Trisha Greenhalgh, Cecil Helman, A Mu’min Chowdhury

Trisha Greenhalgh,
senior lecturer

Cecil Helman,
senior lecturer

A Mu’min Chowdhury,
research fellow

Correspondence to: Dr Greenhalgh p.greenhalgh@ucl.ac.uk

Successful diabetes management requires that we understand the lifestyle, beliefs, attitudes, and family and social networks of the patients being treated.⁽¹⁾ Type 2 diabetes is three to six times as prevalent in British south Asians as in the white British population, and morbidity from its complications is high in these ethnic groups.⁽²⁾

Qualitative methods are particularly useful when the field of research is relatively unexplored and the research question is loosely defined or open-ended.⁽³⁾ With the exception of two recently published studies from Canada,^{(4) (5)} and a small UK study based entirely on individual interviews,⁽⁶⁾ such methods have rarely been used in the study of patients with diabetes. This paper reports a qualitative study into the health beliefs and folk models of diabetes of a community sample of 40 British Bangladeshi patients with diabetes, together with a smaller sample of non-Bangladeshi controls, with a view to exploring their experience of diabetes, identifying psychological, societal and material barriers to optimum health outcomes and, in the longer term, improving services.

Anthropological analysis accepts that there are three levels of cultural behaviour: what people say they do (for example, during an interview), what they are actually observed to do, and the underlying belief system which drives that behaviour (Hall’s "primary level culture"⁽⁷⁾). In addition, consideration must be given to the wider context in which the behaviour takes place. In particular, the British Bangladeshi informants in this study must be viewed as members of an atomistic rural society living as recent immigrants in a socioeconomically deprived urban environment (see box). The challenge of the medical anthropologist is to tap into primary level culture as well as simply recording stated beliefs or observed behaviour.⁽⁹⁾

The frontiers of present day Bangladesh were drawn after the second world war, when British India was partitioned. The Muslim majority of Bengal, along with Sylhet district in the far north east, came to form East Pakistan. In 1971 Bengal seceded from Pakistan and became the separate state of Bangladesh. The country is flat, with a monsoon climate, prone to flooding, and served mainly by inland waterways. The economy is pre-industrial, and most people live in scattered homesteads with an atomistic social organisation (that is, the family is the dominant unit with no effective social organisation or hierarchy beyond the family). The staple crop is rice, and the diet is largely fish, rice, and vegetables. Although about 95% of the population is Muslim, the society contains vestiges of its Buddhist and Hindu cultural roots. In the 1960s and ’70s, large numbers of economic migrants came to Britain, particularly from certain villages in rural Sylhet. Men tended to emigrate several years before their wives followed.

Data from the 1991 census suggest that British Bangladeshis account for about 0.3% of the population of England and Wales,⁽⁸⁾ and about a quarter of the population of Tower Hamlets (East London and City Health Authority; unpublished estimates for 1997 based on projections from 1991 census data).

After gaining approval from local research ethics committees, we recruited patients from three GP practices in East London known to have a high proportion of Bangladeshis on their list. In one practice the GP, and in the others a nurse or an advocate-interpreter based at the practice, spoke Sylheti. Using computerised diabetes registers where available, and otherwise by manual search of case notes, we identified patients with diabetes and approached them to request a tape-recorded interview.

We found in pilot studies that a written invitation alone produced a poor (virtually nil) response from Bangladeshi patients, and that appointments booked several days in advance for an external venue were often not kept. In the definitive study, therefore, we sent a written invitation and followed it with a telephone call with a view to conducting an interview at the informant’s home. In addition, in one practice, we recruited opportunistically—i.e. one of us (AMC) waited at a venue close to the surgery and when eligible patients arrived to book an appointment or collect a prescription, they were asked by the practice receptionist if they would give an interview straight away.

In all cases it was made clear to the potential subjects that there was no obligation to be interviewed. Despite this, 4 of 44 subjects approached agreed to give an immediate interview and an additional 36 agreed to home interview at a later date. In about a third of cases, informants were not at home at the agreed time and the researcher rearranged the interview later by telephone or by calling back in the evening. All those who agreed to give an interview (91% of those approached) eventually did so.

Initial interviews (narrative recording, semi-structured interview, pile sorting and genogram) took between 90 minutes and 3 hours to complete, including regular breaks. Since virtually all informants were unemployed, working in the home, or retired, we did not experience many time constraints in completing the interview, although in seven cases the interviewer revisited on a second occasion.

We interviewed 40 Bangladeshis and a control group of 8 white British and 2 African-Caribbean subjects, all of whom lived in East London and had similar socioeconomic background. Subjects were aged between 24 and 78, with approximately equal numbers of males and females. We used purposive sampling methods (i.e. we intentionally sought to interview subjects with certain characteristics) so as to ensure as wide a range of experiences and attitudes as possible. The main variables that directed our sampling are listed in table 1^(T1).

In practice, our sample as recruited opportunistically reflected a wide range of demographic variables and illness experience. However, some additional effort was required to recruit females since they were less likely to attend the GP surgery. Somewhat surprisingly, virtually all the Bangladeshis on practice diabetes registers had telephones. Only one second generation immigrant with diabetes was identified.

The research methods used are summarised in the box. Since a high proportion of subjects were illiterate, we gave a verbal explanation of the aims of the study along with assurance that there was no obligation to continue with the interview, and then requested confirmation of understanding and consent to proceed, all of which was recorded on tape. Rather than use a standard list of questions, we allowed the subject to tell his or her story in their own words and in no particular order, but we used a checklist of semistructured prompting questions to make sure that the domains in the box were covered at some stage by all subjects.

