Despite evidence of the value of secondary prevention in coronary heart disease (CHD),1 reports from various countries over recent years have shown that its provision is suboptimal in relation to prescribing, monitoring risk factors and managing modifiable risk.2^–6 Many studies have shown that interventions in primary care can lead to improved provision of secondary prevention,2 7^–11 but descriptions of these interventions, which are often complex in nature, have tended to lack detail.12 There is a need for careful description of such complex health service interventions, and of the context in which trials take place,13 in order that healthcare planners, policy makers and clinicians can translate research into practice.

The relevance and importance of knowledge of different systems of healthcare has been illustrated by Macinko et al14 who have shown how the strength of a country’s primary care system is negatively associated with premature mortality from various causes, including heart disease. The primary care system within the UK National Health Service (NHS) is recognised as being strong, with a Starfield score of 29/30.15 Recent developments in its structure and organisation include increasing emphasis on the use of computer technology, disease registers and financial incentives relating to general practitioners’ (GPs’) provision of care for people with chronic disease.16 While evidence suggests that the pay-for-performance programme within the new NHS GP contract of 2004 led to improved provision of clinical care,17 the quality of care was already improving before its introduction.18 In contrast, within the Republic of Ireland (RoI), which has a less strong primary care system (estimated Starfield score 18/30 but no reported formal calculation identified), such developments and the availability of comparative performance data have been very limited.

We are currently engaged in a randomised controlled trial19 of a complex health service intervention, based in primary care and designed to improve secondary prevention for people with CHD in two different healthcare systems, one in the RoI (a mixed private/public system) and the other in Northern Ireland (NI) (the NHS, publicly funded) (table 1). We plan to provide results which allow conclusions that are generalisable to other settings by describing the context of intervention and “usual” care in detail.

In this paper we aim to describe the context of the trial and examine possible differences in baseline management of cardiovascular risk, including lifestyle, between subsamples recruited from the two different healthcare systems, which vary in the strength of their primary care components and in financial incentives for GPs’ performance in disease management.

METHODS

Following development of the intervention described previously,20 21 and a pilot study, we recruited study participants from both the RoI and NI. From Health Board lists in three different centres (Galway and Dublin, RoI; Belfast, NI) we identified general practices which had a practice nurse involved in general patient care and had not participated in our pilot or in “Heartwatch”3 (a limited national pilot programme for secondary prevention in the RoI). In order to identify at least 40 potentially eligible patients per practice, we included only those with a General Medical Services (GMS) list size not less than 700 (in the RoI) or with a NHS list size not less than 1800 (in NI).

Practices in each centre were randomly selected from those thus identified, by an individual independent of the research team, using computer-generated random numbers, and were contacted by telephone to determine their interest in participating. Those who expressed interest were posted study information and offered a practice visit at which their eligibility and commitment to the study were confirmed. If a practice declined to participate, another was selected and invited in the same way, until 16 in each centre were recruited.

Participating practices prepared lists of all patients with known CHD (documented myocardial infarction (MI), coronary artery bypass graft (CABG), angioplasty or angina (confirmed by exercise stress test, isotope test or coronary angiogram)). Those with significant mental or physical illness likely to impair capacity to change lifestyle behaviour or assimilate new information were excluded. For each practice a computer programme was used at a remote location to randomly order potential participants: they were posted an invitation to participate, a reply slip and questionnaires relating to quality of life (SF-12) and lifestyle (validated questionnaires for diet22 and exercise23). Allowing 6 weeks for non-response, during which reminders were posted, patients were invited in sequence from the random order list until 20 in each practice agreed to participate.

Those who consented attended a baseline consultation at their own practice where the practice nurse measured their blood pressure (BP) using an Omron M7 automated validated syphgomanometer (Omron, Kyoto, Japan), waist–hip ratio (WHR) and body mass index (BMI), took a venous blood sample for serum cholesterol assay (non-fasting) and recorded prescribed medications. A research nurse observed the fidelity of the measurement process and accuracy of recording in one randomly selected baseline consultation for each practitioner involved. The research nurse also reviewed, with their consent, participants’ medical records and noted documentation confirming their CHD diagnosis, previous observations of cardiovascular risk factors (BP, lipids, glucose, smoking habit) and health service use within the previous year.

The age and gender of those declining to participate and, for RoI non-participants, eligibility for free medical services were recorded. Data were collected between October 2004 and January 2006. This was within the first year of a new NHS contract for NI GPs under which they returned data to the NHS on 1 April 2005 for payments based on indicators of the quality of the care they provided.

