Valuing the EQ-5D-Y Principles, methods and current research
- 1. Nancy Devlin & Koonal Shah, OHE Valuing Youth Health Methodological issues and current state of research on EQ-5D-Y 2nd EuroQol Academy Meeting Noordwijk 8 March 2017 Valuing the EQ-5D-Y Principles, methods and current research
- 2. 2nd EuroQol Academy Meeting March 2017 2 On behalf of the project teams • Juanma Ramos-Goni • Oliver Rivero-Arias • Simone Kreimeier • Yan Feng • Ben van Hout • Koonal Shah • Nance Devlin
- 3. 2nd EuroQol Academy Meeting March 2017 3 EQ-5D-Y: state of play • EQ-5D-Y published in 2010 (Wille et al; Ravens-Sieberer et al) • By 2015, 40 language versions available; use modest but growing: 34 requests for use in 2015 (cf. 2274 for the 5L) • Demand for use in HTA, but no value sets to support that • Kind et al (2015): VAS valuation of EQ-5D-Y from 3 perspectives: (a) own (b) hypothetical adult (c) hypothetical child. Lower values for (c) than (a) or (b), but no simple relationship between (c) and 3L value sets • Kreimeier et al (2015): DCE and TTO valuation of 3L and Y. Complex interaction between perspective and wording of descriptive system.
- 4. 2nd EuroQol Academy Meeting March 2017 4 So, EQ-5D-Y are we waiting? • Its complicated! • Normative issues (whose preferences should we elicit?) • Perspective issues (whose health states should we elicit the preferences for?) • Wording issues: The Y is a different instrument to the 3L – but can we find a linking function? • Methods issues: VAS? DCE? TTO? Something else? On what basis do we make this choice? – VAS values lower for Y than A – TTO values: reluctance to sacrifice life years may mean values higher for Y than A ‘biases’ values up
- 5. 2nd EuroQol Academy Meeting March 2017 5 Principles • Value sets for the EQ-5D-Y to be based on the stated preferences of the general public (not those of the sub- group whose health is being evaluated) • Reflects HTA concern with the broad allocation of resources across an entire population (2nd Washington Panel, 2017) • Taxpayers and potential patients • eg. NICE: “The valuation of HRQoL… should be based on a valuation of public preferences from a representative sample of the UK population” • But it may still be relevant to know about the preferences of children (as patients) (Versteegh & Brouwer 2016) • And children’s preferences may be relevant in other (non-HTA) uses of the instrument
- 6. 2nd EuroQol Academy Meeting March 2017 6 Strategy re: choice of methods • Value sets required urgently • Concern about TTO • Concern about acceptability of VAS for use of value sets in HTA eg. NICE (2013): values should be ‘trade-off based’ • Pilot: DCE ‘works’ & has merit of being quick and relatively inexpensive; but lacks anchors Therefore, the strategy adopted is: • Undertake DCE studies to rapidly obtain latent scale values • Interim anchoring using EQ-5D-3L values for 33333 • In parallel, test a range of methods for anchoring latent scale values for the EQ-5D-Y at 0 = dead
- 7. 2nd EuroQol Academy Meeting March 2017 7 Anchoring at dead: experimental approaches (a) VAS (b) DCE with duration (DCEd) (c) TTO, adapted to avoid the reluctance to trade. • The scenario to be valued is period of ill health in childhood, followed by a length of time in full health in ‘adulthood’ • It is equivalent to a ‘lag-time TTO’ (Devlin et al 2013) plus a transition from child to adult. (d) Identifying the location of dead within the descriptive system • utilises the approach developed within the Personal Utility Function work (Devlin et al 2016 Plenary paper) Note: whether a state is better or worse than dead depends on its duration. All methods state duration; in (a), (c) and (d) this is fixed at 10 years.