Personal medical history

Psychological reaction to diagnosis of diabetes

Knowledge about causes, complications and treatment objectives in diabetes

Body image and beliefs about physiological and pathological processes

Attitude to dietary restriction

Attitude to physical exercise

Perceived social constraints resulting from diabetes

Satisfaction with current diabetes service

Experience of, and attitude to, health professionals

Disease ranking—Diabetes is ranked against 10 other medical conditions (heart attack, gastric ulcer, flu, asthma, gall stones, back pain, tuberculosis, cancer, stroke, and malaria)

Foods—Raw foodstuffs are grouped into "permitted" and "not permitted" and the classification then discussed

Meal menus—Above exercise is repeated with complete meals

"Preferred" and "healthy" body size—A selection of eight photographs of Bangladeshis (all of similar age and same sex as informant but of varying body mass index) is sorted into "most [aesthetically] preferred" through to "least preferred" and again into "most healthy" through to "least healthy"

Focus groups were convened by asking individual informants if they would attend a daytime or evening group meeting at a local Bangladeshi community centre. Many agreed, but actual attendance was only about 40%, and some groups had to be cancelled because there were fewer than three participants. There were no significant differences between attenders and non-attenders in the variables shown in table 1^(T1).

Interviews with Bangladeshi subjects were conducted in Sylheti, a dialect of Bengali spoken as a first language by all our Bangladeshi subjects. Since Sylheti has no written form, the interviews were simultaneously translated and transcribed by an independent translator, and were all checked by AMC (a Sylheti anthropologist), who listened to the original recording while reading the draft translation. Errors in transcription or translation were simply corrected, but areas of linguistic ambiguity (i.e. where the translator had used one word or phrase but AMC would have chosen a different one) were marked and considered further.

The transcripts of one-to-one and focus group interviews were analysed using NUDIST software. The entire text of the interview was entered onto a computer database and text blocks were then coded into 11 broad categories of statement such as body image, information sources, professional roles, and so on. The software allows analysis by cross-referencing within the different coding categories, and also by text word search in the document or selected sections of it.

The objective of the analysis was to identify constructs—i.e. provisional inferences about primary level culture drawn from statements and observations.⁽¹¹⁾ In particular, we wished to explore the informants’ explanatory models of diabetes – i.e. answers to questions such as "What has happened, why has it happened to me, what should I do about it?" and so on.⁽¹²⁾ By manually reading through the transcripts, we identified potential constructs and used the software search facility to collate relevant statements, mainly using text words and truncation stems (e.g. diabet$ for diabetic, diabetes, etc) and also using our original coding categories. All statements relevant to each construct were then considered together, and the construct modified accordingly. We also searched for incomplete typologies or "silences" in the data—i.e. we noted not just what informants talked about, but also what they omitted to talk about.⁽¹³⁾

Analysis of the translated transcripts was undertaken initially by one researcher (TG) and preliminary impressions offered to AMC, who reflected on these, returning to the original field data where necessary. Frequent discussions were held between all three researchers when constructs were being formulated. An iterative approach to data collection was used, in that the domains sought through prompted narrative and focus group discussions were modified and expanded as the study progressed on the basis of preliminary analysis of interviews already completed.

An important technique for demonstrating the validity of qualitative findings is triangulation—comparing data obtained by one method with similar data obtained by another method.⁽¹⁴⁾ In formulating the constructs in this study, we used a number of additional methods. Thus, for example, a statement from a subject such as "diabetes isn’t a very serious disease" could be compared with that subject’s ranking of diabetes against conditions like flu and heart attack, and by their challenge (or lack of challenge) to a contrary statement made by fellow participant in a focus group.

After developing the constructs discussed here, we presented them to a smaller sample of the subjects to determine whether our interpretation of the initial interviews had been correct. For this, we used two methods, the first being a further set of gender specific focus groups in which we presented our initial constructs and recorded the group’s responses on videotape.

In the second, we developed the new qualitative technique of structured vignette, in which the constructs we were attempting to validate were presented in the form of a story recorded on tape about Mr (or Mrs) Ali, a person with diabetes. The story is first played in full, and then played back slowly, sentence by sentence. After each sentence, the tape is stopped and the subject is asked, "do you agree that this person would have acted in this way / thought this / etc?". A sample paragraph is reproduced in the appendix. The vignette includes a number of deliberately incorrect statements inserted to check that subjects are not simply agreeing with all the statements.

This method was developed to avoid the problems, which have been well-documented previously in non-European cultures,⁽¹⁵⁾ of asking informants to respond to closed questions about their own beliefs or behaviour, which would require them to challenge directly statements made by the interviewer. We performed the structured vignette study on a sample of 18 subjects, and repeated it on 10 of these same subjects after an interval of two months. The internal reliability of the technique was high (overall, 89% of questions received identical answers on repeat interview).

A number of general and diabetes-specific themes derived from the data are listed below. Unless otherwise stated, the results refer to the Bangladeshi informants rather than to the comparison group. Original quotes are designated by informant’s ethnicity (e.g. B=Bangladeshi), gender (e.g. M=male), identification number and line of transcript. Quotes from focus groups are further designated FG.