Baseline data collection was completed prior to randomisation of practices to intervention and control groups in order to minimise potential recruitment bias. Measurement and medical record data were collected initially on paper records and then, with questionnaire data, were entered onto a computer database. Accuracy of data entry was checked for a random 5% sample of participants and, of 1344 items reviewed, inaccuracy was found in 53 (3.9%).

Sample size calculation

A total sample size of 907 patients from 46 practices was estimated to allow for 30% patient attrition, 10% practice attrition and a varying but minimum power of 80% to detect clinically significant differences between intervention and control groups in proportions with BP >140/90 mm Hg and total cholesterol >5.0 mmol/l, in measures of physical and mental well being (SF-12) and in numbers of hospital admissions. The target number of practices was increased to 48 so that equal numbers (16; eight intervention, eight control) would be recruited in each centre.

Data analysis

Computer data were analysed using SPSS v 14. Categorical variables were compared between NI and RoI groups using χ² testing; other measurements were compared using t tests. The extent of variation within practices and centres was explored using cluster analysis. Regression analysis was performed to further examine differences between groups.

Ethics approval

Ethics approval was given for the study by the Irish College of General Practitioners (RoI) and Queen’s University Research Ethics Committee (NI).

RESULTS

Participating practices

In NI, the recruitment rate for practices was 32%: to recruit our target of 16 practices for the trial, 50 were telephoned and invited to participate. In RoI 35% (40/115) agreed, but eight were subsequently excluded because they were unable to identify adequate numbers of patients with CHD (five), encountered staffing problems (two) or had unpredicted building renovations (one). The most frequently cited reasons for declining to participate were workload or lack of time (n = 50) and lack of interest (n = 38).

Participating practices’ list sizes ranged from 1500 to 15 000; mean total list size was 3500 in the RoI (GMS+private) and 4825 in NI (all NHS). The comparative mean national list size for NI was 4897 but was not identified for the RoI. Numbers of GPs (whole time equivalents) in the practices ranged from one to 4.5. In both jurisdictions single handed practices were under-represented: 37.5% (12/32) and 6% (1/16) in our sample compared to 51% and 19% of all practices in the RoI and NI, respectively. In relation to 2004–05 returns for the Quality and Outcomes Framework CHD domain, the participating NI practices scored a mean of 97.5% of available points (range 83.1% to 100%) compared to a mean of 96.6% for all practices in the Health Board areas.

Based on Multiple Deprivation Measure Scores attributed to practice postcodes (NI Statistics and Research Agency data), seven of the 16 NI practices were located within the lowest quintile of socio-economic deprivation in NI (most deprived), two within the highest (least deprived), and the remainder distributed across the other quintiles. No deprivation measures are available for practices in the RoI.

Characteristics of participants

In NI we invited 760 patients to participate and 46% (350) agreed. In the RoI we invited 1035 and 62.6% (648) agreed. However, the eligibility criteria were not confirmed at initial interview for 35 in NI and 60 in the RoI. We thus recruited 903 patients (315 NI; 588 RoI; 631 male (69.9%); mean age 67.5 years). In the RoI 78% of those recruited were GMS eligible. Among those who were invited but did not participate, relatively fewer were male (62.5%, p<0.001), their age distribution (mean 67.5 years) was similar and 81.7% were GMS eligible.

The gender distributions in the RoI and NI subsamples were similar (table 2), but the RoI group was slightly older (mean 68 vs 66 years), with lower levels of educational attainment. Occupational status was similar in both groups. CHD had been diagnosed for slightly longer for NI than RoI participants (mean 9 vs 8 years). The proportion with a previous MI was similar in both subsamples (RoI 52%; NI 47%), but more participants in the RoI had a history of angina (91% vs 85%). Relatively fewer participants in the RoI were known to have diabetes mellitus (15% vs 22%), although the proportion of participants who had ever had their blood glucose levels recorded was greater than in NI.

Significantly, more in the RoI had coronary artery interventions recorded (percutaneous transluminal coronary angioplasty (PTCA) 38% vs 16%; CABG 33% vs 13%; p<0.001). Proportions who had received these interventions varied widely between practices (PTCA 0% to 71.4%; CABG 0% to 62.5%), but no significant associations between interventions and other variables were identified. In particular, there were no significant differences between rates of procedures for GMS and non-GMS participants in the RoI.

