Cost Effectiveness of Support for People Starting a New Medication for a Long-Term Condition Through Community Pharmacies: An Economic Evaluation of the New Medicine Service (NMS) Compared with Normal Practice

  • Rachel A. Elliott
  • Lukasz Tanajewski
  • Georgios Gkountouras
  • Anthony J. Avery
  • Nick Barber
  • Rajnikant Mehta
  • Matthew J. Boyd
  • Asam Latif
  • Antony Chuter
  • Justin Waring
Open Access
Original Research Article

Abstract

Background

The English community pharmacy New Medicine Service (NMS) significantly increases patient adherence to medicines, compared with normal practice. We examined the cost effectiveness of NMS compared with normal practice by combining adherence improvement and intervention costs with the effect of increased adherence on patient outcomes and healthcare costs.

Methods

We developed Markov models for diseases targeted by the NMS (hypertension, type 2 diabetes mellitus, chronic obstructive pulmonary disease, asthma and antiplatelet regimens) to assess the impact of patients’ non-adherence. Clinical event probability, treatment pathway, resource use and costs were extracted from literature and costing tariffs. Incremental costs and outcomes associated with each disease were incorporated additively into a composite probabilistic model and combined with adherence rates and intervention costs from the trial. Costs per extra quality-adjusted life-year (QALY) were calculated from the perspective of NHS England, using a lifetime horizon.

Results

NMS generated a mean of 0.05 (95% CI 0.00–0.13) more QALYs per patient, at a mean reduced cost of −£144 (95% CI −769 to 73). The NMS dominates normal practice with a probability of 0.78 [incremental cost-effectiveness ratio (ICER) −£3166 per QALY]. NMS has a 96.7% probability of cost effectiveness compared with normal practice at a willingness to pay of £20,000 per QALY. Sensitivity analysis demonstrated that targeting each disease with NMS has a probability over 0.90 of cost effectiveness compared with normal practice at a willingness to pay of £20,000 per QALY.

Conclusions

Our study suggests that the NMS increased patient medicine adherence compared with normal practice, which translated into increased health gain at reduced overall cost.

Trial Registration

ClinicalTrials.gov Trial reference number NCT01635361 (http://clinicaltrials.gov/ct2/show/NCT01635361). Current Controlled trials: Trial reference number ISRCTN 23560818 (http://www.controlled-trials.com/ISRCTN23560818/; DOI 10.1186/ISRCTN23560818). UK Clinical Research Network (UKCRN) study 12494 (http://public.ukcrn.org.uk/Search/StudyDetail.aspx?StudyID=12494).

Funding

Department of Health Policy Research Programme.

Key Points for Decision Makers

The New Medicine Service (NMS) appears effective and cost effective compared with normal practice.

Increased patient adherence to their new medicine translated into increased health gain at reduced overall cost that is well below most accepted thresholds for technology implementation.

This is a simple intervention which has been popular with community pharmacists and patients, and is transferable into most therapeutic areas.

Consideration should be given to extending and evaluating the NMS in other potentially beneficial areas, and these results are likely to be transferable into health systems less integrated than the UK NHS.

1 Introduction

Adherence to medication is defined as the extent to which individuals take their medication as prescribed [1]. Non-adherence is commonly reported in key prevalent diseases such as chronic obstructive pulmonary disease (COPD): 33% [2]; schizophrenia: 52% [3]; asthma: 67% [4]; and diabetes mellitus: 78% [5]. Non-adherence causes reduced quality of life, increased hospitalisations and premature deaths [5, 6, 7]. A recent estimate sets the global economic impact at US $285 billion, 57% of the economic impact of suboptimal medicines use [8]. Estimated opportunity cost to the English National Health Service (NHS England) of health gains foregone because of non-adherence is over £930 million per annum in just five diseases [9]: asthma, type 2 diabetes, high cholesterol/coronary heart disease, hypertension and schizophrenia. Improving adherence from current levels to 80% across these five areas would save the NHS £500 million per annum [9].

One effective way to improve adherence focuses on patients who receive a new medicine for a long-term condition, as they often experience problems that can quickly lead to a proportion becoming non-adherent over time [10, 11]. Barber et al. developed an intervention with a theoretical basis in the self-regulatory model (SRM) [10, 12], designed to elicit patients’ experiences with, and concerns about, their new medicine. This intervention significantly reduced non-adherence and was cost effective [13, 14]. Its approach was adopted as government policy [15] and in adapted form was launched as the national New Medicine Service (NMS) in 2011 [15] for people starting a new medicine for asthma/COPD, type 2 diabetes, hypertension or antiplatelet/anticoagulant treatment [16]. The design differs from the original intervention as there are four specified patient groups associated with high rates of avoidable hospital admissions. The original intervention was delivered via a centralised telephone service, whereas NMS is delivered by community pharmacists providing the medicine, either face-to-face or over the telephone. Accredited pharmacies provide NMS, are remunerated for each episode of care and have guidance on how to conduct the intervention [16]. Of 11,495 community pharmacies in England, 10,553 (91.2%) had delivered the NMS to at least one patient between November 2011 and January 2014 [17].

The aim of this study was to evaluate the cost effectiveness of the NMS compared with normal practice in changing medicine-taking behaviour, following published reporting criteria [18]. The incremental cost per extra quality-adjusted-life-year (QALY) generated was determined from the perspective of the funder (NHS England).

2 Methods

A randomised controlled trial (RCT) has assessed NMS effectiveness [19]. At 10 weeks, NMS significantly increased the proportion of patient-reported adherence by 10.2–70.7%, compared with normal practice of 60.5% [20, 21]. Trial design precluded observation of long-term outcomes and costs from changes in adherence. Many benefits of improved adherence are delivered well into the future. Here, we simulated the effect of observed adherence increases on patient outcomes and NHS costs by designing economic models for each drug–disease pair. We developed this method previously in a cross-therapeutic intervention focused on medication errors [22]. Here, we combine the results from the NMS trial with projected harm from non-adherence to generate estimates of patient outcomes and NHS costs (Fig. 1).
Fig. 1

Overview of economic model developed to combine New Medicine Service (NMS) trial results with estimates of harm caused by non-adherence

2.1 Intervention and Comparators, Patient Characteristics and Outcomes

2.1.1 New Medicine Service (NMS) Intervention

NMS begins with the patient’s initial presentation with a prescription for a medicine that is new to them in a community pharmacy. Patients can be referred by their prescriber, self-refer, or the pharmacist can invite the patient to use the service. The intervention consists of a one-to-one consultation 7–14 days later, with a follow-up 14–21 days after that, the whole episode lasting 5 weeks. These are the points where the pharmacist asks about adherence. Outcomes were collected by researchers at 10 weeks.

The primary aim of the intervention, which can be face-to-face or telephone-based (in this study, all follow-up was via telephone) is the patient-centred identification of any problems with the treatment and provision of appropriate support or action [23]. Action may include referring the patient back to their prescriber (Fig. 2).
Fig. 2

New Medicine Service intervention

2.1.2 Normal Practice

Normal practice was the pharmacist’s usual advice. There was no planned follow-up.

2.1.3 Study Outcomes

The RCT primary outcome was self-reported adherence at 10 weeks, considered the minimum time required to demonstrate behavioural change [13].

Patients were contacted by telephone by a researcher and asked about adherence behaviour using the question: “People often miss taking doses of their medicines, for a wide range of reasons. Have you missed any doses of your new medicine, or changed when you take it? (Prompt: when did you last miss a dose?)” [24]. This is the adherence question asked by pharmacists during the NMS intervention and follow-up. The patient was defined as non-adherent if any doses were missed without the advice of a medical professional in the previous 7 days.

Little validation has been carried out for most self-report adherence measures [25]. An existing scale, the Morisky Eight Item Medication Adherence Scale (MMAS-8), validated in hypertension, was used to support our primary outcome measure, and collected via self-completion postal questionnaire [26]. These results are available in [20].

Sample size was determined according to the primary outcome at 10 weeks (see electronic supplementary material and [19, 20, 21]). Sample characteristics and study outcomes are reported in Tables 1 and 2. The characteristics (age, sex, ethnicity, disease area) of the RCT cohort were very similar to the population accessed from the national PharmOutcomes records of 451,222 NMS consultations recorded from October 2011 to 2 December 2013 [20]. With regards to economic deprivation, the median Index of Multiple Deprivation (IMD) rank for England is 16,241, indicating that our study population is slightly more deprived than average.
Table 1

Patient and pharmacy characteristics by treatment arm

Patient characteristics

Normal practice

New Medicine Service

Total

253

251

Antiplatelet/anticoagulant (n = 43, 8.5%)

19 (7.5)

24 (9.6)

Asthma/COPD (n = 117, 23.2%)

58 (22.9)

59 (23.5)

Hypertension (n = 249, 49.4%)

128 (50.6)

121 (48.2)

Type 2 diabetes (n = 95, 18.8%)

48 (19.0)

47 (18.7)

Female (n = 260, 51.6%)

135 (53.4)

125 (49.8)

Age of total cohort (year) n; mean (SD)

253; 59.3 (15.0)

251; 59.5 (15.3)

Total withdrawals by week 10; n (%)

37 (14.6)

16 (6.4)

Economic deprivation based on IMD ranka; mean (SD)

 Pharmacy study sites

10,241.5 (8117.2)

9880.0 (7723.0)

 Study patients

13,708.3 (8546.4)

14,325.5 (8906.8)

COPD chronic obstructive pulmonary disease, n number, SD standard deviation

aIMD: Index of Multiple Deprivation (rank)—each area of England is ranked based on its economic deprivation score. The most deprived area of England is ranked 1. The rank only indicates relative position and does not provide a quantifiable comparison. An area ranked 50 is not twice as deprived as the area ranked 100 (English deprivation rank scores range from 1 to 32482)

Table 2

Adherence results for NMS and normal practice measured using NMS question and MMAS-8 at 10 weeks

ITTa at 10 weeks

Unadjusted probability (95% CI)b

Unadjusted odds ratiob (95% CI)

Adjusted probability (95% CI)c

Adjusted odds ratioc (95% CI)

Adherence NMS (N = 378)

 Normal practice

0.61 (0.54–0.67)

1.58 (1.03–2.42); p = 0.037

0.63 (0.55–0.70)