The desire to understand and explain the onset and experience of illness was often strong. However, it tended not lead to a systematic search for professional or scientific explanations but rather to a reflection on personal experience and the experiences of friends and relatives. Lay sources of information were frequently cited as a major influence on behaviour. In the structured vignette study, 17 of 18 informants agreed that the best way to find out about diabetes was to ask friends and relatives. Medical sources of information were occasionally, but not routinely, cited by the Bangladeshi informants, whereas the white British informants frequently justified statements by reference to a leaflet or a comment by a health professional.

"I heard it from my sisters and brothers or their families that these sort of cuts don’t cure for people with diabetes." (BF40/661)

"Since I have heard from my friend, I have stopped buying sweets and crisps for my children. I give them good foods, fruits, sugar free drinks and brown bread. This is how we live now." (BM30/120)

Whilst strong religious (Muslim) views were held by all the Bangladeshi informants, and explanations often given in terms of "God’s will", such views were usually held in parallel with acceptance of individual responsibility and potential for change. It was perceived as unseemly to complain about one’s lot and the duty of a good Muslim to cope as best one can. This stoical attitude was also strongly evident in the more deeply religious of the non-Bangladeshi informants, two of whom referred to diabetes as a test from God or "a cross I have to bear". This accords with Lambert and Sevak’s observation that the pious in any culture may have more in common with one another than with less religious members of their own cultures, and challenges the view that the Bangladeshis have a particularly fatalistic attitude to illness or that their religious determinism precludes lifestyle modification.

"Medicine is a means recommended by the Prophet’s tradition [sunnat]. God does not Himself come down to relieve you from illness, medicine is his means [uchila]. Having medicine and amulets [tabiz] is recommended acts for us. If you don’t, you will suffer so long you have allotted life [hayat] left." (BM15/368)

Informants displayed a range of different images of the body. Several used the analogy of a machine, believing that the body is "in mint condition" and needs little regular maintenance in youth, but erodes with age and use. Youth and health were usually viewed as virtually synonymous, and physical degeneration and weakness as an inevitable consequence of ageing.

"When the mangoes start turning red on the tree, you know it is time for them to drop. Man too is like this, they have time to go." (BM15/317)

"Once you are 40 eyes tend to give trouble. I am almost 55. So I am expected to have bad eyesight." (BM18/308)

In contrast, Crawford’s study of white women in the USA indicated that "health" for them is not merely the absence of illness but had to be earned by taking positive action in terms of diet and exercise in leisure time. If such behaviour was maintained, health would persist as the body aged.⁽¹⁶⁾

Both men and women chose photographs of large individuals when asked to "pick out the healthiest person". Large body size was generally viewed as an indicator of "more health" and thinness with "less health", but many also perceived that "too much health" (i.e. too large a body size) was undesirable, especially if the body is weakened by diabetes. Reference to the desirability of weight control was widely acknowledged in both individual and group interviews. Hence, health education messages which refer to a "healthy" (meaning thin) body size may be less meaningful to Bangladeshis than ones which speak directly of weight reduction.

"If you have too much of it [rich food], your illness will increase, having it less you will be healthy¼ because of illness less weight is better. They [health professionals at the clinic] explained it to me." (BM36/1145)

Airhihenbuwa has discussed further the phenomenon of immigrants holding simultaneously both "traditional" constructs (deeply rooted values and perceptions drawn from the culture of origin) and "recent" ones (drawn from the host culture and less likely to be enduring in the long term).⁽¹⁷⁾

Illness was generally attributed to events or agents outside the body rather than to primary failure of an organ within it. This model may reflect the predominance of acute infectious illness in the recent cultural history of this group. All informants believed that the primary cause of diabetes, and that of poor diabetic control, was too much sugar in the diet and, to a lesser extent, other dietary factors.

"Every disease has its own characteristic. If you have flu [shardi] and if you catch cold [tanda] it will get worse. Similarly when you have diabetes and take the wrong food it will get worse. Because of this I have to be discerning about food." (BM15/162)

The strong attribution to sugar in the diet, which has been described previously in other cultural groups,^{(18) (19)} may be partly due to the Bangladeshi lay term for diabetes, bahumutra, literally "sugar diabetes". Another important theme was the notion that a Western diet had caused diabetes, which has also been shown to feature strongly in the explanatory models of diabetes in both African Caribbeans in the UK and aboriginal cultures in north America.

Other aetiological factors mentioned by the Bangladeshi informants included heredity (the notion of an agent transmitted through "shared blood", rather than an inherited predisposition), and germs. A minority of informants were confused by the concept of a non-contagious illness.

"Any disease has certain germs. So long as the germ is there, urine will not be normal. When the germ is gone, the colour will be normal again." (BM11/155)

"I thought it is a disease [hinting at contagious nature] and so I took medicine. There is no medicine [implying curative of disease] for this. The tablet we have is a mere help not to produce sugar¼¼ How can sugar [in urine] be a disease? Actually it is not." (BM27/396)

In the structured vignette study, 13 of 18 informants agreed that Mr (or Mrs) Ali could pass diabetes onto his unborn children via the blood. Eight believed that his diabetes had been caused by a germ, but only two agreed that the Alis’ children could catch it from eating off the same plate as their affected parent (a possibility mentioned by one informant in the main interview).

Many informants mentioned physical or psychological stress, either as a perceived cause of diabetes or simply when reporting the experience of daily life—especially in relation to economic difficulties, poor housing, and fear of crime. A survey in Glasgow showed that several circumstances potentially associated with stress—length of working day, low income, crowded housing, liability to attack and perceived lack of social support (in women)—was greater in British Punjabis than in whites,⁽²⁰⁾ and the authors suggested that both stress and socioeconomic deprivation should be considered as potential major risk factors for diabetes in this ethnic group. The patient’s psychological well-being is now established as one of the cornerstones of successful diabetes management,⁽²¹⁾ yet both a definition of such well-being and a valid means of measuring it in a South Asian population (particularly since much stress appears to be social, rather than psychological, in origin)⁽²²⁾ are currently lacking. Further research into this issue is clearly needed.

