Health policy is based on the ideal that all inhabitants should have equal access to health care. US studies have found ethnic differences in the use of health care with lower consumption rates for people from ethnic minorities [1, 2]. European studies on ethnic differences in the use of health care have reported mixed outcomes. Some studies reported higher rates by various ethnic minority groups in comparison with the majority populations with respect to the general practitioner (GP) services in combination with lower rates for outpatient services [3–5]. Other studies reported either no differences in outpatient care use [6, 7] or higher rates among ethnic minorities or immigrant populations [8–12].

Differences in consumption rates could be based upon differences in the incidence of diseases. For some diseases, like diabetes, it has indeed been shown that there are ethnic differences in incidence [13–15]. However, ethnic differences in consumption rates could also have other reasons. For instance, in case of referral of patients by a GP to an outpatient clinic, patients ethnicity might influence the physicians' beliefs about and expectations of patients, and consequently the physicians' actions [16]. There are also indications that, as a result of less effective and satisfying doctor-patient relationship [17], physicians that treat ethnic minority patients are more uncertain in the process of care [18]. Especially in case of language problems, which are common among immigrant populations, the latter might be the case. This could clearly have implications for the referral pattern and the care physicians give. A possible result could be that people from ethnic minorities are more often referred on the basis of vague symptoms, and might therefore less often receive a medical diagnosis. Other possible explanations could be that immigrant patients seek professional medical help more often, not only because they actually do have more health problems, but they also tend to report physical symptoms more often and more expressive. It is suggested that this might be due to the fact that they have a more positive attitude towards care-seeking [19, 20].

Since the number of immigrants in the industrialized countries is growing, it is also of growing importance to obtain data on ethnic differences in the use of health care, referral patterns en diagnoses. E.g., in the Netherlands the proportion of non-western immigrants increased from about 3% in 1980, to 6% in 1990 and more than 10% in 2005. A limitation of previous studies is their reliance on self-reports of health care utilization. Although self-reports have been shown to be a valid estimate of health care utilization across socio-economic strata [21], there is less evidence for cross-cultural validity, especially among some of the larger immigrant groups in the western European countries, Turkish and Moroccan people [22]. Ethnic differences in recall bias, non-response and tendency for giving socially desirable answers, could undermine the validity of self-reported measures [23]. For instance, illiteracy and limited proficiency in the native language, both more prevalent among immigrants, will increase non-response rates. Therefore we have chosen for the use of hospital registration data in order to examine ethnic differences in the use of health care.

The ethnic minority populations in the western European countries mainly exist of immigrants who entered the country in the period between 1955–1985 when there was a severe shortage of people to do the unschooled jobs in these countries (first-generation immigrants). In the case that ethnic differences are found, it would also be worthwhile to know whether these differences also persist in the younger generations or whether the consumption rates in younger generations will be more alike those in the majority population.

Using hospital registration data for an outpatient clinic for internal medicine we investigated the following research questions:

(1) Are there differences between ethnic groups in the use of outpatient health services?

(2) Are there differences in reasons for referral and diagnosis between ethnic groups?

For the referral area of the Erasmus MC, immigrant people have an increased use of the outpatient clinic compared to Dutch people, adjusted for sex and age. The increased use was expressed by relative risks of consultations, which ranged from 1.29 in the Cape Verdean group to 1.82 in the Turkish group. The difference was statistically significant only for Surinamese, Turkish and Moroccan people.

Additional adjustment for socio-economic status hardly changed the estimates (table 2A). The largest decrease in relative risk was observed among Cape Verdeans in the analyses for the municipality of Rotterdam, from 1.99 to 1.88. In analyses in which the first and second immigrant generations were distinguished (table 2B), no difference in the use of health care were observed between the second-generation and the native Dutch citizens. In both areas the increased use can be predominantly ascribed to the first-generation immigrants.

After adjusting for socio-economic status ethnic differences only decreased slightly, indicating that ethnic differences in socio-economic status hardly explained the differences in referral reasons for patients that were referred to the Erasmus MC.

Ethnic differences in risk for 'liver diseases' are partly explained by differences in socio-economic status; after adjusting for socio-economic status the differences in risk became smaller. For 'liver diseases' the risk decreased from 1.96 to 1.75, but retained a borderline significance. For 'analysis without diagnosis' the risk decreased slightly after adjustment for socio-economic status (from 0.62 to 0.67), it retained statistical significance.

Finally, we also analysed ethnic differences in the risk of getting a certain diagnosis given the referral reason and looked for ethnic differences in this relationship. There appeared to be no differences between the ethnic groups under study, except for the category 'gastro-intestinal signs & symptoms', in which immigrant patients were more likely to receive a diagnosis in the category 'liver' (data not shown).