1.67 (1.06–2.62); p = 0.027

1.62 (1.04–2.53)d; p = 0.032

 NMS

0.71 (0.64–0.77)

0.74 (0.67–0.81)

Adherence MMAS-8 (N = 267)

 Normal practice

0.59 (0.51–0.67)

1.74 (1.04–2.90); p = 0.036

0.65 (0.56–0.75)

1.88 (1.06–3.34); p = 0.030

1.77 (0.96–3.28)d; p = 0.068

 NMS

0.72 (0.64–0.80)

0.78 (0.70–0.86)

CI confidence intervals, ITT intention to treat, MMAS-8 Morisky eight-item Medication Adherence Scale, NMS New Medicine Service

aThe ITT cohort was defined as all patients within a randomisation arm with measured outcomes, or who were followed up to the end of the study

bSimple logistic regression model

cMulti-level logistic regression model adjusted for recruiting pharmacy, disease, age, sex and medication count (level 1: patient, level 2: pharmacy)

dModel accounting for multiple imputation of missing data

2.1.4 Intervention and Within-Trial Costs

Costs comprised patient-level intervention costs and healthcare contact over the 10-week follow-up period. Subsequent resource use at follow-up was obtained from patient diaries, 116 normal practice and 122 NMS, consisting of NHS (primary care, secondary care, allied health professionals) and non-NHS costs (community-based practitioners and allied health professionals including community pharmacists). These data were combined with NHS reference costs [27] and Personal Social Services Research Unit (PSSRU) costs [28] (see electronic supplementary material) to derive patient-level total costs. Comparison between treatment arms at patient level was made using a two-sample t-test on the original dataset, or on a bootstrapped dataset, depending on the normality of the distribution of costs [29]. Mean (median, range) total NHS costs for patients in normal practice and NMS are £261 (£121, 0–1669), and £239 (£135, 25–1483), respectively. There was a general trend to reduced NHS costs, statistically non-significant, for the NMS intervention: −£21 (95% CI −59 to 150; p = 0.1281).

2.2 Clinical and Economic Impact of Non-adherence

Six Markov models were developed in TreeAge Pro (TreeAge Software Inc, One Bank Street, Williamstown, MA, 01267, USA). The most commonly prescribed medicine within the four NMS areas was used to inform a model representative of that disease group. Asthma and COPD were modelled separately due to the different natural history of the disease and impact of non-adherence. As hypertension represented over 50% of the cohort, two models were built to reflect the two most common medication groups prescribed: calcium channel blockers and angiotensin-converting enzyme (ACE) inhibitors—index NMS drugs for 34.4 and 24.1% of hypertensive patients, respectively.

The models were hypertension–amlodipine; hypertension–ramipril; asthma-inhaled corticosteroid (ICS)–beclometasone; COPD–tiotropium; diabetes–metformin; anticoagulants–aspirin. Each model had a lifetime horizon (until the age of 100), an annual (hypertension, diabetes, anticoagulant), monthly (COPD), or weekly (asthma) cycle length with half-cycle correction and the UK treasury recommended 3.5% discount rate for both costs and outcomes. Age-related mortality was included in each model.

Each model described the consequences of being adherent to the medicine, compared with non-adherence. Entry age, disease severity, drug prescribed and health status in the models were those in the RCT cohort.

2.2.1 Sources of Data and Model Design

Each model is described in detail in the electronic supplementary material. For all models apart from aspirin we utilised and adapted existing published models to optimise design (amlodipine [30], ramipril [31], beclometasone [32], tiotropium [33], metformin [34]). We derived the aspirin model structure from case–control studies analysing outcomes in large UK cohorts of patients with a first prescription of aspirin [35, 36].

2.2.2 Literature Searches

A literature search was conducted through Medline, Embase and Web of Science using treatment pathway-specific search terms. Databases were searched to the end of 2013. References in English and limited to humans were included. After excluding duplicate records, studies were included if they examined issues on the incidence, prevalence, treatment or resource use of the consequences of non-adherence. Reference lists of the retrieved references were hand-searched.

2.2.3 Transition Probabilities

Data came preferentially from up-to-date UK sources that reflected the NMS trial patient characteristics. The quality of evidence varied for the different models. Data were taken from RCTs such as effect of calcium channel blockers [37] or ACE inhibitors [31] on major cardiovascular consequences in hypertension, and effects of inhaled steroids on asthma control [38]. Observational data were available to populate parts of the metformin and aspirin models. In the metformin model, transition probabilities are based on UKPDS68 equations [34], using data on 3642 (type 2 diabetes) patients from UKPDS (United Kingdom Prospective Diabetes Study) for whom annual data on potential risk factors were available [39]. In the aspirin model, data were taken from case–control studies analysing outcomes in large UK cohorts of patients with a first prescription of aspirin [35, 36]. Where no other primary sources were available, transition probabilities estimated in published models were used, such as transition probabilities for exacerbations in COPD [40].

2.2.4 Identifying the Effect of Non-adherence on Outcomes

Quality of evidence for the effect of non-adherence on outcomes varied widely. Where possible, data on the impact of non-adherence were taken from large long-term cohort studies, such as the impact of non-adherence to antihypertensive treatment on long-term cardiovascular outcomes for the amlodipine [41] and ramipril models [42], and for myocardial infarction/chronic heart disease death [35] and for stroke/transient ischemic attack [36] in the aspirin model. The effect of non-adherence in the metformin model was modelled via a higher level of HbA1c [43], which resulted in estimated higher probabilities of diabetes complications. We used data from a study analysing non-adherence to antidiabetic medications (using MMAS-8 [26]) and HbA1c level in 301 patients with diagnosed type 2 diabetes in the US [43], good adherence (MMAS-8 score ≥3) was associated with 10% lower HbA1c (p = 0.0003).

Little data on the effect of non-adherence on asthma or COPD control was available. We derived the effect of non-adherence on asthma control from baseline data of the SIMPLE study (observational trial of community pharmacy intervention for asthma management), combining adherence to ICS assessed using prescription refill data in the previous 6 months and Asthma Control Test [44] (ACT, 0 ≤ ACT ≤ 25) results [45]. The most relevant results about adherence in COPD were available from Vestbo et al. [7]. Based on the Toward a Revolution in COPD Health (TORCH) study [46], they reported that adherent patients had a 60% [hazard ratio (HR) 0.4; 95% CI 0.35–0.46] lower risk of death and a 44% [relative risk (RR) 0.56; 95% CI 0.48–0.65] lower rate of severe exacerbations [7]. We found no data to inform the effect of adherence on frequency of non-severe exacerbations, so assumed that the HR was equivalent to severe exacerbations.

2.2.5 Health Status

For the amlodipine, ramipril, metformin and aspirin models, utilities were based on EQ-5D data from the Health Survey for England (2003, 2006), adjusted for age, sex and disease status [47]. Asthma utilities were taken from an RCT of ICS, using the Asthma Quality of Life Questionnaire (AQLQ [48]) scores [49]. Utilities for the COPD model were derived from an RCT of tiotropium [33, 50].

2.2.6 Resource Use and Unit Costs

Resource-use data came preferentially from up-to-date UK sources of observation of clinical practice, with disaggregated resource-use data, to allow attachment of current unit prices. If possible, individual patient data were used, with associated measures of mean and variation. If these were not available, point estimates were used, with carefully specified deterministic ranges, and standard methods for allocating distributions to these data were used.

The probability, cost and utility data were assigned beta, gamma and beta distributions, respectively, and are summarised in Tables 3, 4 and 5. Costs are given in year 2014 values.
Table 3

NHS and non-NHS costs for normal practice and NMS intervention

Cost category

Normal practice (n = 116)

Mean cost/£ (n, SE)

NMS (n = 122)

Mean cost/£ (n, SE)

Primary care total

81.6 (111, 5.76)

72.18 (115, 4.99)

GP total

67.7 (100, 5.26)

60.94 (105, 4.34)

 GP contact

59.21 (95, 4.35)

57.13 (98, 4.36)

 GP home visit

3.66 (2, 3.02)

0.7 (1, 0.7)

 GP phone call

4.83 (19, 1.11)

3.11 (13, 0.93)

Nursing total

13.9 (79, 1.53)

11.24 (73, 1.56)

 Nurse contact

12.49 (73, 1.46)

10.73 (72, 1.51)

 Nurse home visit

0.92 (3, 0.60)

0.38 (2, 0.28)

 Nurse phone call

0.49 (7, 0.29)

0.13 (3, 0.08)

Secondary care total

175.54 (53, 28.76)

141.23 (52, 25.79)

 Outpatient

98.85 (47, 16.42)

91.2 (46, 16.19)

 Accident and emergency

2.02 (2, 1.42)

0.96 (1, 0.96)

 Day case

63.01 (17, 16.00)

49.08 (13, 16.62)

 Inpatient

11.66 (1, 11.66)

0 (0, 0)

Allied HCPs (NHS) totala

3.73 (19, 1.13)

1.75 (16, 0.48)

 Allied HCP contact

2.37 (16, 0.66)

1.48 (13, 0.43)

 Allied HCP home visit

1.27 (3, 0.77)

0.15 (1, 0.15)

 Allied HCP phone call

0.08 (2, 0.06)

0.12 (3, 0.07)

NMS intervention

0

24.60

Total NHS cost

260.87 (114, 30.23)*

239.66 (121, 26.61)

Community-based practitioner totalb

4.81 (5, 2.66)

4.71 (2, 4.44)

 Community-based practitioner phone call

0.08 (2, 0.06)

0.08 (1, 0.08)

 Community-based practitioner contact

0.14 (1, 0.14)

0.27 (1, 0.27)

 Community-based practitioner home visit

4.58 (4, 2.54)

4.36 (1, 4.36)

Allied HCPs non-NHS total

7.4 (54, 0.99)

8.69 (64, 1.04)

 Community pharmacist

6.31 (48, 0.93)

7.57 (61, 0.91)

 Other associated HCPs non-NHSc

1.1 (10, 0.35)

1.13 (11, 0.34)

Total non-NHS cost

12.21 (56, 2.86)

13.4 (65, 4.5)

GP general practitioner, HCP healthcare practitioner, NHS National Health Service, NMS New Medicine Service, SE standard error