The diagnosis of diabetes was generally seen as devastating, and the expression "I was spoiled" was used by several informants. Virtually all felt that diabetes was a chronic, incurable condition and a potential threat to life. They feared acute complications (collapse and "dropping dead"), and a few volunteered specific long-term sequelae in the heart, eyes and kidneys. However, all 18 informants in the structured vignette study failed to distinguish between diabetes-related damage to the heart as opposed to the lung, suggesting poor specific knowledge of internal organs or, alternatively, linguistic confusion with the term "lung" (which might mean "contents of the thoracic cavity" rather than a specific organ). Control of diabetes (and therefore reduction in disability and prolongation of life) was felt to lie in restoring the body’s internal balance via taking particular foods and fighting the "germ" with medicine.

Many informants expressed difficulty obtaining food that was both acceptable and palatable.

"I have not cut down on the amount of rice. What else can you eat here [in UK]? I cannot put brown bread in my mouth. I cannot take biscuits." (BF14/54)

Other practical difficulties included confusion over sickness benefits, language barriers when speaking to professionals, especially the use of children as interpreters, and the inability to understand leaflets, either because of the concepts presented or because the leaflets were printed in standard Bengali. Some informants said they were better able to understand the English alphabet (e.g. in road signs or notices) than standard Bengali. In the structured vignette study, 13 of 18 informants agreed that Mr Ali had probably received a leaflet but not found it helpful.

In all societies, food has social, religious and economic significance as well as nutritional value, and the patterns of harvesting, cooking and consuming food are a defining feature of different ethnic groups.⁽²³⁾ Our Bangladeshi informants held firm views about the acceptability and edibility of particular foods, which differ significantly from those held by dieticians.

The underlying food classification, as illustrated by the pile-sorting exercise, was determined by Muslim religious restrictions and also by perceptions of the food’s digestibility and its strength or nourishing power. "Strong" foods, perceived as energy-giving, included white sugar (in solid form), lamb, beef, ghee (derived from butter), solid fat, and spices. Such foods were considered health-giving and powerful for the healthy body but liable to produce worsening of illness in those debilitated by age or illness. They were thought particularly suitable for and desirable for young children, pregnant women and honoured guests. "Weak" foods, preferred in the everyday menu and for the old or infirm, included boiled (pre-fluffed) rice and cereals.

The digestibility of food was considered to be related to the cooking method, with boiled and steamed foods classified as easy to digest and raw, fried or baked foods hard to digest. The former were considered suitable for the elderly, infirm and young while the latter were preferred for healthy adults. Thus, the recommendation to bake or grill foods rather than fry them may not accord with cultural perceptions of digestibility. Excess strong, rich or spicy foods, and especially sugar, were incriminated in the aetiology of diabetes.

"People say the body’s sugar, it is the body’s power [shakti]." (BF14/155)

"I think that the fat and the ‘shakti’ [strength, power] have led to diabetes." (BM05/192)

The structured vignette study showed almost universal agreement that strong foods, solid fat and ghee should be avoided in diabetes. All 18 agreed that Mr Ali should not eat spicy foods because of his diabetes, and that a person with diabetes should eat a different diet from the rest of the family.

Foods were not grouped by informants according to western notions of nutritional content such as carbohydrate or protein. Some informants indicated that body components may be linked to certain foods because of physical similarity. For example, sugar, butter, ghee, body fat, bone marrow, semen, and white vaginal discharge were perceived by some to be the same fundamental entity, because their colour is the same and they all solidify when cool and liquefy when heated. Eight of 18 informants in the structured vignette study thought that molasses (a dark form of raw sugar, liquid at room temperature) was an acceptable substitute for sugar in the diet.

"Sugar is the white substance that is stored in the bone marrow, is it not? From this semen is produced. Since I have diabetes, I have come to think that [it is] because of using the semen more. When the ‘calcium’ inside the bone is exhausted at that time our diabetes starts." (BM05/FG2/197)

This statement brought general agreement in the focus group. These findings are consistent with Lambert’s work on the traditional South Asian "humoral" conceptions of health, which centre on the ecological flow of substances and qualities between the environment, food and the human body.⁽²⁴⁾

In traditional Bangladeshi culture, rice is the staple, and, for some, the only essential food. The term "eating cooked rice" is used synonymously with the term "main meal", and other foods (curries, spices) are taken as an accompaniment to the rice. The instruction to take less rice is rarely accepted. Many informants believed that the same amount of rice could be taken as frequent small meals since it was imbalance, rather than total quantity, that mattered. In the structured vignette study, 16 of 18 informants agreed that Mr Ali’s doctor had underestimated the amount of rice he needed when advising him to cut down, and all 18 agreed he should take biscuits or other snacks between meals to sustain his strength. Only five thought that such snacks could cause any harm.

[AMC: Did you reduce your rice intake?] "No, I did not." [AMC: Why do you think the doctors ask you to take less rice?] "As I said, with a full stomach, it creates pressure on the sugar." (BF40/516)

[AMC: How do you reduce your weight?] "By eating less. To eat a little bit at a time, whenever hungry." (BF08/658)

Potatoes and root vegetables were seen as unhealthy (hard to digest and/or constipating), particularly in diabetes. Seventeen of 18 informants agreed that Mr Ali should avoid vegetables that grow under the ground, including potatoes, because of his diabetes.