Comparison of cardiovascular risk factor measurements

For significantly more RoI than NI participants (37% vs 28%), measured systolic BP was >140 mm Hg (table 2). Mean systolic BP was also higher in the RoI (139 vs 132 mm Hg). Fewer in the RoI had a BP measurement recorded in the previous year (89% vs 97%), but there was no significant difference between the jurisdictions in the proportions whose most recently recorded systolic BP was >140 mm Hg or diastolic BP >90 mm Hg. In the RoI, relatively more had a baseline cholesterol level above 5 mmol/l (24% vs 17%), fewer had a cholesterol level recorded within the previous year (68% vs 89%) and for relatively more than in NI their most recent record was above 5 mmol/l (30% vs 19%). Differences in mean baseline WHR measurements were statistically significant but of little clinical significance. BMI measurements were similar in both groups.

Prescribing and health service use

Similar proportions in both jurisdictions were prescribed statins (84%; 85%) and ACE inhibitors (41%; 48%), but relatively fewer RoI participants (55% vs 70%) were prescribed β blockers (table 3); β blocker prescribing among those with a history of MI was also lower in the RoI (57% vs 76%). Although statistically different, high proportions of both groups were prescribed aspirin (83% vs 77%). Reported compliance was similar for both groups.

Relatively more visits to the GP were recorded for RoI than NI participants (5.6 vs 4.4 in previous year); the opposite was found for visits to the practice nurse (1.6 vs 2.1) (table 3). Those in the RoI also had a higher mean numbers of outpatient visits, hospital admissions and inpatient days in the year prior to baseline observations. Self-reported health service usage confirmed these findings except for outpatient visits.

Lifestyle, behavioural factors and quality of life

Fewer RoI than NI participants (22% vs 84%) had their smoking habit recorded within the previous year or ever (76% vs 99%) (table 4). There was no significant difference in recorded smoking prevalence or in self-reported smoking habit. Similar proportions had never smoked cigarettes (32%; 27%). No significant differences in dietary fat consumption (DINE questionnaire) were found between the groups, but more RoI participants reported a high fibre intake (35% vs 23%). However, similar proportions reported taking at least five portions of fruit and/or vegetables daily (RoI 11%; NI 14%). Levels of exercise (Godin questionnaire) were higher in the RoI: 62% (NI 44%) reported exercising for at least 15 min more than three times weekly. Quality of life, assessed by the SF-12, was better for measures of both physical and mental health within the RoI group.

Cluster and regression analysis

Intracluster correlation coefficients (ICCs) indicated that the only variable showing a notable cluster effect was the distance people lived from hospital, with coefficients of 0.68, 0.44 and 0.21 for practice, centre and healthcare system, respectively. Following regression analysis which took account of differences in age, gender, educational attainment, occupational status, years since CHD diagnosis, history of MI, diabetes mellitus and distance from hospital, significant differences between RoI and NI participants remained in relation to mean systolic and diastolic blood pressures, mean and raised (>5 mmol/l) cholesterol, cardiac interventions, recording of smoking status, fibre intake, physical activity, physical and mental well-being, prescribing and health service usage.

DISCUSSION

Our findings indicate that people with CHD in a strong, publicly funded primary care system (NI),15 which includes arrangements for financial reward for GPs’ performance, have better management of their biophysical risk factors but report less healthy lifestyles and poorer quality of life, than those in a less strong primary care system within a mixed healthcare economy (the RoI). While our findings indicate that control of blood pressure (in NI)4 and cholesterol (in the RoI and NI)4 5 has improved over recent years, there is still room within both healthcare systems for improvement in risk factor management and in people’s adherence to healthy lifestyle behaviours.

Factors influencing professional practice

The higher levels of recording and better control of blood pressure and cholesterol in NI concur with previous conclusions that increasing structural organisation of primary care is associated with increased recording of risk factors16 18 and better management of chronic disease.24 All the NI practices in our study used computer systems for disease registration, recall of patients and clinical data recording. Computer systems were used to a much lesser extent in the RoI practices, amongst which there was little evidence of recall systems existing prior to our study.

