Advertisement

Different profiles of advanced heart failure among patients with and without diabetes mellitus. Findings from the EPICTER study

      Highlights

      • The different scores used to define mortality risk include some variables that are difficult to assess in patients with advanced HF.
      • Clinical profile of advanced HF in patients with DM displayed more comorbidity.
      • Patients with DM and advanced HF also had progression of complications of DM.
      • General criteria of advanced HF prevailed over cardiac ones regarding prognosis.
      • A better characterization will help to avoid overtreatment and improve quality of care of these patients.

      Abstract

      Aim

      This work aims to compare the characteristics of advanced heart failure (HF) in patients with and without type 2 diabetes mellitus (DM) and to determine the relevance of variables used to define advanced HF.

      Patients and methods

      This cross-sectional, multicenter study included patients hospitalized for HF. They were classified into four groups according to presence/absence of advanced HF, determined based on general and cardiac criteria, and presence/absence of DM. To analyze the importance of variables, we grew a random forest algorithm (RF) based on mortality at six months.

      Results

      A total of 3153 patients were included. The prevalence of advanced HF among patients with DM was 24% compared to 23% among those without DM (p=0.53). Patients with advanced HF and DM had more comorbidity related to cardiovascular and renal diseases; their prognosis was the poorest (log-rank <0.0001) though the adjusted hazard ratio by group in the Cox regression analysis was not significant. The variables that were significantly related to mortality were the number of comorbidities (p=0.005) and systolic blood pressure (p=0.024). The RF showed that general criteria were more important for defining advanced HF than cardiac criteria.

      Conclusions

      Patients with advanced HF and DM were characterized by DM in progression with macro and microvascular complications. The outcomes among advanced HF patients were poor; patients with advanced HF and DM had the poorest outcomes. General criteria were the most important to establish accurately a definition of advanced HF, being decisive the evidence of disease progression in patients with DM

      Keywords

      1. Introduction

      Heart failure (HF), the representative end-stage of most heart diseases, has a heterogeneous course that makes it difficult to know with certainty whether a patient is in the advanced stage of the disease [
      • Kavalieratos D
      • Mitchell EM
      • Carey TS
      • Dev S
      • Biddle AK
      • Reeve BB
      • Abernethy AP
      • Weinberger M
      Not the 'grim reaper service'": an assessment of provider knowledge, attitudes, and perceptions regarding palliative care referral barriers in heart failure.
      ]. In advanced HF, therapies such as cardiac resynchronization (CRT) or an implantable cardioverter defibrillator (ICD) may not be suitable or accepted due to disability, comorbidity, or the patient's preferences [
      • Ponikowski P
      • Voors AA
      • Anker SD
      • Bueno H
      • Cleland JG
      • Coats AJ
      • Falk V
      • González-Juanatey JR
      • Harjola VP
      • Jankowska EA
      • Jessup M
      • Linde C
      • Nihoyannopoulos P
      • Parissis JT
      • Pieske B
      • Riley JP
      • Rosano GM
      • Ruilope LM
      • Ruschitzka F
      • Rutten FH
      • van der Meer P
      Authors/Task Force Members; Document Reviewers. 2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: The Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.
      ]. Despite improvements in treatment, advanced HF continues to have substantial morbidity and mortality.
      Physicians who care for patients with advanced HF should be able to recognize the disease's end stage, yet the different validated scores used to calculate mortality risk in patients with HF include some variables that are difficult to assess in the majority of patients with advanced HF (e.g., the 6-minute walk test [
      • Levy WC
      • Mozaffarian D
      • Linker DT
      • Sutradhar SC
      • Anker SD
      • Cropp AB
      • Anand I
      • Maggioni A
      • Burton P
      • Sullivan MD
      • Pitt B
      • Poole-Wilson PA
      • Mann DL
      • Packer M
      The Seattle Heart Failure Model: prediction of survival in heart failure.
      ,
      • O'Connor CM
      • Hasselblad V
      • Mehta RH
      • Tasissa G
      • Califf RM
      • Fiuzat M
      • Rogers JG
      • Leier CV
      • Stevenson LW
      Triage after hospitalization with advanced heart failure: the ESCAPE (Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness) risk model and discharge score.
      ]). Furthermore, it is likely that the vast majority of patients included in the development of risk stratification instruments did not have advanced disease and thus these stratification tools may not be as useful for these patients [
      • Lee DS
      • Austin PC
      • Rouleau JL
      • Liu PP
      • Naimark D
      • Tu JV
      Predicting mortality among patients hospitalized for heart failure: derivation and validation of a clinical model.
      ].
      The prevalence of diabetes mellitus (DM) in patients with HF is rapidly increasing [
      • Lehrke M
      • Marx N
      Diabetes Mellitus and Heart Failure.
      ]. Patients with HF and DM have specific metabolic, neurohormonal, and structural heart abnormalities that could potentially contribute to worse HF outcomes than patients without comorbid DM [
      • Dei-Cas A
      • Fonarow GC
      • Gheorghiade M
      • Butler J
      Concomitant diabetes mellitus and heart failure.
      ]. In light of this, it may be even more difficult—yet more critical—to determine both the prevalence of advanced HF in this population and which variables allow for recognizing advanced disease in order to initiate appropriate treatment and referrals to palliative care (PC) units.
      The EPICTER study was a cross-sectional, multicenter project in Spain whose aim was to calculate the prevalence of advanced HF in hospitalized patients [
      • Fernández-Martinez J
      • Romero-Correa M
      • Salamanca-Bautista P
      • Aramburu-Bodas Ó
      • Formiga F
      • Vázquez-Rodríguez P
      • Conde-Martel A
      • García-García JA
      • Páez-Rubio I
      • López-Reboiro M
      • Sánchez-Sánchez C
      • Arias-Jiménez JL
      Prevalence of advanced heart failure and use of palliative care in admitted patients: Findings from the EPICTER study.
      ]. It incorporated new clinical variables beyond cardiac parameters: a combination of cardiac (National Hospice Organization criteria [
      Medical guidelines for determining prognosis in selected non-cancer diseases
      The National Hospice Organization.
      ]) and general criteria (PALIAR project [
      • Bernabeu-Wittel M
      • Ruiz-Cantero A
      • Murcia-Zaragoza J
      • Hernández-Quiles C
      • Barón-Franco B
      • Ramos-Cantos C
      • Nieto-Martín MD
      • Fernández-López A
      • Fernández-Moyano A
      • Moreno-Gaviño L
      • Ollero-Baturone M
      Precisión de los criterios definitorios de pacientes con enfermedades médicas no neoplásicas en fase terminal. Proyecto PALIAR [Reliability of different criteria in identifying end-of-life trajectory of patients with chronic medical diseases. PALIAR Project].
      ]) were used to define advanced HF.
      The aim of this study is to analyze the characteristics of advanced HF in patients with and without DM and to determine the relevance of cardiac and general criteria in order better define advanced disease, as measured by mortality at six months. We hypothesize that patients with DM have a different profile of advanced HF than patients without DM and that the variables which should be taken into account when defining advanced HF differ depending on the presence or absence of DM.