In Bangladeshi society, feasts, festivals and social occasions are common, culturally important, and centre on eating sweet and rich food. A calculated compromise between dietary compliance and social duty was usually made.

"In such a situation [weddings, parties], I eat whatever is served, but in smaller portions." (BM09/463)

Of our Bangladeshi informants, 9 of the 23 males and none of the 17 females smoked; only a few took paan (chewing tobacco) regularly, and those who did acknowledged that it was harmful and expressed a desire to quit. This contrasts with research undertaken 10 years ago in the same population, in which males (both diabetic and non-diabetic) were found to have a high prevalence of smoking.⁽²⁵⁾ In the structured vignette study, only 4 of 18 informants disagreed that tobacco was harmful.

"Yes, I have [thought of giving up paan], since I saw it in a newspaper. They spoke about cancer. I fear cancer very much. But it is hard." (BF40/795)

Many cultures equate balance with health and imbalance with illness.⁽²⁶⁾ There was a strong and almost universal belief amongst the Bangladeshi informants that both the onset and the control of diabetes depends on the balance of food entering the body and on balanced emission of body fluids such as sweat, semen, urine, menstrual blood, and so on. Excess emission was perceived to deplete the internal stock, low quantity of emissions to indicate inner build-up and putrefaction, and thin quality a weakening, of the internal stock. Weakness (as in diabetes) was perceived to occur as a result of such depletion or weakening.

"5 or 6 years back when I was in hospital, they told me that diabetes turns your blood into water. Makes you weak. You’ll feel dizzy, have more hunger, all these." (BM31/FG2/431)

"When my daughter was born, after the bleeding I had a white discharge. Since then the problem started. When the date for my normal menstruation comes, I bleed very little. Allah will forgive me and keep me healthy. ¼.. Because of this I have this illness." (BF33/118)

Absence of sweating (due to the cold British climate and lack of physical labour) on immigration to the UK was commonly cited as a cause of diabetes and a reason why the condition improved or disappeared on return to hot countries. In the structured vignette study, 14 of 18 informants agreed that if Mr Ali returned to Bangladesh his diabetes might be cured.

"Our Bangladesh is a hot country and the food is different from here. At that time [when in Bangladesh] we had dry fish curry and leaf vegetables, the body sweated and we have laboured. So we did not have diabetes." (BM38/284)

A small preliminary study in African-Caribbeans in south London also found cold and the absence of sweating to be a predominant explanatory model for diabetes.

In western society, exercise takes place (as sport) in a structured, ritualised and highly organised context. Exercise in the context of health and fitness seemed to have little cultural meaning for the Bangladeshi informants, even though informants often recalled specific advice on this topic from their doctor. Exercise was viewed as potentially exacerbating illness or physical weakness. The association between sweating (see above) and leisure-time exercise was not made by any informant, but ritual Muslim prayers (namaz) were often cited as a worthy and health-giving form of exercise.

Conflicting perceptions of exercise are supported by the structured vignette study, in which 9 of 18 informants agreed that physical exercise would make Mr Ali weaker and more tired, but none agreed with the statement that "Mr Ali thinks that exercise will not make him fitter."

Some informants gave physical or material constraints to taking exercise. In particular, many of the female subjects rarely left the house, apparently through fear of physical attack. Some informants lived in high-rise flats with no working lift, and some commented on the absence of parks, dirty pavements, and street crime.

"This exercise [praying] is a very big exercise. Daily five times you do the namaz. There is a full exercise." (BM04/FG2/505)

The Sylheti language has no expression for physical activity which has the same connotations of vitality, improvement in body condition, social desirability and inherent "moral" value as the word "exercise". Sports and games are not generally pursued by adults in Bangladesh⁽²⁷⁾ or by Bangladeshis in the UK.⁽²⁸⁾ The closest translation for the word "exercise" is beyam, a word of obscure etymology. Interestingly, the prefix bey in Sylheti often has negative connotations—for example, beyaram (meaning illness, literally "no comfort") or beytamiz (poor etiquette, literally "no manners"), and we were struck by the lack of positive connotations accorded to the concept by our Bangladeshi informants.

[AMC: Do you walk?] "Walk. Can come round the whole of Brick Lane." [AMC: You can, but do you?] [no reply] [AMC: Do you walk in connection with your work or for exercising purposes?] "When required." [AMC: Not for exercise?] "No." (BM15/234)

In contrast, white British and African-Caribbean informants often reported that they took, or ought to take, physical exercise either for its own sake or as a means of achieving diabetic control:

"When my glucose levels are high up, I exercise in my room—jump, jump, jump on the floor." (WF45/1167)

The doctor was viewed as a busy, authoritative and knowledgeable person who rarely makes mistakes and has full understanding of the conditions s/he treats. Several informants felt that the doctor’s instructions should always be obeyed, and 12 of 18 in the structured vignette study agreed that "Mr Ali’s doctor [GP] knows everything about diabetes"; twelve also agreed that it would be impertinent for Mr Ali to ask the doctor any questions.

