ABC | Volume 112, Nº6, June 2019

Original Article Fernandes et al Lifestyle and costs of medicine use Arq Bras Cardiol. 2019; 112(6):749-755 Table 1 – Most frequently bought medicines according to anatomical therapeutic chemical code Anatomical Therapeutic Chemical Code Types of medicine Number of medicines bought Digestive tract and metabolism 18 33 Blood and Blood-forming organs 2 2 Cardiovascular system 31 42 Dermatological 1 1 Genito-urinary system and sex hormones 9 25 Hormones, except sexual and insulin 10 10 Systemic anti-infective agents 1 1 Anti-neoplastics and immune modulators 1 2 Muscle-skeletal system 7 9 Nervous system 28 38 Antiparasitics 1 2 Respiratory system 3 3 Sensory organs 1 1 Other 1 3 Overall 114 172 Covariates Covariates were data collected via questionnaire (sex [male or female], date of birth [chronological age estimated using the difference between birthday and date of assessment] and formal schooling [in years]). Clinical data also were evaluated (body fatness [dual-energy X-ray absorptiometry], systolic and diastolic blood pressure respectively). Researchers performed the clinical measures in university facilities with controlled temperature and followed standardized procedures. Statistical analyses Descriptive statistics were undertaken using mean, 95% confidence intervals (95%CI) and proportions as appropriate. Due to non-parametric distribution (attested by Kolmogorov- Smirnov test), the costs of medicine use were converted into base-10 logarithms. Both the Pearson correlation and linear regression were conducted to assess the relationship between the costs of medicines and the independent variables. In the former approach, Pearson correlation (expressed as standardized coefficients [“ r ” values]) analyzed the relationship of the costs of medicine use with lifestyle behaviors (sleep quality, PA, SB at work, smoking and alcohol consumption) and covariates (sex, age, schooling, blood pressure and body fatness) separately. In the latter, linear regression models (expressed as unstandardized coefficients [ β values]) were fitted to examine the relationship between the costs of medicine use and lifestyle behaviors controlling for all covariates. For each approach, four models were fitted based on different specifications of lifestyle behaviors ([A] only baseline values, [B] only follow-up values, [C] difference between follow-up and baseline and [D] sum of baseline and follow-up), to explore the differential relationship these specifications may present. Diagnosis of multicollinearity and homoscedasticity were assessed and the linear regression models were considered adequately fit. All analyzes were performed using BioEstat (version 5.2) and the significance level was set at p-value < 0.05. Results At baseline, themean age of the samplewas 51.7±7.1 years, ranging from 40 to 68 years (Table 2). Alcohol was consumed on average 2.1 days per week, while 5.1% of the sample were smokers. Expenses on medicine use were reported by 52.5% of the sample. During 12-months of follow-up, 62 subjects bought 172 medicines (Table 2), representing an overall cost of US$ 3,087.01 for the entire sample. There was no missing data. PA decreased significantly from baseline to follow‑up (p-value = 0.024), while the score for SB at work (p-value = 0.396), sleep quality (p-value = 0.951) and alcohol consumption (p-value = 0.100) remained stable between baseline and follow-up. In the bivariate analysis, costs of medicine use were negatively related to PA baseline (r = -0.194; p-value = 0.035), PA follow-up (r = -0.281; p-value = 0.002), but positively related with sleep quality baseline (r = 0.299; p-value= 0.001) and sleep quality follow-up (r = 0.315; p-value = 0.001), and age baseline (r = 0.274; p-value = 0.003). Gender, education, SB at work, alcohol consumption and smoking were not significantly related with costs of medicine use. There was no interrelationships among the lifestyle behaviors. In the multivariate model considering lifestyle behaviors at baseline (Model-A), sleep quality and body fatness were positively related to higher 12-months medicine costs, while alcohol consumption was negatively related to it. Model-A explained 19.1% of all variance in the outcome (Table 3). In the multivariate model considering lifestyle behaviors at follow-up (Model-B), only sleep quality was positively related to higher 12-months medical costs. Model-B explained 21.9% of all variance in the medicine costs. 751