      2. Subjects, materials, and methods

      The EPICTER study (“Epidemiological survey of advanced heart failure”) was a cross-sectional, multicenter project that included consecutive patients admitted for HF in 74 Spanish hospitals. It included public and private hospitals regardless of size. Patients were enrolled in two intervals (summer and winter). To avoid bias, all hospitals started collecting data on the same day (June 1 and November 30, 2016) and all patients admitted to the cardiology or internal medicine ward, intensive care unit, or other departments were included. The inclusion criteria were: 1) patients older than 18 years; 2) admission to the hospital before 8:00 a.m. on the day of data collection; and 3) HF as the main cause of hospitalization: acute HF, acute pulmonary edema, Killip class III-IV acute coronary syndrome, or cardiogenic shock. The exclusion criteria were: 1) patients attended to the emergency department but not yet admitted and 2) patients who did not provide informed consent. Researchers at each hospital examined patients who met the inclusion criteria every day. Recruitment continued until the mandatory number of patients to be included at each hospital, which was pre-determined according to the number of beds at the hospital, were selected.
      All inpatients were given treatment and medical care according to clinical guidelines and their physicians’ clinical judgment. Patients were classified depending on whether they met criteria for advanced HF as well as according to the presence of DM on their medical history, with no distinction between either, type 1 or type 2. Patients’ vital status at six months was verified by researchers at each center. In order to verify this endpoint, local medical record databases were consulted, then the patient was contacted, and then, if necessary, relatives were contacted. The study was conducted pursuant to the Declaration of Helsinki. Ethical approval (Virgen Macarena University Hospital Ethics Committee, internal code 0942-N-15; November 24, 2015) was obtained before starting enrolment. All patients signed an informed consent form prior to inclusion. More information about the sample size and variables included in the project is available in a previously published work [
      • Fernández-Martinez J
      • Romero-Correa M
      • Salamanca-Bautista P
      • Aramburu-Bodas Ó
      • Formiga F
      • Vázquez-Rodríguez P
      • Conde-Martel A
      • García-García JA
      • Páez-Rubio I
      • López-Reboiro M
      • Sánchez-Sánchez C
      • Arias-Jiménez JL
      Prevalence of advanced heart failure and use of palliative care in admitted patients: Findings from the EPICTER study.
      ].