"How could you not trust your doctor? You go there to save your life—don’t expect him to kill me. He will give me good medicine, advice and so on, so that I am cured or my sickness is reduced." (BF40/662)

In contrast, both white British and African-Caribbean informants were openly assertive and critical of health professionals:

"Well, I think they could communicate a bit more with us—with the patients—the doctors and that. I know they haven’t got a lot of time but what’s five minutes, or a couple of minutes, just to stand there and explain to you?" (WF50/593)

Nurses were sometimes viewed in a traditional, caring and technical role, but were sometimes recognised as providers of information and advice. Nine of 18 informants in the structured vignette study agreed that the nurse’s job is to weigh patients and test urine, but not to give medical advice. Males and females were equally split on this question.

Informants generally tested their urine regularly and all who did so appeared to understand the significance of a change in the colour of the test strip. In the structured vignette study, 7 of 18 informants agreed that if Mr Ali’s strips were clear, he no longer had diabetes, which probably indicated a linguistic confusion, since the Sylheti term "sugar diabetes" implies "sugar in the urine".

Most informants appeared to believe that in the absence of symptoms, diabetes is well controlled. The need for regular surveillance when asymptomatic was rarely acknowledged, and only one of 18 informants in the structured vignette study thought that Mr Ali should ever visit the doctor if he did not feel ill. Preventive care was not well understood.

[AMC: So why do you check your eyesight?] "They feel itchy and I get headaches." [So when you are in pain you go for an eye test, but not otherwise?] "No, I don’t." (BF23/788)

"He [doctor] explained to me and said before complications start, start wearing glasses. This is because your eyes are all right. The diabetes may affect either your eyes or your feet. So if you take the glasses, your eyes may be spared." (BM09/279)

This study addressed an important and previously underexplored area in health research. We used a wide range of qualitative techniques on a sample that is likely to have included the least acculturated members of British Bangladeshi society, since we recruited from practices with Bangladeshi general practitioners, nurses or advocates, we required neither literacy (in any language) nor spoken English or Bengali for participation in the study (indeed, 24 of the 40 informants spoke only the Sylheti dialect), and the response rate for the individual interviews was high (91%). Furthermore, our main fieldworker was an experienced anthropologist who has worked with this community for 25 years and speaks Sylheti as a mother tongue.

The sample does, however, have limitations. In recruiting from primary care, we have not accessed those who do not seek or receive western medicine in any form, or those who receive all or most of their medical care in the secondary sector. We only recruited one second generation Bangladeshi, probably for demographic reasons (diabetes onset is usually in late middle age, and few current second generation immigrants are beyond young adulthood). We did not assess any measure of diabetic control in our informants (such as HbA1, which was inconsistently recorded in patients’ case notes) so we were unable to relate individual perceptions or experiences to level of control.

In interpreting our data, we have deliberately not assumed an all-encompassing and internally consistent picture of "Bangladeshi culture". Although the differences in body image and illness maps demonstrated here are of considerable anthropological interest, we believe that our findings support the findings of Lambert and Rose (in relation to coronary heart disease prevention programmes), that the similarities in health beliefs and health-related behaviours (for example, failed attempts to lose weight or give up smoking) between minority groups and the host culture are often understated and may be of more practical significance than their differences.⁽²⁹⁾

A recurring theme in this research was that of the Bangladeshis’ structural and material barriers to improving their health. Poor housing, unsafe streets, and financial hardship were at least as significant in preventing certain outcomes (such as taking regular exercise) as religious restrictions or ethnic customs. As another group of researchers in this field recently concluded, "what could be interpreted as ‘ethnic differences’ could also be interpreted in the context of material existence, anchored in class relations¼."⁽³⁰⁾ It is not within the remit of this paper to expand on the profound socioeconomic disadvantage of many British Bangladeshis, nor on the literature linking poverty with health inequalities in general,⁽³¹⁾ but the importance of this factor as a barrier to health gain should not be ignored.

It has been argued previously that health education that concords with people’s "lay epidemiology" and folk models is more likely to lead to changes in behaviour than that which appears to contradict such models. Airhihenbuwa and colleagues, in the context of AIDS prevention, have exposed the fallacious assumption that health education is merely a matter of determining "deficiencies" in knowledge and meeting those deficiencies with educational material such as leaflets, teaching seminars, or mass media programmes. Rather, educators must centralize the cultural experiences of those who have hitherto been marginalized.^{(17) (32)} Given that the Bangladeshis in this study indicated a high regard for oral explanations from informal sources (friends, relatives, and other patients with diabetes), we feel that the potential for learning via oral sources within Bangladeshi culture is very high.

Hence, rather than designing an education programme to be delivered externally to rectify "deficiencies" in knowledge or "incorrect" behaviour,⁽³³⁾ we suggest that health promotion programmes attempt to build on those beliefs, attitudes, and behaviours that already exist within Bangladeshi culture which promote good diabetes control, prevent complications, and improve quality of life, and on addressing the practical barriers to positive health behaviours such as non-availability of particular foodstuffs. We have listed in the box examples of constructs which, though not universally held, are sufficiently prevalent in Bangladeshi culture to form the starting point for successful culturally sensitive health education and promotion.

