Se ha denunciado esta presentación.
Se está descargando tu SlideShare. ×

Rasch analysis of the Dermatology Life Quality Questionnaire (DLQI)


Eche un vistazo a continuación

1 de 24 Anuncio

Más Contenido Relacionado

Similares a Rasch analysis of the Dermatology Life Quality Questionnaire (DLQI) (20)

Más reciente (20)


Rasch analysis of the Dermatology Life Quality Questionnaire (DLQI)

  1. 1. Rasch analysis of the Dermatology Life Quality Questionnaire (DLQI) James Twiss and Stephen McKenna Galen Research Ltd, Manchester, UK Email:
  2. 2. Objectives • To discuss good scale development methodology in the context of Rasch analysis • Example - Dermatology Life Quality Index (DLQI)
  3. 3. The DLQI • 10-item generic dermatology Patient Reported Outcome (PRO) measure • Scored 0-30. High scores=worse • Used with 30 different skin conditions • Available in 55 languages • Used for treatment selection in the UK
  4. 4. DLQI development • Uses patient reports of problems • 49 identified ‘aspects’ condensed into 10 items • Designed to fit on one side of A4 • Items phrased to include additional aspects
  5. 5. The DLQI - content Sample DLQI items: Over the last week… how embarrassed or self conscious have you been because of your skin? how much has your skin interfered with you going shopping or looking after your home or garden? how much has your skin made it difficult for you to do any sport? how much of a problem has the treatment for your skin been, for example by making your home messy, or by taking up time?
  6. 6. The DLQI – response format • Four point response format – Very much – A lot – A little – Not at all • 8 items also have a ‘Not relevant’ option
  7. 7. Psychometric properties • Classical psychometric properties adequate (Basra et al, 2008) • One previous Rasch analysis compared 6 language versions (Nijsten and Meads, 2007) – Overall misfit to the model – Misfit in 3 individual countries – DIF by country for every item
  8. 8. Study aims • Rasch analysis of DLQI data from two patient groups: - atopic dermatitis - psoriasis • Relate results to development methods used for DLQI
  9. 9. Analyses Rasch analysis employed RUMM2020 • Overall fit to model • Individual Item fit • Response threshold order • DIF by age and gender • DIF by disease (AD vs psoriasis) • Item-trait coverage
  10. 10. Sample Psoriasis (n = 146) Atopic Dermatitis (n = 146) Gender (%) Male 73 (50) 73 (50) Female 73 (50) 73 (50) Age (Years) Mean (SD) 44.4 (14.7) 45.5 (16.6) Range 66 (17-83) 62 (20-82) Duration (years) Mean (SD) 20.9 (13.5) 28.2 (17.5) Range 67 (2-69) 76 (0-76)
  11. 11. DLQI scores Psoriasis Atopic dermatitis Mean (SD) 8.8 (6.7) 6.1 (4.6) Median (IQR) 7 (3.0-12.3) 5 (3-8) Range 29 (0-29) 26 (0-26)
  12. 12. Overall fit to the Rasch model Initial fit statistics (partial credit model) Item-trait interaction PSI Items Persons Mean SD Mean SD 0.85 0.98 -0.30 0.81 0.01 -0.81
  13. 13. Item fit Item description Location Fit residual Chi2 p value 1. itchy, sore, painful or stinging -1.61 -0.03 3.4 0.50 2. embarrassment/self consciousness -0.52 -1.97 15.0 3. interferes with shopping/looking after home/garden 0.005* 0.99 -1.18 3.8 0.44 4. influences choice of clothes -0.54 -0.14 4.4 0.35 5. affects social/leisure activities 0.32 12.7 6. affect ability to do sport -0.02 0.29 3.0 0.56 7. prevents working/studying 0.25 0.23 12.6 8. creates problems with partner/close friends/relatives -2.69 0.01 0.01 0.59 -0.89 2.1 0.72 9. causes sexual difficulties 0.89 -1.23 2.2 0.70 10. problems with treatment -0.36 -0.47 3.5 0.48
  14. 14. Problems with items • 2 - Over the last week, how embarrassed or self conscious have you been because of your skin? • 5 - Over the past week, how much has your skin affected any social or leisure activities? • 7 - Over the past week, how much has your skin prevented you working / studying?
  15. 15. DIF analysis Item description Uniform DIF Non-uniform DIF itchy, sore, painful or stinging Disease embarrassment/self consciousness Age group/Gender interferes with shopping/looking after home/garden Disease Disease influences choice of clothes Gender*/ Disease affects social/leisure activities Disease affect ability to do sport Gender Gender prevents working/studying Gender/Disease* creates problems with partner/close friends/relatives causes sexual difficulties problems with treatment Age group*
  16. 16. Example of DIF by disease Prevents working or studying Person locations (logits) Expected value Psoriasis Atopic Dermatitis
  17. 17. Example of DIF by gender Influences the clothes you can wear Expected value Person locations (logits) Females Males
  18. 18. Response thresholds
  19. 19. Response options • Four point response format  Very much  A lot  A little  Not at all Source of problem for items 4, 7 and 8 Source of problem for items 6 and 9
  20. 20. Item Map Clustering of items Poor coverage of persons
  21. 21. Conclusions • DLQI – Pre-Rasch PRO • Rasch analysis highlighted several fundamental problems with DLQI • Problems probably result from inadequate scale development methodology • Concern for: - clinical trials - treatment decisions
  22. 22. Good PRO Design Item Reduction • Based on sound theoretical model • Cognitive debriefing interviews • Patient survey & application of Rasch analysis to data Draft 2 Draft 3 Scale Evaluation • Scaling properties • Classical psychometrics Draft 4 Item Generation • Patient Interviews • Qualitative analysis Draft 1
  23. 23. References Finlay AY, Khan GK (1994). Dermatology Life Quality Index (DLQI)-a simple practical measure for routine clinical use. Clin Exp Dermatol 19: 210-6. Basra MKA, French R, Gatt RM, et al(2008). The dermatology Life Quality Index 1994-2007: a comprehensive review of validation and clinical results. Br J Dermatol 159: 997-1035 Nijsten T, Meads DM, de Korte J et al (2007). Cross-Cultural Inequivalence of Dermatology-Specific Health related Quality of Life Instruments in Psoriasis Patients. Journal of Investigative Dermatology 127: 2315-2322 Nijsten T, Meads DM, McKenna SP (2006). Dimensionality of the dermatology life quality index (DLQI): a commentary. Acta Derm Venereol 86:284-5; author reply 285-6.
  24. 24. Overall fit to the model Item-Trait interaction PSI Unidimensionality (CI) Items Persons Mean SD Mean SD 0.01 0.85 -0.81 0.98 -0.30 0.81 0.03 (0.01 – 0.06)

