Abstract
Design: The health status of a representative sample randomly selected from the population register of two Bologna Health Authorities in Northern Italy was measured using the EuroQoL questionnaire. A postal survey was performed among a sample of 1622 adults. Single questions were added on patient age, gender, marital status, education, occupation and smoking status. Objectives and Methods: We propose a new approach for modelling the EQ-5D dimensions in order to identifying perceived health impact of life-style determinants. We created a new variable with four ordinal categories based on different levels of severity (from low to high): “no problem” in any dimension, a moderate problem in only one dimension, at least two moderate problems in any dimension, and at least one extreme problem. The use of these categories is appropriate to test a specific regression model for categorical dependent variables: the Ordered Regression Model (ORM). Analysis were conducted with STATA 8.2. Results: Descriptive analysis suggested that age, gender, marital status, education, smoking behaviour, comorbidity and BMI can explain the variability of the perceived health status measured by EQ-5D index and the VAS. Applying ORM, odds ratio from the four dependent strata of the outcome are summarised by the OR of having at least one extreme problem versus having any problem. Using a confidence level of 95%, the significant covariates were: gender, age, education, comorbidity and BMI. The results of the model indicate that women have 1.87 times the odds of a bad health status perception vs. a good perception than man. High education level increases (54%) the probability of observing no health problem with respect to a medium education. An interesting result concerns the Body Mass Index. Obese people (with no health problems) have 2.64 times the odds of a good perceived health related quality of life, than people who have normal BMI; but when a obese person has one or more than one health problem, he/she is about 3 times more likely to have a low health state perception. Conclusion: This is the first attempt to perform an Ordered Regression Model in order to identifying perceived health impact of life-style determinants. A notable feature of the ORM model is its ease of interpretation: it provides a summary ratio of the odds of having a bad perceived health status that is independent of the level of severity used to classify the health status.