      2.1 Advanced HF criteria

      The criteria for determining if a patient had advanced HF or not were based on a combination of cardiac and general criteria. The National Hospice Organization criteria [
      Medical guidelines for determining prognosis in selected non-cancer diseases
      The National Hospice Organization.
      ] (cardiac criteria) were used to select patients with an expected life expectancy of less than six months who could benefit from specialized PC; it is based on some specific criteria for patients with HF. They include: 1) New York Heart Association (NYHA) class III-IV; 2) left ventricular ejection fraction (LVEF) less than 20%; 3) intractable angina with HF; 4) symptoms despite optimal treatment; 5) contraindication for transplant, implantation of devices, coronary revascularization or valvular replacement; and 6) presence of refractory supraventricular and ventricular arrhythmias.
      General criteria used for the PALIAR project (general criteria) were used for determining to what extent heart disease impacts the patient's life [
      • Bernabeu-Wittel M
      • Ruiz-Cantero A
      • Murcia-Zaragoza J
      • Hernández-Quiles C
      • Barón-Franco B
      • Ramos-Cantos C
      • Nieto-Martín MD
      • Fernández-López A
      • Fernández-Moyano A
      • Moreno-Gaviño L
      • Ollero-Baturone M
      Precisión de los criterios definitorios de pacientes con enfermedades médicas no neoplásicas en fase terminal. Proyecto PALIAR [Reliability of different criteria in identifying end-of-life trajectory of patients with chronic medical diseases. PALIAR Project].
      ]. Although this instrument was not specifically designed for patients with HF, has been internally and externally validated for advanced non-oncologic chronic diseases [
      • Bernabeu-Wittel M
      • Murcia-Zaragoza J
      • Hernández-Quiles C
      • Escolano-Fernández B
      • Jarava-Rol G
      • Oliver M
      • Díez-Manglano J
      • Ruiz-Cantero A
      • Ollero-Baturone M
      • Researchers PALIAR
      Development of a six-month prognostic index in patients with advanced chronic medical conditions: the PALIAR score.
      ,
      • Gómez-Aguirre N
      • Fuertes-Ruiz D
      • Gracia-Tello B
      • Clemente-Sarasa C
      • Artajona-Rodrigo E
      • Cabrerizo-García JL
      • de Escalante-Yangüela B
      • Bueno-Castel MC
      • Velilla-Marco J
      • Díez-Manglano J
      External validation of the PALIAR index for patients with advanced, nononcologic chronic diseases.
      ]. Its items include: 1) estimated survival of less than 6 months, as assessed by the attending physician at the time of admission; 2) acceptance of a palliative approach by the family or the patient; 3) at least one of the following criteria: a) evidence of disease progression, b) three or more emergency department visits within the past six months, c) unintentional weight loss (more than 10%), d) functional impairment (as assessed by the Barthel Index with dependence in at least three activities of daily living or the Pfeiffer Short Portable Mental Status Questionnaire with more than three errors), and e) consensus among several physicians that the patient had advanced HF.
      Patients were considered to have advanced HF if they met at least one cardiac criterion and three general criteria. Patients were considered to have DM if it was recorded in their medical history or if they received treatment with oral hypoglycemic agents, insulin, or a combination of both. Four groups were created based on the presence/absence of DM and presence/absence of advanced HF (Fig. 1). Group one included patients without advanced HF and without DM. Group two included patients without advanced HF and with DM. Group three included patients with advanced HF and without DM. Group four included patients with advanced HF and with DM.

      2.2 Statistical analysis

      Qualitative variables were expressed as absolute values (percentage) and the chi-square test was used to compare these variables among groups. Quantitative variables were expressed as median (interquartile range, IQR). ANOVA was used to compare these variables among groups in the case of normal distribution and Welch's t-test was used in the event of unequal variance. The Kruskal-Wallis test was used for non-parametric variables with similar variance (as assessed by Levene's test) whereas ANOVA and Welch's t-test were used for non-parametric variables with unequal variance. A Kaplan-Meier curve was created to analyze all-cause mortality among the four groups and additionally to secondary analyze cardiovascular mortality and readmissions (log-rank test). Two Cox regression models were built to analyze predictive variables according to all-cause mortality and cardiovascular mortality. The variables included apart from the groups were, age, the number of comorbidities, the left ventricular ejection fraction, hemoglobin, creatinine, systolic and diastolic blood pressure, and the NT-ProBNP measured at admission.
      To analyze the importance of the different variables used to categorize patients according to advanced HF criteria (cardiac and general), we grew a random forest algorithm. Random forests (RF) are a classification and regression method based on the aggregation of a large number of decision trees [
      • Breiman L
      Random forests.
      ]. Along with the variables selected, we performed two algorithms according to the presence/absence of DM. In the first algorithm (patients without DM), we included all subjects belonging to groups one and three. In the second algorithm (patients with DM), we included all subjects belonging to groups two and four. Split selection was carried out based on the decrease in Gini impurity (DGI), the procedure followed in the most commonly used type of RF. For both algorithms, each sample was randomly divided into a training set, which included 70% of patients, and a test set, which included the remaining 30%.
      This version of RF is used in the ‘randomForest’ package of R [
      • Liaw A
      • Wiener M
      Classification and Regression by randomForest.
      ]. For this RF, two types of variable importance measures (VIMs) were considered: the mean DGI (decrease Gini) and the unscaled permutation-based importance measure. Both are implemented in the ‘importance’ function of the ‘randomForest’ package. Further details on VIMs can be consulted in a previously published work [
      • Strobl C
      • Boulesteix AL
      • Zeileis A
      • et al.
      Bias in random forest variable importance measures: illustrations, sources, and a solution.
      ].
      All analyses were conducted using R version 3.3.2 (R Core Team 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/).

