Inequity of access to investigation and effect on clinical outcomes: prognostic study of coronary angiography for suspected stable angina pectorisBMJ 2008; 336 doi: https://doi.org/10.1136/bmj.39534.571042.BE (Published 08 May 2008) Cite this as: BMJ 2008;336:1058
- Neha Sekhri, clinical research fellow1,
- Adam Timmis, professor of clinical cardiology1,
- Ruoling Chen, senior lecturer in epidemiology2,
- Cornelia Junghans, research fellow in epidemiology2,
- Niamh Walsh, statistician3,
- Justin Zaman, clinical research fellow in epidemiology2,
- Sandra Eldridge, professor of biostatistics3,
- Harry Hemingway, professor of clinical epidemiology2,
- Gene Feder, professor of primary health care4
- 1Cardiac Directorate, Barts and the London NHS Trust, London
- 2Department of Epidemiology and Public Health, University College London Medical School, London
- 3Centre for Health Sciences, Barts and the London, Queen Mary’s School of Medicine and Dentistry, Queen Mary, University of London, London
- 4Academic Unit of Primary Health Care, University of Bristol, Bristol BS8 2AA
- Correspondence to: G Feder
- Accepted 26 March 2008
Objectives To determine whether coronary angiography for suspected stable angina pectoris is underused in older patients, women, south Asian patients, and those from socioeconomically deprived areas, and, if it is, whether this is associated with higher coronary event rates.
Design Multicentre cohort with five year follow-up.
Setting Six ambulatory care clinics in England.
Participants 1375 consecutive patients in whom coronary angiography was individually rated as appropriate with the Rand consensus method.
Main outcome measures Receipt of angiography (420 procedures); coronary mortality and acute coronary syndrome events.
Results In a multivariable analysis, angiography was less likely to be performed in patients aged over 64 compared with those aged under 50 (hazard ratio 0.60, 95% confidence interval 0.38 to 0.96), women compared with men (0.42, 0.35 to 0.50), south Asians compared with white people (0.48, 0.34 to 0.67), and patients in the most deprived fifth compared with the other four fifths (0.66, 0.40 to 1.08). Not undergoing angiography when it was deemed appropriate was associated with higher rates of coronary event.
Conclusions At an early stage after presentation with suspected angina, coronary angiography is underused in older people, women, south Asians, and people from deprived areas. Not receiving appropriate angiography was associated with a higher risk of coronary events in all groups. Interventions based on clinical guidance that supports individualised management decisions might improve access and outcomes.
We do not know whether potential inequities in the management of cardiovascular disease have consequences in terms of prognosis. Equitable access to health care is a widely accepted goal for health services internationally and a key principle of the NHS.1 Most2 3 4 but not all5 6 studies suggest that older people,7 women,8 ethnic minorities,9 and those who are socioeconomically deprived10 have less access to effective interventions for stable angina or acute coronary events.
