Coronary heart disease is the most common cause of death in the United Kingdom, responsible for around 114 000 deaths in 2003.1 Coronary heart disease is also associated with substantial costs to both individuals and the healthcare system: treatment of cardiovascular disease has been estimated to account for 10–15% of total healthcare costs in developed countries.2 A recent study estimates the cost of coronary heart disease in the UK National Health Service (NHS) at £3859 million (€4905 million) in 2004, in addition to substantial non-health-related costs.3

Secondary prevention in patients with coronary heart disease aims to prevent further events after coronary heart disease has manifested itself and several measures have been shown to be effective in clinical trials. These include lifestyle measures such as smoking cessation, dietary changes and increasing physical activity; and medical measures such as blood pressure and lipid control and the use of aspirin.4 Secondary prevention has been shown to be effective in improving quality of life, functional status and in reducing hospital admissions in patients with coronary heart disease.5 In the United Kingdom a variety of secondary prevention methods are recommended by the National Institute for Health and Clinical Effectiveness (NICE).6 There is also evidence to suggest that many forms of secondary prevention may be cost-effective; examples include angiotensin-converting enzyme (ACE) inhibitors,7^–10 β-blockers after myocardial infarction,11 12 statins for the secondary prevention of coronary heart disease13 14 and the use of aspirin.15

However, despite the demonstrated effectiveness and cost-effectiveness of many secondary prevention strategies, there is potential to increase the use of secondary prevention for coronary heart disease in primary care.16 17 One effective method of improving outcomes for patients in primary care is prompted care in nurse-run primary care clinics.18 19 A recent study has shown that such a disease management programme using community cardiology nurse-led clinics for patients with coronary heart disease and heart failure can be effective in improving both clinical outcomes and quality of life.20 This paper builds upon this work by presenting the results of an economic analysis conducted alongside the clinical trial with the aim of determining the effect of a nurse-led disease management programme on health service costs. Additionally, by combining cost data with data on quality of life, we aimed to assess the cost-effectiveness of these nurse-led clinics to determine if they represent an efficient use of health service resources.

METHODS

Clinical study

The effectiveness results used in the economic analysis were obtained from a cluster randomised trial of a nurse-led disease management programme in primary care. Full details are given elsewhere.20 In brief, 20 practices were randomised to receive either conventional care or a nurse-led disease management programme. There were 505 patients in the intervention group and 658 patients in the control group. Patients in practices in the control group received usual general practitioner (GP) and practice nurse care, but these practices were additionally offered the facility to refer patients for open access echocardiography. In the intervention group, one of two peripatetic nurse specialists trained in the management of heart failure held weekly clinics at the surgery for coronary heart disease and heart failure patients. Patients were included if a diagnosis of coronary heart disease (angina or past medical history of myocardial infarction) or heart failure was specifically recorded or suggested by medication prescribed. The nurse intervention included patient assessment, confirmation of diagnosis by investigations, medication management and titration, home visits for housebound patients with heart failure and liaising with primary and secondary care. Patients with a presumed diagnosis of heart failure were assessed clinically and had an electrocardiogram. Patients were referred for an echocardiogram on the basis of signs, symptoms and an abnormal electrocardiogram.21

Resource use data

Both baseline and follow-up data were collected from general practice records at the end of the study period for all patients recruited including any who failed to return follow-up questionnaires. The period of data collection was for the year before randomisation as well as the follow-up year. A structured data-collection form was prepared to collect detailed information regarding the use of health service resources related to coronary heart disease. We collected data on all coronary heart disease-related medicines prescribed in both periods including frequency, dosage and duration of prescribing. In addition, we collected information on all visits to the GP and practice nurse and all visits to the nurse-led clinics. Visits to the hospital outpatient department were recorded, including whether it was a new appointment or a follow-up visit, the hospital to which the referral was made and details of the consultation or any investigations. Also collected were cardiac-related hospital admissions including the length of stay, name of hospital, speciality and type of admission (emergency, elective or day case). We recorded the use of relevant blood tests (urinalysis and electrolytes, thyroid function test, liver function test and full blood count). To supplement information obtained from practice notes the clinic nurses also recorded all home visits made to participants in the intervention group who were unable to attend clinic sessions. At the end of the follow-up period we distributed a questionnaire to participants; this solicited information on travel costs incurred during a typical visit to the GP and included questions relating to mode and duration of travel, distance for car travel and any fares or parking fees incurred.