Our findings add support to evidence which points to the value of financial incentives in promoting high quality care.16 17 GPs in the RoI did not have the same financial incentives to record and monitor care for people with CHD as did those working in the NHS in NI. However, variations in quality of care in different countries depend not only on structural inputs but also on healthcare policy25 and individual physician and patient factors.26

NI participants had relatively more visits with their practice nurse and fewer with their GP than those in the RoI. It has been suggested that the 2004 contract has encouraged GPs to delegate routine tasks, such as measuring risk factors, to nurses. This may allow the GP to spend more time with the patient, but it may also diminish the holistic and caring aspects of their role.16 Healthcare policy which seeks to drive high quality care must consider the potential conflict in requiring health professionals to measure their performance and address their patients’ perspectives simultaneously.16

Lifestyle risk factors

Despite evidence of better management of measured risk factors and better organisational systems for primary care in NI, RoI participants reported better adherence to healthy lifestyle advice. They reported eating healthier diets, with more fibre, and taking more exercise than those in NI. Reasons for this are unclear: health professionals in both jurisdictions are known to offer similar advice and public health bodies deliver similar messages, including shared television advertising. Ongoing qualitative research with patients who experienced the intervention delivered within the SPHERE study19 may clarify the reasons for these differences.

Quality of life

The SF-12 is a well validated quality of life measure for people with CHD.27 Our finding that both physical and emotional measures of quality of life were better in the RoI concurs with a recent survey28 which found that older people in the RoI viewed their health more positively than did those in NI. Differences between populations in their interpretation of symptoms and use of health services may result from cultural differences, relating to health beliefs and attitudes, or may reflect ease of access to care.13

Interventions and prescribing

Inequalities in access to coronary revascularisation have previously been reported in relation to socio-economic status, age and geographic location.29 The extent of difference in levels of cardiac interventions between our RoI and NI participants was unexpected. Our work does not allow explanation of this apparent inequality in provision of cardiac related services, which we suggest may deserve further study.

Our findings relating to prescribing compare very favourably with other recent reports,2 30 but they suggest that opportunity to improve this aspect of performance remains. Explanation of differences observed between NI and RoI participants in respect of proportions prescribed aspirin and β blockers is difficult: the absence of incentives for practitioners to monitor treatment may contribute to these.

Strengths and weaknesses of the study

There can be difficulties in comparing healthcare systems in different countries.17 Strengths of our study are that the guidance targets for CHD management and the ways dates, investigations and measurement units are recorded are similar for all participants, allowing direct comparisons. We used minimal inclusion and exclusion criteria in selection and described participants in relation to those who did not participate: we hope thereby that the intervention is being tested in a population to which the research findings may be applied readily in practice. In order to avoid recruitment bias, we stratified the allocation of practices to intervention and control groups within centres and collected baseline data prior to allocation. Quality checking of our data has shown a high level of accuracy.

External validity

We believe that our study has high external validity: we have reported the numbers approached, recruited and analysed.31 The study setting is described in as much detail as possible since descriptive data relating to RoI practices are limited, but we suggest that our participating practices are among those which are best organised and most likely to provide the highest standards of care and monitoring of performance. In NI our participating practices’ mean list size is similar to the national average, as is their performance within the Quality and Outcomes Framework for CHD.

We initially planned to gather data relating to the diagnoses of non-participants, but this was subsequently considered unethical. We may have identified a sample in the RoI which has more severe illness and higher use of health services than that identified in NI because of incomplete registers existing in the RoI prior to our study: while similar percentages in both NI and the RoI had a history of MI and of angina, the RoI sample had a higher prevalence of PTCA and CABG. Having had an intervention may have prompted practitioners’ identification of these patients and thus caused over-representation within their lists of patients with CHD. However, it might be expected that this would also have promoted the recording of biophysical measures monitoring their care. Nevertheless, all participants, both in NI and the RoI, fulfilled the same inclusion criteria and the process of invitation, following randomly ordered lists, avoided selection bias. The patients included in our evaluation are representative of the non-respondents in respect of age and GMS eligibility; approximately two-thirds of both samples were male, which reflects the distribution of CHD within the wider population.

CONCLUSION

It appears that a primary care system with better provision of infrastructure and arrangements for pay-for-performance is associated with higher quality of care for people with CHD in terms of control of blood pressure and cholesterol. However, this is not associated with healthier lifestyles or better quality of life. These findings should contribute to the design of optimal primary healthcare policy and the evolving NHS Quality and Outcomes Framework.16 Healthcare systems are complex, as are the effects of interventions designed to improve people’s health, and it must also be recognised that medical practice exists within a complex social environment.32 Further exploration of possible reasons for differences between healthcare systems in professionals’ and patients’ engagement in strategies to reduce cardiovascular risk is needed. Factors which appear to contribute to better implementation of secondary prevention in one healthcare system may not have the same impact in another: the gathering of baseline data against which to measure change following the introduction of new initiatives in healthcare is important.