A Differential Item Functioning Analysis of the EQ-5D in Cancer

• Published in: Value in health 2016-12, Vol.19 (8), p.1063-1067
• Main Authors: Smith, Adam B., PhD, Cocks, Kim, PhD, Parry, David, MSc, Taylor, Matthew, PhD
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Age Factors; Aged; Aged, 80 and over; Cancer; Carcinoma, Non-Small-Cell Lung - psychology; Classification; Clinical trials; Differential equations; differential item functioning; EQ-5D; Female; Health economics; Health status; Humans; Internal Medicine; Lung cancer; Lung Neoplasms - psychology; Male; Middle Aged; Mobility; Oncology; Prostate; Prostate cancer; Prostatic Neoplasms - psychology; Psychometrics; Quality of Life; Questionnaires; Randomized Controlled Trials as Topic; Rasch; Sex Factors; Statistical analysis; Surveys and Questionnaires - standards; Young Adult
• ISSN: 1098-3015; 1524-4733
• DOI: 10.1016/j.jval.2016.06.005

Abstract Objectives To determine whether differential item functioning (DIF) was present in the EuroQol five-dimensional questionnaire (EQ-5D) used in cancer (non–small cell lung cancer and prostate cancer). Methods The Partial Credit Model was applied to the three-level version of the EQ-5D with data obtained from four randomized controlled trials in prostate cancer and non–small cell lung cancer completed at baseline before treatment (N = 2213). DIF was assessed across cancer type (two levels), sex (two levels), and age group (three levels) using Mantel-Haenszel chi-square statistics and evaluated against the Educational Testing Service classification rules. Results The presence of DIF was determined in 14 of 25 (56%) potential DIF contrasts in all the EQ-5D domains. Although mostly the DIF was categorized as either negligible (3 of 25 [12%]) or medium (7 of 25 [28%]), large DIF was observed in 4 of the 25 contrasts (16%). The mobility domain, in particular, showed consistently large DIF across cancer type, sex, and age. Conclusions Given the use of the instrument in health status assessments across conditions and interventions, these results may have significant implications for the EQ-5D in health economic evaluations. Further research is warranted to determine whether these results hold for other cancers.