A cost-effectiveness analysis of the Prediabetes Intervention Package (PIP) in primary care: a New Zealand pilot programme

Diabetes imposes significant personal and economic costs.1–2 These include the costs for the management of diabetes itself, and the associated complications.3 As diabetes is a long-term condition, costs accrue over time. Identifying people at risk of developing diabetes and preventing, or at least delaying, progression to diabetes is therefore worthwhile from a personal, health system and societal perspective.4–6

Clinical trials have shown that lifestyle interventions can prevent progression from prediabetes to type 2 diabetes.7–8 Generally, these interventions were resource intensive, employing a relatively expensive specialist workforce, such as dietitians and exercise physiologists and considered too expensive to implement in real-world settings.9 More affordable, sustainable approaches are required, such as utilising primary care nurses.6,10

Few published studies have utilised primary care nurses to deliver diabetes prevention lifestyle interventions6 and, to our knowledge, none have published economic evaluations. The New Zealand-based Prediabetes Intervention Package (PIP) in primary care study was a six-month pragmatic non-randomised pilot study [ACTRN12615000806561] designed to address workforce capacity and cost challenges.11 The aim of the intervention was to provide those with prediabetes and a body mass index (BMI) >25kg/m2 and their whānau (family group) with an understanding of the principles of healthy eating to empower them to make healthy dietary choices and facilitate weight loss. The aim of this study was to estimate the cost-effectiveness of the primary care-based PIP lifestyle intervention.

Methods

A cost-effectiveness analysis of the PIP study was conducted from a health funder perspective on an intention-to-treat basis. Costs associated with the programme intervention were considered, with costs and per kilogram (kg) weight change analysed at an individual participant level. Incremental cost-effectiveness ratios (ICERs) comparing the difference in cost between PIP and usual care and the difference in weight between the groups were calculated. Three alternative cost scenarios were modelled.

In brief, the PIP study, which has been previously described,11 was conducted in general practices and community settings in two neighbouring provincial cities in the Hawke’s Bay region. In New Zealand, primary medical care is delivered by general practitioners (GPs) in mostly group practices. Almost all general practices have government capitation funding with varying levels of patient co-payment, and most belong to a primary health organisation (PHO), which is responsible for providing essential primary healthcare services to an enrolled population. Four intervention practices were located in one city and four control practices in the other.

Participants

Eligible participants were non-pregnant adults aged ≤70 years with newly diagnosed prediabetes according to the New Zealand diagnostic criteria (HbA1c 41–49mmol/mol (5.9–6.6%) or fasting plasma glucose 6.1–6.9mmol/L)12 with a BMI above 25kg/m2, not prescribed Metformin and able to communicate in English. Recruitment occurred between August 2014 and April 2015. All participants provided informed written consent. The study was approved by the Northern A Health and Disability Ethics Committee, New Zealand (Ethics reference: 14/NTA/114).

Intervention

The intervention, informed by a literature review of lifestyle interventions and behaviour change theory, sought to provide participants and their family/whānauwith an understanding of healthy eating principles and enhance empowerment around dietary choices. The six components were:

1. Primary care nurse and community nurse intervention training and dietitian support.
2. Individualised patient dietary assessment, goal setting and dietary advice at an initial 30 minute dietary session followed by 15-minute appointments at 2–3 weeks, three months and six months for reassessment, support and advice.
3. Key messages and consistent individualised opportunistic reminders based on participants’ three dietary goals, which were recorded in the patient management system.
4. Nutritionally supportive general practice environment where pamphlets, magazines and posters provided appropriate and consistent dietary messages.
5. Referral to a community-based group nutrition education course consisting of six weekly sessions of 1–1.5 hours each.
6. Written patient resources, the main one being the Diabetes New Zealand booklet, Diabetes and healthy food choices.13

Usual care

Primary care nurses at control practices provided prediabetes dietary advice in their usual way according to the 2013 national Prediabetes Advice interim recommendations.14 This typically consisted of unstructured advice using routinely available dietary pamphlets. Patients were followed up at intervals deemed appropriate depending on the goals and plan agreed with their nurse, usually at 3–6 months.

Physical activity

All participants were given standard physical activity advice, that is, 30 minutes of physical activity of moderate intensity on most, if not all, days of the week.

Data

Data collected as part of routine primary care practice included demographic and medical details, lifestyle information (smoking, alcohol, diet and physical activity), blood pressure, anthropometric measures (height, weight and waist circumference) and laboratory measures (HbA1c, total cholesterol, HDL-cholesterol and triglycerides).

