bateman eq5 d_for uk rasch user group 2014
DESCRIPTION
Presented at the UK Rasch User Group these slides include the in my view, interesting evidence of DIF by clinical setting that I think actually contributes to the validity - but this is debatable. Ie note the curve for pain - much more readily endorsed than among rehab patients. Anyway I'll be interested in your feedback & comments.TRANSCRIPT
Andrew Bateman PhD MCSPOliver Zangwill Centre for Neuropsychological Rehabilitation
Clinical Lead for NeuroRehab in CCS
Affiliated Lecturer, Dept of Psychiatry, University of Cambridge
Health Education East of England Quality Improvement Fellow
Clinical Use of the Euroqol EQ5D-5L in Community Rehabilitation and Musculoskeletal Physiotherapy services:
Item ordering, Item Bias and Disordered Thresholds
outline
Overview of eq5d-5l – what it is an why I’m using it
My experiences and some findings from our dataset
Use of RMSEA
Reasons for selecting EQ5D a)it was in the OF, b)NHS England also still promotingc)Chartered Soc Physio push
CGs now responsible for commissioningservices, to include outcomes in contract?
How to collate data?How to analyse data?
“Overarching indicator 2.21 Domain 2 seeks to capture how successfully the NHS is supporting people with long-term conditions to live as normal a life as possible. The overarching indicator ‘health-related quality of life for people with long-term conditions’ allows the Secretary of State for Health to understand whether health-related quality of life is improving over time for the population with long-term conditions”.
The EQ5D-5L
Describes a health state “today”
Health related quality of life
& favoured tool inHealth economics
Five questions plus
an overall health“thermometer”
UK
11111 1.000
11112 0.879
11113 0.848
11114 0.635
11115 0.414
11121 0.837
11122 0.768
11123 0.750
11124 0.537
11125 0.316
11131 0.796
11132 0.740
11133 0.725
11134 0.512
11135 0.291
11141 0.584
11142 0.527
11143 0.513
A list of 3127 values is available.
The values are set on a scale where 1 = full health through to 0 = death(minus values are states worse than death e.g. coma or severe intractable pain)
So a Health State 11111 = 1 (Full health)
55555 = -0.594 (State worse than death)
The previous example: 2 3 4 2 1 = 0.516(or as a percentage of full health 51.6%)
Using the EQ-5D-5L Value Sets
Portion of EQ-5D-5L Value Set
Demographics of the 3293 Patients incuded in analysis:Group (17-60) (61-80) (80+)REHAB 299 882 725MATS 561 279 44MSK 408 153 14
Gender %MREHAB 44 41 34MATS 45 (ALL AGES)MSK 38 (ALL AGES)
rehab
MSK patients
item threshold plot illustrates different ordering of item difficulties in Rehab and MSK serviceThe order of item difficulties reflect the priorities of community rehabilitation –ie people seeking support to return to Usual Activities and improve Mobility
Rehab person item thresholds: 94% targettingMean scores per age group are sig different (F=20.12, p<.001), Younger group reporting worse “qol”
MATS N=902, 98% targetting, Age differences, eldest ANOVA F=3.86, p=0.02Oldest group (small n)reporting worst qol,
RMSEA a supplementary statistic
http://www.rasch.org/rmt/rmt254d.htmIts formula can be shown to equal:
for large sample size, RMSEA is used as a supplementary fit statistic, where a value of <0.02 is indicative of fit" [Tennant ibid]
√[(χ2/df - 1) /(N - 1)] 0.0422 REHAB (n=1883)
0.0424 MATS (n=884)
0.0361 MSK (n=576)
0.0539 ALL (n=3293)
0.0529 all (un rescored n= 3331)
Future analysesLocality differences, link to staffing density?Comparison between health conditionsRepeated measurement data calibrated for
age effectsAutomate analysis into performance
dashboardHealth economics (but “there’s no F in
utility”?)Refs http://www.scoop.it/t/eq-5d
So problem #1 is that we haven’t really sorted out NHS outcome
measurement (yet)We agree about importance of proms & prems Importance of data collation system in CCSEMPHASISED in the NHS Outcomes Framework
but analysis approach not establishedNeed for ongoing collaborations with
psychometric and statistical colleagues
Conclusion
Some problems (with bias and thresholds) but good targetting
Generic PROMS like EQ5D useful for thinking about prioritising services to meet needs of patients
Collaboration for Leadership in Applied Health Research and Care
Thank you for your attention!
When shall we meet next?
OZC training events – see flyer – get on distribution list
Keep in touch!Twitter @ozcboss
LinkedInScoop.it