Article Text

Download PDFPDF

Ethnic differences in healthcare-seeking behaviour and management for acute chest pain: secondary analysis of the MINAP dataset 2002–2003
  1. Y Ben-Shlomo1,
  2. H Naqvi2,
  3. I Baker3
  1. 1
    Department of Social Medicine, University of Bristol, Bristol, UK
  2. 2
    Bristol Primary Care Trust, Bristol, UK
  3. 3
    Department of Social Medicine, University of Bristol, Bristol, UK
  1. Professor Y Ben-Shlomo, Department of Social Medicine, University of Bristol, Bristol BS8 2PR, UK; y.ben-shlomo{at}


Objective: To examine whether there are ethnic differences in the healthcare-seeking behaviour and management of patients with chest pain.

Design: Prospective cohort of patients attending accident and emergency departments with chest pain.

Setting: Hospitals in England and Wales from 1 January 2002 to 31 December 2003.

Participants: Patients with chest pain.

Main outcome measures: Whether patients arrived by ambulance, whether they received thrombolysis and the time it took from symptom onset to arrive at hospital and receive thrombolysis.

Results: South Asian patients were less likely to arrive by ambulance (age and sex adjusted odds ratio 0.64, 95% CI 0.60 to 0.69, p<0.001) regardless of admission diagnosis. Overall, they were more likely to receive thrombolysis (adjusted multivariable odds ratio 1.19, 95% CI 1.10 to 1.30, p<0.001) and the difference was more marked if they had non-specific ECG changes for heart disease rather than definite evidence of a myocardial infarction. There was no evidence of an important clinical delay in South Asians receiving thrombolysis after arrival at hospital.

Conclusions: There are ethnic differences in healthcare-seeking behaviour and the way doctors manage South Asians with chest pain. The relative underuse of ambulances by South Asians may either reflect cultural differences or geographical proximity to hospitals. Doctors may have a lower threshold for giving thrombolytic therapy to South Asian men with chest pain possibly because they are aware of the increased risk of coronary heart disease in this population.

Statistics from

It is important that healthcare provided to ethnic minorities is of similar quality to that given to white patients.1 Most past research has suggested that ethnic minorities receive less good quality care although this has been challenged by a recent study from the United States,2 which found that African Americans received better quality chronic and preventive care, but equivalent levels of acute care. In the United Kingdom, several publications have suggested that South Asian patients with chest pain may have less equitable access to cardiac care than white patients. They wait longer for specialist referral,3 and are less likely to receive a stress test4 or thrombolysis.5 6 One occupational cohort found little evidence for differential access to revascularization,7 but a more detailed clinical cohort with far larger numbers of South Asian patients found that they were less likely to receive revascularization after adjusting for clinical appropriateness.8

Understanding the reasons for such variations in care is complex and requires disentangling patient-related factors (sociocultural differences in healthcare-seeking behaviours), doctor-related factors (differential thresholds for diagnosing or initiating therapy) and the doctor–patient interaction (how patients express symptoms and how doctors respond). We have previously shown that there are sociocultural differences in how South Asian and African Caribbean patients respond to a hypothetical case vignette describing chest pain.9 10 In those studies, ethnic minority patients were more likely to state they would seek urgent medical care for a “grey” history of anginal-type symptoms. We recently undertook an audit of ischaemic heart disease patients registered in primary care practices that served a large South Asian community.11 As part of a self-completion questionnaire, we noted that South Asian patients were less likely to report calling 999 if they experienced chest pain lasting more than 15 minutes.12 This observation, if confirmed, suggests that there may be sociocultural differences in responding to chest pain. We used the MINAP dataset for 2002–2003 to test the hypothesis that South Asian patients with chest pain would be less likely to attend by ambulance and examined whether this would influence the likelihood and/or delay in receiving thrombolytic therapy.


