METHODS

Setting

The study was set in three general practices in Greater Manchester and focused on the four ethnic minority communities with the largest populations (table 1). Practices 1 and 2 were in Tameside, to the east of Manchester, where many people who would describe themselves as south Asian (Indian, Pakistani, or Bangladeshi) were believed to be registered. Practice 3 was in central Manchester where we expected a high proportion of registered patients who would describe themselves as African Caribbean. Tameside and Glossop and Central Manchester local research ethics committees granted ethical approval.

Sample size

We stratified the study group by sex and the following age groups: 16–44; 45–64; 65–74; 75+. We aimed at obtaining as complete a picture as possible of the ethnic minority population of the three practices, which we hoped would subsequently be generalisable. If we had managed to study 750 people from each age-sex band and each ethnic group, we would have enabled a prevalence of 5% to be detected with a precision of 1% and with 80% power. However the size and age structure of the ethnic minority populations prevented us from achieving this. The chosen practices had relatively large proportions of people registered from ethnic minorities compared with their neighbouring practices, but numbers of people from ethnic minority communities were still too small to achieve the required sample size, especially among the older age groups. We still aimed at obtaining a full picture of the ethnic minority population of the practices, but the sample size was considerably reduced. The study group was still stratified into the four age groups and all age groups were studied. Numbers in the older age groups were small, but all people believed to be from ethnic minority communities in these age groups were mailed.

Subject selection

In practices 1 and 2 we identified all people aged 16 or over whose name suggested south Asian ethnicity; this method has been used before.¹⁷ Because of the age structure of the population, we excluded 50% of those aged under 45 by using computer generated random numbers. In practice 3 the practice staff identified people who might classify themselves as African Caribbean and we included all of those aged 16 or over. Lists of subjects were checked to exclude those who had recently died and general practitioners were asked to identify people unsuitable for the study.

Questionnaire

We mailed the study group a questionnaire in English which included questions about the presence of musculoskeletal pain in the past month lasting more than one week; the sites of any musculoskeletal pain; the presence of pain in most joints; physical function measured by the modified Health Assessment Questionnaire (mHAQ)¹⁸; access to health care; demography including ethnicity according to the 1991 census categories (table 1). The mHAQ comprises eight questions and the overall scores range from 0 (no disability) to 3 (severe disability). Categories for self definition of ethnicity were chosen to enable comparison with previous studies and with routinely collected data. A personally addressed covering letter from the subject's general practitioner endorsing the study and encouraging participation was sent with the questionnaire. A reply paid envelope was included. People who did not respond were sent a reminder and then a repeat questionnaire.

Postal response rates among south Asian populations may not be high,¹⁷ although interviewers can achieve high response rates.¹⁴ We combined the low cost coverage of postal surveys with the high response rates of interviews. Thus, in Tameside those who did not respond to the second questionnaire were visited by an interviewer who administered orally a standardised validated translation of the questionnaire in the subject's preferred language. Translations of the English questionnaire were made in Gujarati, Urdu, Punjabi, Bangla, and Sylheti. Each language had a standardised translation, which was back-translated to assure validity. We did not ascertain the proportion of the study group who had not responded initially because of language or literacy problems, but levels of knowledge of English and literacy are relatively low among the elderly and among women in some ethnic minority communities.^19,20 We compared the two administration methods by interviewing 55 people who had previously responded by post. The results supported the decision to analyse postal and interview responses together. In practice 3, the study group all spoke English and no community interviewers were available. Therefore a sample of people who had not responded to the postal questionnaire were telephoned, where possible, but no visits were made. Response rates were calculated based on the study group less those who had died or moved away and included in the denominator those who could not be traced. The case notes of a sample of people who had not responded to postal, interview or telephone questionnaires were selected by stratified random sampling from each practice. We noted the numbers of consultations and referrals, presence of comorbidity, and drugs used. These results were compared with the survey data provided by the responders.