There are still important uncertainties that hamper the transition from “talk” about inequitable access to “action” and the development of policies to reduce inequities.11 Firstly, in patients with stable angina investigators rarely measure the effect on cardiac events or mortality of variation in management.12 Prognostic studies of patients early in the course of chronic coronary disease are lacking. Secondly, most studies that have found inequitable access to services are based on patients with acute events13 or identified in hospital outpatient settings,14 although many presented years earlier in primary care. Few studies have identified patients at an early stage of presentation in the context of primary or ambulatory care. Thirdly, most studies have examined access to treatments. Yet investigations such as coronary angiography are a prerequisite for revascularisation and some decisions about medical management. Finally, most previous studies have not taken account of the appropriateness of coronary angiography15 or the confounding of comorbidity in decisions about investigations. We have previously shown that appropriateness criteria for revascularisation developed by the Rand-UCLA method predictedoutcomes in a cohort of 2552 patients presenting for angiography.16
We assessed patients in rapid access chest pain clinics who were deemed appropriate for coronary angiography to determine whether angiography was underused inpatients who were older, women, from a south Asian ethnic group, or from economically deprived areas; and whether underuse was associated with higher rates of coronary events. In the United Kingdom, most people identified in general practice with recent onset stable chest pain are assessed in rapid access chest pain clinics. These ambulatory care clinics are run by cardiology teams and accept same day referrals from family physicians of patients with recent onset chest pain who have not previously been investigated or diagnosed with coronary disease.17
We used a multicentre cohort of consecutive patients attending these clinics and independently assessed their appropriateness for angiography using criteria that we developed using the Rand-UCLA consensus method.16
Data sources and definitions
Individual clinical data on 10 634 consecutive people attending six rapid access chest pain clinics were electronically recorded from 2 January 1996 to 31 December 2002 on identical databases, details of which have been reported previously.18 An independent panel deemed 1375 of these patients as appropriate for angiography. Figure 1⇓ shows the origins of the study sample. The study size was determined by the availability of systematically collected comprehensive baseline data in the six clinics using the same electronic record system. The clinics were sited in Blackburn, Burnley, Kingston, Manchester, Newham, and Oldchurch. Clinical data collected at attendance included age, sex, ethnicity, duration of symptoms, character of chest pain, smoking status, history of hypertension, diagnosis of diabetes, resting and exercise electrocardiograms (ECG), pulse rate, systolic blood pressure, drugs, and follow-up plan on discharge. At the end of the consultation the clinician recorded the cause of chest pain (angina or non-cardiac chest pain).
During the consultation the clinician who assessed the patient in the clinic ascribed ethnicity as Asian, white, black, or other. “Asian” was used for patients of Indian, Pakistani, Sri Lankan, and Bangladeshi origin, and we have referred to this category as “south Asian” in this paper. In our analysis we compared white and south Asian patients. In the UK south Asian people are at particularly high risk of coronary artery disease.19 We did not have the power to analyse differences in access and outcomes of other ethnic minority groups (African and Afro-Caribbean) because of their relatively low incidence of coronary disease and small numbers of patients in our cohort.
We used fifths of the 2001 census Townsend index, a score of material deprivation20 based on four variables (unemployment, overcrowding, car ownership, and home ownership). In our analysis we compared the most deprived (highest fifth) with the other four combined.
Appropriateness ratings for angiography
Before undertaking this study, we defined the appropriateness of angiography with two expert panels using a modification of the Rand/UCLA method of systematically combining evidence with expert opinion.21 We identified 13 clinical descriptors that influence the decision to undertake angiography on people with suspected or confirmed angina: age (<40, 40-49, 50-59, 60-74, 75-84), sex, typicality of symptoms, severity of symptoms (Canadian Cardiovascular Society (CCS) class), drugs for symptoms (submaximal, maximal), coronary risk factors (low, medium, high), previous acute coronary syndrome, timing of acute coronary syndrome (within past year, more than a year ago), resting ECG (normal, abnormal), findings on exercise ECG (none, normal, abnormal, very abnormal), previous result of angiography, (abnormal, normal), timing of angiography (within past year, more than a year ago), previous revascularisation. We used clinically meaningful combinations of these factors to define specific clinical indications that we grouped in six broad clinical presentations.
Each panel consisted of five cardiologists, five family physicians, and one cardiothoracic surgeon. Panellists judged appropriateness for angiography on a 9point scale. Scores of 7-9 indicated an appropriate investigation, when benefit from subsequent treatment was judged to outweigh harm. Panellists were invited to base their ratings on evidence from peer reviewed research and were provided with a literature review with evidence tables and graded strength of evidence.22 They were asked to rate purely on the basis of clinical benefit and harm, without consideration of financial or workload constraints. Panel members carried out the first round of rating independently and then had the opportunity to change their ratings in the light of a panel discussion.
We matched indications and their associated ratings to patients from the six clinics, based on the 13 clinical individual clinical characteristics and initial investigations listed above, with the exception of CCS class. Patients were included in our analysis if either panel had deemed their indications appropriate for angiography.