The MINAP database collects data from all hospitals in England and Wales about the care and management of patients with acute coronary syndromes (ACS). It is a joint project between the Royal College of Physicians and the British Cardiac Society. It is managed by the Royal College of Physicians Clinical Effectiveness and Evaluation Unit on behalf of the Healthcare Commission.13 We were provided with a data extract for all patients who entered into the MINAP dataset between 1 January 2002 and 31 December 2003. In theory this should capture 100% of all hospital patients who presented with an ACS but there was no validation study to estimate the completeness of data capture. Subjects who were already in hospital at the time or who were transferred from another hospital were dropped from the analysis (8471 subjects) leaving 162 516 observations.

Outcomes measures

In response to our hypothesis, we examined three outcomes: (1) arriving at hospital by ambulance in response to an emergency 999 call; (2) whether thrombolytic therapy was provided; and (3i) time from symptom onset to hospital arrival, (3ii) time from arrival to receiving thrombolysis and (3iii) total time from symptom onset to receipt of thrombolytic therapy (kindly derived by MINAP team; Dr J Birkhead). We combined subjects who came by ambulance either by calling 999 directly or through their general practitioner as we were interested to see if there were ethnic differences in arrival by ambulance regardless of how the ambulance was called. We also undertook further subgroup analyses to examine if there were any differences in ambulance attendance that was prompted by the patient directly calling 999 or when this was done on their behalf by their general practitioner. We derived whether thrombolysis was given by using a combination of three variables (thrombolysis given, where thrombolysis was given and the time from arrival to thrombolysis). The subject was deemed to have received thrombolytic therapy if there was a positive response to the first question, and a non-missing valid response to the latter two variables. We dropped negative times for the onset to arrival or onset to thrombolysis outcomes as they were not possible. There were 344 valid negative times for hospital to thrombolysis because thrombolysis was started in the ambulance before arrival at hospital, but these had to be reset to a small positive value (10 minutes) to enable statistical transformation given the skewness of the data. We also dropped time values that we felt were too extreme, eg 2.13 hours from ambulance call to hospital arrival (>95th percentile). Some times were extremely short and we reclassified these to what we felt were more realistic times, eg onset of chest pain to arrival time of less than 20 minutes was reclassified as 20 minutes (0.7% altered). As the time variables were all positively skewed, we used the natural log of each time variable in our analyses.

Exposure and potential confounding variables

Our main exposure was ethnicity, which was based on the healthcare professional classifying subjects into five groups (Caucasian, black, Asian, oriental, other). As we were specifically interested in comparing South Asian with white Caucasian subjects we dropped subjects who were classified as black, oriental or “other”. This analysis is therefore based on 150 330 observations. We also included the following variables as potential confounders as they could be related to both ethnicity and how subjects sought healthcare or received it. These were age group (20–54 years, 55–64 years, 65–74 years, 75+ years), gender, past history of myocardial infarction (MI), angina, hypertension, diabetes and smoking status.

Statistical methods

We used logistic regression analysis to calculate the odds ratio (OR; 95% confidence interval (CI), p value) for our dichotomous outcomes (arriving by ambulance; receiving thrombolysis) and linear regression for our continuous outcomes by ethnicity adjusting for a number of other covariates; age group, sex, co-morbidity (past diagnosis of MI, angina, diabetes). We undertook two different sets of stratified analyses. In relation to calling an ambulance, which is predominantly determined by the experience and interpretation of chest pain, we used the doctors’ initial diagnosis on admission as at this stage the clinician will base this on the clinical history and ECG changes but will not usually be aided by biochemical results. These were aggregated into the following groups: definite or probable MI; ACS; chest pain of unknown cause or other diagnosis. We used a different stratification strategy in relation to thrombolysis, as the clinical decision to administer thrombolytic therapy is predominantly guided by evidence of ST elevation as shown on the ECG, although there are grey areas, for example when the patient has left bundle branch block. For this analysis we dropped patients with a normal ECG and classified patients as showing ST elevation (indicative of a MI) or other ECG abnormalities consistent with some evidence of heart disease (eg T-wave inversion). Because of gender differences in the management of ischaemic heart disease,14 we also examined whether the results differed by gender and tested for an ethnicity–gender interaction. The main analyses were performed only on those with complete data on ethnicity, but we also undertook a sensitivity analysis by imputing missing values. We used multiple imputation by chained equations and created four copies of the dataset with imputed values. Parameter estimates (OR and regression coefficients) were then averaged across datasets to give a single value and “Rubin rules” were used to allow for the between and within-imputation components of variation.15


