Assessment of HRQoL in patients with eating disorders by the beta-binomial regression approach

Authors

  • Inmaculada Arostegui PhD,

    Corresponding author
    1. Departamento de Matemática Aplicada y Estadística e Investigación Operativa, Universidad del País Vasco, Leioa, Bizkaia, Spain and CIBER Epidemiología y Salud Pública (CIBERESP) Spain
    • Universidad del País Vasco UPV/EHU, Departamento de Matemática Aplicada y Estadística e Investigación Operativa, Apartado 644, Bilbao 48080, Spain
    Search for more papers by this author
  • Angel Padierna MD,

    1. Servicio de Psiquiatría, Hospital Galdakao-Usansolo, Galdakao, Bizkaia, Spain
    Search for more papers by this author
  • Jose M. Quintana MD, PhD

    1. Unidad de Investigación, Hospital Galdakao-Usansolo, Galdakao, Bizkaia, Spain and CIBER Epidemiología y Salud Pública (CIBERESP) Spain
    Search for more papers by this author

Abstract

Objective:

To study the influence of clinical variables on health-related quality of life (HRQoL) among women with eating disorder (ED) using beta-binomial regression (BBR) to analyze scores on the Short-Form 36 (SF-36) as dependent variable.

Method:

Female patients diagnosed with ED completed the SF-36 at the beginning of the study and after 2 years of treatment. Sociodemographic and clinical information was recorded. For the multivariate analysis, we used BBR models to identify factors that influence SF-36.

Results:

Questionnaires were completed by 193 women at baseline and 158 (82%) after 2 years of treatment. Anxiety, depression, and the severity of ED, explained scores in most domains of the SF-36 at baseline. The main predictor of HRQoL after 2 years of follow-up was the HRQoL in the same domain at baseline. However, depression, anxiety, and duration of symptoms at baseline also significantly influenced HRQoL after 2 years of treatment in some domains. Higher levels of anxiety or depression, longer duration of symptoms, and poorer SF-36 scores at baseline were associated with worse HRQoL after 2 years of treatment and follow-up.

Discussion:

BBR models provide understandable results for clinicians and can be used in multivariate models with HRQoL dependent variables. © 2009 by Wiley Periodicals, Inc. Int J Eat Disord 2010

Ancillary