Methods of analysis

Data were analysed by Microsoft Access and SPSS. Ethnicity was assigned according to the questionnaire response. The category of African Caribbean was created by combining Black-Caribbean and Black-Other with people describing themselves as Black British or Black European. Physical disability was defined as an mHAQ>0. Body mass index was calculated by the formula weight (kg)/height (m)². The Townsend score²¹ by census enumeration district was used to assign a measure of social deprivation to subjects based on their area of residence. This was based on quintiles of the population of Greater Manchester. Confidence intervals for prevalence were calculated using a formula for proportions. Logistic regression was carried out using the dependent variables of age, sex, body mass index, ethnicity, and social deprivation. The results from the different ethnic groups were compared with those from the local white population from the recent study in Tameside, which used similar methodology and which was undertaken by the same research team.⁷

RESULTS

The sample size was 2117 (table 2) and 689 (33%) mailed questionnaires were returned by post; 148 people declined to participate. After interviews in practices 1 and 2 and telephoning in practice 3, and once people who had died or moved away were removed from the denominator population, response rates were 73% in practice 1, 77% in practice 2, and 47% in practice 3. Table 3 shows the self defined ethnicities of respondents. Over 95% of practice 1 respondents described themselves as Indian, Pakistani, or Bangladeshi, while the vast majority of practice 2 respondents were Bangladeshi. African Caribbean, Black African, and Black British people constituted 87% of practice 3 respondents. Eighty seven per cent of respondents were aged under 65. The median age of respondents was lower for all ethnic groups than for the comparative white population, and the median age of men was higher than that of women, except among Indian subjects (table 4). It was not possible to determine response rates for individual ethnic minority groups, because the ethnicity of non-responders was not known. However, response rates in practices 1 and 2 reflect the South Asian community and practice 3 the African Caribbean community.

The primary care case notes of 93 questionnaire responders and 69 non-responders were examined. Comorbidity was present in 20 (22%) responders and 21 (30%) non-responders. Sixteen (17%) responders were prescribed drug treatment, compared with 19 (28%) non-responders. The median number of primary care consultations for musculoskeletal disease in the past year was the same for responders and non-responders in each practice. Past rheumatology referrals had taken place for two (2%) responders and three (4%) non-responders, past orthopaedic treatment for 21 (23%) responders and 12 (17%) non-responders, and past physiotherapy referrals for 11 (12%) responders and three (4%) non-responders.

Analysis of the results from the 55 people who completed both a mailed questionnaire and a face to face interview showed similar results for the two methods. Overall pain prevalence was the same by mail and by interview, although for 6/10 pain questions, more people reported pain in the postal questionnaire and for 6/8 disability questions more people reported disability at interview.

The results of the whole study were examined by age band and compared with results from the recent study of the white population using the same methodology.⁷ One hundred and two people reported musculoskeletal pain but not in one of the specified areas. Comparison of crude prevalence is only presented for people aged 16–64, because the sample sizes in the older age groups were small. However, those aged 65 or over are included in other analyses. The crude prevalence of musculoskeletal pain among people aged 16–44 ranged from 29% for Bangladeshi women to 53% for African Caribbean men, compared with 36% for the comparative white population (table 5). Among people aged 45–64, crude musculoskeletal pain prevalence was higher in all ethnic groups than in white subjects. Prevalence ranged from 63% among Indian men to 89% among Pakistani women, compared with 53% for white men and 55% for white women (table 6).

Pain in multiple areas was studied under two categories: (a) those people who agreed that they had had “pain in most joints” lasting more than one week in the past month; (b) the number of individual joint pain areas reported by the subject.

All ethnic groups had a considerably higher crude prevalence of “pain in most joints” than did the white population and this reached statistical significance in most cases (table 6). For example, white women aged 45–64 had a crude prevalence of 8% (95% CI 6 to 10%), compared with 37% (15 to 59%) for Pakistani women and 44% (30 to 58%) for Indian women. The proportion of people from ethnic minority communities reporting pain in three or more separate joint areas also tended to be greater than among white subjects (table 7). Twenty one per cent of white women aged 45–64 reported pain in three or more areas, compared with 46% of African Caribbeans, 52% of Indians, 63% of Pakistanis, and 28% of Bangladeshis.

Disability prevalence measured by an mHAQ>0 was not significantly different among ethnic minority groups aged 16–44, compared with the white population (table 5). Among those aged 45–64, disability prevalence tended to be higher among ethnic minority groups, but this only reached statistical significance for Indian men, Bangladeshi men, Indian women, and Bangladeshi women (table 6).

Logistic regression was used to compare the groups with the white population, controlling for age, sex, body mass index, and area of residence. The odds ratio for any musculoskeletal pain was only significantly raised among Pakistanis, although the odds ratio for an mHAQ>0 was significantly raised among Indians, Pakistanis, and Bangladeshis (table 8). The odds ratios for pain in specific joint areas tended to be slightly raised among ethnic minorities compared with white subjects. For example for back pain the odds ratios were African Caribbeans 1.2 (95% CI 0.8 to 1.5), Indians 1.6 (1.3 to 1.9), Pakistanis 1.5 (1.2 to 1.8), Bangladeshis 1.5 (1.2 to 1.8). For knee pain the odds ratios were African Caribbeans 1.2 (95% CI 0.8 to 1.5), Indians 2.0 (1.7 to 2.4), Pakistanis 1.7 (1.2 to 2.1), Bangladeshis 1.1 ( 0.8 to 1.5).