Outcomes

In the pilot study, after adjustment, there were positive differences in the intervention group compared to the control group for most of the clinical and laboratory measures (see Coppell et al 2017 for further detail11) including for weight (kg) and glycated haemoglobin (HbA1c). Participants who completed the six-month intervention lost a mean 1.3kg while those in the control group gained a mean 0.9kg (p<0.001); HbA1c decreased by a mean 1.3mmol/mol in the intervention group and increased by a mean 0.5mmol/mol in the control group (p<0.096). As the only statistically significant outcome in the pilot study was weight, the outcome measure used in this economic evaluation was change in weight in kilograms (kgs) at six months.

Measurement of costs

Costs were calculated in New Zealand dollars ($US1=$NZ1.48, February 2019) for the intervention and control groups at the participant level, using data from the Health Hawke’s Bay Primary Health Organisation for the 2014–15 year. All resources needed to deliver the pilot intervention programme including training of practice and community nurses, dietetic support for practice nurses, written patient resources and pamphlets, magazines and posters for waiting rooms were included. Intervention costs included the total time cost for the practice nurses based on four visits (90 minutes), and patient resources such as the Diabetes New Zealand diet and diabetes booklet.13 Usual care costs included the time cost for two practice nurse visits at baseline and six months. Overhead costs were excluded as the administration and overhead cost associated with two to four visits for a relatively small number of participants spread over eight practices was considered to be minimal. Research-specific costs such as the time taken to obtain informed written consent were excluded.

Cost per participant was based on the number of visits, except where costs were fixed, in which case costs were apportioned regardless of the number of visits. For instance, the community nutrition programme consisted of six sessions of 60–90 minutes’ duration, and was funded under a fixed price contract. As the total cost was based on the number of expected participants, the cost was allocated equally to the 85 intervention participants irrespective of attendance.

Three alternative scenarios were costed where the costs of nurse training and/or community education were reduced, given the cost of delivering the intervention programme in routine practice is likely to be lower than the cost of the pilot programme. For example, the set-up costs associated with training practice nurses to implement the intervention are high initially, but over time these costs will reduce as refresher training replaces full training for practice nurses familiar with the programme. Furthermore, attendance at community education sessions was shown to be relatively low (approximately 50% attendance), and a differently structured, less costly programme could be offered.

Statistical analysis

Cost-effectiveness analysis was undertaken from a health funder perspective, following intention-to-treat principles. Missing data were imputed using multiple imputation with chained equations with the imputation model estimating weight at six months based on age, sex, Māori ethnicity, baseline weight and family history of diabetes. Following each imputation, a bootstrap sample was taken with 10,000 samples used to obtain means, medians, and 95% confidence intervals (using percentiles), and to construct incremental cost-effectiveness planes and cost-effectiveness acceptability curves. This analysis was repeated for the three alternative scenarios where costs were reduced. Discounting was not used as the duration of the trial was less than one year. Analyses were conducted using Stata version 15.1 and R 3.5.1 using the BCEA package (version 2.2.6).15

Results

The demographic characteristics and diabetes-related co-morbidities for the 157 participants enrolled at baseline are shown in Table 1. The two groups were similar. Almost one-third self-identified as Māori and 40% had a family history of diabetes.

Table 1: Demographic characteristics and diabetes-related co-morbidities of participants at baseline prior to receiving the intervention.

‍

Data are % unless otherwise indicated. †For continuous measures, two-sample t-test where residuals were normally distributed, assuming or not assuming equal variance based on Levene’s test, and Mann-Whitney-Wilcoxon otherwise; for categorical variables, Chi-squared test where at least 80% of cells have expected counts of 5 or above, Fisher’s Exact test otherwise.

These data are a subset of those presented elsewhere (11) in “Table 2 Demographic characteristics and diabetes-related co-morbidities of participants at baseline and six months”.

Table 2: Costs per participant in 2015 NZ$ for the intervention and control groups.

‍

*Based on participants attending all assessments. The mean cost is lower in Table 3 as costs per participant were adjusted based on the number of assessments received.
+Based on each practice nurse managing six patients in a six-month period.

Not all participants completed all assessments. Of the 72 participants attending control practices, 66 participants had a six-month assessment, and of the 85 participants attending intervention practices, 14 participants had a baseline assessment only, four participants had a baseline and three-month assessment, and 67 participants had all three assessments.

The cost per participant for the intervention and control groups are detailed in Table 2. The per participant cost for the intervention was NZ$349.28. The practice nurse training and dietitian support cost of NZ$106.14 included nurse time costs, cost for the community educators and dietitian, administration, resource cost and room hire, and assumed each nurse managed six patients in a six-month period. This cost is expected to reduce to NZ$58.44 in subsequent intakes as refresher training is substituted for full training and dietetic support. The cost per participant for the control group was NZ$24.75.

