Making Value Based Pricing A Reality: Issue Panel Moderator: Meindert Boysen Panelists: John Brazier, Roberta Ara and Werner Brower ISPOR 16 th Annual European Congress 2 6 November 2013, The Convention Centre in Dublin, Eire
Value based pricing: wider considerations There is a basic NHS cost per QALY threshold Costs and QALYs (through weighting) to take into account: diseases with greater burden of illness as reflected in QALY loss from a condition greater therapeutic innovation and improvement (size of QALY gain) wider societal benefits (e.g. productivity and carer time) Basic threshold adjusted to reflect the opportunity cost of displaced activities weighted using same methods Price negotiated on the basis of the cost per weighted QALY compared to the new threshold (from 2014) Comparing new and displaced treatments in VBP: Expression as an adjusted cost per QALY threshold Adjustment to c/q threshold: New drug Other use (?) 25,000 * 1+ 30%+ 0.1 = 24,138 1+ 20%+ 0.25 X OR Cost:.. 50k ( 50k displaced) Cost / QALY:.. 25k (measured ICER) 25k (centre of threshold range) ->QALYs gained:.. 2 2 lost BoI weight: +30% +20% ->Weighted QALYs: 2.6 2.4 WSBs, :. 12,000 30,000 ->WSBs, QALYs: 0.2 worth 0.5 worth -> Total Benefits: 2.8 QALYs worth gained 2.9 QALYs worth lost = < X
Elicitation of societal preferences for Burden of Illness, Therapeutic Improvement and End of Life from a UK online panel John Brazier DH PRU in Economic Evaluation of Health and Care Interventions (EEPRU), University of Sheffield Donna Rowen, Clara Mukuria, Sophie Whyte, Anju Keetharuth, Aki Tsuchiya, Phil Shackley Health Economics and Decision Science, ScHARR, University of Sheffield Arne Risa Hole Economics Department, University of Sheffield Acknowledgements: Angela Robinson (University of East Anglia) and Gavin Roberts (DH) Outline of presentation Value based pricing: BOI, TI and EOL Methods Main results Weights for use in DH framework Discussion
Elicitation of societal preferences Discrete choice experiment (DCE) survey using online UK panel to elicit societal preferences for: Burden of illness (QALY loss from condition) Therapeutic improvement (size of QALY gain from treatment) End of life (e.g. NICE weights QALY gain more where expected survival is 24 months and survival gain 3 months or more) 100% Conceptual framework Normal population Health Dead Today Life expectancy from today Normal life expectancy
100% Conceptual framework Normal population Health Health without treatment Dead Today Without treatment Life expectancy without treatment Life expectancy from today Normal life expectancy 100% Conceptual framework Normal population Health gain Health Health without treatment Dead Today Treatment gain Without treatment Life expectancy without treatment Survival gain Life expectancy from today Normal life expectancy
Main survey design Internet panel sample allows for large numbers, collection fast Survey content Introduction video played 2 practice and 10 real DCE questions 9 questions asking general attitudes assessed in survey 17 questions on you and your health and understanding Design 4 normal life expectancies (5, 20, 40, 80 years) Both small and large starting point and gains in health and survival 580 pairs selected using D efficient design. Impossible scenarios not included 58 card blocs in total across 4 normal life expectancies 1
FEEDBACK Modelling U=f(QALY gain, QALY gain squared, EOL or BOI) Estimation by conditional logit regression model Dependent variable = Choice patient group A or patient group B Estimated for pooled data and each of the 4 separate normal life expectancies Basic additive model: V = β 1 QALY + β 2 QALY 2 + β 3 BOI (or EOL) Where a positive β 2 would suggest TI
