The data used in the analysis of health-related quality of life were obtained from the 2006 Health Survey of Catalonia (CHS), which was used to gather information on the general health of non-institutionalized adults (16 or more years of age). The CHS microdata were provided to the authors by the Department of Health of the Generalitat (Government) of Catalonia. Microdata are provided for research purposes with the only requirement to complete an application form that explains the aim of the research.
Participants in the survey spent most of the year residing in family dwellings that were their habitual residences. Individuals were excluded if they resided in group homes or were hospitalized at the time of the survey. The CHS is representative for each one of the 37 health areas existing in Catalonia with a maximum estimation error of
5%. The CHS was made jointly by the Department of Health and the Catalan Institute of Statistics. A random multistage stratified sample was obtained with two stages, first health region and second municipality. The CHS is also representative of regional populations by sex and age groups and it was conducted by specialized interviewers through personal interviews. Individuals interviewed were selected by simple random extraction process without replacement of the Population Register of Catalonia. To avoid sample loss between the theoretical and effective people interviewed five people were selected for each person interviewed for possible substitution according to a strict protocol to replace losses
]. Adult responses were obtained from 15,926 persons.
The ESCA used the EQ-5D, a well-known generic HRQOL instrument, which consists of five dimensions: mobility, self- care, usual activities, pain/discomfort and anxiety/depression. HRQOL is measured on three levels in regard to functional state (no health problems, some health problems and extreme health problems), resulting in 243 aggregate combinations. Participants were surveyed on the five dimensions of EQ-5D and each observation was translated to a single health score using the Spanish time trade-off (TTO) value set
]. The Spanish value set have scores ranging from −0.653 to 1, where 1 corresponds to a perfect state of health and 0 corresponds to death.
CHS also provides additional variables that could be associated with health-related quality of life: socio-demographic factors (age, gender, level of educational), previously diagnosed diseases or chronic conditions (vascular disease, rheumatic disease, digestive diseases, mental illness, respiratory disease, diabetes mellitus, musculoskeletal diseases), risk factors (hypertension and hypercholesterolemia) and negative health experiences (undergoing hospitalization) and lifestyle (smoking, alcohol intake).
Given the nature of the dependent variable, we perform a multivariate analysis for identifying variables for predicting HRQOL scores reported by the health survey participants. Our empirical strategy starts with a basic model construct (Model 1
) that uses control variables such as age, sex, previously diagnosed disease, and health problems (diabetes mellitus, risk of vascular disease, vascular disease, musculoskeletal disease, digestive disease, mental illness, other diseases, and report of an accident in the last 12
months). The variable “risk of vascular disease” receives a score of 1 if the person is obese (BMI
), has hypertension or abnormal cholesterol levels; otherwise the score is 0.
As an abundant literature indicates, hypertension, obesity and hypercholesterolemia represent a higher vascular risk, for both non-diabetic population and for people with diabetes
]. Although we have no clinical measures on levels of blood pressure or cholesterol, data collected provide information on whether a person has been diagnosed with hypertension or hypercholesterolemia and self referred height and weight. The inclusion of this variable can provide interesting results on the influence of these factors of vascular risk in the QOL as other studies have shown
The variable “vascular disease” is assigned a score of 1 if the respondent has been diagnosed with or has ischemic heart disease or has had an embolism. Data on other less common vascular diseases has not been collected in this survey. Model 2, which is built upon Model 1, adds a host of demographic factors such as marital status and level of education. Model 3, in turn, provides additional data on lifestyle (excessive drinking, smoking, illegal drug use).
Due to the continuous nature of the dependent variable, we performed regression models with heteroskedasticity robust least squares estimate by applying the Eicker-White Heteroskedasticity Consistent Covariance Matrix Estimate
] so that the inference procedures are valid even where the error terms have non constant variance among survey participants.
Furthermore, the two-fold objective of this study led us to construct two versions of each model. In the first version (A), previously diagnosed DM is used as an explanatory variable in the multivariate analysis on the dependent variable where we contrast if diagnosed DM is associated with a lower health-related quality of life. The second version (B) identifies persons who have been diagnosed of diabetes but have neither had not previously been diagnosed with any other vascular disease (ischemic heart disease or embolism) and do not exhibit any reported risk factors (obesity, high cholesterol, hypertension) and compares them to diabetic patients who are at risk for vascular disease and those who have had or have been diagnosed with vascular disease. In addition, control and comparison variables have been introduced by including data on persons who do not have diabetes but are at risk for vascular disease and others who do not have diabetes but have had or have previously been diagnosed with vascular disease. So, Model B can be interpreted as analysis of subgroups.
Lastly, a detailed analysis is performed on each of the dimensions that comprise the research instrument for assessing quality of life (mobility, self-care, usual activities, pain/discomfort and anxiety/depression). The probability of reporting a problem as moderate or extreme for each of the dimensions of the EQ-5D is therefore analyzed using the same explanatory variables from Model 3 as control variables. Given the nature of each of the variables studied (dichotomous variables that receive a score of 1 if the individual identifies a problem in the dimension under study or a score of 0 where no problem is indicated) the analysis is carried out using discrete choice models of the probit type.