* Mean difference in costs: £21.11 (95% CI −59.01 to 100.24; p = 0.1281)

aAllied HCPs (NHS) include podiatrists, phlebotomists

bCommunity-based practitioners include social workers

cAllied HCPs (non-NHS) include dentists, opticians, chiropractors

Table 4

Summary of probabilities in the Markov models in the adherent and non-adherent groups for each of the six models (for full details of parameter derivation for each model, see electronic supplementary material, Table 16)

Parameter

Estimate and source

1. Hypertension–amlodipine model

P [stroke for adherent patient]

Age- and sex-dependent risk from ASCOT study [30]

P [MI/fatal CHD for adherent patient]

P [stroke for non-adherent patient]

P [stroke for adherent patient] × effect of non-adherence (HR)

P [MI/fatal CHD for non-adherent patient]

P [MI/fatal CHD for adherent patient] × effect of non-adherence (HR)

Effect of non-adherence, HR

Intermediate vs high adherence: HR 1.39 [41]

P [MI/CHD (non-fatal MI or fatal CHD) being fatal CHD]

Age- and sex-dependent risk from ASCOT study [30]

P [stroke being fatal]

P [death for MI survivors]

P [death for stroke survivors]

P [death from all other causes]

Age and sex dependent [73]

2. Hypertension–ramipril model

P [fatal cardiovascular event (MI, stroke,coronary event, heart failure, other cardiovascular events, first events) for adherent patient]

0.0068 [31]

P [fatal cardiovascular event for non-adherent patient]

0.0083, P [fatal cardiovascular event for adherent patient] × effect of non-adherence (HR)

P [non-fatal MI for adherent patient]

0.0041 [31]

P [non-fatal stroke for adherent patient]

0.0075 [31]

P [non-fatal MI for adherent patient]

0.0050, P [non-fatal MI for adherent patient] × effect of non-adherence (HR)

P [non-fatal stroke for adherent patient]

0.0092, P [non-fatal stroke for adherent patient] × effect of non-adherence (HR)

P [death for MI survivors]

Age, sex, number of years from MI dependent [74], personal communication

P [death for stroke survivors]

Age, sex, number of years from stroke dependent [75], personal communication

Effect of non-adherence, HR

HR 0.81 (95% CI 0.67–0.98), adjusted for age and sex (base-case scenario) [42]

P [death from all other causes]

Age and sex dependent [73]

3. Asthma–beclometasone model

 

Adherent patients [38, 76]

Non-adherent patients [38, 76]

P [successful control → sub-optimal control]

0.1563 × (1 − p)a

0.3710 × (1 − p) [45]a

P [successful control → primary care exacerbation]

0.0135 × (1 − p)a

P [successful control → secondary care exacerbation]

0.0054 × (1 − p)a

P [sub-optimal control → successful control]

0.1394 × (1 − p)a

P [sub-optimal control → sub-optimal control]

0.8322 × (1 − p)a

P [sub-optimal control → primary care exacerbation]

0.0174 × (1 − p)a

P [sub-optimal control → secondary care exacerbation]

0.0109 × (1 − p)a

P [secondary care exacerbation → successful control]

0.2000 × (1 − p)a

P [secondary care exacerbation → sub-optimal control]

0.2000 × (1 − p)a

P [secondary care exacerbation → primary care exacerbation]

0.4000 × (1 − p)a

P [secondary care exacerbation → secondary care exacerbation]

0.4000 × (1 − p)a

P [death from all other causes]

Age and sex dependent [73]

4. COPD–tiotropium model

Probability of exacerbation

Adherent patients [7, 77]

Non-adherent patients [7, 77]

Moderate COPD → exacerbation

0.051

0.089

Moderate COPD → severe exacerbation given an exacerbation occurs

0.097

0.165

Severe COPD → exacerbation

0.075

0.129

Severe COPD → severe exacerbation given an exacerbation occurs

0.136

0.229

Very severe COPD → exacerbation

0.096

0.164

Very severe COPD → severe exacerbation given an exacerbation occurs

0.192

0.316

Effect of non-adherence on exacerbations rate, HR

44% (HR 0.56; 95% CI 0.48–0.65); lower rate of severe exacerbations for adherent patients [7]

Effect of non-adherence on death rate, HR

60% (HR 0.4; 95% CI 0.35–0.46); lower risk of death for adherent patients [7]

P [moving between chronic health states ] year 1, subsequent years

[77]

5. Diabetes–metformin model

P [fatal first diabetes complication (MI, CHF, stroke, renal failure, amputation) for adherent patient]

Patient characteristic and HbA1c-dependent value from UKPDS68 [34] and UKPDS34 [39]

P [fatal first diabetes complication for nonadherent patient]

P [non-fatal first diabetes complication (MI, CHF, stroke, renal failure, amputation, blindness, IHD) for adherent patient]

P [non-fatal first diabetes complication for nonadherent patient]

P [second non-fatal diabetes complication after the first complication]

P [death from all other causes]

Age and sex dependent [73]

Effect of non-adherence

HbA1c multiplier for non-adherent patient: 1.105 (95% CI 1.047–1.166)

HbA1c for non-adherent patient (age) = HbA1c for adherent patient (age) × 1.105 [43]

6. Antiplatelets/anticoagulants–aspirin model

 

Adherent patient

Non-adherent patient

P [event-free → non-fatal MIb]

0.0056

0.0086 [35]

P [event-free → non-fatal strokeb]

0.0019

0.0028 [36]

P [event-free → fatal MI/CHDb]

0.0015

0.0023 [35]

P [event-free → fatal strokeb]

0.0002

0.0003 [36]

P [non-fatal MI → death]

Dependent on age, year after the first MI, sex [74]

P [non-fatal stroke → death]

Dependent on age, year after the first stroke, sex [75]

P [death from all other causes]

Age and sex dependent [73]

CHD coronary heart disease, CHF congestive heart failure, CI confidence intervals, COPD chronic obstructive pulmonary disease, HbA1c glycosylated haemoglobin, HR hazard ratio, IHD ischaemic heart disease, MI myocardial infarction, p probability, SE standard error, UKPDS United Kingdom Prospective Diabetes Study

aTransition probabilities between asthma states from [76, 38] with mortality (P[death from all other causes]) incorporated (p)

bProbabilities for three age groups, 50–64 years, 65–74 and 75–84 years, respectively. Probabilities calculated from incidence rates reported. In the case of non-adherent patients, incident rates adjusted by the effect of non-adherence (reported rate ratios for events, comparing non-adherence vs adherence)

Table 5

Summary of utilities and costs for the Markov models in the adherent and non-adherent groups for each of the six models (for full details of parameter derivation for each model, see electronic supplementary material, Table 17)

Health state

Utility weights

Mean cost/patient (£; 2014 values)

1. Hypertension–amlodipine

Well

Age- and sex-dependent, no cardiovascular event [47]

Mean annual cost of medication (amlodipine): 13.4 [78, 79]

Non-fatal MI

Age- and sex-dependent + MI history

Utility decrement (MI) added [47]

1st year: 5704.6

≥2nd year: 986.7 [78, 79, 80, 81]

Non-fatal stroke

Age- and sex-dependent + stroke history

Utility decrement (stroke) added [47]

1st year: 4161.8

≥2nd year: 770.9 [78, 79, 80, 81]

2. Hypertension–ramipril

Well

Age- and sex-dependent no cardiovascular event [47]

Mean annual cost of medication (ramipril): 95.8 [78, 79]

Non-fatal MI

Age- and sex-dependent + MI history

Utility decrement (MI) [47]

1st year: 5787

≥2nd year: 1069

Non-fatal stroke

Age- and sex-dependent + stroke history

Utility decrement (stroke) added [47]

1st year: 4244

≥2nd year: 853

3. Asthma–beclometasone

Successful control

0.900 [49, 82]

13.4 [32, 81]

Sub-optimal control

0.842 [49, 82]

34.9 [32, 81]

Asthma exacerbation

Primary care-managed: 0.57

Hospital-managed: 0.33 [83]

Primary care-managed: 105.6

Hospital-managed: 2013.1 [76, 81]

4. COPD–tiotropium

Moderate COPD

0.787 [84, 85]

46.53 per month [86]

Severe COPD

0.750 [84, 85]

79.32 per month [86]

Very severe COPD

0.647 [84, 85]

125.13 per month [86]

COPD exacerbation

Non-severe decrement: 0.01

Severe decrement: 0.042 [50, 87]

Non-severe: 75.97

Severe: 1372 [86]

5. Diabetes–metformin

Well

Age- and sex-dependent, no cardiovascular event [47]

Mean annual cost of medication (metformin): 8.05 [78, 79]

Other diabetes health states

Utility decrement [88]

Fatal event; non-fatal event 1st year; non-fatal event ≥2nd year [81, 88, 89]

 IHD

−0.090

N/A; 2916.4; 963.8

 MI

−0.055

1477.7; 5624.0; 926.0

 CHF

−0.108

3252.8; 3252.8; 1140.2

 Stroke

−0.164

4338.9;3440.0; 650.1

 Amputation

−0.280

11,200.4; 11,200.4; 646.9

 Blindness

−0.074

N/A; 1469.0; 622.0

 Renal failure

−0.263

32,452.5; 32,452.5; 32,452.5

6. Antiplatelets/anticoagulants–aspirin

Event-free

Age- and sex-dependent + utility decrement for MI/stroke history [35, 47]

1510.9 [35, 47, 79–82]

Non-fatal MI

Age- and sex-dependent + utility decrement for MI [47] [35]

1st year after MI: 6662.5

≥2nd year: 1597.1 [35, 47, 78, 79, 80, 81]

Non-fatal stroke

Age- and sex-dependent + utility decrement for stroke [36, 47]

1st year after stroke: 4593.5

≥2nd year: 1817.5 [36, 47, 78, 79, 80, 81]

CHF congestive heart failure, COPD chronic obstructive pulmonary disease, IHD ischaemic heart disease, MI myocardial infarction; N/A not applicable

2.3 Incremental Analysis

In the base case, we applied probabilities of adherence to each model, estimated from 10-week trial results, for the primary adherence outcome measure (Table 2). Adjusted probability of adherence in the NMS group was 74%, probability of adherence in normal practice was 63%, and the odds ratio (OR) (NMS vs normal practice) was 1.67 (1.06–2.62; p = 0.027). Adherence was assumed to stay the same in both arms over the time horizon of the model.