The adjusted odds ratios for “pain in most joints” were considerably higher than 1 for all groups comparing them with the white population and all were statistically significant. Figures ranged from 2.2 (95% CI 1.8 to 2.6) for African Caribbeans to 5.1 (4.6 to 5.6) for Pakistanis (table 8). This tendency for increased prevalence of pain in several areas is also reflected in the odds ratios for reporting pain in three or more separate areas. Odds ratios for pain in one or two areas were not significantly different from 1 among ethnic minority groups compared with the white population. However, the odds ratios for three or more areas were significantly greater than 1 among all four ethnic minority groups (table 9).

DISCUSSION

We have demonstrated differences in the prevalence of musculoskeletal symptoms among different ethnic minority groups in a population study that achieved a high response rate in the south Asian community. It is reasonable to equate this practice based study with the population, because the great majority of people from ethnic minority communities are known to be registered with primary care practices.¹⁹ Joint pain lasting for more than one week in the past month was slightly more prevalent among ethnic minorities. Pain reported in “most joints” and pain in three or more individual sites were markedly more prevalent among the ethnic minorities than in the white population. The small increase in prevalence in individual joint areas among ethnic minorities might be explained as being part of a picture of widespread pain rather than indicating differences in specific joints. Few studies of the prevalence of musculoskeletal symptoms among ethnic minority populations in the UK have been published. Comparison is possible with results from other countries, although methodology and case definitions differ and the focus of some studies has been radiological abnormality rather than symptoms. Bremner et al noted a similar prevalence of radiological osteoarthritis between a white British population and one from rural Jamaica, although they reported that symptoms and incapacity were lower among Jamaicans.⁹ Gibson et al found similar levels of joint disease in Pakistan and in white European populations.¹⁰ In the USA, levels of self reported arthritis have been found to vary little by ethnicity,²² and musculoskeletal disability was similar in African-American and white populations.²³ A telephone study of acute back pain in North Carolina, USA, found a slightly lower prevalence in non-white subjects.²⁴ Our results are compatible with these studies in that we have shown only small differences in overall pain prevalence between ethnic minority and white populations. As far as we know, our finding of markedly increased prevalence of “pain in most joints” has not previously been reported.

The difficulties associated with any study of this type mean that some caution must be expressed. Extrapolation outside the four ethnic groups studied is not possible. The response rate was far higher among south Asian subjects than among the African Caribbean community and, inevitably, the sample sizes in most age-sex groups were small, given the population age structure of ethnic minorities in the UK. There was good equivalence between interviews and questionnaires. Precise estimates of the effects of the administration method used and non-response were not possible, but there was no evidence of large systematic differences that might have altered prevalence estimates or odds ratios greatly. Also, case note analysis showed that responders and non-responders had a reasonably similar profile. We have used responders as the denominator population for prevalence estimates and so assumed that responders and non-responders had a similar prevalence of pain. We might have used the entire study group as the denominator, assuming that non-responders had no pain, but because pain prevalence between responders and non-responders appears similar, this would have underestimated the overall pain prevalence. However, it is still possible that there were some systematic differences between the different samples. Self reported pain is likely to have major cultural influences, but it is the symptoms that are presented in clinical consultations and differences in symptomatology which are worthy of note, even if the underlying objective morbidity is similar. Using the Townsend score to represent socioeconomic deprivation has limitations and other socioeconomic factors may contribute to the results. However, despite its limitations, this study presents important evidence of clinically significant differences among ethnic groups.

Further research examining other ethnic minority populations in other locations would be valuable. It would also be of great interest to look more closely at the relation between symptoms and objective measures of morbidity, and at the phenomenon of widespread musculoskeletal pain among people from ethnic minority communities. There are several possible explanations for the differences in pain reporting by different ethnic groups. These include pain thresholds and the experience of pain, the effect of change of culture and migration, and mental health issues. Widespread pain might, for example, be an indication of mental distress and, possibly, migration, rather than ethnicity, is a key factor. A review of published reports of these issues has recently been carried out with an emphasis on South Asian subjects.¹⁵ The authors of the review call for further work to investigate the extent and nature of musculoskeletal disease among ethnic minorities. Our study has made a contribution to a study of this issue, but further work is needed to assess specific healthcare needs, whether or not the actual needs are for musculoskeletal care, and then to deal with those needs.