      3. Results

      A total of 3,153 patients were included. The median age (IQR) of the sample was of 81 (13) years and there were slightly more women (50.6%) than men. Seven hundred and thirty-nine patients (23.4%) met the criteria of advanced HF. A total of 1,434 (45.5%) patients had a prior DM diagnosis. Of them, 344 (24%) met the criteria of advanced HF compared to 23% of those without DM (p=0.53).
      Table 1 shows the main characteristics of patients according to group. Patients with advanced HF and DM had the highest prevalence of previous HF and atrial fibrillation/flutter (p<0.001). Comorbidities were more frequent in this group, notably those related to renal and cardiovascular disease, such as stroke and peripheral arterial disease (p<0.00). In regard to clinical manifestations, patients with advanced HF were the most symptomatic overall. Among patients with DM, both nausea and erratic pains were more frequent, but the differences were not significant (p=0.08 and p=0.09, respectively). On the contrary, delusions were more frequent in patients without DM, though this finding was also not significant (p=0.07). Finally, the therapeutical approach as palliative care was more used in the groups of advanced HF, being the patients with DM who were administered nitroglycerine and vasoactive amines the most.
      Table 1Baseline characteristics according to the groups.
      Heart failureAdvanced heart failure
      VariableTOTAL SAMPLE (N=3153)GROUP 1 (No DM) N=1324GROUP 2 (DM) N=1090GROUP 3 (No DM) N=395GROUP 4 (DM) N=344P
      P: P-value of ANOVA or Kruskal-Wallis tests for quantitative variables among the groups depending on the normality, and Chi-square test for qualitative ones. COPD: Chronic obstructive pulmonary disease; CKD: Chronic kidney disease; DBP: diastolic blood pressure; DM: Diabetes mellitus; HF: Heart failure; IQR: interquartile range; LVEF: Left ventricular ejection fraction; MI: Myocardial infarction; NT-ProBNP: N-terminal pro-brain natriuretic peptide; SBP: Systolic blood pressure.
      Age81 (13)81 (15)79 (13)86 (9)83 (9.2)<0.0001
      Sex (female)1596 (50.6)651 (49.2)527 (48.3)237 (60)181 (52.6)<0.0001
      SBP (mmHg)124 (30)123 (30)129 (30)120 (30)122 (35)<0.0001
      DBP (mmHg)68 (18)70 (19)69.5 (19)65 (17)66.5 (17)
      Previous MI1005 (31.8)341 (25.7)399 (36.6)108 (27.3)157 (45.6)<0.0001
      Valvular heart disease1328 (42.1)578 (43.6)405 (37.1)188 (47.6)157 (45.6)<0.0001
      Previous HF2269 (71.9)859 (64.9)779 (71.5)324 (82.2)307 (89.2)<0.0001
      Atrial fibrillation/Flutter1788 (56.7)747 (56.4)577 (52.9)242 (61.3)222 (64.5)<0.0001
      Comorbidities
      Hypertension2677 (84.9)1031 (77.9)1003 (92)328 (83)315 (91.6)<0.000
      COPD812 (25.8)304 (22.9)295 (27.1)111 (28.1)102 (29.6)0.01
      Cerebrovascular disease671 (21.3)248 (18.7)212 (19.4)102 (25.8)109 (31.7)<0.0001
      Peripheral arterial disease493 (15.6)149 (11.2)208 (19.1)60 (15.2)76 (22.1)<0.0001
      CKD1480 (46.9)479 (36.2)563 (51.6)199 (50.4)239 (69.5)<0.0001
      Cancer480 (15.2)198 (14.9)140 (12.8)82 (20.7)60 (17.4)0.001
      Cognitive impairment573 (18.2)184 (13.9)155 (14.2)129 (32.6)106 (30.8)<0.0001
      Chronic liver disease176 (5.6)65 (4.9)69 (6.3)24 (6.1)18 (5.2)0.46
      Charlson Com. Index3 (3)3 (3)4 (2)3 (3)5 (2)<0.0001
      Laboratory data
      Hemoglobin (g/dl)11.4 (2.8)11.8 (3)11.3 (2.6)11.1 (2.3)10.9 (2.5)<0.0001
      Glucose (mg/dl)122 (66.5)107 (39)150 (90.7)110 (38)150 (87)<0.0001
      Creatinine (mg/dl)1.3 (0.8)1.2 (0.7)1.3 (0.8)1.3 (0.8)1.5 (1)<0.0001
      Sodium (mEq/L)139 (6.7)139 (6)139 (5)139 (7)138 (7)0.09
      NT-proBNP (pg/ml)4381 (8096)3526 (7007)3910 (7069)5691 (9857)6689 (15143)<0.0001
      Troponin T (ng/ml)37.4 (73.5)30.8 (65.7)38.8 (48.2)34 (114.9)61.3 (165.3)0.2
      LVEF (%)55 (24)55 (23)55 (25)50 (28)47 (26)<0.0001
      Symptoms
      Dyspnea1604 (50.8)523 (39.5)412 (37.8)358 (90.6)311 (90.4)<0.0001
      Anxiety660 (20.9)181 (13.7)131 (12)184 (46.6)164 (47.7)<0.0001
      Insomnia713 (22.6)214 (16.2)141 (12.9)186 (47.1)172 (50)<0.0001
      Chest pain387 (12.3)127 (9.6)96 (8.8)85 (21.5)79 (23)0.029
      Nausea233 (7.4)60 (4.5)43 (3.9)60 (15.2)70 (20.3)<0.0001
      Erratic pains554 (17.6)160 (12.1)116 (10.6)137 (34.7)141 (41)<0.0001
      Delusions305 (9.7)59 (4.4)43 (3.9)120 (30.4)83 (24.1)<0.0001
      Outcomes
      Mortality at six months753 (23.9)239 (18)185 (17)183 (46.3)146 (42.4)<0.0001
      Palliative care referral148 (4.7)16 (1.2)20 (1.8)65 (16.4)47 (13.7)<0.0001
      Emergency room consultation998 (75.4)390 (29.4)358 (32.8)117 (29.6)133 (38.7)0.006
      All-cause readmissions (median, IQR)1 (1)0 (1)1 (2)0 (1)1 (2)0.001
      Therapy as palliative care for heart failure
      Furosemide infusions514 (16.3)162 (12.2)133 (12.2)115 (29.1)104 (30.2)<0.0001
      Nitroglycerine317 (10)85 (6.4)90 (8.3)65 (16.4)77 (22.8)<0.0001
      Vasoactive amines165 (5.2)45 (3.4)39 (3.6)33 (8.3)48 (13.9)<0.0001
      Vaptans11 (0.34)2 (0.1)3 (0.3)6 (1.5)0<0.0001
      Opioids541 (17.2)105 (7.9)89 (8.2)191 (48.3)156 (45.3)<0.0001
      Benzodiazepines691 (21.9)202 (15.3)163 (15)167 (42.3)159 (46.2)<0.0001
      Phenothiazines246 (7.8)54 (4.1)31 (2.8)92 (23.3)69 (20.1)<0.0001
      Ultrafiltration23 (0.7)7 (0.5)5 (0.5)6 (1.5)5 (1.4)<0.0001
      Non-invasive ventilation172 (5.4)66 (5)42 (3.8)30 (7.6)34 (9.9)<0.0001
      low asterisk P: P-value of ANOVA or Kruskal-Wallis tests for quantitative variables among the groups depending on the normality, and Chi-square test for qualitative ones.COPD: Chronic obstructive pulmonary disease; CKD: Chronic kidney disease; DBP: diastolic blood pressure; DM: Diabetes mellitus; HF: Heart failure; IQR: interquartile range; LVEF: Left ventricular ejection fraction; MI: Myocardial infarction; NT-ProBNP: N-terminal pro-brain natriuretic peptide; SBP: Systolic blood pressure.