Each model was populated with probability, cost and health status data. This allowed the generation of the outcomes and costs in a cohort who were adherent, and in a cohort who were non-adherent to the medicine. The adherence for each drug–disease pair at 10-week follow-up in the NMS and normal practice arms were combined with the appropriate disease-drug-specific model. Using these models, we generated the difference in patient outcome and costs between NMS and normal practice for each disease–drug pair. Probabilistic estimates of costs and outcomes were derived, the analysis generating 5000 iterations, using Monte Carlo simulation for each disease–drug pair.

The incremental costs and outcomes associated with each disease–drug pair were incorporated additively into the economic model to allow derivation of the total incremental impact of the NMS intervention costs and outcomes for all six disease–drug pairs. At this point the NMS intervention costs were added.

Deterministic and probabilistic incremental economic analyses were carried out. The incremental cost-per-QALY generated by NMS over normal practice was calculated using the following equation:
$$({\text{Cost}}_{\text{NMS}} {-}{\text{Cost}}_{\text{Normal practice}} )/({\text{QALY}}_{\text{NMS}} {-}{\text{QALY}}_{\text{Normal practice}} ).$$

Utilising Microsoft Excel, we used 5000 Monte Carlo simulations to obtain the incremental cost-effectiveness ratio (ICER) distribution. Negative ICERs are difficult to interpret and often arise when one of the interventions is either ‘dominant’ (more effective, less costly) or ‘dominated’ (less effective, more costly). It is not possible to tell this from the ICER itself. We report the proportion of ICER estimates in each of the four quadrants of the cost-effectiveness plane. We present mean ICERs for all results, indicating for negative ICERs whether the intervention is dominant or dominated.

Cost-effectiveness acceptability curves (CEACs) [51] were constructed to express the probability that NMS is cost effective as a function of the decision maker’s ceiling cost-effectiveness ratio (λ) [52].

2.4 Sensitivity Analysis

Deterministic analysis was conducted using the MMAS-8 adherence measure, for which probabilities of adherence were 78 and 65% in the NMS and normal practice groups, respectively, with OR of 1.88 (1.06–3.34), (Table 2).

The deterministic analysis was repeated to determine the effect of reducing the effect size, by reducing the adherence in the NMS arm, keeping the probability of adherence in the current practice arm unchanged. The difference in adherence between NMS and normal practice that would be required to attain an ICER of £20,000 per QALY was determined.

The probabilistic analysis was repeated in the disease-specific subgroups.

2.5 Model Validation

Validity testing (conceptual model, input data, assumptions, model outcomes) was carried out iteratively as part of the development of the model throughout the project, with general practice, clinical pharmacy and health economics experts on the project team and the independent advisory panel [53]. This was carried out as multiple ‘walk-throughs’ and review of specific written summaries of model structure, inputs and outcomes. There is no comparable model of a cross-therapeutic intervention to assess adherence. However, cross validity of individual models was maximised by using published models to derive a model for each disease where possible. The computerised individual and composite models were developed by LT and GG and examined by RAE, who has built a composite model in a previous study. Models were only accepted if there were no illogical or illegal inputs or outputs.

3 Results

3.1 Incremental Analysis

Tables 6 and 7 summarise the lifetime costs and outcomes derived from each disease-drug-specific model, their relative contribution to the economic model, and the overall results for the incremental analysis of NMS versus normal practice.
Table 6

Results from individual models and incremental economic analysis of NMS versus normal practice: deterministic analysis

Model

Percentage of NMS cohort

Mean cost (£)

Mean QALY

Incremental

ICER (£/QALY)

NMSa

Normal practice

NMS

Normal practice

Cost (£)

QALY

Amlodipine

25.3

1496.9

1512.0

14.22

14.17

−15.1

0.04

−338.0

Ramipril

24.1

2925.4

2922.9

16.37

16.30

2.6

0.07

37.9

Aspirin

8.5

22,881.6

22,830.1

10.04

10.03

51.5

0.01

5151.0

Beclometasone

17.5

71,539.9

72,432.2

16.56

16.54

−892.3

0.02

−44,614.0

Tiotropium

5.7

10,508.6

10,250.3

6.99

6.85

258.3

0.14

1845.2

Metformin

18.8

15,285.7

15,279.8

9.55

9.53

5.9

0.02

293.0

Overall

100

19,013.2

19,151.8

13.49

13.45

−138.6

0.04

−3166.1

ICER incremental cost-effectiveness ratio, NMS New Medicine Service, QALY quality-adjusted life-year

aIncorporating cost of intervention equal to £24.6

Table 7

Results from individual models and incremental economic analysis of NMS versus normal practice: probabilistic sensitivity analysis

Model

Percentage of NMS cohort

Mean cost (95% CIb)/£

Mean QALY (95% CIb)

Incremental

ICER (£/QALY)

NMSa

Normal practice

NMS

Normal practice

Cost, £ (95% CIb)

QALY (95% CIb)

Amlodipine

25.3

1230.2 (498.9–2201.1)

1242.4 (481.9–2265.8)

13.96 (11.63–16.36)

13.90 (11.59–16.31)

−12.2 (−119.7 to 37.4)

0.06 (−0.04 to 0.23)

−1996.2

Ramipril

24.1

2946.4 (2280.0–3876.5)

2943.9 (2264.8–3882.8)

16.35 (13.93–18.82)

16.28 (13.87–18.76)

2.5 (−40.8 to 24.0)

0.07 (0.00 to 0.18)

−42.8

Aspirin

8.5

22,870.4 (17,855.9–30,396.3)

22,816.6 (17,822.8–30,300.3)

10.12 (8.94–11.28)

10.10 (8.93–11.26)

53.82 (23.9 to 108.3)

0.01 (0.00 to 0.04)

5585.7

Beclometasone

17.5

75,418.05 (21,472.3–253,610.9)

76,279.6 (21,475.9–257,102.1)

16.53 (15.61–17.14)

16.50 (15.58–17.13)

−861.5 (−4414.6 to 269.8)

0.02 (0.00 to 0.07)

−48,996.6

Tiotropium

5.7

11,184.9 (3150.8–25,321.4)

10,987.5 (3072.7–24,732.0)

7.27 (4.33–10.01)

7.14 (4.22–9.93)

197.4 (−234.9 to 979.9)

0.13 (0.00 to 0.39)

1819.1

Metformin

18.8

15,079.2 (9400.5–29,221.8)

15,073.5 (9409.0–29,152.1)

9.55 (8.77–10.31)

9.52 (8.74–10.29)

5.7 (−54.0 to 73.9)

0.03 (0.00 to 0.07)

1564.4

Overall

100

20,201.8 (9843.1–50,750.5)

20,345.8 (9840.2–51,749.6)

13.48 (12.59–14.39)

13.43 (12.53–14.34)

−144.0 (−768.7 to 72.9)

0.05 (0.00 to 0.13)

−2638.4

CI confidence interval, ICER incremental cost-effectiveness ratio, NMS New Medicine Service, QALY quality-adjusted life-year

aIncorporating cost of intervention equal to £24.6

b95% CIs were obtained from 2.5% and 97.5% percentiles for costs, QALYs and ICERs in probabilistic sensitivity analysis

NMS generated a mean of 0.04 more QALYs per patient than normal practice, at a mean reduced cost of −£139, with probabilistic means of 0.05 (95% CI 0.00–0.13) and −£144 (95% CI −769 to 73). Therefore, NMS dominates normal practice, with an ICER of –£3166 (probabilistic mean −£2638). The probability that NMS dominates normal practice is 0.78. NMS has a high probability (0.96) of cost effectiveness compared with normal practice at a willingness to pay of £20,000 for one QALY (see Figs. 3, 4).
Fig. 3

Incremental cost-effectiveness plane: New Medicine Service (NMS) intervention versus normal practice. 5000 iterations in composite and medicine-specific models. In composite model, cost and QALY in NMS and normal practice arms are calculated as weighted cost and QALY from each medicine-specific model. Iterations were ordered by the index reflecting the strength of NMS effect (incorporating adherence effect from medicine-specific model and trial effect of NMS on adherence). QALY quality-adjusted life-year

Fig. 4

Cost-effectiveness acceptability curve for New Medicine Service (NMS) intervention versus normal practice. This graph demonstrates the probability of cost effectiveness at a range of decision-maker ceiling willingness to pay for the NMS intervention overall. See electronic supplementary material (Section 4, Fig. 7) for Cost-effectiveness acceptability curves when only one disease group is considered at a time. WTP willingness to pay

3.2 Sensitivity Analysis

The results were robust to changing adherence outcome. When MMAS-8 was used to estimate changes in adherence, the incremental QALY was 0.06, incremental cost was −£164, with an ICER of −£2953 (see electronic supplementary material). The threshold analysis demonstrated that the effect size of NMS compared with normal practice would need to be reduced to an OR of 1.01 to derive an ICER of £20,000 per QALY or above (see electronic supplementary material).