• Diabetes is caused by sweet things, a Western diet, and stress

• Diabetes is chronic and incurable, but its effects can be lessened by changes in lifestyle

• Dietary modification is essential for diabetes control, and effort must be made to prepare special food for the family member with diabetes

• A person with diabetes should aim to lose weight if overweight

• Physical labour which produces sweat is beneficial to health

• Sugar, fatty food, and solid fat (including ghee derived from butter) are harmful

• Complications may occur if diabetes is poorly controlled

• Poor diabetic control can be detected by change in the colour of the urine testing strip

Many informants in this study (both Bangladeshi and non-Bangladeshi) displayed mutually contradictory beliefs, attitudes and behaviours—such as, for example, the concurrent use of herbal or homeopathic remedies with Allopathic medicines, or the statement that the sole cause of diabetes is eating sugar, followed some time later by the statement that diabetes is caused by stress. Stainton Rogers has argued that individuals are active weavers, not passive users, of explanations about illness, and that most "lay beliefs" do not fit easily into linear, predefined and relatively constant categories as set down in any particular biomedical, psychological or sociological paradigm. Rather, the individual lives constantly with, and draws upon, multiple realities and multiple paradigms.^{(17) (34)} It is important that health education and health policy is not based on an "objective" but superficial assessment of health beliefs, for example, responses to closed questions in a limited questionnaire or contextualised interview.

Table 2^(T2) draws on a framework developed by Daniel and Green to identify perceptual, structural and reinforcing factors that influence specific behavioural outcomes in health promotion.⁽³⁵⁾ We have used the recommendation for regular low intensity physical exercise as an example of a desired behavioural priority for people with diabetes. As Table 2 shows, many of the constructs identified in our fieldwork have direct implications for educators working on an individual or public health level. In addition, however, this framework highlights both the broader social and political context within which behaviour change in minority ethnic groups must be placed, and the danger of assuming that "non-compliance" with such lifestyle advice is always attributable to "cultural factors".

We all seek to explain bodily functions and illness in terms of things that are familiar and predictable.⁽³⁶⁾ Failure to ascertain differences in body image and "illness maps" between patients and health professionals will make communication difficult and successful educational intervention impossible.⁽³⁷⁾ In biomedical and psychological models, the success of preventive strategies is said to depend on how far patients are willing to accept responsibility for their own health-related behaviour and initiate changes in lifestyle,⁽³⁸⁾ which in turn depends fundamentally on their beliefs about the aetiology of illness.⁽³⁹⁾

Those who take an anthropological perspective acknowledge that the physical symptoms of disease and distress, and the way individuals react to the experience of illness, are heavily dependent on cultural norms and expectations, and these factors must therefore be understood and addressed at a societal level before a particular health intervention is directed at the private individual.^{(40) (41)} The need for more "culturally sensitive" services is now officially recognised by both clinicians⁽⁴²⁾ and those who plan and fund services,⁽⁴³⁾ although the meaning of the term "cultural sensitivity" is hotly debated.^{(44)(45)(46)(47)}

Lambert and Sevak have challenged the growing popularity of qualitative research methods for the investigation and documentation of culturally specific characteristics of particular minority groups, since such studies may produce a static, structuralist view of culture which in turn may unwittingly lead to the victims of ill health being blamed for their "different" beliefs and lifestyles and the potential for change overlooked.

Other writers, while recognising that the well-intentioned but naïve research into ethnicity and health conducted to date has failed to produce significant material benefits for minority groups, argue that rather than abandoning all such research, more refined approaches to the dynamic interactions between culture, socioeconomic status and health experience are urgently needed^{(48) (49)}

Contributors: TG conceptualised and supervised the study, helped with fieldwork, analysed and interpreted the data, and wrote the paper. AMC performed the fieldwork and data entry and analysed and interpreted the data. CH provided general advice and contributed to analysis and interpretation of the data. TG is guarantor for the paper.

Funding: The salaries of TG and AMC for this study were covered by a Health Services Research Grant from the Wellcome Trust.

Conflict of interest: None.

1. Bradley C, Gamsu DS, for the Psychological Well-being Working Group of the WHO/IDF St Vincent Declaration Action Programme for Diabetes. Measures of psychological well-being and treatment satisfaction developed from the responses of people with tablet-treated diabetes. Diabet Med 1994;7:510-6.
2. Greenhalgh PM. Diabetes in British South Asians—nature, nurture and culture. Diabet Med 1997;11:10-4.
3. Helman CG. Research in primary care—the qualitative approach. In: Norton PG, Stewart M, Tudiver F, Bass MJ, Dunn E, eds. Primary care research: traditional and innovative approaches. London: Sage Publications, 1991:105-24.
4. Gittelsohn J, Harris SB, Burris KL, Kakegamic L, Landman L, Sharma A, et al. Use of ethnographic methods for applied research on diabetes among the Ojibway-Cree in Northern Ontario. Health Educ Q 1996;23:365-82.
5. Grams G, Herbert C, Heffernan C, Calam B, Wilson MA, Grzybowski S, et al. Haida perspectives on living with non-insulin-dependent diabetes. Can Med Assoc J 1996;155:1563-8.
6. Kelleher D, Islam S. "How should I live?" Bangladeshi people and non-insulin dependent diabetes. In: Kelleher D, Hillier S, eds. Researching cultural differences in health. London: Routeledge, 1996:220-37.
7. Hall ET. Beyond culture. New York: Anchor Books, 1977.
8. Balarajan R, Soni Raleigh V. The ethnic populations of England and Wales: the 1991 census. Health Trends 1992;24:113-6.
9. Helman C. The application of anthropological methods in general practice research. Fam Pract 1996;13:S13-6.
10. Kitzinger J. The methodology of focus groups: the importance of interaction between research participants. Sociol Health Illness 1994;16:104-21.
11. Hayes BE. How to measure empowerment. Qual Progress 1994;Feb:41-6.
12. Helman C. The explanatory model. In: culture, health and illness. 3rd ed. Oxford: Butterworth-Heinemann, 1994:111-4.
13. Silverman D. Qualitative methodology and sociology: describing the social world. Aldershot: Gower, 1985.
14. Denzin NK, Lincoln YS, eds. Handbook of qualitative research. London: Sage Publications, 1994.
15. Helman C. Culture, health and illness. 3rd ed. Oxford: Butterworth-Heinemann, 1994.
16. Crawford R. A cultural account of "health": control, release and the social body. In: McKinley JB, ed. Issues in the political economy of health care. London: Tavistock, 1984:60-103.
17. Airhihenbuwa CO. Developing culturally appropriate health programs. In: Health and culture: beyond the Western paradigm. London: Sage Publications, 1995:25-43.
18. Pierce MB. Non-insulin dependent diabetes and its complications—beliefs, perceptions and prospects for risk reduction [thesis]. London: University of London, 1997.
19. Garro L. Individual or societal responsibility? Explanations of diabetes in an Anishaabe (Ojibway) community. Soc Sci Med 1995;40:37-46.
20. Williams R, Bhopal R, Hunt K. Coronary risk in a British Punjabi population: comparative profile of non-biochemical factors. Int J Epidemiol 1994;23:28-37.
21. Bradley C, Gamsu DS, for the Psychological Well-being Working Group of the WHO/IDF St Vincent Declaration Action Programme for Diabetes. Guidelines for encouraging psychological well-being. Diabet Med 1994;11:510-6.
22. Brunner E. Stress and the biology of inequality. BMJ 1997;314:1472-6.