The ethnic distribution of the subjects was as follows: 118 323 subjects were classified as Caucasian, 5486 Asian and 26 521 subjects were unclassified (16.3% of the total dataset). The basic descriptive details are shown in table 1 for all participants and by ethnicity. Asian subjects were more likely to be male, younger, receive thrombolysis, have had a past history of MI, hypertension, diabetes and have an admission diagnosis of acute MI or an ACS. They were less likely to have arrived by ambulance. The proportion of missing data varied from sex (0.6%) to diabetes status (16.3%). We examined whether the pattern of missingness was associated with ethnicity. Asian subjects were more likely to have missing data for gender, age group, diabetes, and smoking status. In contrast they were less likely to have missing data on arrival by ambulance and thrombolytic therapy.

Table 1 Basic descriptives of MINAP database

In a logistic model adjusting for age group and sex (model 1), Asians were less likely (OR 0.64, 95% CI 0.60 to 0.69) to arrive at hospital by ambulance (table 2). This was true regardless of the initial diagnosis on admission. Adjustment for a past history of MI, angina, diabetes, hypertension and smoking status in general slightly increased the ethnic differences. The differences were more marked for women than men (OR for women 0.57, 95% CI 0.49 to 0.66; OR for men 0.67, 95% CI 0.61 to 0.73; p value for ethnicity and gender interaction 0.05). The majority of patients who came by ambulance called this themselves (79.7%). The relative ethnic differences were, however, more marked when we compared ambulance arrivals that were self-initiated compared with those in which the patient had originally called the general practitioner who subsequently ordered the ambulance (OR 0.72, 95% CI 0.67 to 0.77 for direct ambulance call; OR 0.29, 95% CI 0.25 to 0.34 for general practitioner ambulance call).

Table 2 Multivariable regression models for whether patients arrived by ambulance and time from symptom onset to hospital arrival by ethnicity

The geometric mean time between symptom onset and arrival at hospital for all patients regardless of diagnosis was 3.10 hours (Caucasians 3.07, South Asians 3.07). The data on time shown in tables 2 and 3 are on a natural log scale, thus making the coefficients very small. It is helpful to multiply the regression coefficients by 100 to derive the symmetric percentage differences or sympercents.16 The coefficient of 0.01 for onset to arrival for all patients is thus equivalent to Asians taking a 1s% increase in mean time. Alternatively, the coefficients can be back-transformed to give the ratio of times. Overall, there were no ethnic differences for time from onset to arrival, but Asians arrived faster when the diagnosis was MI and slower when it was ACS or other chest pain.

Table 3 Multivariable regression models for receiving thrombolysis, time from symptom onset and arrival at hospital to receipt of thrombolysis by ethnicity

Although approximately half of the patients (44%) received thrombolysis, this varied greatly by whether there was evidence of ST elevation. Only 9% of patients receiving thrombolysis did not have ST elevation, whereas 14% of those not receiving thrombolysis did. Overall, there was no marked ethnic difference in receiving thrombolysis but after adjustment and on stratified analyses Asians were more likely to receive thrombolysis (table 3). This effect was more marked for patients with non-specific ECG changes (OR 1.84). An interaction test between ethnicity and ECG features showed that these differences were unlikely to be due to chance (p = 0.001). Stratified analyses by gender also showed that ethnic differences were more marked for men than women (eg OR for all male patients 1.21, p<0.001; OR for all female patients 0.92, p = 0.20; test for interaction between sex and ethnicity <0.001).