The disease-specific sensitivity analysis showed that if the NMS intervention targeted one of the disease areas only, NMS generated more QALY gain than normal practice in all models. NMS generated lower lifetime costs than normal practice for amplodipine–hypertension and asthma models. Higher costs in the NMS arm were generated for ramipril–hypertension, diabetes, COPD and aspirin models. In all cases, higher lifetime costs were the effect of reduced mortality for adherent patients (deaths from COPD exacerbations and from cardiovascular events). The mean ICERs were hypertension only (amlodipine plus ramipril): −£115; asthma only: −£44,614; COPD only: £1845; diabetes only: £293; aspirin only: £5151 (Table 8). Targeting individual disease areas with NMS has a probability over 0.90 of cost effectiveness compared with normal practice at a willingness to pay of £20,000 per QALY (Fig. 4).
Table 8

Incremental effectiveness analysis of NMS intervention versus current practice in specific disease groups (base-case adherence outcome)

 

Incremental

ICER, £/QALY

% ICERs in each quadrant

Probability C/E at λ < £20 K

Cost/£

QALY

SEa

NE

SW

NW

Base case

−138.6

0.04

−3166.1

78.5

17.9

0.5

3.0

0.96

Hypertension onlyb

−6.5

0.06

−115.5

46.9

47.1

0.0

6.0

0.93

Asthma only

−892.3

0.02

−44,614.0

88.6

7.1

0.2

4.2

0.93

COPD only

258.3

0.14

1845.2

13.1

84.3

1.1

1.5

0.97

Diabetes only

5.9

0.02

293.0

40.7

57.2

0.0

2.1

0.97

Aspirin only

51.5

0.01

5151.0

0.0

96.8

0.0

3.1

0.93

C/E cost effective, CI confidence interval, COPD chronic obstructive pulmonary disease, ICER incremental cost-effectiveness ratio, NMS New Medicine Service, QALY quality-adjusted life-year

aProbability of dominance

bProbabilities of dominance are 50.0 and 36.2% and probabilities of cost effectiveness at λ < £20,000 are 83.7 and 98.0% for amlodipine and ramipril models, respectively. Probabilistic means with 95% CI for hypertension patients are −£4.5 (95% CI −69.4 to 28.4) and 0.06 (95% CI −0.01 to 0.20) for difference in costs and QALYs, respectively; ICER (£/QALY) was −£39,236.1 (95% CI −4049.2 to 5096.9), with a median of −£28.9

3.3 Model Validation

Where possible, outputs from individual models were compared with published model outputs. The output of four of our individual models was comparable with published studies. The amlodipine model generated 14.3 QALYs in the adherent group, similar to another UK hypertension model [54] and a German model [55]. The ramipril model generated 16.6 QALYs in the adherent group, similar to another ACEI model [56]. The diabetes model generated 9.62 QALYs in the adherent group, similar to UKPDS68 [34]. The aspirin model generated 10.07 QALYs in the adherent group. There were no lifetime horizon models available for comparison, but a 10-year model generated 8.2–8.4 QALYs [57]. We were not able to find models with time horizons similar to our asthma and COPD models, so relied on feedback from clinical experts regarding model outcome validity.

4 Discussion

This economic evaluation suggests that NMS will deliver better patient outcomes than normal practice at overall reduced costs to the NHS in the long term. In the short term, extra costs incurred by remunerating community pharmacists were absorbed by small reductions in other NHS contact-related costs.

Interventions to improve adherence were assessed in a recent Cochrane review as “mostly complex and not very effective”, and in need of better design [58]. Most strategies to improve medicines have been costly and atheoretical with little evidence of cost effectiveness [59, 60]. In contrast, NMS is an effective, relatively simple intervention with a strong theory base, transferable across diseases and healthcare settings, and which we estimate here to be cost effective for the NHS when compared with normal practice.

4.1 Strengths and Limitations

The current UK evaluative framework requires a cost per QALY to compare the value for money of different healthcare interventions. Patient adherence is an intermediate process indicator assumed to lead to changes in patient outcomes, but criticised as insufficient to demonstrate patient benefit [58]. Our analysis has moved beyond adherence, giving an estimate of clinical and economic impact of the intervention, and is an important development.

Pharmacist-led interventions often reflect their cross-therapeutic role around prescribing safety and patient adherence, bringing a significant challenge to the evaluative framework, which is historically disease-specific. Evaluations of pharmacist-led interventions often rely on generic process measures such as errors [61, 62], medication changes [63] or patient adherence [14], and tend to report ICERs such as cost per adherent patient or cost per error avoided [14, 61]; or utilise assumptions about the level of disutility incurred [64]. In a previous study examining the economic impact of a pharmacist-led information technology-based intervention (PINCER) to reduce medication errors in general practices, we developed a novel approach where economic models were developed for each of six errors, and generated a cost per QALY [22]. We applied this method in the economic evaluation of NMS. Therefore, this study differs from most other economic evaluations in this area, and we have been able to generate cost-per-QALY statistics to inform decision making.

The effect size of an absolute 10% improvement in adherence from normal practice to NMS is similar to that reported in the original work by Clifford [65]. Although the effect sizes here might initially be considered small, we anticipate significant benefits for two reasons. First, large numbers of patients have experienced the service. Second, we suggest that this is a conservative effect size, given probable patient recruitment bias, use of self-report of adherence, and the assumption that all the patients in the intervention arm actually received the NMS.

There is no gold standard for measuring patients’ medicines adherence. Each approach has limitations. More than one adherence measure should be used to provide an internal check on validity [1]. In this study, we chose two self-report measures. Prescription filling was not an option for routine adherence monitoring in England due to lack of interoperability between community pharmacy and general practice systems. Although self-report tends to return a higher rate of medication adherence (+15%) than some objective measures, it correlates with objective clinical measures [66]. It is possible to minimise biases through confidential interview [67], as is carried out as part of NMS; normalising non-adherence by recognising the challenges of taking regular medications; avoiding negative or positive questions which may encourage a biased response; and asking about a missed dose in the few days or a week prior to data collection rather than months or years [68].

A key limitation is the paucity of data upon which to base the estimates of economic impact of adherence in the individual disease–drug pairs, particularly the link between adherence and outcome. The wide range around the point estimates of cost effectiveness reflects the uncertainty in some of the individual adherence models.

Weaknesses in the models centre on assumptions made. It is assumed that the incremental effect of NMS compared with normal practice on adherence is the same over a patient’s lifetime. Published estimates of persistence to new hypertension medicines at 1 year are around 48% [69, 70] and 51% [71] dropping from 6-month rates of 65–68%. This suggests that adherence is likely to drop in the cohort not having NMS at the beginning of their treatment. Our study suggests that the effect of NMS appears to be an absolute increase in adherence of 10% at 10 weeks post-initiation of the new medicine. It is not clear whether this effect will:
  1. (a)

    disappear, such that there is no difference in adherence at 6 or 12 months;

     
  2. (b)

    be maintained, such that adherence in both the NMS and control groups drops over time at the same rate, so the current benefit is maintained; or

     
  3. (c)

    initiate a change in the patient’s motivation or ability to adhere that leads to sustained adherence to the medicine such that the usual drop in adherence over time is prevented.

     

In the absence of any evidence to support which of these scenarios reflects reality, the economic analysis assumes scenario (b) for the basis of extrapolation of effect size. Sensitivity analysis suggested that the OR had to be reduced from 1.67 to 1.01 for NMS to stop being cost effective at a ceiling willingness to pay of £20,000 per QALY.

Apart from general limitations associated with the use of modelling, specific model limitations in our study include the use of the same effect of NMS on adherence for each disease-specific model since the trial was powered to analyse the effect of NMS in the entire NMS trial population. However, the effect of NMS on adherence may differ between disease groups and, in this case, the results of disease-specific models would be different than those assuming the same effect of NMS on adherence. The effects of adherence incorporated in the models were observed in the studies with different time horizons, from 1 year to longer (with maximal follow-up 4–5 years), while in the model we assume that effect of adherence is kept over a lifetime. We did not incorporate adverse event states in the models that may affect cost effectiveness (e.g. in older people, antihypertensive drugs may increase the risks of falls).

4.2 Implications for Policy and Practice

From inception of the NMS to the end of August 2016, 3.59 million consultations have been claimed for with over 820,000 in the year 2015/16 [90]. From the results of this economic evaluation, this suggests £75.4 million short-term savings to the NHS, £517.6 million long-term cost savings to the NHS and 179,500 QALYs gained.

The research presented above suggests that the NMS is cost effective for each disease population than normal practice, with high (above 97%) probability of cost effectiveness at a willingness to pay set at £20,000. On the basis of this evidence, it is recommended that this service continue to be commissioned in the future.

Where there is evidence suggesting therapeutic areas with significantly poor adherence, especially when non-adherence has significant effect on outcomes, consideration should be given to expanding the NMS. Potential areas might include conditions where medicines can have early adverse effects that subside over time such as anti-depressants.

5 Conclusions

This study suggests increased health gain with NMS over normal practice at a cost per QALY well below most accepted thresholds for technology implementation [72]. This intervention could be extended to other groups of medicines. The findings are likely to have applicability to other healthcare systems, including those based on insurance.

5.1 Data Availability Statement

Several datasets were used for this analysis: (1) efficacy, patient-reported outcomes and healthcare resource utilisation data collected in the NMS randomised controlled trial (RCT). The patient-level data are not publicly available, but the results of the trials have been presented in several publications. The trial results supporting the findings of this analysis are available within the article and its electronic supplementary material. (2) The six individual models use RCT, observational data and estimation tools from multiple sources, for which references are provided in the article. (3) Cost data used in the model were obtained from referenced publicly available sources. (4) The model was developed in Data TreeAge and is not publicly available, but is available from the authors upon request.

Notes

Acknowledgements

This study was commissioned and financed by the Department of Health. In addition, financial support was received from the Comprehensive Local Research Networks (CLRNs) to allow sites to receive service support costs as a result of participation in the study. The CLRNs involved included Trent, South Yorkshire, Leicester, Northamptonshire and Rutland (LNR) and the North Central London Research Consortium. Assistance with operationalising the study was provided by the Primary Care Research Networks. The study team expresses its gratitude to all of the study participants, the patients, the pharmacists and pharmacy companies and GPs, without whose participation this study would not have been possible. The study was monitored and supported by the independent NMS Evaluation Advisory Group (NEAG) chaired by Prof. Nick Mays, LSHTM. The study design and delivery was assisted by Dr. Tracey Thornley, Richard Harris, and Sara Garfield. Pharmacy recruitment was assisted by Dr. Tracey Thornley, Richard Harris, Mimi Lau (Numark), and Alistair Buxton (PSNC). Additional patient perspectives and review were provided by Ember Vincent and Clancy Williams. The study delivery was supported by Angela Wu, Rory Constable, Beatrice Odugbemi, Chris Macleod, Edward Lees-Manning, Julia Graichen, Katharine Wells, Lina Bader, Linda van Eikenhorst, Melanie Lynn, Oluwagbemileke Ojeleye, Stefanie Sickinger, Xiao-Jun Xu and Zoe Lim. Advice and support with data analysis was provided by Dr. Sarah Armstrong, Dr. Sarah Rodgers, Dr. Koen Putman, Dr. Anna Murphy and Dr. Amanj Baker. We would like to thank the following researchers for providing data used in the economic evaluation: Ramon Luengo-Fernandez, University Research Lecturer, Health Economics Research Centre, University of Oxford; Dr. Anna Murphy, Consultant Respiratory Pharmacist at University Hospitals of Leicester NHS Trust; Luis Alberto García Rodríguez, Director, CEIFE—Centro Español de Investigación Farmacoepidemiológica Madrid, Spain; Kate Smolina, Unit of Healthcare Epidemiology and British Heart Foundation Health Promotion Research Group, University of Oxford. Loraine Buck has been the NMS study administrator and has contributed significantly to the execution of all aspects of the research programme. Commissioned assistance was provided by Alison Taylor (transcriber), Alphagraphics—Peter Johnson (production of study diaries), Ambrow Ltd (study website), Graham Watson (database design), Marita Smith (transcriber), Kath’s Keying Services Limited (transcriber) Petrina Salema (study logo design and web graphics). National data relating to NMS service provision was provided by Health Information Exchange (to March 2013) and by myhealthplace (from April 2013). The team are grateful to Prof. Donald E. Morisky, Department of Community Health Sciences, UCLA School of Public Health, for granting copyright permission to use the MMAS-8 adherence tool for this research. The University of Nottingham acted as the study sponsor and indemnifies the study. The team specifically wish to acknowledge the advice and support of Paul Cartledge, Angela Shone, Samantha Bateman and Kristy Angell.