Unit cost data

We costed the resource use information obtained in a number of ways. The costs of named drugs were obtained using the British National Formulary.22 We calculated a cost per day using a combination of this source and the resource data obtained from practice notes. This was multiplied by the duration of drug use to estimate a total cost for the relevant time period. For GP and practice nurse contacts we used costs from a published source23—these were £19 and £9 per contact, respectively. For the clinic nurses, salary plus on-cost information was obtained from the study. To this we added annual training costs, overheads and capital overheads relating to an appropriate grade of specialist nurse practising in primary care.23 We also included equipment and training costs that related specifically to providing the clinics; these costs were obtained from the study records. For equipment and training that would be expected to have a lifespan of greater than one year we estimated an annual equivalent cost using a discount rate of 6%.24 Information was obtained from the study nurses on workload and working patterns to elicit the number of clinic and home contacts made per year and the proportion of time spent on these activities and also on non-patient contact activities. This information was combined with the costing information to estimate a cost per contact for clinic and home visits, these were £21.16 (€27) and £26.45, respectively. For outpatient visits we obtained a description of the referral from the patient notes and costed these referrals by using NHS reference costs.25 These report healthcare resource groups (HRGs), which are groups of clinically similar procedures that require similar levels of resource use. Costs were obtained for the year 2002–3 (year 2002–3 HRGs were used as appropriate HRG codes were not available for the year 2003–4) and these were inflated to 2003–4 prices using a healthcare specific inflator.23 A local clinical expert checked the descriptions collected against the HRG code to ensure that appropriate costs were used. For inpatient stays we obtained data on the trust, speciality, type of admission (planned, emergency or day case) and the length of stay. HRG costs were obtained from the local NHS trust and these were used for day case and for inpatient stays. In addition to the HRG costs the trust has a cost per excess bed day for longer than average stays. To allow for this we added a cost per excess bed day to each stay that was longer than the local NHS trust’s average length of stay (excluding all excess bed days). Costs for blood tests were obtained from the local NHS trust. For the cost of travel borne by patients we used information obtained from the patient questionnaires for the cost of fares and parking fees. For car travel we obtained an estimate of distance travelled and multiplied this by an estimated cost per mile of car travel.26 For time costs associated with travel we costed the respondents’ estimates of the time taken from Department of Transport published values.27 These values were published for the year 1998–9 and were inflated using the retail price index. All costs are in 2003–4 UK pounds sterling.

Quality of life

Quality of life was measured using the EuroQol.28 The instrument was sent to all subjects in both the control and the intervention groups at baseline and at the one year follow-up period. The EuroQol values were used to obtain quality of life scores and hence quality adjusted life years (QALY). This was carried out by assuming a linear relation between the values obtained at baseline and one year follow-up. For all participants who had died during the follow-up year we assigned a value of zero for the proportion of the follow-up year for which the participant was not alive, as the EuroQol assigns this value to death by definition.

Analysis

We carried out statistical analyses according to CONSORT guidelines for cluster-randomised trials. The intra-cluster correlation coefficients (ICC) along with the 95% confidence intervals (CI) were calculated for the EuroQol and QALY scores. Mixed effects regression models were used to analyse the effect of the intervention on EuroQol and QALY scores, with clustering incorporated as a random effect. The generalised least squares estimation technique was adopted with adjustments for the unbalanced cluster sizes. We also used the biased corrected and accelerated bootstrapped estimation technique to obtain the bootstrapped p values. For all models the bootstrap was replicated 2000 times.

