Objectives Introduce natriuretic peptide (NP) case study Data - PDF document

Economic Evaluation of Health and Social Care Interventions Policy Research Unit Value of Implementation: Getting cost-effective technologies into practice CASE STUDY 2: NP testing for diagnosing suspected heart failure Sophie Whyte Rita Faria Stephen Palmer Simon Dixo n Simon Walker Mark Sculpher Objectives • Introduce natriuretic peptide (NP) case study • Data availability issues • Methods applied: • Diffusion curves • Multi-period analysis • Challenges for value of implementation 2 1

Background • In 2010 NICE released a clinical guideline recommending that natriuretic peptide (NP) testing in patients with suspected heart failure without previous myocardial infarction can accelerate diagnosis of heart failure and also avoid unnecessary echocardiography. • The framework for the evaluation of the value of implementation activities is applied to this recommendation for NP testing. 3 Data availability issues Cost effectiveness data The economics of the diagnostic section of the NICE CG108 was informed by the HTA report by Mant et al 2009. Model had to be adapted for this project as it used a pathway based on MICE score rather than history of MI as used in CG108. The HTA used wrong comparator (‘do nothing’ rather than ECG) hence CE is likely to be overestimated; hence results will overestimate the value of implementation activities. Utilisation data: Derived from NHS Atlas of variation in diagnostic services. Maximum utilisation rate was assumed to be 8.6 tests per 1,000 practice population per year (the 4 level of utilisation achieved in over 10% of PCTs). 2

Data availability issues Implementation interventions: It was not possible to estimate the effectiveness of the London BNP initiative as utilisation data was only available from two trusts. Hence estimates of effect size from the O’Brien et al study were used (5% base case, 9% scenario analysis). Population size: Data on incidence of heart failure vary considerably between sources. (e.g. General Practise Research Database (GPRD) data 22,542 cases; The Hillingdon Heart Failure Study 59,000 cases; Sheffield heart failure clinic 55,000 cases; A clinical expert estimates 70,000 new cases; Data from Hospital Episode Statistics combined with clinical opinion suggests 23,000- 40,000 cases.) Population with suspected heart failure was then estimated using a prevalence of HF in persons presenting with suspected HF of 33%. 5 Methods: Diffusion curves Diffusion curves were estimated based on historic data to produce predictions of future utilisation using the following approach: • A S shaped curve (of the form f(t)=1/(1_exp(-at+k))) was fitted to two data points for 2010 and 2012. • Utilisation in 2012 was based on estimates from the Diagnostics Atlas data. • Utilisation in 2010 was based on the NICE costing template expert clinical opinion (without the NICE CG108 approximately 30% of patients currently receive a BNP or NTproBNP test and approximately 90% currently receive an ECG). • Assumption: for these 30% utilisation was 50% of optimal maximum utilisation i.e. 15%. 6 3

Methods: Diffusion curves 1.0 0.9 0.8 0.7 0.6 Utilisation 0.5 0.4 Utilisation at 25% 2010, and 63% 2012 0.3 Utilisation at 25% 2010, and 51% 2012 0.2 Utilisation at 15% 2010, 0.1 and 51% 2012 - 2010 2012 2014 2016 2018 2020 Time (years) 7 Methods: Diffusion curves Predicted diffusion of utilisation with and without an implementation intervention 1.0 0.9 0.8 0.7 Utilisation 0.6 0.5 0.4 Utilisation at 15% 2010, and 51% 2012 0.3 Utilisation at 15% 2010, and 51% 2012 + intervention in 2012(+5%) 0.2 Utilisation at 15% 2010, and 51% 2012 0.1 + intervention in 2012(+9%) - 2010 2012 2014 2016 2018 2020 Time (years) 8 4

Methods: Multiperiod ana • In addition to a static population analysis a multi-period analysis was undertaken. • The multi-period analysis calculated the total value of the implementation activity over a 10 year period. • Future costs and QALYs accrued were discounted at a rate of 3.5% in line with the NICE methods guide. • This analysis assumes that the increase in utilisation of the intervention does not change the population size. • The population presenting with suspected HF was assumed to increase at a rate of 4.2% per annum, based on Hospital Episode Statistics data. 9 Methods: Multiperiod analy Multi-period analysis: Predictions for 10 years, including discounting, WTP=£20000 Population Predicted Predicted Expected Value Expected Value Value of presenting utilisation utilisation Perfect Current Year of Actual of Perfect Implementation with without with Implementation value Implementation Implementation Activity suspected HF* intervention intervention 2012 210,000 51% 56% 100% £163.2 £179.2 £320.1 £16.0 2013 218,820 72% 77% 100% £230.9 £247.0 £322.2 £16.1 2014 228,010 86% 91% 100% £279.0 £295.2 £324.4 £16.2 2015 237,587 94% 99% 100% £306.1 £322.4 £326.6 £16.3 2016 247,566 97% 100% 100% £320.0 £328.8 £328.8 £8.8 2017 257,963 99% 100% 100% £327.3 £331.1 £331.1 £3.7 2018 268,798 100% 100% 100% £331.7 £333.3 £333.3 £1.6 2019 280,087 100% 100% 100% £334.9 £335.5 £335.5 £0.6 2020 291,851 100% 100% 100% £337.5 £337.8 £337.8 £0.3 2021 304,109 100% 100% 100% £340.0 £340.1 £340.1 £0.1 TOTAL £2,971 £3,051 £3,300 £80 All costs are presented in £millions, all costs are discounted at annual rate of 3.5% from 2013 onwards *Assumes HF incidence increasing by 4.2% per annum 10 5

Methods: Multiperiod an 11 Multi-period analysis: predictions for 10 years, including discounting, WTP=£20,000 300,000 £50 £45 250,000 £40 demonstrates how £35 £millions discounted the value of the 200,000 £30 implementation Population activity accrues 150,000 £25 over a 10 year £20 period 100,000 £15 £10 50,000 £5 - £0 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Year Population presenting with suspected HF (assumes HF incidence increasing by 4.2% per annum) Predicted utilisation without intervention Predicted utilisation with intervention Value of Implementation Activity Challenges for value of implementation • What can be done when incremental costs and QALYs related to clinical guidelines compared to current care are not available? • What are the challenges of getting routine data on diffusion? • Does diffusion theory apply to healthcare technologies? How can we predict future utilisation? • How to determine the optimal utilisation rate? • How to incorporate the effectiveness of implementation activities in the diffusion curve? • What is the value of doing multiperiod analysis? What are the challenges? 12 6