Our sample comprised white or south Asian patients with chest pain and no known coronary heart disease presenting to rapid access chest pain clinics and deemed appropriate for coronary angiography by either panel. We matched the patient’s characteristics (age; typical, atypical, or non-cardiac symptoms of chest pain; cardiac risk factors; results of resting and exercise ECGs) to the panels’ ratings to define whether they were appropriate for coronary angiography.
Missing baseline data
In the total cohort of 10 634 patients, 801 had incomplete baseline data (7.5%). We did not include these patients in the main analysis; a sensitivity analysis showed that the rates of coronary deaths and non-fatal events for patients with angina and non-cardiac chest pain with missing baseline data were not significantly different from those with angina and non-cardiac chest pain and complete baseline data.
Follow-up of patients and outcome measures
Over 99% of patients were successfully matched at the Office for National Statistics and the NHS-wide clearing system who informed us of the date and ICD-10 (international classification of diseases, 10th revision) coded cause of death and hospital discharge, respectively. Average follow-up for the cohort was three years, until the end of 2003. Analysis of receipt of angiography was truncated three years after the index clinic visit. Data on use of coronary angiography were obtained from the NHS-wide clearing system. We analysed a single combined end point of death from coronary heart disease (ICD I20-I25) and admission to hospital because of acute coronary syndromes, including acute myocardial infarction (I21-I23) and unstable angina (I20.0-I20.9, I24.0, I24.8 and I24.9). We used the primary discharge diagnosis after hospital admission to define non-fatal events in these analyses. To define a group of patients without major comorbidity we identified those without any hospital admission for non-coronary reasons within a year of the clinic visit.
We performed all analyses in STATA version 8.
Receipt of angiography
We fitted univariate and multivariable Cox’s regression models to estimate rates of receipt of angiography by age group, sex, ethnic group, and deprivation. The multivariable models included the clinic in which patients were assessed as a random variable. The proportional hazards assumption of the final Cox model was tested by calculating Schoenfeld residuals. We used severity of symptoms to define appropriateness ratings. As Canadian Cardiovascular Society (CCS) angina class was not recorded in this cohort, we assumed all patients were CCS class I-II (mild symptom severity). We chose this assumption because it would underestimate the need for angiography. In a further analysis we made the opposite assumption, coding the patients as class III-IV.
To measure the effect of underuse of angiography on morbidity and mortality in these demographic subgroups we fitted univariate and multivariable random effects Cox’s regression models to estimate the hazard of non-fatal acute coronary syndrome or death from coronary heart disease within five years of clinic visit according to whether or not patients underwent coronary angiography within three years of the index clinic attendance. To address the possibility of confounding by indication—those who did not undergo the procedure might have had too high a coronary risk—we adjusted for all other demographic variables, secondary prevention medication, and result of exercise ECG. The regression models were fitted for each demographic subgroup separately (for example, men separately), and the hazard ratios presented refer to the risk of an event related to not undergoing angiography. The final model of risk of a coronary event incorporated a propensity score for individuals that, in addition to the above variables, included diabetes, hypertension, hypercholesterolaemia, and duration and character of symptoms. The propensity score is the probability of receiving treatment for a patient with specific prognostic factors and is a scalar summary of all measured confounders.23 We assessed differences in the event rates between subgroups with an interaction test.24
Patients with severe comorbid conditions might not have been referred for angiography, though they fulfilled the panels’ criteria for appropriateness. To control for this possible confounding, in an additional analysis we excluded all patients with admissions for non-coronary conditions (all ICD chapters apart from I for circulatory disease, R00-03 for symptoms and signs involving the circulatory system, and R07.4 for unspecified chest pain) in the year after the index attendance.