Owing to the problem of highly positively skewed cost data, we used the generalised linear mixed effects modelling approach,29 30 rather than linear regression on transformed data, to estimate the mean costs for the randomisation groups. The generalised modelling approach is flexible with a wide class of distributions and link functions and leads to straightforward inference about the mean cost. We compared the performance of model fits with gamma and inverse Gaussian distributions with identity link function in generalised estimating equation (GEE) platform. The goodness of fit and model comparisons was based on the minimum sum of squares of residuals approach. The estimated costs are based on model-based linear prediction.

The quality of life scores and the costs were analysed with adjustments for the following patient characteristics: diabetes, angina or a history of myocardial infarction or confirmed heart failure, age and gender. For cost variables we also analysed with adjustments for cost in the previous year. All analyses were carried out using Stata (Release 8). The economic evaluation used a cost-utility approach as we estimated the cost per additional QALY generated in the intervention group compared to the control group. This form of analysis was appropriate as the disease management programme could influence the amount of secondary prevention provided and hence may have effects on both the expected length of life and also its quality. All costs and benefits are calculated in the one-year follow-up period of the study. We did not discount costs and benefits as these all occurred within a one year time frame. We adopted a patient perspective for outcomes. Costs were measured from the perspective of both the NHS and patients themselves (travel costs). We did not measure costs associated with time off work (productivity costs) as the majority of subjects were elderly (mean age 70) and were unlikely to be working. To illustrate the uncertainty relating to our cost-effectiveness results we calculated a cost-effectiveness acceptability curve, which estimates the probability that the results are cost-effective at different values of a QALY.31

RESULTS

The patient characteristics of the two groups are given in table 1. The groups are very similar in age but the intervention group had a higher proportion of male participants. The proportions of subjects who had angina or myocardial infarction were also similar between groups, as was the proportion of respondents with heart failure at baseline. However, the intervention group had a higher increase in the number of patients with confirmed heart failure between baseline and follow-up, an increase of approximately 6% versus 2%. The control group had more patients with diabetes, though this difference was more marked at baseline than at follow-up. The costs of coronary heart disease relevant medicines are given in table 2. The mean values for total costs were higher in the intervention group (p = 0.05). However, the cost of individual medicines was only statistically significantly higher in the intervention group for the statins. The usage of NHS services and also the usage of blood test in the control and intervention groups are given in table 3. This shows the mean number of contacts with the disease management programme, both in the clinic and for home visits. It can also be seen that the disease management programme group had higher numbers of contacts with other NHS services. Table 4 shows that the disease management programme was associated with higher costs for all resource categories.

The results of the quality of life estimates using the EuroQol are presented in table 5. As evident from the ICC estimates and the corresponding confidence intervals, the cluster effect is negligible as far as these quality of life scores are concerned. Baseline values are similar; however, at follow-up quality of life had increased slightly in the intervention group and decreased slightly in the control group. We calculated QALYs from these values and these are also given in table 5. The difference in QALY value between the two groups was 0.03 but this was not statistically significant. The cost-effectiveness of the clinic service in the follow-up year is given in table 6, which shows the estimated extra costs incurred in the clinic arm compared to the control group and also the estimated extra benefits gained by this group. From this we were able to estimate the cost per QALY gained in this group at £13 158.

DISCUSSION

Our research shows that the nurse-led clinics generate gains in QALYs at a comparatively modest cost, within the range that is usually considered cost-effective by NICE. For example, Raftery et al in an examination of NICE decisions found that interventions with a cost per QALY of less than £30 000 were recommended.32 The uncertainty around these estimates of cost per QALY was explored by means of a cost-effectiveness acceptability curve (fig 1). From this it can be seen that the probability of the disease management programme being cost-effective increases with the value placed upon a QALY. If a QALY was valued at around £30 000 then there would be approximately a 90% probability that the disease management programme would be cost-effective.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Cost-effectiveness acceptability curve for the disease management programme. QALY, quality adjusted life year.