There is a higher number of new patient referral visits of Surinamese, Turkish and Moroccan immigrants, living in the referral area of the Erasmus MC, compared to native Dutch people than could be expected from their relative distribution in the population. In Rotterdam municipality the five largest ethnic minority groups all demonstrate a higher use of the outpatient care facilities. This increased use can be predominantly ascribed to the first-generation immigrants; second-generation immigrants do not appear to have an increased use of health care services. Immigrant patients who visited the outpatient clinic were more likely to be referred because of 'gastro-intestinal signs & symptoms' and less likely to be referred because of 'diffuse and ambiguous signs'. Regarding ethnic differences in diagnosis, we noted an increased risk of 'liver related diagnosis' and a decreased risk of 'analysis without diagnosis' for immigrant patients.

We have to consider a few limitations of the current study. Although over 4000 new patients were registered in the hospital, numbers for those eligible for the study were small for some ethnic groups, and especially for second-generation immigrants. Therefore, not all research questions could be examined for the ethnic groups separately, nor could the first-generation be distinguished from second-generation immigrant for all research questions. For our second research question it was necessary to aggregate all ethnic groups to one 'immigrant' group. The aggregation was justified by the outcomes of table 2, in which all ethnic groups show a deviated use of health care in the same range and direction. A similar limitation concerns the aggregating of codes of referral reasons and diagnosis. In the results of research question 2 no ethnic differences were found for the referral reason 'risk factor vascular disease'. However, looking in more detail shows large differences between the ethnic minority groups for more specific referral reasons. Surinamese and Cape Verdean patients often are referred with the most prevalent underlying risk of vascular diseases, namely hypertension. The same holds with regard to ethnic differences in diagnosis: we found no different risk of diagnosis 'risk factor vascular diseases' regarding ethnicity. But the underlying codes showed that diabetes mellitus was significantly more prevalent among referred immigrant patients and dyslipidemia more common among Dutch patients. Odds ratios have to be interpreted in a relative sense, because they were calculated for the closed group of patients that visit the outpatient clinic of Erasmus MC. An apparent lower odds ratio might be the result of higher rates in other groups of diagnosis.

In the second place ethnicity is based on countries of birth. Although this is a well-accepted definition [6, 26–28] we were unable to address ethnic variations within immigrant groups. Differences in the use of health care may have been more differentiated within certain ethnic minority groups, especially for the ethnically diverse Surinamese and Antillean/Aruban population.

We approximated socioeconomic status at the individual level by making use of mean neighbourhood incomes, a variable at the ecological level. This measure may not be equally good for all ethnic groups. In some groups, the place of residence is determined by the mean socioeconomic status of a neighbourhood, whereas in others it is predominantly determined by the ethnic composition of a neighbourhood. In that case, neighbourhood income may be a less valid indicator of socioeconomic status. For Antilleans this does not seem to be the case, however for Turks, Moroccans and Surinamese a somewhat larger proportion (5 to 15%) of the population belonged to the lowest income quintile according to the measure at the ecological level than according to the measure at the individual level. This means that the place of residence of Turks, Moroccan and Surinamese may be more strongly determined by factors other than neighbourhood income. As the discrepancy was fairly small, the influence of the differential validity on the outcomes of this study would be limited [25].

Besides Erasmus MC, there are four more hospitals in Rotterdam that offer health care services. Differences in preference for Erasmus MC could have introduced the differences in health care use. For at least the referral area, this seems hardly the case. A survey among general practitioners in the referral area reported a slightly different referral pattern among different ethnic groups to Erasmus MC and other hospitals in Rotterdam (unpublished data). General practitioners send immigrant patients more often than Dutch patients to the internal medicine outpatient's care of Erasmus MC. The difference is (at most) 5% and cannot explain the increased use of 80% by immigrant patients. Additional support for our assumption that potential differences in referral patterns (due to preferences or the reputation of the Erasmus MC) between ethnic groups in Rotterdam municipality, have had little influence on the outcomes of our study comes from the analysis of the ethnic differences in referral reasons for both areas separately (data not shown). The findings at least indicate that there are no ethnic differences in referral reasons between the referral area and Rotterdam as a whole. Herewith a correct inference for the population of Rotterdam municipality is deduced, since the assumption for representativeness of the patient sample seems to be supported.

Remarkable is that the ethnic differences in likelihood of being referred are higher when focussing on Rotterdam municipality than when focussing on the referral area. It is uncertain whether this can be attributed to the prevalence of certain diseases, which require special care. Erasmus MC is also a university hospital and delivers tertiary medical care.