      3.1 Outcomes

      In regard to prognosis, patients with advanced HF and DM had the worst outcomes in terms of mortality at six months (log-rank<0.0001, Fig. 2), but the HR as measured by an adjusted Cox regression analysis was not significant compared to the remaining groups (Fig. 2). However, with regards to cardiovascular mortality (supplementary figure 1), groups three and four had the poorest prognosis (group 3, HR 2.15, 95%CI 1.2-3.9, p=0.014; group 4, HR 2.8, 95%CI 1.5-5.1, p<0.00). In patients in group four (advanced HF and DM), the variables that were significantly related to mortality were comorbidities (p=0.005) and systolic blood pressure (p=0.024). With respect to cardiovascular mortality (supplementary figure 1), other significant variables in the model were the age (p=0.005), the number of comorbidities (p=0.038), the LVEF (p=0.0038) and DBP (p=0.002). Additionally, these patients had the highest rate of emergency room visits compared to the remaining groups (p=0.006) as well as compared to patients with advanced HF without DM (p=0.01). Finally, Table 1 display the number of readmissions by group. The groups of patients with DM had significanttly more readmissions than patients without DM (p=0.001).
      Fig. 2
      Fig. 2Kaplan-Meier curves according to patient group (left) and box-plot of the adjusted Cox regression analysis. DBP: diastolic blood pressure; LVEF: left ventricular ejection fraction; SBP: systolic blood pressure.

