ABC | Volume 114, Nº1, January 2019

Original Article Figueiredo et al. Poor quality of life in heart failure outpatients Arq Bras Cardiol. 2020; 114(1):25-32 and younger age were predictors of physical symptom status after three months. The great unanswered question is how these variables interact to worsen patients' QoL, and how much they compromise the therapeutic approach in the various degrees of impairment of ventricular function. Thus, we collected sociodemographic, clinical variables, anxiety and depression symptoms, medications in use, previous hospitalization and left ventricular ejection fraction (LVEF) to investigate which factors are associated and interact to get worse quality of life of outpatients with heart failure. Methods This study was approved by the Ethics Committee of the Hospital Universitário Clementino Fraga Filho under the protocol number 104/2010. The patients signed a consent form to participate in this study, which included an observational, cross-sectional and descriptive series of consecutive cases. Patients allowed to participate had HF with reduced LVEF < 40 10 and New York Heart Association (NYHA) functional class I to IV and they were aged ≥ 20 years. Patients with HF caused by valvular dysfunction or reversible causes, and patients who were unable to be interviewed due to psychiatric syndromes, cognitive impairment assessed clinically, and hearing loss were excluded. All participants were recruited at the HF outpatient clinic of the Cardiology Department of the Federal University of Rio de Janeiro from March 2011 to September 2012. One hundred and twenty patients were interviewed individually and a sample of 99 patients of both sexes met the criteria for inclusion. All participants sociodemographic and clinical questionnaires, the Brazilian version of the Minnesota Living with Heart Failure Questionnaire (MLwHF), 11,12 and the Hospital Anxiety and Depression Scale (HADS). 13,14 The sociodemographic questionnaire considered age, sex, monthly family income (dollar), education (years) categorized in illiterate, education 1 (<5 yrs), education 2 (6‑12 yrs), education 3 (>12 yrs), marital status (married) and family support. The clinical questionnaire considered NYHA functional class, comorbidities such as atrial fibrillation, chronic renal failure, diabetes mellitus and arterial hypertension; current use of drugs such as betablockers, spironolactone, angiotensin-converting-enzyme inhibitor (ACE), angiotensin AT1 receptor blocker (ARB), nitrate, hydralazine; previous hospitalization and LVEF. The MLwHF is a structured QoL questionnaire for patients with HF and it has been translated and validated for the Brazilian population. 11,12 The questions are related to how the patient felt during the 30 days before completing the questionnaire. The MLwHF is made up of 21 questions that address the perception of the physical (strongly correlated with dyspnea and fatigue), emotional (correlated with emotional and social aspects) and general well-being (correlated with financial issues, the side effects of medication and lifestyle), and their scores vary from 0 to 40, 0 to 25 and 0 to 40, respectively. A higher score reflected a worse QoL. The HADS was developed specifically for use in medically ill populations. It is based on mood, depression and anhedonia and excludes physical symptoms such as sleep disturbance, fatigue and body pain, which can be confused with symptoms of other diseases. The HADS is made up of 14 questions, each with four possible answers, and consists in two subscales – anxiety and depression – of seven items each. The responses refer to how the patients felt in the last seven days, and the sum of each subscale varies from 0 to 21. It has been translated and validated in a Brazilian version, using a cutoff of ≥ 8 in samples of medically ill patients. 13,14 For the analysis, this cutoff was used as an indication of depression and anxiety, which are referred to in this study as “depression and anxiety symptom”. This variable was dichotomized into “possible anxiety” (8 to 11) and “probable anxiety” (12 to 21), and the same for “possible depression” and “probable depression”. The outcome variables were the MLwHF dimensions, namely, total score, physical, emotional and general welfare. The independent variables were the sociodemographic variables, clinical variables, anxiety symptoms, depression symptoms, current drugs, previous hospitalization and LVEF. Statistical analysis The continuous variables were presented as mean ± standard deviation (variables normally distributed); or median, first quartile and third quartile (non-normal variables). Data normality was tested using the Shapiro-Wilk normality test. The comparison between the NYHA I/II and NYHA III/IV groups was made using unpaired t-Student test, for normal continuous variables, Wilcoxon rank sum test, for non-normal continuous variables, and Exact Binomial Test, for categorical variables. Total score, physical, and the emotional and general dimensions of the MLwHF were the outcome variables. The association of the variables described above with the outcomes and dimensions of QoL were evaluated using a parametric beta regression model and a non-parametric regression tree. 15 Beta regression is a new model recently developed by the Brazilian authors Silvia Ferrari and Francisco Cribari, used when the outcome is a continuous variable that varies in the interval (0,1). The regression tree, apart from its predictive power and its easy visual interpretation, is also extremely useful to find possible interactions between predictive variables, including in situations of unexpected interactions, as in our case. Its final nodes result in the boxplot 16,17 of the outcome variables. The Betareg package 18 of the R software 19 was used. Values of p < 0.05 and 0.05 < p < 0.10 were considered statistically significant and clinical significance, respectively. Data of 99 patients were analyzed; two did not have the LVEF data, and three did not have data of monthly family income. The missing data were imputed considering the MCAR (missing completely at random) as the missing mechanism of these data. 20 To facilitate the reading of the figures, the MLwHF scales were adjusted to vary from 0 to 100 (0, 1) and the scores were inverted so the highest scores would be equivalent to better QoL. Results Table 1 describes the characteristics of the sample dichotomized according to the NYHA functional class. Lower LVEF and lower MLwHF, in all dimensions, were 26