Author contributions

RAE designed and led the economic analysis, led drafting of the manuscript, was study principal investigator and contributed to the analysis of the RCT. LT designed and built the economic models, including collection of input data on transition probabilities, utilities and costs; carried out the economic analysis (including method and estimation of overall cost effectiveness); contributed to leading and design of the economic analysis; contributed to the COPD model; and contributed to the analysis of the RCT. GG designed and built the COPD model, including collection of input data on transition probabilities, utilities and costs; contributed to the economic analysis and carried out intervention and within-trial costing; and contributed to the analysis of the RCT. MJB led the delivery of the RCT including design of data collection tools and oversight of data collection. AJA contributed to model design and reviewed all models for clinical face validity. NB was involved in intervention design and costing and informed and reviewed the economic analysis for practice face validity. AL was involved in the delivery of the RCT and informed and reviewed the economic analysis for practice face validity. RM led the statistical design and analysis of the RCT and derived the estimates of effect for the economic evaluation. AC provided patient and public input to the project and interpretation of the economic evaluation. JW led the process evaluation of NMS and informed and reviewed the economic analysis for practice and policy relevance. All authors were involved in the drafting of the manuscript. Prof. RE is the guarantor and affirms that the manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Compliance with Ethical Standards

Ethical approval

This report is independent research commissioned and funded by the Department of Health Policy Research Programme (Grant no. PR-IC-0711-10010) (‘Understanding and Appraising the New Medicines Service in the NHS in England’–PRP 029/0124). The views expressed in this publication are those of the authors and not necessarily those of the Department of Health. The RCT (ClinicalTrials.gov Trial reference number NCT01635361; http://clinicaltrials.gov/ct2/show/NCT01635361) had full ethical approval.

Conflict of interest

Rachel A. Elliott, Lukasz Tanajewski, Georgios Gkountouras, Anthony J. Avery, Nick Barber, Rajnikant Mehta, Matthew J. Boyd, Asam Latif, Antony Chuter and Justin Waring have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years; and no other relationships or activities that could appear to have influenced the submitted work.

Supplementary material

40273_2017_554_MOESM1_ESM.docx (596 kb)
Supplementary material 1 (DOCX 596 kb)