For the subset of patients who received thrombolytic therapy, the geometric mean time between hospital arrival and therapy was 0.63 hours (Caucasians 0.63, South Asians 0.66). The overall time between onset and therapy for those who received thrombolytic therapy was 3.34 hours (Caucasians 3.34, South Asians 3.15). There was no marked evidence of any delay for South Asian patients in receiving thrombolysis from symptom onset. There was some evidence that South Asian patients waited longer from hospital arrival to receiving thrombolysis, even with ST elevation, although this was mostly explained by co-morbidity and smoking status as the results were markedly attenuated after adjustment for the other covariates (p = 0.13). A repeat analysis using the original raw data produced similar results.

We undertook a sensitivity analysis using imputed values for ethnicity. This did not qualitatively alter our conclusions. The OR for arrival by ambulance were very similar (eg age and sex adjusted OR for all patients 0.67, 95% CI 0.59 to 0.77, p<0.0001), whereas for thrombolysis they were very similar or slightly attenuated (eg age and sex-adjusted OR for other chest pain 1.78, 95% CI 1.21 to 2.59, p = 0.003). Similarly, the results for time from onset to arrival, onset to thrombolysis and hospital arrival to thrombolysis were little changed after imputation (data not shown).


This is the first large study in the United Kingdom to observe clear ethnic differences between South Asians and Caucasians in their response to chest pain. The main differences relate to how patients with chest pain choose to get to hospital and whether they receive thrombolysis. South Asians were less likely to arrive at hospital by ambulance compared with their white counterparts and this difference was more marked for women compared with men. This could not be explained by the seriousness of the symptoms as the same pattern was seen regardless of the initial clinical diagnosis. South Asians were slightly more likely to receive thrombolytic therapy, but there was an interaction between ethnicity and ECG changes so that more marked ethnic differences were seen for non-specific ECG features. This suggests that the decision to give thrombolytic therapy is unaffected by the patient’s ethnicity when there is strong objective evidence from the ECG but when the ECG evidence is more dubious South Asians were more likely to receive thrombolysis. This may be because clinicians are aware that young South Asian men are at increased risk of heart disease compared with their white and female counterparts,17 and/or they think South Asian patients present with more atypical histories.6 The fact that there was an interaction between ethnicity and gender for receiving thrombolysis suggests that the first explanation may be more relevant as the other proposed reason is likely to apply equally to male and female South Asians.

Reassuringly, in our study there was no evidence that South Asians waited longer to receive thrombolysis from symptom onset. South Asians overall waited slightly longer from hospital arrival to receipt of thrombolysis even when the clinician thought they had an MI. This could be because they were less likely to have come by ambulance and therefore have an ECG on arrival. Alternatively, the clinicians may have found it took longer to get a convincing history of cardiac pain. The time difference, however, was small in absolute terms and it is debatable whether it is clinically relevant, despite statistical significance. This is similar to evidence from the United States where African Americans, Hispanic individuals and Asian/Pacific Islanders all had slightly longer waiting times (between two and seven minutes) for receiving thrombolytic therapy from hospital arrival.18 This study used elegant multilevel models to demonstrate that some of the differences reflected between-hospital characteristics although there still remained ethnic differences within hospitals that could not be explained by clinical factors.

Relatively few studies have examined how ethnicity may determine healthcare-seeking behaviours and mode of presentation. A small study of Bangladeshi patients in east London6 found that these patients had a more atypical history and took longer to receive thrombolysis. The influence of a past cardiac history is unclear as it is sometimes associated with slower arrival times19 and in other cases it either makes no difference or shortens the time to hospital arrival.20 21 Our results suggest that the increased prevalence of MI, hypertension and diabetes did not have a major confounding influence on the results.