      3.2 Random forest algorithm

      The prediction error rate (out-of-bag error) used for assessing the prediction performance of the random forest was 25% (that is, the observations that were not used to construct a tree). The model performed well on the training and test sets, yielding an accuracy of 0.76.
      Fig. 3 shows the importance of the variables selected. Overall, variables of the general criteria (PALIAR project) better defined advanced heart failure as measured in terms of mortality than those of the cardiac criteria (National Hospice Organization criteria).
      Fig. 3
      Fig. 3Importance of variables in patients with and without diabetes mellitus according to random forest algorithm. Left: Importance of variables according to mean decrease in accuracy and mean decrease in Gini. Further explanation is available in the article and in reference 12. Right: multi-way importance plot. The size of the points reflects the number of nodes split on the variable. The top ten variables are highlighted in blue. The x-axis shows the mean minimal depth calculated using top trees (minimal depth for a variable in a tree equals the depth of the node which splits on that variable and is closest to the root of the tree. If it is low, then many observations are divided into groups based on that variable). The y-axis shows the times in a root. As is typically the case, there is a negative relationship between the x- and y-axes.
      There were few differences between patients with and without DM. Evidence of disease progression may be a decisive variable in patients with DM whereas emergency department visits may be important in patients without DM. Variables such as intractable angina with HF were not used in the decision trees.