The finding that the disease management programme increased total NHS costs differs slightly from previous literature. Raftery et al33 presented an analysis of the costs of nurse-led community cardiology clinics in a follow-up of the study reported by Campbell et al19 and found that the only significant differences in costs were increased costs of prescribing and the intervention itself, they found non-significant decreases in inpatient stays. However, this study had some important differences compared to the present study. Their study only included patients with CHD while our study included patients with CHD and/or chronic heart failure. They randomised at an individual and not at a practice level, and the nurse intervention included two rather than four visits in the first year. In addition, the provision of the service was slightly different between the two with the present study using CHD nurse specialists and Campbell et al using a mixture of health visitors, district nurses and practice nurses. The CHD nurse specialists in the present study also had the ability to refer individuals to echocardiograms. In contrast to Raftery et al, we found that the total NHS costs had increased in the intervention group. This was largely driven by inpatient stays, which accounted for nearly 60% of the total increase in costs between the two groups. However, the cost of providing the clinics and the extra costs of outpatient appointments were also important, accounting for approximately 20% and 10% of the additional cost, respectively. It is interesting to note that the cost of the clinics was only a small component of the total extra cost of introducing the nurse-led clinics. This shows the importance of comprehensive costings when evaluating any new service. This increase in all coronary heart disease-related costs implies that, as well as promoting increased use of secondary prevention, the disease management programme also facilitated the referral of individuals to outpatient clinics, and possibly through this mechanism, to inpatient stays.

The results found here are supported by the results of the clinical evaluation of this service,20 that found improvements for the nurse-led group in quality of life (as measured by the SF-36), with statistically significant improvements in the physical functioning, vitality and social functioning dimensions. Improvements in quality of life were also seen in some dimensions of the Seattle Angina Questionnaire. A significantly higher proportion of patients in the nurse-led clinic group had adequate control of cholesterol, which may be related to our finding of higher costs of prescribing statins. It was also found that there were significantly higher numbers of individuals with previously undiagnosed heart failure (36.5% of all heart failure patients in the intervention group compared to 10% in the control group) who had their diagnosis confirmed by echocardiography.

The potential weaknesses of this present study are similar to those reported for the main clinical study.20 These were related to the fact that all practices were drawn from one locality and that we only had 43% participation. In addition, values for the EuroQol were taken from postal surveying of participants and we therefore did not achieve quality of life estimates from all participants. Higher response rates would have improved the estimation and precision of quality of life estimates. Also, the duration of follow-up was limited to one year. Therefore, it is unclear whether the additional costs that we have detected for the nurse-led clinics would persist over time. Possibly, the nurse-led clinics expedite referrals that would happen anyway; this would mean that true costs may be similar over a longer follow-up period. Conversely, the clinics may be causing the referral of individuals who would not otherwise be referred. If the extra care that these individuals received was clinically effective then this process would be expected to lead to persistently higher costs but also persistent increases in quality of life and decreases in mortality. In addition, it is unclear whether the benefits achieved would also persist over time. However, in the previously mentioned economic evaluation by Raftery et al, clinical benefits achieved at one year were sustained at four years of follow-up.33 These factors imply that both a long-run economic model looking at the cost-effectiveness of nurse-led clinics and also studies that involve longer term follow-up of patients managed by nurse-led clinics would be useful pieces of future research.

There are a number of strengths of the present study. We obtained estimates of resource use from patient GP notes and hence can have more confidence in the veracity of these data than if obtained from patient recall via surveys. Appropriate statistical techniques allowed us to take account of both clustering and differences in patient characteristics between the samples. Practices were randomised so we were able to estimate the additional costs and benefits of the intervention compared to a valid control group who received standard care. This allowed us to estimate the additional costs of the nurse-led clinics in terms of changes in coronary heart disease care received as well as just in terms of the cost of providing the clinics.

The cost-effectiveness study presented here was carried out alongside a clinical trial that showed that a disease management programmes can improve the care of patients with coronary heart disease and presumed heart failure in primary care.20 The cost-effectiveness study reported here showed that the disease management programme also appeared to increase the use of health services. However, these extra costs are compensated for by an apparent (though not statistically significant) increase in estimated QALYs. The cost-effectiveness acceptability curve indicated a high probability (approximately 90% at a value of a QALY of £30 000) that the service would be cost-effective. Although there are issues with the short-term nature of the follow-up in this study, we believe the present research provides evidence to support the cost-effectiveness of disease management programme run by specialist nurses.