The summary results of the economic evaluation (and the three cost scenarios) are presented in Table 3.

Table 3: Cost-effectiveness results (2015 $NZ).

‍*Calculated following multiple imputation and bootstrapping.

The mean cost per participant was NZ$24.07 for usual care and NZ$344.43 for the intervention, giving a mean difference of NZ$320.36 between groups. The median weight change after imputation for the usual care group was a 0.79kg (95% CI -0.02, 1.63) weight gain and for the intervention group a 1.08kg (0.01, 2.06) weight loss (median difference of 1.87 kg, 95% CI 0.54, 3.15 kg), attenuating the difference from completers only. The ICER for the pilot study was NZ$170.90 (95% CI 100.37, 553.93) per 1kg of weight loss.

When the community education costs are halved (all else being the same), the mean cost per participant is NZ$250.31 (Scenario A). When refresher training replaces full training (all else being the same), the mean cost per participant is NZ$296.73 (Scenario B), and when both refresher training replaces full training and the community education costs are halved, the mean cost per participant is NZ$202.61 (Scenario C).

The point estimate and bootstrapped estimates of incremental cost and incremental weight change for each scenario are shown in the cost-effectiveness planes in Figure 1A. The ICER for the lowest-cost scenario (Scenario C) is NZ$75.57 lower than the full-cost scenario (Scenario A) at NZ$95.33 ($56.12–$308.36). The cost-effectiveness acceptability curves for each scenario are shown in Figure 1B. If the willingness to pay for a 1kg reduction in weight is NZ$250 for example, the PIP intervention has a 0.81 probability of being cost-effective which increases to 0.96 for the lowest-cost scenario.

Figure 1: Cost-effectiveness results for the PIP programme.

(a): Incremental cost-effectiveness plane. Larger solid point = estimates used to calculate the ICER; grey points = bootstrapped replicates of incremental costs and incremental weight loss. (b): Cost-effectiveness acceptability curves of reducing weight (kgs) by one unit. Solid black line = probability that the PIP is cost-effective; dashed and dotted lines = probability of Scenarios A, B and C being cost-effective in practice. Costs are reported in 2015NZ$.  

Discussion

Diabetes is a global epidemic that accrues significant costs to the health system.1–2 Clinical trials that have shown lifestyle interventions can prevent progression from prediabetes to type 2 diabetes7–8 are cost-effective in the short- and long-term,4 and in many cases probably cost-saving in the longer term.16–18 How these clinical trial lifestyle interventions are translated into real-world settings influences the cost and effectiveness of such programmes.16

The PIP lifestyle intervention pilot programme was implemented in busy general practices in New Zealand. The cost-effectiveness evaluation, from a health funder perspective, showed a mean cost of NZ$170.90 (95% CI 100.37, 553.93) per 1kg of weight loss with a lower-cost scenario estimated at NZ$95.33 (95% CI 56.12, 308.36). The probability of being cost-effective at a willingness to pay of NZ$250 per 1kg of weight loss was 0.81 and 0.96, respectively. The probability of the intervention being cost-effective increases significantly if costs are reduced. For example, at an acceptability threshold of NZ$150 for a 1kg reduction in weight, the PIP intervention has a 0.34 probability of being cost-effective, whereas the lowest-cost scenario has a probability of 0.84. As a weight loss of 1kg reduces the risk of diabetes among those with prediabetes by 16%,19 any weight loss is likely to be value for money in terms of preventing future costs associated with the treatment of diabetes and its complications.

The mean cost per intervention participant was NZ$344.43, of which 55% was for the community group nutrition programme. Given almost half of the intervention participants did not attend the community group education, the programme could be modified to be less costly, thereby reducing per patient cost. The health professional group who delivers the lifestyle programme also influences programme cost-effectiveness with physicians and specialists such as dietitians typically costing more.4,20–21 Primary care nurses have infrequently been employed to deliver diabetes prevention programmes, yet they have the potential to deliver a cost-effective diabetes prevention programme.22 In the PIP study, specifically trained practice nurses competently delivered the programme and did not require ongoing support from a dietitian after their first 6–12 months of delivering the intervention.11 Further, ongoing support from healthcare professionals has been shown to be effective in helping patients maintain weight loss.12,24 As primary care nurses typically have an ongoing trusting relationship with their patients,6,25 and many of those with prediabetes have co-morbidities requiring treatment,11 primary care nurses are in an ideal position to provide ongoing guidance and support for weight loss and maintenance.