Marginal rate of substitution The marginal rate of substitution between BOI and QALY (or EOL and QALY) provides a measure of the weight of BOI in terms of QALY gain equivalents e.g. MRS 1 = β 3 /β 1 MRS 2 = β 3 /(β 1 + 2*β 2 QALY) So MRS 2 varies by size of QALY Sample Main results (1) 3669 respondents (55% response rate) Similar age, but more females and unemployed respondents and less healthy than UK norm Practice questions PQ1 Majority chose larger QALY gain (90.7 92.5%) PQ2 No evidence of preference for higher BOI (46.8% 54.3%)
Regression results VARIABLES All 5 yrs 20 yrs 40 yrs 80 yrs QALY 0.276*** 3.641*** 0.751*** 0.404*** 0.171*** QALY_sq 0.004*** 0.709*** 0.037*** 0.014*** 0.002*** BOI 0.017*** 0.12*** 0.000 0.039*** 0.005** VARIABLES All 5 yrs 20 yrs 40 yrs 80 yrs QALY 0.281*** 3.229*** 0.761*** 0.400*** 0.175*** QALY_sq 0.004*** 0.602*** 0.037*** 0.014*** 0.002*** EOL 0.609*** 0.607*** 0.375*** 0.576*** 0.314*** Regression results: Overview of results QALYs matter but at a decreasing rate no support for TI BOI matters but is weak and inconsistent EOL is significant Coefficients change for different variants of normal life expectancy
Weights for BOI Model (1): Assuming the value of a QALY is constant MRS (1) of 1 more unit of BOI is 0.040 QALYs Model (2) Allowing value of a QALY to vary Warning: This is additive and not proportionate to the size of QALY gain QALY gain MRS (2) 0.05 0.063 0.1 0.063 0.5 0.063 1 0.064 2 0.066 5 0.073 10 0.087 20 0.141 Limitations Limited range of characteristics (e.g. no age) Online data collection Additive design Robustness many respondents may have continued to make the mistake of assuming the profiles were for them even after feedback Identified respondents who chose a profile with smaller QALY gain and lower BOI but larger number of lifetime QALYs Once these were excluded (n=2247) then BOI coefficients were all positive, significant and larger than for the whole sample Weights choice of variant and specification
DH model: impact of BOI on ICER Code Disease (pharma ICDs, n/a's deleted). Cancer in blue. BoI BoI weight (QALY loss (QALY loss x pp) tariff) INPUTS Cost per QALY displaced 25,000 C22 Liver cancer 10.70 54% 0.00 35,926 Value of a QALY 60,000 C25 Pancreatic cancer 9.97 50% 0.00 35,073 BoI tariff, % add per QALY of loss 5% C34 Lung cancer 9.68 48% 0.00 34,729 Weighting on WSBs F20 Schizophrenia 7.62 38% 0.00 32,317 G35 Multiple sclerosis 6.18 31% 0.00 30,630 BoI for displaced QALY (QALYs lost pp) 1.36 C92 Myeloid leukaemia 6.15 31% 0.00 30,600 WSBs of displaced QALY (QALYs' worth) 0.07 G20 Parkinson's disease 4.60 23% 0.00 28,782 copied from Template Calc by ICD v5.2 C90 Myeloma 4.45 22% 0.00 28,617 J43 Emphysema and COPD 3.80 19% 0.00 27,849 BoI weighting on displaced QALY 7% C64 Kidney cancer 3.75 19% 0.00 27,794 WSBs per QALY for displaced QALY 0.00 F30 Depression 3.63 18% 0.00 27,656 M05 Rheumatoid arthritis 2.83 14% 0.00 26,710 Total weighting for displaced QALY 0.07 E11 Diabetes 2.68 13% 0.00 26,539 J45 Asthma 1.86 9% 0.00 25,581 G30 Alzheimer's disease 1.68 8% 0.00 25,367 F03 Dementia 1.68 8% 0.00 25,367 displ (average displaced QALY) 1.36 7% 0.00 25,000 G40 Epilepsy 1.32 7% 0.00 24,953 C18 Colon cancer 1.28 6% 0.00 24,902 I26 Embolisms, fibrillation, thrombosis 1.16 6% 0.00 24,760 C61 Prostate cancer 1.06 5% 0.00 24,639 I21 Acute myocardial infarction 1.00 5% 0.00 24,571 I64 Stroke 0.83 4% 0.00 24,371 NB, figures for whole ICD population. Actual populations for e.g. NICE cancer drugs likely to have much higher BoI maybe 10 30 lost QALYs WSBs Adjusted threshold C53 Cervical cancer 0.60 3% 0.00 24,102 C50 Breast cancer 0.55 3% 0.00 24,050 A40 Streptococcal septicaemia 0.38 2% 0.00 23,843 J30 Allergic rhinitis 0.30 1% 0.00 23,751 M81 Osteoporosis 0.28 1% 0.00 23,730 K50 Irritable Bowel Syndrome 0.26 1% 0.00 23,706 