23 Helman C. Diet and nutrition. In: Culture, health and illness. 3rd ed. Oxford: Butterworth-Heinemann, 1994:37-62.

23. Lambert H. The cultural logic of Indian medicine: prognosis and etiology in Rajastani popular therapeutics. Soc Sci Med 1992;34:1069-76.
24. McKeigue PM, Marmot MG, Syndercombe Court YD, Cottier S, Rahman S, Riemersma RA. Diabetes, hyperinsulinaemia and coronary risk factors in Bangladeshis in east London. Br Heart J 1988;60:390-6.
25. Helman C. Balance and imbalance. In: Culture, health and illness. 3rd ed. Oxford: Butterworth-Heinemann, 1994:21-4.
26. Chowdhury AM. Household kin and community in a Bangladesh village [thesis]. Exeter: University of Exeter, 1986.
27. Health Education Authority. Health and lifestyles survey: black and minority ethnic groups in England. London: HEA, 1994.
28. Lambert H, Rose H. Disembodied knowledge? Making sense of medical science. In: Irwin A, Wynne B, eds. Misunderstanding science: making sense of science and technology within everyday life. Cambridge: Cambridge University Press, 1996:65-83.
29. Anderson JM, Wiggins S, Rajwani R, Holbrook A, Blue C, Ng M. Living with a chronic illness: Chinese-Canadian and Euro-Canadian women with diabetes—exploring factors that influence management. Soc Sci Med 1995;41:181-95.
30. Calman KC. Equity, poverty, and health for all. BMJ 1997;314:1187-91.
31. Airhihenbuwa CO, Di Clemente RJ, Wingwood GM, Lowe A. HIV/AIDS education and prevention among African-Americans: a focus on culture. AIDS Educ Prev 1992;4:267-76.
32. Becker MH. The tyranny of health promotion. Public Health Rev 1986;14:15-25.
33. Staiton Rogers W. Explaining health and illness: an exploration of diversity. Hemel Hempstead: Harvester Wheatsheaf, 1991.
34. Danial M, Green LW. Application of the precede-proceed model in prevention and control of diabetes: a case illustration from an aboriginal community. Diabet Spectrum 1995;8:80-123.
35. Helman C. The body image in health and disease: exploring patients’ maps of body and self. Patient Educ Counselling 1995;26:169-75.
36. Helman CG. Communication in primary care: the role of patient and practitioner explanatory models. Soc Sci Med 1985;20:923-31.
37. Stott NCH, Rollnick S, Rees MR, Pill RM. Innovation in clinical method: diabetes care and negotiating skills. Fam Pract 1995;12:413-8.
38. Pill R, Stott NCH. Concepts of illness causation and responsibility: some preliminary data from a sample of working class mothers. Soc Sci Med 1982;16:43-52.
39. Krause I-B. Sinking heart—a Punjabi communication of distress. Soc Sci Med 1989;29:563-75.
40. Douglas M. Natural symbols—a explorations in cosmology. London: Penguin, 1973.
41. Bhopal RS, White M. Health promotion for ethnic minorities: past, present and future. In: Ahmad WIU, ed. "Race" and health in contemporary Britain. Buckingham: Open University Press, 1993:137-66.
42. Balarajan R, Raleigh VS. Ethnicity and health: a guide for the NHS. London: Department of Health, 1993.
43. Ahmad WIU. The trouble with culture. In: Kelleher D, Hillier S, eds. Researching cultural differences in health. London: Routeledge, 1996:190-219.
44. Donovan J. We don’t buy sickness—it just comes. Aldershot: Gower Press, 1986.
45. Townsend P, Davidson N. Inequalities in health: the Black report. London: Penguin, 1982.
46. Lambert H, Sevak L. Is "cultural difference" a useful concept? In: Kelleher D, Hillier S, eds. Researching cultural differences in health. London: Routeledge, 1996:124-59.
47. Smaje C. Health, race and ethnicity: making sense of the evidence. London: King’s Fund Institute, 1995.
48. Kelleher D. A defence of the use of the terms "ethnicity" and "culture." In: Kelleher D, Hillier S, eds. Researching cultural differences in health. London: Routeledge, 1996:69-90.