- 8. 2nd EuroQol Academy Meeting March 2017 8 Administered via internet survey • Background/screening questions • Self-reported health using EQ-5D-Y • Instructions (emphasising that that the states describe problems being experienced by a 10 year old child) • 15 paired comparison tasks (DCE experimental design) • 1 fixed paired comparison task (with dominant option) • Debrief questions (standard EQ-VT statements) • Further debrief question (specific to adult vs. child issue) • Further background questions Latent scale DCE study
- 9. 2nd EuroQol Academy Meeting March 2017 9 Latent scale DCE design Two-step approach 1. Initial design for soft launch data collection a) Prepared candidates with overlap in two dimensions; no dominant pairs; no repetitions; level balance cut-off b) Simulated 1,000 designs each including 150 pairs; using D-efficiency measure based on main effects model + all 2-way interactions, best designs were kept c) Blocked design into 10 blocks (15 tasks each) by minimising the variance of the level balance between blocks 2. Updated Bayesian design based on priors from the soft launch data a) Used the same set of candidates as in step 1a b) Simulated 1,000 designs each including 150 pairs’ stored both D-error estimates - one related to a main effects model and one related to main effects + all 2-way interactions c) Plotted results and selected on this basis (see figure) d) Blocked design into 10 blocks (15 tasks each) by minimising the variance of the level balance between blocks
- 10. 2nd EuroQol Academy Meeting March 2017 10 DCE task screenshot
- 11. 2nd EuroQol Academy Meeting March 2017 11 22233 • Some problems walking about • Some problems washing or dressing • Some problems doing usual activities • A lot of pain or discomfort • Very worried, sad or unhappy Fixed pair 11122 • No problems walking about • No problems washing or dressing • No problems doing usual activities • Some pain or discomfort • A bit worried, sad or unhappy
- 12. 2nd EuroQol Academy Meeting March 2017 12 Q: Do you think your choices would have been different if you had been asked to imagine that the health problems were being experienced by you, rather than a 10 year old child? o Yes – at least some of my choices would have been different o No – my choices would have been exactly the same o Don’t know Adult/child debrief question
- 13. 2nd EuroQol Academy Meeting March 2017 13 • UK public sample (n=1,000; representative in terms of selected observable characteristics) • Initial data collection – early February • Design updated based on data from n=127 • Main launch – early March • Data collection now complete Data collection
- 14. 2nd EuroQol Academy Meeting March 2017 14 Administered via specially adapted version of EQ-VT • Self-reported health using EQ-5D-3L • Standard background questions • Ranking task • VAS tasks • Lag-time TTO tasks • DCE with duration tasks • ‘Location of dead’ tasks based on PUF approach • Debrief questions • Further background questions Anchoring study
- 15. 2nd EuroQol Academy Meeting March 2017 15 • Respondents answer questions both about their own health (using EQ-5D-3L), and about the health of a 10 year old child (using EQ-5D-Y) • Two-arm design: random half of sample starts with adult questions and proceeds to child questions; ordering reversed for the other half Anchoring study
- 16. 2nd EuroQol Academy Meeting March 2017 16 • Rating of 33333… VAS
- 17. 2nd EuroQol Academy Meeting March 2017 17 • …followed by rating of ‘Dead’ VAS
- 18. 2nd EuroQol Academy Meeting March 2017 18 • Valuation of 22222… Lag-time TTO
- 19. 2nd EuroQol Academy Meeting March 2017 19 • …followed by 33333 Lag-time TTO
- 20. 2nd EuroQol Academy Meeting March 2017 20 • Seven paired comparisons DCE with duration
- 21. 2nd EuroQol Academy Meeting March 2017 21 Anchoring study DCE design Six-step approach 1. Prepared a set of candidates with overlap in two dimensions; no dominant pairs; no repetitions; level balance cut-off 2. Simulated 2,000 designs each including 42 pairs; Using D-efficiency measure based on main effects model, all pairs from best 20 designs were extracted 3. Based on priors of from the latent scale DCE soft launch, choice probabilities for the pairs from step 2 were estimated 4. Using these estimated probabilities, pairs were organised into categories and durations were attached to states accordingly (e.g. longer durations for states with lower probabilities of being chosen) 5. Based on Bansback et al. model, where the time is an interaction, 2,000 designs were simulated; best was selected based on D-efficiency measure 6. Blocked the design into 6 blocks (7 tasks each) by minimising the variance of the level balance between blocks
- 22. 2nd EuroQol Academy Meeting March 2017 22 • Preliminary ranking exercise Location of dead tasks
- 23. 2nd EuroQol Academy Meeting March 2017 23 Location of dead tasks
- 24. 2nd EuroQol Academy Meeting March 2017 24 • Up to five paired comparison tasks Location of dead tasks
- 25. 2nd EuroQol Academy Meeting March 2017 25 • If respondent chooses B, a new (‘better’) health state is presented in option A • If respondent chooses A, a new (‘worse’) health state is presented in option A • Selection of states determined by respondent’s earlier response to ranking task • Five choices in total, unless respondent considers 33333 to be better than dead Location of dead tasks
- 26. 2nd EuroQol Academy Meeting March 2017 26 Location of dead tasks
- 27. 2nd EuroQol Academy Meeting March 2017 27 Debrief questions
- 28. 2nd EuroQol Academy Meeting March 2017 28 Debrief questions (2)
- 29. 2nd EuroQol Academy Meeting March 2017 29 • EQ-VT testing in progress – nearly ready • Pilot / cognitive interviews (n=10) to take place in March/April • Interviews undertaken by moderators with qualitative research backgrounds • Debrief questions designed to probe how respondents interpret the tasks, where they struggle and how the survey could be improved • Further adaptation of the EQ-VT and interview materials may be required depending on findings from the pilot • Main stage interviews (n=300) to take place in May/June Data collection
- 30. 2nd EuroQol Academy Meeting March 2017 30 • 10 year old child • Which do you prefer? • Slight adjustment to SC and UA labels Some framing decisions
- 31. 2nd EuroQol Academy Meeting March 2017 31 To enquire about additional information and analyses, please contact Professor Nancy Devlin at ndevlin@ohe.org To keep up with the latest news and research, subscribe to our blog, OHE News Follow us on Twitter @OHENews, LinkedIn and SlideShare Office of Health Economics (OHE) Southside, 7th Floor 105 Victoria Street London SW1E 6QT United Kingdom +44 20 7747 8850 www.ohe.org OHE’s publications may be downloaded free of charge from our website. Thank you for listening