Notas del editor

  • Aim: to discuss how problems identified during Rasch analysis can be used to illustrate scale development issues
  • One of the most popular HRQoL measures in Dermatology
    NICE guidelines & Health Technology Assessment (HTA)
    Developed in 1994 before the wide use of Rasch analysis in the area of health outcomes
  • Here in lies several of the problems.
    Too much condensing of information. Too many ideas are squeezed into each item. Items should be simple, clear and contain only one idea. This is important from the perspective of Rasch analysis. Failure to follow this produces scales which have built in error.
    No theoretical model
  • The content of the DLQI mainly covers issues related to functioning
    There is one emotional related item
    You can also see from this the items try to cover several areas in one item
  • One may argue that items should not be included if they are not relevant to some of the patients
    A side issue is that individuals who answered not relevant were given the same score as those that answered ‘not at all’
  • Psoriasis sample from the Manchester Psoriasis Service at Hope Hospital
    Atopic Dermatitis sample from the National Eczema society
  • Mean fit residual for item-person interaction was extreme indicating misfit to the Rasch model.
    Shows the degree of divergence between the expected and actual data for each item. Transformed into a Z-score.
    Unidimensionality t-test was fine – suggesting that there was not multiple factors within the scale – Everett Smith (2002)
  • Items 2, 5 and 7 all have some indication of problems
  • Item 2 – Why does this item not fit the Rasch model? – Is it because this is the only emotional item? It is possible that it just doesn’t fit with the other items
    Item 5 – This is a very poorly worded item. The target audience would not use this kind of language. This may not be the reason for the item problem but it certainly doesn’t help. Items should be simple and in a language that is easy for respondents to relate to.
  • It is unclear exactly why this item showed DIF by working or studying.
    This does, however, suggest that scores for participants with different kinds of skin conditions should not be combined
  • DLQI 4 – obvious gender bias with the item
    If data were Rasch analysed during the development stage then some of these issues could have been addressed
    Category probability curve
  • Several items with disordered response thresholds

    Response option 2 ‘a lot’ is also poorly defined in all items
  • Response threshold issues were due to poor response option choices
    ‘Not relevant’ responders were scored the same as ‘not at all’ responders
    Results could underestimate problems with scoring of the response options
    Ambiguous nature of some of the items may have contributed to some of the problems
  • This another issue which could again have been highlighted at the development stage
  • DLQI scores cannot be compared across diseases without splitting of items
    The DLQI lacks items covering mild levels of HRQoL in dermatological conditions
    Problematic scaling issues are due to fundamental problems with the DLQI development
  • Based on these findings it has been shown how failure to follow principals of good questionnaire design can lead to inbuilt error into the scale
    Simple and clear items, good measurement spread, items that are relevant, good response options, free from DIF