References

  1. 1.
    Horne R, Barber N, Weinman J, Elliott RA, Morgan M, Cribb A. Concordance, adherence and compliance in medicine taking: a scoping exercise. Report for the National Co-ordinating Centre for NHS Service Delivery and Organisation. London: R & D; 2006. http://www.netscc.ac.uk/hsdr/files/project/SDO_FR_08-1412-076_V01.pdf. Accessed 10 June 2015.
  2. 2.
    Marsden E, Cubbin I, McAlavey A. An investigation into how poor compliance traditionally associated with corticosteroid therapy in asthma and chronic obstructive pulmonary disease can be improved to enhance long-term management and patient care. Int J Pharm Pract. 2009;17(S2):B55–6.Google Scholar
  3. 3.
    Llorca PM. Patient compliance in schizophrenia and the impact on patient outcome. Psychiatr Res. 2008;161:235–47.CrossRefGoogle Scholar
  4. 4.
    Cerveri I, Locatelli F, Zoia MC, Corsico A, Accordini S, Marco R. International variations in asthma treatment compliance. Eur Respir J. 1999;14:288–94.CrossRefPubMedGoogle Scholar
  5. 5.
    Ho PM, Rumsfeld JS, Masoudi FA, McClure DL, Plomondon ME, Steiner JF, et al. Effect of medication nonadherence on hospitalization and mortality among patients with diabetes mellitus. Arch Intern Med. 2006;166(17):1836–41.CrossRefPubMedGoogle Scholar
  6. 6.
    Ho PM, Spertus JA, Masoudi FA, Reid KJ, Peterson ED, Magid DJ, et al. Impact of medication therapy discontinuation on mortality after myocardial infarction. Arch Intern Med. 2006;166(17):1842–7.CrossRefPubMedGoogle Scholar
  7. 7.
    Vestbo J, Anderson JA, Calverley PM, Celli B, Ferguson GT, Jenkins C, et al. Adherence to inhaled therapy, mortality and hospital admission in COPD. Thorax. 2009;64(11):939–43. doi:10.1136/thx.2009.113662.CrossRefPubMedGoogle Scholar
  8. 8.
    IMS Institute for Healthcare Informatics. Advancing the responsible use of medicines: applying levers for change. USA: Parsipanny; 2012.Google Scholar
  9. 9.
    Trueman P, Lowson K, Blighe A, Meszaros A, Wright D, Glanville J, et al. Evaluation of the scale, causes and costs of waste medicines. YHEC/London School of Pharmacy: London; 2010. https://core.ac.uk/download/pdf/111804.pdf. Accessed 10 June 2015.
  10. 10.
    Barber N, Parsons J, Clifford S, Darracott R, Horne R. Patients’ problems with new medication for chronic conditions. Qual Saf Health Care. 2004;13(3):172–5.CrossRefPubMedPubMedCentralGoogle Scholar
  11. 11.
    Ereshefsky L, Saragoussi D, Despiegel N, Hansen K, Francois C, Maman K. The 6-month persistence on SSRIs and associated economic burden. J Med Econ. 2010;13(3):527–36. doi:10.3111/13696998.2010.511050.CrossRefPubMedGoogle Scholar
  12. 12.
    Leventhal H, Cameron LD. Behavioral theories and the problem of compliance. Patient Educ Couns. 1987;10:117–38.CrossRefGoogle Scholar
  13. 13.
    Clifford S, Barber N, Elliott R, Hartley E, Horne R. Patient-centred advice is effective in improving adherence to medicines. Pharm World Sci. 2006;28(3):165–70.CrossRefPubMedGoogle Scholar
  14. 14.
    Elliott RA, Clifford S, Barber N, Hartley E, Horne R. The cost effectiveness of a pharmacy advisory service to improve adherence to medicines. Pharm World Sci. 2008;30:17–23.CrossRefPubMedGoogle Scholar
  15. 15.
    Department of Health. Pharmacy in England: Building on strengths—delivering the future http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/PublicationsPolicyAndGuidance/DH_083815. 2008. Accessed 12 Aug 2011.
  16. 16.
    Pharmaceutical Services Negotiating Committee. New Medicines Service. 2011. http://www.psnc.org.uk/pages/nms.html. Accessed 12 Aug 2011.
  17. 17.
    NHS Business Services Authority. Complete new medicines service (NMS) data. London. 2014. http://psnc.org.uk/services-commissioning/advanced-services/nms/service-evaluation-nms-statistics-andpharmoutcomes-data/. Accessed 10 June 2015.
  18. 18.
    Husereau D, Drummond M, Petrou S, Carswell C, Moher D, Greenberg D, et al. Consolidated health economic evaluation reporting standards (CHEERS) statement. Pharmacoeconomics. 2013;31(5):361–7. doi:10.1007/s40273-013-0032-y.CrossRefPubMedGoogle Scholar
  19. 19.
    Boyd M, Waring J, Barber N, Mehta R, Chuter A, Avery A, et al. Protocol for the New Medicine Service Study: a randomized controlled trial and economic evaluation with qualitative appraisal comparing the effectiveness and cost effectiveness of the New Medicine Service in community pharmacies in England. Trials. 2013;14(1):411.CrossRefPubMedPubMedCentralGoogle Scholar
  20. 20.
    Elliott RA, Boyd M, Waring J, Barber ND, Mehta R, Chuter A et al. Understanding and Appraising the New Medicines Service in the NHS in England (029/0124)’ A randomised controlled trial and economic evaluation with qualitative appraisal comparing the effectiveness and cost effectiveness of the New Medicine Service in community pharmacies in England. University of Nottingham. 2014.Google Scholar
  21. 21.
    Elliott RA, Boyd MJ, Salema N-E, Davies J, Barber N, Mehta RL, et al. Supporting adherence for people starting a new medication for a long-term condition through community pharmacies: a pragmatic randomised controlled trial of the New Medicine Service. BMJ Qual Saf. 2015;. doi:10.1136/bmjqs-2015-004400.Google Scholar
  22. 22.
    Elliott RA, Putman KD, Franklin M, Annemans L, Verhaeghe N, Eden M, et al. Cost effectiveness of a pharmacist-led information technology intervention for reducing rates of clinically important errors in medicines management in general practices (PINCER). Pharmacoeconomics. 2014;. doi:10.1007/s40273-014-0148-8.Google Scholar
  23. 23.
    Avery AJ, Rodgers S, Cantrill JA, Armstrong S, Cresswell K, Eden M, et al. A pharmacist-led information technology intervention for medication errors (PINCER): a multicentre, cluster randomised, controlled trial and cost-effectiveness analysis. Lancet. 2012;379:1310–9. doi:10.1016/S0140-6736(11)61817-5.CrossRefPubMedGoogle Scholar
  24. 24.
    Pharmaceutical Services Negotiating Committee, NHS Employers. NHS community pharmacy contractual framework 2011/12 service developments—latest information. August 2011.Google Scholar
  25. 25.
    Garfield S, Clifford S, Eliasson L, Barber N, Willson A. Suitability of measures of self-reported medication adherence for routine clinical use: a systematic review. BMC Med Res Methodol. 2011;11:149. doi:10.1186/1471-2288-11-149.CrossRefPubMedPubMedCentralGoogle Scholar
  26. 26.
    Morisky DE, Ang A, Krousel-Wood M, Ward HJ. Predictive validity of a medication adherence measure in an outpatient setting. J Clin Hypertens. 2008;10:348–54.CrossRefGoogle Scholar
  27. 27.
    National Health Service Executive. NHS reference costs 2012–13. 2013. https://www.gov.uk/government/publications/nhs-reference-costs-2012-to-2013. Accessed 20 June 2014.
  28. 28.
    Personal Social Services Research Unit (PSSRU). Unit Costs of Health and Social Care. 2012. http://www.pssru.ac.uk/project-pages/unit-costs/2012/. Accessed 20 June 2014.
  29. 29.
    Briggs A, Gray A. The distribution of health care costs and their statistical analysis for economic evaluation. J Health Serv Res Policy. 1998;3(4):233–45.CrossRefPubMedGoogle Scholar
  30. 30.
    Lindgren P, Buxton M, Kahan T, Poulter NR, Br DahlÃf, Sever PS, et al. The lifetime cost effectiveness of amlodipine-based therapy plus atorvastatin compared with atenolol plus atorvastatin, amlodipine-based therapy alone and atenolol-based therapy alone: results from ASCOT1. Pharmacoeconomics. 2009;27(3):221–30.CrossRefPubMedGoogle Scholar
  31. 31.
    Wing LMH, Reid CM, Ryan P, Beilin LJ, Brown MA, Jennings GLR, et al. A comparison of outcomes with angiotensin-converting—enzyme inhibitors and diuretics for hypertension in the elderly. N Engl J Med. 2003;348(7):583–92. doi:10.1056/NEJMoa021716.CrossRefPubMedGoogle Scholar
  32. 32.
    Steuten L, Palmer S, Vrijhoef B, Van Merode F, Spreeuwenberg C, Severens H. Cost-utility of a disease management program for patients with asthma. Int J Technol Assess Health Care. 2007;23(2):184–91.CrossRefPubMedGoogle Scholar
  33. 33.
    Price D, Asukai Y, Ananthapavan J, Malcolm B, Radwan A, Keyzor I. A UK-based cost-utility analysis of indacaterol, a once-daily maintenance bronchodilator for patients with COPD, using real world evidence on resource use. Appl Health Econ Health Policy. 2013;11(3):259–74. doi:10.1007/s40258-013-0021-5.CrossRefPubMedPubMedCentralGoogle Scholar
  34. 34.
    Clarke PM, Gray AM, Briggs A, Farmer AJ, Fenn P, Stevens RJ, et al. A model to estimate the lifetime health outcomes of patients with Type 2 diabetes: the United Kingdom prospective diabetes study (UKPDS) outcomes Model (UKPDS no. 68). Diabetologia. 2004;47(10):1747–59. doi:10.1007/s00125-004-1527-z.CrossRefPubMedGoogle Scholar
  35. 35.
    Rodríguez LAG, Cea-Soriano L, Martín-Merino E, Johansson S. Discontinuation of low dose aspirin and risk of myocardial infarction: case-control study in UK primary care. BMJ. 2011;. doi:10.1136/bmj.d4094.PubMedCentralGoogle Scholar
  36. 36.
    Garcia Rodriguez LA, Cea Soriano L, Hill C, Johansson S. Increased risk of stroke after discontinuation of acetylsalicylic acid: A UK primary care study. Neurology. 2011;76(8):740–6.CrossRefPubMedGoogle Scholar
  37. 37.
    Dahlöf B, Sever PS, Poulter NR, Wedel H, Beevers DG, Caulfield M, et al. Prevention of cardiovascular events with an antihypertensive regimen of amlodipine adding perindopril as required versus atenolol adding bendroflumethiazide as required, in the Anglo-Scandinavian Cardiac Outcomes Trial-Blood Pressure Lowering Arm (ASCOT-BPLA): a multicentre randomised controlled trial. Lancet. 2005;366(9489):895–906. doi:10.1016/S0140-6736(05)67185-1.CrossRefGoogle Scholar
  38. 38.
    Kavuru M, Melamed J, Gross G, Laforce C, House K, Prillaman B, et al. Salmeterol and fluticasone propionate combined in a new powder inhalation device for the treatment of asthma: a randomized, double-blind, placebo-controlled trial. J Allergy Clin Immunol. 2000;105(6 Pt 1):1108–16.CrossRefPubMedGoogle Scholar
  39. 39.
    UK Prospective Diabetes Study. UKPDS) Group. Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34. Lancet. 1998;352(9131):854–65.CrossRefGoogle Scholar
  40. 40.
    Oostenbrink JB, Rutten-van Mölken MPMH, Monz BU, FitzGerald JM. Probabilistic markov model to assess the cost-effectiveness of bronchodilator therapy in COPD patients in different countries. Value Health. 2005;8(1):32–46. doi:10.1111/j.1524-4733.2005.03086.x.CrossRefPubMedGoogle Scholar
  41. 41.
    Mazzaglia G, Ambrosioni E, Alacqua M, Filippi A, Sessa E, Immordino V, et al. Adherence to antihypertensive medications and cardiovascular morbidity among newly diagnosed hypertensive patients. Circulation. 2009;120(16):1598–605. doi:10.1161/circulationaha.108.830299.CrossRefPubMedGoogle Scholar
  42. 42.
    Nelson R, Reid CM, Ryan P, Willson K, Yelland L. Self-reported adherence with medication and cardiovascular disease outcomes in the Second Australian National Blood Pressure Study (ANBP2). Med J Aust. 2006;185:487–9.PubMedGoogle Scholar
  43. 43.
    Krapek K, King K, Warren SS, George KG, Caputo DA, Mihelich K, et al. Medication adherence and associated hemoglobin A1c in type 2 diabetes. Ann Pharmacother. 2004;38(9):1357–62. doi:10.1345/aph.1D612.CrossRefPubMedGoogle Scholar
  44. 44.
    Quality Metric. The asthma control test. GlaxoSmithKline. 2012. http://www.asthmacontroltest.com/. Accessed 15 Mar 2012.
  45. 45.
    Murphy AL. The community pharmacy SIMPLE approach to asthma management. Regional Innovation Fund (RIF) Project Evaluation Report. 2012.Google Scholar
  46. 46.
    Vestbo J. The TORCH (towards a revolution in COPD health) survival study protocol. Eur Respir J. 2004;24(2):206–10.CrossRefPubMedGoogle Scholar
  47. 47.
    Ara R, Brazier J. Health related quality of life by age, gender and history of cardiovascular disease: results from the Health Survey for England. Discussion Paper: Universities of York, Sheffield, Leeds. 2009.Google Scholar
  48. 48.
    Juniper E, Guyatt G, Ferrie P, King D. Development and validation of a questionnaire to measure asthma control. Eur Respir J. 1999;14(4):902–7.CrossRefPubMedGoogle Scholar
  49. 49.
    Bateman ED, Bousquet J, Keech ML, Busse WW, Clark TJH, Pedersen SE, et al. The correlation between asthma control and health status: the GOAL study. Eur Respir J. 2007;29(1):56–62. doi:10.1183/09031936.00128505.CrossRefPubMedGoogle Scholar
  50. 50.
    Rutten-van Mölken MMH, Hoogendoorn M, Lamers L. Holistic preferences for 1-year health profiles describing fluctuations in health. Pharmacoeconomics. 2009;27(6):465–77. doi:10.2165/00019053-200927060-00003.CrossRefPubMedGoogle Scholar
  51. 51.
    Fenwick E, Byford S. A guide to cost-effectiveness acceptability curves. Br J Psychiatry. 2005;187(2):106–8. doi:10.1192/bjp.187.2.106.CrossRefPubMedGoogle Scholar
  52. 52.
    Fenwick E, Claxton K, Sculpher MJ. Representing uncertainty: the role of cost effectiveness acceptability curves. Health Econ. 2001;10:779–87.CrossRefPubMedGoogle Scholar
  53. 53.
    Vemer P, Corro Ramos I, van Voorn GA, Al MJ, Feenstra TL. AdViSHE: a validation-assessment tool of health-economic models for decision makers and model users. Pharmacoeconomics. 2016;34(4):349–61. doi:10.1007/s40273-015-0327-2.CrossRefPubMedGoogle Scholar
  54. 54.
    Montgomery AA, Fahey T, Ben-Shlomo Y, Harding J. The influence of absolute cardiovascular risk, patient utilities, and costs on the decision to treat hypertension: a Markov decision analysis. J Hypertens. 2003;21(9):1753–9. doi:10.1097/01.hjh.0000084708.87421.d1.CrossRefPubMedGoogle Scholar
  55. 55.
    Gandjour A, Stock S. A national hypertension treatment program in Germany and its estimated impact on costs, life expectancy, and cost-effectiveness. Health Policy. 2007;83(2–3):257–67. doi:10.1016/j.healthpol.2007.01.003.CrossRefPubMedGoogle Scholar
  56. 56.
    Nordmann AJ, Krahn M, Logan AG, Naglie G, Detsky AS. The cost effectiveness of ACE inhibitors as first-line antihypertensive therapy. Pharmacoeconomics. 2003;21(8):573–85.CrossRefPubMedGoogle Scholar
  57. 57.
    Lamotte M, Annemans L, Evers T, Kubin M. A multi-country economic evaluation of low-dose aspirin in the primary prevention of cardiovascular disease. Pharmacoeconomics. 2006;24(2):155–69.CrossRefPubMedGoogle Scholar
  58. 58.
    Nieuwlaat R, Wilczynski N, Navarro T, Hobson N, Jeffery R, Keepanasseril A, et al. Interventions for enhancing medication adherence. Cochrane Database Syst Rev. 2014. doi:10.1002/14651858.CD000011.pub4.PubMedGoogle Scholar
  59. 59.
    Elliott RA, Barber N, Horne R. Cost-effectiveness of adherence-enhancing interventions: a quality assessment of the evidence. Ann Pharmacother. 2005;39:508–15.CrossRefPubMedGoogle Scholar
  60. 60.
    National Institute for Health and Clinical Excellence. Medicines adherence: involving patients in decisions about prescribed medicines and supporting adherence. Clinical Guidelines: CG76. 2009.Google Scholar
  61. 61.
    Avery AJ, Rodgers S, Cantrill JA, Armstrong S, Cresswell K, Eden M, et al. A pharmacist-led information technology intervention for medication errors (PINCER): a multicentre, cluster randomised, controlled trial and cost-effectiveness analysis. Lancet. 2012;379(9823):1310–9.CrossRefPubMedGoogle Scholar
  62. 62.
    Klopotowska JE, Kuiper R, van Kan HJ, de Pont AC, Dijkgraaf MG, Lie AHL, et al. On-ward participation of a hospital pharmacist in a Dutch intensive care unit reduces prescribing errors and related patient harm: an intervention study. Crit Care (London, England). 2010;14(5):R174.CrossRefGoogle Scholar
  63. 63.
    Denneboom W, Dautzenberg MGH, Grol R, De Smet PAGM. Treatment reviews of older people on polypharmacy in primary care: Cluster controlled trial comparing two approaches. Br J Gen Pract. 2007;57(542):723–31.PubMedPubMedCentralGoogle Scholar
  64. 64.
    Karnon J, Campbell F, Czoski-Murray C. Model-based cost-effectiveness analysis of interventions aimed at preventing medication error at hospital admission (medicines reconciliation). J Eval Clin Pract. 2009;15:299–306.CrossRefPubMedGoogle Scholar
  65. 65.
    Clifford S. Evaluation of a patient-centred, pharmacist-delivered intervention to improve patients’ adherence to medication. Psychol Health. 2005;20:48–9.Google Scholar
  66. 66.
    Murri R, Ammassari A, Gallicano K, Luca AD, Cingolani A, Jacobson D, et al. Patient-reported nonadherence to HAART is related to protease inhibitor levels. J Acquir Immune Defic Syndr. 2000;24(2):123–8.CrossRefPubMedGoogle Scholar
  67. 67.
    Butler JA, Peveler RC, Roderick P, Horne R, Mason JC. Measuring compliance with drug regimens after renal transplantation: comparison of self-report and clinician rating with electronic monitoring. Transplantation. 2004;77(5):786–9.CrossRefPubMedGoogle Scholar
  68. 68.
    Lehmann A, Aslani P, Ahmed R, Celio J, Gauchet A, Bedouch P, et al. Assessing medication adherence: options to consider. Int J Clin Pharm. 2014;36(1):55–69. doi:10.1007/s11096-013-9865-x.CrossRefPubMedGoogle Scholar
  69. 69.
    Caro JJ, Salas M, Speckman JL, Raggio G, Jackson JD. Persistence with treatment for hypertension in actual practice. Can Med Assoc J. 1999;160(1):31–7.Google Scholar
  70. 70.
    Vrijens B, Vincze G, Kristanto P, Urquhart J, Burnier M. Adherence to prescribed antihypertensive drug treatments: longitudinal study of electronically compiled dosing histories. BMJ. 2008;336(7653):1114–7. doi:10.1136/bmj.39553.670231.25.CrossRefPubMedPubMedCentralGoogle Scholar
  71. 71.
    Morgan SG, Yan L. Persistence with hypertension treatment among community-dwelling BC seniors. Can J Clin Pharmacol. 2004;11(2):e267–73.PubMedGoogle Scholar
  72. 72.
    McCabe C, Claxton K, Culyer AJ. The NICE cost-effectiveness threshold: what it is and what that means. Pharmacoeconomics. 2008;26(9):733–44.CrossRefPubMedGoogle Scholar
  73. 73.
  74. 74.
    Smolina K, Wright FL, Rayner M, Goldacre MJ. Long-term survival and recurrence after acute myocardial infarction in England, 2004 to 2010. Circ Cardiovasc Qual Outcomes. 2012. doi:10.1161/circoutcomes.111.964700.PubMedGoogle Scholar
  75. 75.
    Luengo-Fernandez R, Gray AM. Rothwell PM, null. A population-based study of hospital care costs during 5 years after transient ischemic attack and stroke. Stroke J Cereb Circ. 2012;43(12):3343–51.CrossRefGoogle Scholar
  76. 76.
    Price MJ, Briggs AH. Development of an economic model to assess the cost effectiveness of asthma management strategies. Pharmacoeconomics. 2002;20(3):183–94.CrossRefPubMedGoogle Scholar
  77. 77.
    Rutten-van Mölken MMH, Oostenbrink J, Miravitlles M, Monz B. Modelling the 5-year cost effectiveness of tiotropium, salmeterol and ipratropium for the treatment of chronic obstructive pulmonary disease in Spain. Eur J Health Econ. 2007;8(2):123–35. doi:10.1007/s10198-007-0039-4.CrossRefPubMedPubMedCentralGoogle Scholar
  78. 78.
    Pharmaceutical Services Negotiating Committee. Evaluation of Evidence Provided by PharmOutcomes New Medicines Service Data. 2013.Google Scholar
  79. 79.
    British Medical Association, Royal Pharmaceutical Society of Great Britain. British National Formulary 66th Edition. September (66) ed. London: BMJ Group and RPS Publishing. 2013.Google Scholar
  80. 80.
    Lindgren P, Buxton M, Kahan T, Poulter NR, Dahlöf B, Sever PS, et al. Economic evaluation of ASCOT-BPLA: antihypertensive treatment with an amlodipine-based regimen is cost effective compared with an atenolol-based regimen. Heart. 2008;94(2):e4. doi:10.1136/hrt.2007.127217.CrossRefPubMedGoogle Scholar
  81. 81.
    Department of Health. HSHC 2013. Health service cost index, annual summaries. 2013. http://www.info.doh.gov.uk/doh/finman.nsf/Newsletters?OpenView&Start=13.1&ExpandView. Accessed 10 June 2015.
  82. 82.
    Briggs AH, Bousquet J, Wallace MV, Busse WW, Clark TJH, Pedersen SE, et al. Cost-effectiveness of asthma control: an economic appraisal of the GOAL study. Allergy. 2006;61(5):531–6. doi:10.1111/j.1398-9995.2006.01038.x.CrossRefPubMedGoogle Scholar
  83. 83.
    Lloyd A, Price D, Brown R. The impact of asthma exacerbations on health-related quality of life in moderate to severe asthma patients in the UK. Prim Care Respir J. 2007;16(1):22–7.CrossRefPubMedGoogle Scholar
  84. 84.
    Rutten-van Mölken MPMH, Oostenbrink JB, Tashkin DP, Burkhart D, Monz BU. Does quality of life of copd patients as measured by the generic euroqol five-dimension questionnaire differentiate between copd severity stages? Chest. 2006;130(4):1117–28. doi:10.1378/chest.130.4.1117.
  85. 85.
    Dolan P, Gudex C, Kind P, Williams A. A social tariff for EuroQol: results from a UK general population survey: Centre for Health Economics, University of York; 1995.Google Scholar
  86. 86.
    Hertel N, Kotchie RW, Samyshkin Y, Radford M, Humphreys S, Jameson K. Cost-effectiveness of available treatment options for patients suffering from severe COPD in the UK: a fully incremental analysis. Int J Chronic Obstr Pulm Dis. 2012;7:183–99. doi:10.2147/copd.s29820.CrossRefGoogle Scholar
  87. 87.
    Hettle R, Wouters H, Ayres J, Gani R, Kelly S, Lion M, et al. Cost-utility analysis of tiotropium versus usual care in patients with COPD in the UK and Belgium. Respir Med. 2012;106(12):1722–33. doi:10.1016/j.rmed.2012.09.006.CrossRefPubMedGoogle Scholar
  88. 88.
    Diabetes Trial Unit. UKPDS Outcomes Model. University of Oxford. 2013. http://www.dtu.ox.ac.uk/outcomesmodel/. Accessed 10 June 2015.
  89. 89.
    Clarke P, Gray A, Legood R, Briggs A, Holman R. The impact of diabetes-related complications on healthcare costs: Results from the United Kingdom Prospective Diabetes Study (UKPDS Study No. 65). Diabet Med. 2003;20(6):442–50. doi:10.1046/j.1464-5491.2003.00972.x.CrossRefPubMedGoogle Scholar
  90. 90.
    Pharmaceutical Services Negotiating Committee. NMS statistics. http://psnc.org.uk/funding-and-statistics/nhs-statistics/nms-statistics/. Accessed 21 Jan 2017.