J10 Influenza 0.19 1% 0.00 23,628 L40 Psoriasis 0.19 1% 0.00 23,628 E66 Obesity 0.18 1% 0.00 23,619 M45 Ankylosing spondylitis 0.11 1% 0.00 23,527 zero (zero BoI, WSB condition) 0.00 0% 0.00 23,404 Capturing Wider Social Benefits to inform VBP approach Roberta Ara DH PRU in Economic Evaluation of Health and Care Interventions (EEPRU), University of Sheffield John Brazier, Simon Dixon, Donna Rowen, Monica Hernandez, Ben Kearns, Ben van Hout Health Economics and Decision Science, ScHARR, University of Sheffield Acknowledgements: Gavin Roberts (DH) www.eepru.org.uk
Wider social benefits; looking at measurement & valuation of Formal care (FC) Informal care (IC) Productivity effects (PE) Examine how FC, IC & PE vary by EQ 5D & ICD Estimate informal care effects using EQ 5D and ICD 10 N HODaR >59.5k Patient characteristics: General population Recently discharged from hospital Informal care: Number of days received informal care in last 6 weeks Quality of life: EQ 5D Quality of life today Health status: ICD 10 Used first digit (chapter)
Days needed informal care in the last 6 weeks by EQ 5D score EQ 5D score Mean days (s.d.) N 1 1.46 ( 5.39) 13,268 0.75 EQ 5D <1 3.84 ( 9.43) 13,128 0.5 EQ 5D <0.75 10.87 (15.64) 21,515 0.25 EQ 5D <0.5 19.15 (18.46) 2,374 0 EQ 5D <0.25 19.95 (18.45) 4,940 0.25 EQ 5D <0 27.65 (17.05) 4,106 EQ 5D < 0.25 32.90 (16.54) 181 Number of days needed informal care EQ 5D for sample
Days needed informal care in the last 6 weeks Days Observed Frequency Observed (%) 0 33272 55.91 55.55 1 7 8887 14.94 15.64 8 14 3615 6.07 10.14 15 21 2388 4.01 6.32 22 28 971 1.64 4.00 29 35 1542 2.60 2.60 36 41 592 1.01 1.51 42 8245 13.85 4.26 All 59512 100 100 Explanatory variables: ICD, EQ 5D, age, gender Predicted, Zero inflated negative binomial, constant inflation (%) Days needed informal care in the last 6 weeks Number of days Age EQ 5D Female Male Respiratory (J) 50 1 1.70 1.07 55 0.6 7.52 5.57 60 0.5 10.11 7.92 65 0.3 15.90 13.88 Musculoskeletal (C) 50 1 1.89 1.19 55 0.6 8.37 6.21 60 0.5 11.25 8.82 65 0.3 17.70 15.45
Estimate mean days off work sick using EQ 5D and ICD 10 HODaR Understanding Society N >30k >27k Survey design prospective X sectional Patient characteristics: General population discharged within previous 6 weeks Health status: ICD 10 self reported 17 broad definitions Quality of life & recall period: EQ 5D today SF36 (or SF12) previous 4 weeks previous 4 weeks Absence from paid employment due to ill health & recall period: Off work sick previous 6 weeks previous week number of days off yes/no for absence Method STEP 1: HODaR E(SF6D) = f(eq5d, age, gender, ICD group) STEP 2: Understanding Society p(ows) = f(sf6d, age, gender) STEP 3: HODaR E(DOWS) = f(age, gender, ICD group) STEP 4: Calculate Weight E(DOWS) using p(ows)
Relationship between EQ 5D and SF 6D (Step 1, HODAR) Relationship between QoL & absence from paid employment due to ill health (Step 2, U.Society)
Number days off work sick by EQ 5D & ICD Health condition Age Sex EQ 5D P(OWS) DOWS OWS Group (ICD code) Disorders glucose regulation 18 M 1 0.27% 0.026 Group 14 (E29) 18 M 0.594 67.12% 6.623 65 M 1 0.51% 0.075 65 M 0.594 79.76% 11.58 Gastrointestinal problems 18 M 1 0.27% 0.031 Group 51 (A04) 18 M 0.594 67.74% 7.712 65 M 1 0.53% 0.089 65 M 0.594 80.21% 13.43 Malignant melanoma breast 18 F 1 0.48% 0.09 Group 04 (C50) 18 F 0.594 78.67% 14.66 65 F 1 0.93% 0.254 65 F 0.594 87.69% 24.037 Limitations Recall periods Double/triple mapping not ideal Switching between datasets not ideal Zero inflated negative binomial? Neither surveys capture total absence due to ill health eg early retirement due to condition, long term sickness etc Ongoing/planned research HIPPO new HODAR Includes both SF12 & EQ5D Recall for work absence is 1 week (and 4 weeks)