      4. Discussion

      Our analysis revealed that the prevalence of advanced HF was very similar among patients with and without DM. However, differences in characteristics mean that these patients have different profiles, which is in line with our hypothesis.
      Patients with advanced HF and DM had more comorbidity and a higher rate of previous HF and atrial fibrillation/flutter. On the other hand, patients with advanced HF without DM had a higher rate of valvular heart disease. Conditions commonly found in patients with DM include hypertension, obesity, hyperlipidemia, and chronic kidney disease [
      • Iglay K
      • Hannachi H
      • Joseph-Howie P
      • Xu J
      • Li X
      • Engel SS
      • Moore LM
      • Rajpathak S
      Prevalence and co-prevalence of comorbidities among patients with type 2 diabetes mellitus.
      ], yet interestingly, in this analysis, patients with DM and advanced HF had a significant predominance of vascular comorbidities (coronary artery disease, peripheral arterial disease, and stroke). These conditions were likely related to a more advanced stage of HF as well as DM, which led to a higher rate of vascular complications, in contrast to patients who had DM but who did not meet the criteria for advanced HF. This finding has not been reported previously and could have an influence on the importance of variables such as evidence of disease progression observed in the random forest algorithm. It is known that DM, independent from its metabolic effects such as insulin-resistance, impacts on prognosis of ambulatory HF [
      • Cittadini A
      • Salzano A
      • Iacoviello M
      • Triggiani V
      • Rengo G
      • Cacciatore F
      • Maiello C
      • Limongelli G
      • Masarone D
      • Perticone F
      • Cimellaro A
      • Perrone-Filardi P
      • Paolillo S
      • Mancini A
      • Volterrani M
      • Vriz O
      • Castello R
      • Passantino A
      • Campo M
      • Modesti PA
      • De-Giorgi A
      • Monte IP
      • Puzzo A
      • Ballotta A
      • D'Assante R
      • Arcopinto M
      • Gargiulo P
      • Sciacqua A
      • Bruzzese D
      • Colao A
      • Napoli R
      • Suzuki T
      • Eagle KA
      • Ventura HO
      • Marra AM
      • Bossone E
      T.O.S.CA. Investigators
      Multiple hormonal and metabolic deficiency syndrome predicts outcome in heart failure: the T.O.S.CA.
      ] but also precipitates the development of atherosclerosis, including in coronary vessels. In advanced DM, the progression of vascular disease affects ventricular function, which can worsen and result in the progression of HF [
      • Bauters C
      • Lamblin N
      • Mc Fadden EP
      • Van Belle E
      • Millaire A
      • de Groote P
      Influence of diabetes mellitus on heart failure risk and outcome.
      ]. Additionally, advanced HF is related to marked insulin resistance [
      • Swan JW
      • Anker SD
      • Walton C
      • Godsland IF
      • Clark AL
      • Leyva F
      • Stevenson JC
      • Coats AJ
      Insulin resistance in chronic heart failure: relation to severity and etiology of heart failure.
      ], which will affect the progression of DM.
      Another difference found was a higher rate of chronic kidney disease (CKD) in patients with DM. It was more prevalent in patients with DM and without advanced HF (51.6%) than in patients without DM and with advanced HF (50.4%). CKD is a major microvascular complication in type 2 DM. Data from the UK and the USA show that around 20% and 40%, respectively, of patients with DM develop CKD [
      • Dean J.
      Organising care for people with diabetes and renal disease.
      ,
      • Bailey RA
      • Wang Y
      • Zhu V
      • et al.
      Chronic kidney disease in US adults with type 2 diabetes: an updated national estimate of prevalence based on Kidney Disease: Improving Global Outcomes (KDIGO) staging.
      ]. Patients with advanced HF and DM had the highest rate of CKD. In an advanced diabetic state, progressive fibrosis and proteinuria worsen as a result of many factors related to endothelial dysfunction [
      • Karnib HH
      • Ziyadeh FN
      The cardiorenal syndrome in diabetes mellitus.
      ], which are also implicated in the progression of heart disease. Therefore, this association could be useful for determining advanced disease in these patients.
      In regard to clinical manifestations, differences among advanced HF patients with and without DM were not significant. However, there was greater presence of nausea and erratic pain in patients with advanced HF and DM that could be related to the higher prevalence of autonomic neuropathy among these patients [
      • Low PA
      • Benrud-Larson LM
      • Sletten DM
      • Opfer-Gehrking TL
      • Weigand SD
      • O'Brien PC
      • Suarez GA
      • Dyck PJ
      Autonomic symptoms and diabetic neuropathy: a population-based study.
      ].
      In our study, patients with advanced HF had a poor prognosis and patients who also had DM had the poorest prognosis. As previous studies have shown, outcomes for patients with HF and DM are poor, especially in patients with ischemic HF [
      • Dries DL
      • Sweitzer NK
      • Drazner MH
      • Stevenson LW
      • Gersh BJ
      Prognostic impact of diabetes mellitus in patients with heart failure according to the etiology of left ventricular systolic dysfunction.
      ], which is the predominant etiology in this sample, similar to what has been described in previous research [
      • Stratton IM
      • Adler AI
      • Neil HA
      • Matthews DR
      • Manley SE
      • Cull CA
      • Hadden D
      • Turner RC
      • Holman RR
      Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study.
      ].
      In regard to advanced HF criteria, it is well-known that HF has a heterogeneous course, an aspect of the disease that makes it more difficult to specify whether a patient has advanced disease. It is common for patients with HF and their attending physician to remain hopeful regarding life expectancy in the final stages of this disease. This is reflected in the low rate of palliative care referrals among advanced HF patients (with and without DM) observed in our study.
      There are very few specific tools that help identify the terminal stage in patients with HF. Most are general indexes such as the CARING criteria (Cancer, Admissions > or = 2, Residence in a nursing home, Intensive care unit admission with multiorgan failure, > or = 2 Noncancer hospice Guidelines), which has a high sensitivity and specificity for death at one year [
      • Fischer S
      • Gozansky WS
      • Sauaia A
      • Min SJ
      • Kutner JS
      • Kramer A
      A practical tool to identify patients who may benefit from a palliative approach: the CARING Criteria.
      ] but was not specifically designed for patients with HF. The most commonly used criteria for terminality in patients with HF are those of the National Hospice Organization [
      Medical guidelines for determining prognosis in selected non-cancer diseases
      The National Hospice Organization.
      ] and of the European Society of Cardiology [
      • Crespo-Leiro MG
      • Metra M
      • Lund LH
      • Milicic D
      • Costanzo MR
      • Filippatos G
      • Gustafsson F
      • Tsui S
      • Barge-Caballero E
      • De Jonge N
      • Frigerio M
      • Hamdan R
      • Hasin T
      • Hülsmann M
      • Nalbantgil S
      • Potena L
      • Bauersachs J
      • Gkouziouta A
      • Ruhparwar A
      • Ristic AD
      • Straburzynska-Migaj E
      • McDonagh T
      • Seferovic P
      • Ruschitzka F
      Advanced heart failure: a position statement of the Heart Failure Association of the European Society of Cardiology.
      ], though the latter may be excessively focused on heart dysfunction. In addition, there are other models for predicting mortality specific to patients with HF [
      • Brophy JM
      • Dagenais GR
      • McSherry F
      • Williford W
      • Yusuf S
      A multivariate model for predicting mortality in patients with heart failure and systolic dysfunction.
      ,
      • Levy WC
      • Mozaffarian D
      • Linker DT
      • Sutradhar SC
      • Anker SD
      • Cropp AB
      • et al.
      The Seattle Heart Failure Model. Prediction of Survival in Heart Failure.
      ] that are also focused on heart disease, but they have important limitations in elderly patients, especially when predicting short-term mortality. It is precisely for these reasons that we chose to use criteria from two different instruments: the National Hospice Organization and the PALIAR Project scales [
      • Bernabeu-Wittel M
      • Ruiz-Cantero A
      • Murcia-Zaragoza J
      • Hernández-Quiles C
      • Barón-Franco B
      • Ramos-Cantos C
      • Nieto-Martín MD
      • Fernández-López A
      • Fernández-Moyano A
      • Moreno-Gaviño L
      • Ollero-Baturone M
      Precisión de los criterios definitorios de pacientes con enfermedades médicas no neoplásicas en fase terminal. Proyecto PALIAR [Reliability of different criteria in identifying end-of-life trajectory of patients with chronic medical diseases. PALIAR Project].
      ].
      In patients with advanced HF, it is likely that age along with the burden of comorbidities and unspecified symptoms lead to certain general criteria being more important instead of criteria more specific to HF. This once again points to the need to design a more accurate scale for these patients. A recently published work described how the care of terminally ill patients with HF was characterized by aggressive use of medical care and invasive techniques [
      • Sacco A
      • Morici N
      • Villanova L
      • Viola G
      • Lissoni B
      • Forni L
      • Mazza U
      • Oliva F
      Withdrawal of active treatments in terminally ill heart failure patients.
      ]; a better characterization of these patients would help avoid overtreatment and improve quality of care.
      One strength of this study is that the cohort is prospective and unselected, which lends trustworthiness and representativeness to the study, given that it analyzes real-world data. However, our work has several limitations. First, there are no studies that have used both general and cardiac criteria to define advanced HF. Second, in the six-month follow-up period, it was not specified whether a patient died in the hospital or after being discharged. Third, some variables that could play a role in defining advanced HF or determining prognosis were missing. For example, the baseline treatment, the presence of right HF or in the case of patients with DM, levels of glycated hemoglobin or albuminuria, time since diagnosis, or treatment with antidiabetic agents were not recorded. Fourth, although specialized palliative care was available in most hospitals, the low referral rate may be explained by the absence of specialists in some of the centers. Lastly, the random forest model was trained and tested on our data set. Validation with external data is necessary in order to confirm our findings.

