bech, P., Gudex, C., & Johansen, K. S. (1996). The WHO (Ten) Well-Being Index: Validation in diabetes. Psychotherapy & Psychosomatics, 65, 183 – 190.
442
T. Greenhalgh et al.
Bernheim, J. L. (1999). How to get serious answers to the serious question: ‘‘How have you been?’’: Subjective quality of life (QOL) as an individual experiential emergent construct. Bioethics, 13, 272 – 287. Blaxter, M. (1994). Black and minority ethnic groups in England: Health and lifestyles. London: Health Education Authority. Bowden, A., & Fox-Rushby, J. A. (2003). A systematic and critical review of the process of translation and adaptation of generic health-related quality of life measures in Africa, Asia, Eastern Europe, the Middle East, South America. [Review]. Social Science & Medicine, 57, 1289 – 1306. Boynton, P. M., Wood, G. W., & Greenhalgh, T. (2004). A hands on guide to questionnaire research part three: Reaching beyond the white middle classes. British Medical Journal, 328, 1433 – 1436. Bradley, C., & Speight, J. (2002). Patient perceptions of diabetes and diabetes therapy: Assessing quality of life. Diabetes/Metabolism Research Reviews, 18 Suppl 3, S64 – S69. Bruner, J. (1990). Acts of meaning. Cambridge: Harvard University Press. Charon, R. (2001). Narrative medicine: Form, function, and ethics. Annals of Internal Medicine, 134, 83 – 87. Chowdhury, A. M., Helman, C., & Greenhalgh, T. (2000). Food beliefs and practices in British Bangladeshis with diabetes: An ethnographic analysis. Medical Anthropology, 7, 209 – 226. Donovan, J. (1986). We don’t buy sickness—it just comes. Aldershot: Gower Press. Eisemann, M., Richter, J., Bauer, B., Bonelli, R. M., & Porzsolt, F. (1999). Physicians’ decision-making in incompetent elderly patients: A comparative study between Austria, Germany (East, West), and Sweden. International Psychogeriatrics, 11, 313 – 324. Garratt, A., Schmidt, L., Mackintosh, A., & Fitzpatrick, R. (2002). Quality of life measurement: Bibliographic study of patient assessed health outcome measures. British Medical Journal, 324, 1417. Green, J., & Thorogood, N. (2004). Qualitative methods for health research. London: Sage. Greenhalgh, T., Collard, A., & Begum, N. (2005). Sharing stories: Complex intervention for diabetes education in minority ethnic groups who do not speak English. British Medical Journal, 330, 628 – 631. Greenhalgh, T., Griffiths, T., Thomas, K., Wheeler, J., & Raleigh, V. (2001). Reducing inequalities in diabetes. Report of the Expert Reference Group for the National Service Framework for Diabetes. London: Department of Health. Greenhalgh, T., Helman, C., & Chowdhury, A. M. (1998). Health beliefs and folk models of diabetes in British Bangladeshis: A qualitative study. British Medical Journal, 316, 978 – 983. Greenhalgh, T., & Hurwitz, B. (1999). Narrative based medicine: Why study narrative? British Medical Journal, 318, 48 – 50. Hendrickson, S. G. (2003). Beyond translation . . . cultural fit. Western Journal of Nursing Research, 25, 593 – 608. Herdman, M., Fox-Rushby, J. A., & Badia, X. (1997). ‘Equivalence’ and the translation and adaptation of healthrelated quality of life questionnaires. Quality of Life Research, 9, 316 – 322. Herskovits, M. J. (1950). The hypothetical situation: A technique of field research. Southwest Journal of Anthropology, 6, 32 – 40. Joppke, C., Morawska, E., & Layton-Henry, Z. (2003). Towards assimilation and citizenship: Immigrants in the liberal nation-states. London: Palgrave. Kaiser, H. (1974). An index of factorial simplicity. Psychometrika, 39, 31 – 36. Kind, P., Dolan, P., Gudex, C., & Williams, A. (1998). Variations in population health status: Results from a United Kingdom national questionnaire survey. British Medical Journal, 316, 736 – 741. Lambert, H., & Sevak, L. (1996). Is ‘cultural difference’ a useful concept. In D. Kelleher & S. Hillier (Eds.), Researching cultural differences in health. London: Routledge. Love, M. B., & Thurman, Q. (1991). Normative beliefs about factors that affect health and longevity. Health Education Quarterly, 18, 183 – 194. Muller, J. (1999). Narrative approaches to qualitative research in primary care. In B. F. Crabtree & W. L. Miller (Eds.), Doing qualitative research (2nd ed., pp. 221 – 238). London: Sage. Parker, G., Mahendran, R., Yeo, S. G., Loh, M. I., & Jorm, A. F. (1999). Diagnosis and treatment of mental disorders: A survey of Singapore mental health professionals. Social Psychiatry and Psychiatric Epidemiology, 34, 555 – 563. Paterson, C., & Britten, N. (2000). In pursuit of patient-centred outcomes: A qualitative evaluation of the ‘Measure Yourself Medical Outcome Profile’. Journal of Health Services Research & Policy, 5, 27 – 36. Richardson, B., Kitchen, G., & Livingston, G. (2003). Developing the KAMA instrument (knowledge and management of abuse). Age & Ageing, 32, 286 – 291. Salant, T., & Lauderdale, D. S. (2003). Measuring culture: A critical review of acculturation and health in Asian immigrant populations. Social Science & Medicine, 57, 71 – 90.
