The disease burden morbidity assessmenta validation study

  1. WIJERS, IRENE GEURTJE MARIA
Supervised by:
  1. Maria Joao Bettencourt Pereira Forjaz Director
  2. Manuel Franco Tejero Co-director

Defence university: Universidad de Alcalá

Fecha de defensa: 04 July 2017

Committee:
  1. Antonio Sarría Santamera Chair
  2. Carmen Rodríguez Blázquez Secretary
  3. Gloria Fernández Mayoralas Committee member
Department:
  1. Cirugía, Ciencias Médicas y Sociales

Type: Thesis

Abstract

Introduction: Population aging is a process that is taking place all over the world. One of the consequences is the increase in the prevalence of chronic conditions, and therefore also the co-existence of them, so-called comorbidity or multimorbidity. Different instruments exist to assess multimorbidity, and the choice of instrument depends on the study context and outcomes of interest. The Disease Burden Morbidity Assessment (DBMA) is a self-report questionnaire in which participants rate the disease burden caused by a number of medical conditions. It was designed and validated to be associated with patient-centered outcomes. However, a validation following psychometric or clinimetric methodology had not been performed yet. Objectives: The objectives of this thesis were to validate the DBMA according to the Classical Test Theory(CTT) (Study 1), to assess known-groups, convergent and predictive validity (Study 2) and to perform a Rasch analysis of the scale (Study 3). Methods: Data were used from the Ageing in Spain Longitudinal Study, Pilot Survey (ELES-PS). The DBMA consists of a list of 21 chronic medical conditions. Participants are asked for every condition whether they have it and if so, to what extent it interferes with their everyday life on a scale from 1 (not at all) tot 5 (a lot). Scores are summed to obtain a measure of disease burden. In the first study, psychometric properties of the scale (feasibility, acceptability, scaling assumptions, reliability and construct validity) were analyzed. Dimensionality was assessed through an exploratory factor analysis. In Study 2, known-groups validity for sex and age groups (< 75 years vs. ¿75 years) was assessed. For convergent validity, a multivariate linear regression model was used to evaluate differences in DBMA scores as a function of age and sex, patient-centered outcomes and utilization outcomes. For predictive validity, the association with four-year mortality was assessed using a Cox model and Kaplan-Meier curves. In the Rasch analysis, test of fit to the Rasch model, reliability, unidimensionality, response dependency, category structure, scale targeting and differential item functioning (DIF) were studied in an iterative way. Construct validity of the linear measure provided by the Rasch analysis was subsequently assessed. Results: In the CTT analysis, satisfactory feasibility and acceptability were found, except for large floor effects (>50%) as well as a skewed distribution (skewness=1.8). Item-total corrected correlation ranged 0.10-0.49, item homogeneity index was 0.09, and Cronbach¿s alpha was 0.72. Disease burden correlated strongly with physical functioning and perceived health, and moderately with depression and quality of life. Exploratory factor analysis extracted 5 factors, explaining 44% of the variance. The known-groups analysis in Study 2 found higher disease prevalences and also higher disease burden per present condition for women. The same differences were found for age groups but less pronounced. In the multivariate regression, sex, perceived health, physical functioning, quality of life, affect balance and primary/outpatient care utilization were significantly associated with the DBMA. The Cox model displayed a hazard ratio of 1.07 and the Kaplan-Meier curves showed lower survival rates in participants with higher DBMA scores. In the Rasch analysis, items needed to be rescored by collapsing response categories to achieve an adequate fit to the Rasch model. Reliability (person separation index) was low. The scale was unidimensional and neither response dependency nor relevant DIF were found. Relative precision analysis showed that the linear measure discriminated better between age groups than the original raw score, but for sex no difference was found.