MAPPING THE SF-36 TO EQ-5D-3L IN RANDOMIZED TRIAL- THE EMOCAR STUDY (MAY 2011 – APRIL 2016)
Author(s)
Cabout E1, Kabeshova A1, Le Teurnier Y2, Launois R1
1REES France, Paris, France, 2CHU Nantes, Nantes, France
Presentation Documents
OBJECTIVES: Cost-utility analyses need a measure to summarize the quality of life in a single index. A single score is often obtained by the EQ-5D questionnaire which is adaptable to each country with score tariffs METHODS: The EMOCAR study was a French cohort of patients with carotid endarterectomy where 904 adults completed both EQ-5D and SF-36 questionnaires. Ordinary least squares (OLS) and multinomial logit models were implemented to predict the overall score and the response to each of the 5 categories of EQ-5D with two types of specifications: item-based and summary score-based of SF-36. The performance of each model was determined by the estimated average score, the absolute and mean squared errors as well as the percentage of errors. RESULTS: The model including SF-36 items using OLS method has the most accurate predictions comparing to other models. This mapping performance was indicated by a significant prediction of EQ-5D score of 0.7327 and by MAE (0.116), MSE (0.026), R² (0.67) and the percentage of an absolute error > 0.1 (43.71%). The model including SF-36 dimensions scores predicted an EQ-5D score significantly different from the EQ-5D observed score. OLS models predicted better EQ-5D score of younger subgroups and multinomial model with SF-36 items older subgroups. CONCLUSIONS: Our models suggest that models using OLS method have good performance for mapping SF-36 onto EQ-5D. Low scores are underpredicted with both models which is already present in the existing literature.
Conference/Value in Health Info
2018-11, ISPOR Europe 2018, Barcelona, Spain
Value in Health, Vol. 21, S3 (October 2018)
Code
PRM167
Topic
Methodological & Statistical Research
Topic Subcategory
PRO & Related Methods
Disease
Cardiovascular Disorders