The results of this study regarding the use of health care differ from the results of Stronks et al., who reported no differences in the use of outpatient care [6], likewise using registration data. An explanation could be that they addressed outpatient care clinics comprising of several types of specialists, while we made a restriction to internal medicine. Immigrant patients are known to have a higher incidence of several diseases and syndromes, which are referred to the internal medicine clinic (i.e., diabetes, liver diseases and gastro-intestinal complaints). Diseases referred to other outpatient care clinics probably are more equally distributed among different ethnic groups[8]

The results of our study are in agreement with the results of some other studies. Both Manna [29] and Weide and Foets [19] reported an increased risk for immigrant patients for referral with 'signs & symptoms gastro-intestinal'. Some of our results however, differ from the results of other studies. Other studies have reported that reasons for medical consultation among immigrants patient's are more often misunderstood or perceived as not being appropriate by the physician, and that the diagnostic process among immigrant patients might be more complicated because of language barriers, other concepts of disease, and other expressions of pain or other symptoms [30–33] Possibly this could lead to more referrals for indefinite or ambiguous signs and immigrant patients would be more likely to end up in the category 'analyses without diagnosis', but we found the opposite: less immigrant patients came to the outpatient clinic with 'indefinite ambiguous signs' and compared to Dutch patients they have a lower risk for the category 'analysis without diagnosis'. Differences in domains of health care under study may explain the dissimilarity of their results with ours, as these other investigations mainly focussed on general practitioners or on health care in general. Given the Dutch system, where general practitioners are the gatekeepers to most other health services, including the outpatient services, health complaints perceived as inappropriate might have been filtered out by the general practitioner effectively.

A possible explanation for the higher use of outpatient care among immigrants might be a direct reflection of a higher incidence and prevalence of certain diseases. We did not have information about health status, but previous studies have reported a higher incidence of infectious diseases [34], hypertension [35, 36], circulatory diseases [37–39], diabetes [40–42], and worse health status in general [8, 43] among immigrant groups. Despite these higher incidences, we cannot rule out that referral rates for immigrant groups are inappropriately low or inequitable. Another explanation for higher use could be different styles/patterns in referring immigrant patients and Dutch patients to the outpatient care. Uitewaal [44] reported that more diabetes patients from Turkish descent than native Dutch diabetes patients were referred to the outpatient care. Moreover, immigrant patients asked more for referrals to outpatient clinics, instead of analysis or treatment by the general practitioner[45] It is known that immigrant patients seek professional medical help more often, not only because they actually do have more health problems, but they also tend to report physical symptoms more often. It is suggested that this might be due to the fact that they have a more positive attitude towards care-seeking [20, 19] and they have different beliefs concerning health and illness. [46] However we did not find evidence for ethnic differences in mismatch between referral and diagnosis, general practitioners can cause differences in referrals between immigrant and Dutch patients, when communication with immigrant patients is less effective than in consultations with Dutch patients, there is more misunderstanding and also more non-compliance. [47] These explanations could also contribute to the interpretation of the finding that the increased use of health care services predominantly can be ascribed to the first-generation immigrants. Compared to the first-generation, immigrants of the second-generation generally have a higher education, better language skills and have better control of their lives [20]. Thus, second-generation immigrants could become more alike to Dutch patients and their health care use will become more similar. While first-generation immigrants directly benefited from the more favourable socio-economic, public health and health-care conditions in the Netherlands compared with their country of origin, they are not yet affected by the higher risks of diseases associated with prosperity. [25] In the future, next generations immigrants, will be exposed to new risks similar to the risks of the native Dutch. Old risks, like higher risk for infections, will be substituted for risks more comparable to the native Dutch. Concerning first and second generation immigrants, duration of residence might explain the difference in health care use. However, there is no different mean time of residence between first- and second generation of the ethnic groups under study. Most first generation immigrants came to the Netherlands decades ago as labor workers.

Besides the differences in health care use between native Dutch and ethnic minority groups, there also appear to be differences among the ethnic minority groups themselves. Additional analyses showed that Cape Verdean immigrants have a statistically significant lower use of health care than Surinam, Turkish and Moroccan immigrants. Further research is needed to explore why Cape Verdean immigrants are more similar to the native Dutch population regarding health care use.

Because our data are limited to one particular outpatient care unit and moreover to a university hospital, we must be cautious in generalizing the results and the hypotheses generated by this study require further study.

We conclude that especially first-generation immigrants make significantly more use of the outpatients' care in internal medicine. Ethnic differences might decrease as the share of first-generation immigrants decreases. Concerning this point, it is warranted to monitor the risks of diseases associated with prosperity in the future among immigrant groups. Ethnic differences in referral reasons and diagnosis might be based on a higher prevalence of diseases. Further study is needed to establish this. We found no evidence that the increased use is based on referrals for non-medical reasons, on the contrary. As long as the increased use of outpatient health care is related to ethnic background and the generation of the immigrants rather than to socio-economic status, health professionals have to take ethnicity into account in their daily medical practice. Moreover, they should take the main differences in prevalence of diseases among immigrants into account during the consultations.