There are no comparable published economic evaluations of prediabetes lifestyle intervention programmes in New Zealand. Internationally, although not directly comparable to the PIP study due to its size and employment of private weight loss providers, the programme that is most similar to the New Zealand-based PIP programme is the NHS Diabetes Prevention Programme in England, a structured 9–12 month prediabetes lifestyle intervention programme.5 An impact analysis of this programme to determine cost implications found the per patient medium-end average cost was £270 per participant enrolled (high end £350; low end £155),5 which was deemed cost-effective. This equates to an approximate medium-end cost of NZ$518 per participant, higher than the per participant cost of NZ$344.43 in the PIP study. As mentioned, although the two studies are not directly comparable, it suggests the PIP programme could potentially be viewed as cost-effective by NHS-standards.

Other alternatives for encouraging weight loss are commercial programmes such as Weight Watchers and Jenny Craig. A within-trial cost-effectiveness analysis of a randomised controlled trial comparing Weight Watchers to standard care in populations in Australia, the UK and Germany concluded that relative to standard care, Weight Watchers was cost-effective over one year from a health sector perspective.26 In Australia the cost per 1kg of weight loss was US$122 (NZ$181), which is higher than the PIP programme per participant cost of NZ$170.90. Similarly, when Weight Watchers, Jenny Craig and three weight management pharmaceutical products were compared in a US-based study,27 Weight Watchers was the most cost-effective with an ICER of US$155 (NZ$230) per 1kg weight loss (based solely on subscription costs). The ICER for Jenny Craig’s programme (including subscription costs and incremental food costs) was US$338 (NZ$501) per 1kg of weight loss. Although Weight Watchers was considered cost-effective, there are several important differences between commercial programmes and the PIP programme. First, people who choose commercial weight loss programmes are willing to join and able to pay. It is likely therefore that they are different to those who participated in the PIP programme, many of whom had very low food budgets. Second, Weight Watchers predominantly uses group-based sessions delivered at set times in a public setting by trained peers, and clinical advice is not given or available.

Economic studies that have included future healthcare costs show people who participated in an intervention lifestyle programme used significantly less healthcare resources than people who received standard care.28 Indeed, a lifestyle intervention can be cost-saving. For example, the Diabetes Prevention Program (DPP) 10-year follow-up study found direct medical costs (including emergency department visits, outpatient and hospital admissions) were lowest in the lifestyle intervention group compared with the control and Metformin groups, and were in fact cost-saving.29 Lifestyle changes reduce the impact, or risk of developing, a number of co-morbidities such as cardiovascular disease, and some cancers.30 As co-morbidities were common among PIP participants,11 eg, hypertension (50%) and dyslipidaemia (40%), it is likely the ICER is an over-estimate as it does not include the potential future lower health system costs, on average, for the intervention group. Also not quantified in this study, but captured in the qualitative study25 are the positive consequences that could impact on future healthcare costs; for example, one intervention participant improved their asthma control, and for others, their positive lifestyle changes extended to other family members.31 Valuing and including these wider health benefits would further enhance the cost-effectiveness of the programme.

This economic evaluation used actual costs, rather than estimates, and results were based on the realities of primary care, where patient attendance is not 100%. For instance, in this study about half of the participants attended the group community education sessions and 15% of participants did not attend their final six-month appointment. Evaluating a lifestyle programme conducted in a real-world setting highlights what works well and what could be changed, enabling different cost scenarios to be modelled for future roll-out.

Studies have shown that lifestyle programme intervention effects persist after the trial period.29 Therefore, the six-month time horizon for this study may have underestimated the benefits of the PIP programme. Similarly, this evaluation did not include societal costs, indirect medical costs or future health use costs or the flow-on benefits to participants’ families,31 which may also underestimate the cost-effectiveness. As this was a pilot study, the sample size was not sufficient for sub group (eg, age, ethnicity) analyses to be undertaken. Although multiple imputation can mitigate the effect of missing data, we do not know if there was a difference in the likelihood of excess weight gain between intervention and control group participants who dropped out of the study. As data were collected during primary care consultations in a busy environment, there was insufficient time to collect additional data to calculate QALY gains, limiting comparisons with other studies using QALYs. The lack of a published willingness to pay threshold in New Zealand means it is difficult to determine whether resulting ICERs would be considered value for money by the New Zealand health funder.

In conclusion, this study indicates that the six-month Prediabetes Intervention Package in primary care programme as implemented in Hawke’s Bay is likely to be a cost-effective weight loss strategy for preventing or delaying progression to type 2 diabetes in people with prediabetes, with additional health gains beyond diabetes prevention.

‍