Story-based scales
443
Salomon, J. A., Tandon, A., & Murray, C. J. (2004). Comparability of self rated health: Cross sectional multicountry survey using anchoring vignettes. British Medical Journal, 328, 258. Schutzwohl, M., Jarosz-Nowak, J., Briscoe, J., Szajowski, K., Kallert, T., & Eden Study Group. (2003). Inter-rater reliability of the Brief Psychiatric Rating Scale and the Groningen Social Disabilities Schedule in a European multi-site randomized controlled trial on the effectiveness of acute psychiatric day hospitals. International Journal of Methods in Psychiatric Research, 12, 197 – 207. Scientific Advisory Committee of the Medical Outcomes Trust. (2002). Assessing health status and quality-of-life instruments: Attributes and review criteria. Quality of Life Research, 11, 193 – 205. Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics (3rd ed.). New York: Harper & Row. Wade, D. T. (2003). Outcome measures for clinical rehabilitation trials: Impairment, function, quality of life, or value? American Journal of Physical Medicine & Rehabilitation, 82, S26 – S31. Winter, R. (1986). Fictional-critical writing: An approach to case study research by practitioners. Cambridge Journal of Education, 3, 175 – 182.
Anxiety dimension
Depression dimension
Score
2
17
9
19
Mrs Miah has had diabe- Between Mrs Miah and Mrs Hussein tes for some years. Today, she feels most nervous and anxious. She often feels afraid for no reason at all. She gets upset and panicky, and often feels she is falling apart. She is never calm, and is unable to sit still. She has trouble getting to sleep and is restless during the night.
11
Between Mrs Ali and Mrs Ali developed Mrs Uddin diabetes last year. Today, she feels most downhearted. She feels that people do not need her and she no longer plays a role in her family or her local community. Her life feels empty. She cries on most days, and there is very little in her life that she enjoys.
1
3
4
10
25
21
Between Mrs Hussein Mrs Hussein has had and Mrs Rahman diabetes for some years. On some days she feels nervous and anxious, but on other days she feels fairly calm. Occasionally she feels afraid or panicky. She sometimes finds she can sit still and relax, although she does become restless at times. She sleeps well on some days but on other days she has some problems getting to sleep.
27
Between Mrs Uddin Mrs Uddin developed and Mrs Ahmed diabetes last year. Today, she occasionally feels downhearted but she knows that she still has a fairly important role in her family and her community. On some days she feels that life is empty but on other days she feels somewhat fulfilled. She enjoys some aspects of life but she does cry frequently because of her situation.
Appendix 1. Diabetes well-being scale, showing frequency of responses in our sample.
16 (continued )
Mrs Rahman has had diabetes for some years. She very rarely feels nervous or anxious, and she does not feel unduly afraid of anything. She is not a panicker. She is easily able to sit still and relax, and she has no trouble at all sleeping.
25
Mrs Ahmed developed diabetes last year. Today, she almost never feels downhearted. She plays a very important role in her family and her community, and is always busy. A lot of people depend on her. She enjoys most aspects of her life, and only cries on rare occasions.
5
444 T. Greenhalgh et al.
2
14
18
Between Mrs Begum Mrs Begum developed and Mrs Lais diabetes a few months ago. Today, she feels she has no energy at all. She describes her mood as sluggish and dull. She seems to be constantly tired, and even after a night’s sleep she does not wake up refreshed.