A range of psychological factors, such as fatalism, health beliefs, perceived threat and concern about troubling others, have been suggested as important reasons why patients delay in seeking hospital care.20 21 There are relatively few data to determine whether such factors differ by ethnicity and their relative importance. In a previous population-based study that used hypothetical case vignettes of angina rather than MI, South Asians were both more fatalistic about health but also perceived more threat from the chest pain scenario.9 They were also more likely to consult other family members and lay networks, which has been postulated to result in a delay. Overall, however, they reported a greater probability of seeking immediate healthcare.9

One possible explanation for our results, consistent with the above, is that South Asians are either more reluctant to call an ambulance and/or choose to make their own way to hospital as they may live closer to acute hospitals than Caucasian patients. Geographical proximity has previously been shown to influence access to cardiological services.22 The perceived nature of the pain and the use of social networks may determine their response. When the episode of pain is severe they seek care more rapidly and arrive faster than their Caucasian counterparts. When the pain is less severe, shorter lasting, repeated or atypical, as in ACS or non-cardiac chest pain, they may seek the advice of other family members before making their own way and therefore take longer to arrive at hospital.


There are several important limitations to our results. First, we have only been able to categorise patients into a heterogeneous group labelled “Asians”23 and we assume that that most of these patients are South Asians. This is far from ideal and within this category there are diverse religious, cultural and social groups who may not all respond in the same way to the experience of chest pain. The derivation of the ethnicity code was also based on staff classification rather than self-report. We have no data on the validity of staff versus self-classification but given the very broad grouping that has been used it is likely that there will be a high degree of concordance. Some indirect support for the validity of the classification is the much higher probability of reporting diabetes in the Asian compared with the Caucasian group, which is what one would have expected. It is unlikely that South Asian patients would have been classified as Caucasian or vice versa and if they had this would have, if anything, attenuated our associations. It is possible that some South Asian patients, for example those of east African or Caribbean origin may have been classified as “black” or “other” but these were not included in our analyses. Another problem was the degree of missing data. This varied considerably but was in general similar in proportion for both ethnic groups (other than the age group variable). Our sensitivity analyses using imputed data on ethnicity made relatively little difference to the results, and it is hard to imagine why ethnicity would be systematically underrecorded depending on whether the patient came by ambulance or received thrombolysis.

Public health importance

It is generally reassuring to observe that South Asians are if anything more likely to receive thrombolysis, given past concerns about inequitable access to cardiac services, and do not wait longer for treatment from symptom onset. We do not believe that the modest ethnic differences in receiving thrombolysis after arrival at hospital will have any serious impact on prognosis. The impact of not arriving at hospital by ambulance is harder to assess. The benefits of ambulance arrival may include a quicker diagnosis as a prehospital ECG may already show ST elevation, the provision of supportive therapy and most importantly the opportunity to treat a life-threatening arrhythmia. In our dataset, less than 0.5% of patients received thrombolysis before hospital arrival, either by a doctor at home or in the ambulance. If there is an increasing trend for paramedics to give therapy earlier, however, then there may be additional benefits of calling an ambulance.

We suspect that the lower threshold for giving South Asian patients thrombolytic therapy, especially for atypical chest pain, reflects clinical awareness that South Asian men are at greater risk of MI, especially in populations with large ethnic minorities. This may be of some concern as the decision to administer thrombolytic therapy should be based on the clinical history, ECG evidence and, if possible, biochemical evidence of infarction. South Asians may therefore be more likely to have inappropriate “overtreatment” with the potential for adverse events associated with therapy.


In this large national dataset, South Asian patients with chest pain are more likely to make their own way to hospital rather than call an ambulance and are more likely to receive thrombolytic therapy especially if they are male. Qualitative studies, both with patients and doctors are required to understand the reasons behind these ethnic differences. In general, there is no evidence overall that South Asians are disadvantaged or receive less equitable healthcare.


These data were provided by MINPAP (the National Audit of Myocardial Infarction Project), Clinical Effectiveness and Evaluation Unit, Royal College of Physicians of London. The authors would like to thank Dr John Birkhead for his helpful advice about the MINAP data, and Professor Nish Chaturvedi and the two reviewers for helpful comments on the manuscript.



  • Competing interests: None.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.