Receipt of coronary angiography by appropriate patients
Table 1⇓ shows the characteristics of 1375 patien ts according to whether or not they underwent angiography. Overall, 69% of patients deemed appropriate for angiography did not undergo this investigation. Table 2⇓ shows hazard ratios by demographic groups for receipt of angiography by those deemed appropriate for this procedure after the index clinic visit. Multivariable analysis shows that people aged over 65, women, south Asian patients, and those in the most deprived population fifth were less likely to receive coronary angiography compared with those aged under 50, men, white patients, and those living in the less deprived fifths, respectively. The inequitable pattern of receipt was not explained by comorbidity in patients who did not receive angiography; it persisted when we excluded from the analysis the 252 patients with a non-coronary admission in the year after the index visit to the chest pain clinic. In an additional analysis, when patients were assumed to have more severe angina (Canadian Cardiac Society class III-IV) we found similar results: multiple adjusted hazards ratios 0.64 (0.35 to 1.17) in 65 and older, 0.36 (0.30 to 0.44) in women, 0.53 (0.37 to 0.76) in south Asians, and 0.64 (0.43 to 0.96) in the most deprived fifth.
Coronary events by receipt of coronary angiogram in appropriate patients
For the whole cohort, prognosis of patients deemed appropriate for coronary angiography and who received the procedure was better than those who did not (hazards ratio 1.71, 95% confidence interval 1.24 to 2.34). Table 3⇓ shows univariate and multivariable hazard ratios (fig 2)⇓ for a coronary event over five years of follow-up according to whether or not patients had undergone angiography in the first three years. Hazard ratios for a coronary event were >1 for all demographic groups, with a significantly increased risk for younger people, older people, men, women, white people, and the less deprived fifths. Tests for interaction showed no significant difference between hazard ratios for the three age groups (P=0.203 and 0.223), between men and women (P=0.493), between white people and south Asians (P=0.426), and between the most deprived and less deprived population fifths (P=0.077). The worse prognosis for appropriate patients who did not undergo angiography persisted when we excluded those with non-coronary admissions in the year after index clinic attendance (table 4).⇓
Despite the strong emphasis on fairness of access to cardiac investigation in the UK national service framework for coronary heart disease published in 2000, we found that older people, women, and south Asians who were deemed appropriate for coronary angiography were significantly less likely to receive the investigation. Deaths from coronary heart disease and admissions for unstable angina and myocardial infarction were more common in patients deemed appropriate for coronary angiography but who had not received it. The prognostic validity of appropriateness ratings for coronary angiography was apparent in all demographic subgroups, with hazard ratios for a coronary event >1for appropriate patients who did not undergo angiography, although the difference was not significant for south Asians.
Strengths and weaknesses
The key contributions of this study are the robust measurement of inequity in use of coronary angiography, taking into account the appropriateness of the investigation and comorbidity; and demonstration of the detrimental effect of this inequity on prognosis. This has important implications for clinical decision making in the first consultation patients with stable chest pain have with cardiac specialists. Our findings are probably applicable to patients attending chest pain clinics throughout the UK as the risk profile and prognosis of the whole cohort is comparable with that in other studies in chest pain clinics.25 26 Appropriateness ratings for coronary angiography can inform better decisions by clinicians. We tested specific ratings against conventional guidelines in a web based trial of investigative decisions for patients with chest pain and found that the ratings significantly improved the ordering of exercise electrocardiography and coronary angiography.22
Our analysis of referral and coronary outcomes from rapid access chest pain clinics was multicentred and based on a clinical population with uniformly collected data at the patient level, including ethnicity. We used independent and reliable27 measures of coronary events and deaths from National Statistics and the NHS-wide clearing system. These clinics are ideal settings for investigating access to angiography and the outcomes of patients with incident angina.28 A major strength of the study is the precise attribution of need for coronary angiography using appropriateness ratings. The sample size allowed robust comparison between demographic and clinical subgroups in terms of receipt or non-receipt of angiography and, for most subgroup comparisons, of coronary events. Other strengths include the use of data at the patient level on non-coronary admissions to test the effect of potential indication bias.