1
Between Mrs Zahan and Social activities dimension Mrs Zahan is known to Mrs Nessa have mild diabetes. She is most unhappy and is dissatisfied with her personal life. She greatly regrets the kind of life she has led and does not feel that she has adjusted at all well to her present situation. She has very little enthusiasm to tackle her
Energy dimension
Score
Appendix 1. (Continued ). 4
31
Between Mrs Nessa and Mrs Nessa is known to Mrs Khanum have mild diabetes. There are some things about her life situation that she regrets, but she feels there are a few things to be pleased about. She thinks that she copes reasonably well with her life. However, things get her down occasionally. Her
17
Between Mrs Lais and Mrs Lais developed Mrs Hamed diabetes a few months ago. Today, her energy level is less than it used to be, but she still feels able to get on with her life. On some days she feels quite sluggish, but on other days she is more lively. In general, she feels neither exhausted nor energetic. Although at times she wakes up feeling tired, she often gets a refreshing night’s sleep.
3
(continued )
Mrs Hamed developed diabetes some months ago. Today, she has as much energy as she did before she was told she had diabetes. She describes herself as a very active person, and she almost always wakes up feeling refreshed and ready for an active day.
10
Mrs Hamed developed diabetes some months ago. Today, she has as much energy as she did before she was told she had diabetes. She describes herself as a very active person, and she almost always wakes up feeling refreshed and ready for an active day.
5
Story-based scales 445
Score
Appendix 1. (Continued ).
23
22
10
3 level of enthusiasm for her daily tasks varies, and although on some days she can face problems and new situations confidently, there are other times when she would rather not have to make any decisions at all.
2
daily tasks or make new decisions, and starts to worry whenever she has to cope with problems or changes in her life. There seems to be nothing interesting happening in her life.
1
16
4
19
5
446 T. Greenhalgh et al.
Result
Upbringing of children
Result
Food (shopping)
Result
Food (ethnic choices)
Result
Food (religious restrictions)
15
Between Ali and Uddin
2
53
The Lais family speak only Sylheti at home; educate their daughters separately; children attend Mosque to study the Quran every evening
20
The Zahan family buy all meat, bread, biscuits and fruit from Bangladeshi shop
25
12
The Moin children speak Sylheti to the parents but English among themselves. They have an occasional home tutor to learn Quran
Between Lais and Moin
Mr Uddin buys meat from Bangladeshi shop but goes to English fishmonger and uses Sainsburys for fruit and vegetables
The Rahman family eat mostly Bangladeshi dishes but at weekends have pizza etc. They serve Bandgladeshi and English food to their guests 33
10
28
10
4
6
Between Moin and Ali
6
Between Uddin and Khanum
4
Between Rahman and Karim
2
The Uddin family eat Between Uddin and mostly halal meat at Ahmad home but allow children to eat at McDonalds
3
28
Between Zahan and Uddin
20
Between Miah and The Miah family eat strictly Bangladeshi Rahman dishes at home and serve this kind of food to their guests
57
The Ali family always adhere strictly to halal meat
1
Appendix 2. Diabetes cultural adherence scale (factor 1), showing frequency of responses in our sample.
5 (continued )
The Ali children attend a mixed school; they hope that their daughter will go to university. Parents do not insist on Sylheti at home
4
Mr Khanum usually buys all the shopping from Sainsburys
4
Mrs Karim makes all sorts of food from cookery books, both English and Bangladeshi.
The Ahmad family buy frozen chicken and duck from Sainsburys; Mr Ahmad eats beef from work canteen. 2
5
Story-based scales 447
2
62
6
The Miah parents expect Between Miah and Khanum to find a Bangladeshi marriage partner for their son from friends in UK or back home. They would not allow intermarriage and would expect daughter-in-law to live with them. All their friends are Bangladeshi
1
4
11
1
The Khanum parents also Between Khanum and Ahmad hope to find their son a marriage partner but want him to be happy and would not insist that their daughter-inlaw live at home. They accept that if their son is to get on in life he will have English friends.
3
Note: the stories in the above summary table have been abbreviated for clarity; the full text of the stories is obtainable from the authors.
Result
Marriage and social ties
Appendix 1. 2. (Continued ).
6
The Ahmad parents will allow their daughter to go on a trip round Europe with male and female friends from college. They wish their daughter to become a doctor and understand that she may fall in love with an English doctor
5
448 T. Greenhalgh et al.