Copyright information

© The Author(s) 2017

Open AccessThis article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors and Affiliations

  • Rachel A. Elliott
    • 1
  • Lukasz Tanajewski
    • 2
  • Georgios Gkountouras
    • 2
  • Anthony J. Avery
    • 3
  • Nick Barber
    • 4
  • Rajnikant Mehta
    • 5
  • Matthew J. Boyd
    • 2
  • Asam Latif
    • 6
  • Antony Chuter
    • 7
  • Justin Waring
    • 8
  1. 1.Manchester Centre for Health Economics, Room 4.318, 4th floor, Jean Mcfarlane Building, Division of Population Health, Health Services Research and Primary Care, School of Health Sciences, Faculty of Biology, Medicine and HealthThe University of ManchesterManchesterUK
  2. 2.Division of Pharmacy Practice and Policy, The School of PharmacyUniversity of NottinghamNottinghamUK
  3. 3.Primary Care Research, Division of Primary Care, School of Medicine, Queen’s Medical CentreUniversity of NottinghamNottinghamUK
  4. 4.Emeritus Professor of PharmacyUCL School of PharmacyLondonUK
  5. 5.Research Design Service, East Midlands (RDS EM), School of Medicine, Queen’s Medical CentreUniversity of NottinghamNottinghamUK
  6. 6.School of Health Sciences, Faculty of Medicine and Health Sciences, Queen’s Medical CentreUniversity of NottinghamNottinghamUK
  7. 7.Patient and Public RepresentativeHaywards HeathUK
  8. 8.Organisational Sociology and Improvement Science, Centre for Health Innovation, Leadership and LearningNottingham University Business School, Jubilee Campus, University of NottinghamNottinghamUK

Personalised recommendations