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 significantly more common in NYHA III and IV. On the other hand, family support, absence of comorbidities such as diabetes mellitus, atrial fibrillation, chronic renal failure and the non-use of spironolactone, nitrate or furosemide were significantly more frequent in NYHA I and II. Results of the beta regression analysis are shown in Table 2. It is noteworthy that the predictor variable NYHA functional class has a statistically significant association with all outcome variables, showing that patient’s QoL is affected in all its dimensions by physical symptoms of HF. Anxiety symptoms were also associated with all outcomes except the general well-being dimension. Depression symptoms were associated with the emotional well-being dimension, showing that psychological symptoms affected patients’ QoL in this sample. Previous hospitalization worsened the QoL of outpatients with HF regarding emotional and physical aspects. The current use of medications as betablockers, ACE and furosemide were associated with poor MLwHF in different dimensions. The general well-being dimension of MLwHF decreased with lower monthly family income of the patients. It is worth mentioning that all beta regression models showed a coefficient of determination R2 of about 50% (Figure 1: A-50, B-48, C-56, D-44), which means that the prognostic variables explain 50% of the variation of the outcome or dependent variables. The regression tree, the nonparametric model, also evaluated the association of the set of predictor variables with each outcome variable (Figure 1). The results of this model are consonant with the beta model seen in Table 2 and give further information about interactions between the independent variables. Figure 1 illustrates the negative effects of a worse NYHA functional class in all MLwHF dimensions. Anxiety symptoms contributed directly or indirectly to lower total score and poorer emotional well-being in MLwHF. The same was observed for previous hospitalization variable in the emotional well-being dimension, indicating an interaction with NYHA functional classes I and II. In the general well‑being dimension, the NYHA classes I and II were associated with poorer MLwHF if patients were younger than 60 years old. Due to inversion of the score system, the interpretation of the regression trees considers that higher values, for each outcome variable, correspond to a better QoL. In other words, we could exemplify that the best quality of life was observed in: NYHA functional class I patients without anxiety symptoms (Figure 1A: total score, 14 patients, 14.1%), NYHA functional class I patients (Figure 1B: physical well-being dimension, 19 patients, 19.2%), NYHA I or II patients with no anxiety symptoms or previous hospitalization (Figure 1C: emotional well-being dimension, 27, 27.3%), and NYHA I or II patients older than 59 years old (Figure 1D: general well‑being dimension, 32 patients, 32.3%). We have created a pseudo-coefficient of determination R2 for the regression tree resulting in coefficients around 30% and 40% (Figure 1 A 42% , B 29%, C 44%, D 29%). Discussion The originality of the study consists of the research methods applied to the research question. The regression tree is a nonparametric regression, useful for prediction, and to obtain data not only from the relevant variables, but also from the relevant interactions between these variables. In the usual regression, we can obtain the relevant variables, but we must define what interactions we consider relevant. In the tree of regression, theDataMining (Machine Learning) algorithmverifies, from the existing data, which variables and which interactions are the most important, providing a better understanding of the complex relationships observed in clinical practice. This study found an association of ventricular function (represented by NYHA functional class), anxiety symptoms, male sex, age younger than 60 years old, lower education level, lower monthly family income, recurrent hospitalization and comorbidities (such as arterial hypertension and ischemic heart diseases), current use of medications (such as betablockers, ACE and furosemide) with poorer QoL of HF outpatients. One study 21 explored the perceptions of QoL in HF patients to assess limitations in their daily lives caused by symptoms, happiness, and relationships, and pointed out that QoL was affected not only by negative physical, psychological, social, and economic status, but also by positive physical, psychological and social status, and behavior, although it was not modified by education combined with self-management intervention. The present study revealed the association of the predictor variable NYHA functional class III-IV with all outcome variables, and of anxiety and depression symptoms with two of the three outcome variables in the beta model. Also, our study shows that even a lower NYHA functional class (I-II), associated with anxiety symptoms and previous hospitalization, may worsen the QoL in the total score and emotional well-being dimension. HF may be the most devastating chronic disease, as it affects people’s QoL in several dimensions. In fact, other authors have provided evidences that improvements in self-management skills may enhance outcomes of HF patients. 22 The awareness of one’s own diagnosis has a profound impact on the patients. Mulligan et al. 8 showed that, addressing patients’ mood and beliefs about their illness and its treatment, improved physical and emotional dimensions of MLwHF and promoted an improvement by 55% and 43% in anxiety and depression symptoms, respectively, assessed by HADS, in the early period after the diagnosis of HF. Reduced anxiety symptoms were associated with reduced perception of the severity of HF and of control due to the treatment of the disease. Reduced depression symptoms was attributed to the improvement of NYHA functional class, reduction in symptoms and perceived severity of HF, and increased confidence in the treatment. The present study confirms these findings that depression and anxiety symptoms contribute to a worsening of QoL related to physical symptoms of HF. It also confirms the fact that when HF patients are less symptomatic and have no anxiety symptoms and recurrent hospitalizations, patient’s QoL is better. In another study, 23 less social support and greater depressive symptoms independently predicted poorer QoL. Also, patient’s perspective on family functioning and autonomy support, along with family knowledge about HF, influenced psychological outcomes of depressive symptoms and emotional QoL of patients with HF. 24 These studies corroborate our results, since married people with family 28