EXAMINING ASSOCIATION BETWEEN PATIENT-REPORTED OUTCOMES AND CLINICAL OUTCOMES USING PROBIT REGRESSION ANALYSIS

Author(s)

Xiao Xu, PhD, Principal, Health Economics and Outcomes Research1, John Conlon, PhD, Director, Statistics Consultancy Group2, Gergana Zlateva, PhD, Director, Global Outcomes Research3, Tyler Knight, MS, Senior Research Analyst11Covance Market Access Services Inc, Gaithersburg, MD, USA; 2 Covance, Inc, Princeton, NJ, USA; 3 Pfizer, Inc, New York, NY, USA

OBJECTIVES: The purpose of this study is to examine the impact of visual acuity (VA) on bilateral neovascular age-related macular degeneration (nvAMD) patients’ general quality of life (QOL). Instead of the traditional correlation/linear regression analysis, a novel approach using PROBIT regression on cumulative QOL rates was employed.METHODS: Data from a cross-sectional, multi-country (France, Germany, Spain, UK and Canada) burden of nvAMD was examined (n=401 nvAMD patients). As part of the telephone interview, patients completed EQ-5D as the assessment of general quality of life (QOL). Associations between VA and EQ-5D were examined using traditional Pearson correlation, analysis of variance (ANOVA) test of EQ-5D score as a function of VA level (normal, mild, moderate, severe, and near blindness), and PROBIT regression. In the PROBIT regression, the decimal VA was ordinalized into categories centered at 0.02, 0.04, 0.06, etc., and cumulative QOL rates were calculated at each of these levels.RESULTS: Pearson correlation and one-way analysis of variance tests showed no significant association between VA and EQ-5D.  From the PROBIT regression, the inflection points were determined and the mean and standard deviation of the underlying normal distribution were estimated. The inflection points identify the VA levels at which association between QOL and VA may be expected and at what level no association would occur.  The mean and SD demonstrate where the relationship was linear. While the Pearson correlation showed a 0.01 correlation coefficient between EQ-5D and VA, PROBIT regression showed a 0.89 correlation coefficient, with the association being linear between VA range of 0.02 and 0.40.CONCLUSIONS: The association of VA levels and EQ-5D is affected by many extraneous variables that make it difficult to identify the strength of the association.  The use of PROBIT regression gives researchers an important methodological tool to strengthen their understanding of such associations.

Conference/Value in Health Info

2008-11, ISPOR Europe 2008, Athens, Greece

Value in Health, Vol. 11, No. 6 (November 2008)

Code

OR1

Topic

Clinical Outcomes, Methodological & Statistical Research

Topic Subcategory

Clinical Outcomes Assessment, PRO & Related Methods

Disease

Multiple Diseases

Explore Related HEOR by Topic


Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×