A potential limitation in this study is the broad classification of ethnicity and its source in the labelling of patients by clinicians who entered the information on the database. The validity of these categories, whatever their source, depends on the research question.29 We were able to answer questions about inequity relating to a broad category of ethnicity that maps on to census classifications. It is justified to pose these questions because previous research in the UK suggests inequitable access to some cardiac services using broad ethnicity categories30 and also because discrimination by clinicians against ethnically heterogeneous minority groups might be an explanation for inequity.31 Furthermore, in an analysis of 33 patients in chest pain clinics, Zaman (J Zamman, personal communication) found a high agreement (κ of 0.77) between ethnic group according to the patients and the cardiologist.
A second limitation stems from the use of a deprivation measure, the Townsend score, that is based on ward rather than individual level socioeconomic status, and therefore the analysis is less precise and prone to the ecological fallacy: ascription of average deprivation across a ward to an individual who lives in that ward. On the other hand, Britton and colleagues,6 used individual level data and found no inequity by socioeconomic status or ethnicity.
Finally, we did not assess potential inequities related to age, ethnicity, or deprivation in referral of patients with chest pain from primary care into specialist services.32
Relevance of findings
Why do these inequities in use of coronary angiography arise? They could reflect differential referral of appropriate patients,33 although in a previous study differential referral for treatment proved not to be the cause of differences in rates of coronary revascularisation.14 Qualitative research in chest pain clinics did not show any marked differences in clinical assessment by age or sex, although this did differ between south Asians and whites, largely because of language barriers.34 Some variation in use of coronary angiography might result from patients’ choices not to undergo the procedure, which could vary systematically by demographic group.35 There is a role for further qualitative research to elucidate why patients who are deemed appropriate for coronary angiography do not receive it.
Over and above inequalities between groups in use of angiography, it is striking that there is underuse across our whole cohort, with less than a third of appropriate patients receiving it. This is consistent with the findings of the Euro Heart study from outpatient cardiology services in 34 countries.12 We think it unlikely that the appropriateness ratings for coronary angiography ascribed to the cohort are invalid, thus overestimating the proportion of patients who should have the procedure. The association between non-receipt of angiography with increased coronary events is external validation of the ratings and this association extends to those patients of uncertain appropriateness with a hazard ratio of 1.97 (1.17 to 3.34).
Our findings and those of previous studies13 should inform policy to deal with sources of inequity in the provision of cardiac services. The prospective validity of appropriateness ratings for cardiac investigation developed with the Rand/UCLA method means that they are a robust basis for guidance to clinicians in rapid access chest pain clinics. Application of that guidance to decision making about individual patients might help to address inequitable access to angiography for some groups of patients with recent onset chest pain. The potential of appropriate ratings for specific patients to improve investigative decisions for patients with chest pain22 now needs to be tested in real clinical settings. Appropriateness ratings incorporated into decision support systems might have a wider application for investigative and treatment decisions in other clinical settings.
What is already known on this topic
Potential inequities in the management of cardiovascular disease might have consequences in terms of prognosis
Access to cardiac investigation early in the course of chronic coronary disease might be inequitable
Few studies on inequitable use of investigations consider appropriateness of the procedures
What this study adds
In patients with angina of recent onset deemed appropriate for angiography, older patients, women, south Asians, and those living in the most deprived wards were less likely to undergo this procedure
Older patients, women, south Asians, and those from the most deprived areas deemed appropriate for coronary angiography who did not receive it had a higher rate of coronary events than those who did
In memory of Sarah Cotter (1949-2004), who worked on an earlier version of this analysis.
Contributors: All authors contributed to the design of this study. NS, NW, CJ, and RC carried out the analyses. GF wrote the first draft of the paper and is guarantor, and all authors contributed to revisions.
Funding: NHS service delivery and organisation research and development programme.
Competing interests: None declared.
Ethical approval: Trent multicentre research ethics committee and all local research ethics committees.
Provenance and peer review: Not commissioned; externally peer reviewed.