      5. Conclusions

      Patients with advanced HF and DM have a different clinical profile than patients with advanced HF without DM. It was characterized by advanced DM with macro- and microvascular complications, mainly CKD, with trend to different clinical manifestations (nausea and erratic pains). The outcomes of advanced HF patients were poor and patients with advanced HF and DM had the worst prognosis. Finally, variables of the general criteria (PALIAR project) better defined advanced heart failure both in patients with DM as without it, being the evidence of disease progression a key variable in patients with DM whereas the emergency department visits were decisive in the case of patients without DM.

      Declaration of Competing Interest

      The authors of the present manuscript have nothing to disclose.

      Acknowledgments

      We would like to extend our gratitude to all investigators who form part of the EPICTER study (Appendix 1).

      Appendix B. Supplementary materials

      APPENDIX A

      T. Choucino-Fernández; A.B. Porto-Pérez; P. Piñeiro-Parga; C. Pedrosa-Fraga; R. Suárez-Gil; J.J. González-Soler; P. López-Mato; A. Latorre-Díez; A. Gómez-Gigirey; L. Ferreira-González; M. Sánchez-Cembellin; M. Gallego-Villalobos; J.P. Rugeles-Niño; E.E. Rodríguez-Avila; A. González-Franco; C. Guerra-Acebal; A. Sebastián-Leza; J. Monte-Armenteros; G. Frutos-Muñoyerro; C. Clemente-Sarasa; J. Díez-Manglano; C. Josa-Laorden; I. Torres-Courchoud; N. Gómez-Aguirre; R. Jordana-Camajuncosa; L.E. Cajamarca-Calva; I. Torrente-Jiménez; A. Serrado-Iglesias; L. M. Ceresuela; R. Salas-Campos; J. Delás-Amat; A. Brasé-Arnau; I. Petit-Salas; V. Romaní-Costa; A. Expósito-López; C.E. Sabbagh-Fajardo; J. Recio-Iglesias; C. Alemán-Llansó; J.M. Suriñach-Caralt; J.C. Trullás-Vila; A. Armengou-Arxe; S. García-Torras; J.L. Morales-Rull; C. Solé-Felip; A. Lacal-Martínez; M. Otero-Soler; A. Muela-Molinero; M. Carrera-Izquierdo; P. Arribas-Arribas; L. Inglada-Galiana; Á. Ruiz de Temiño; Á. Silva-Vázquez; L. Fuentes-Pardo; M. García-García; E. Piniella-Ruiz; B. Pérez-Alves; S. Gonzalo-Pascua; J. Marrero-Francés; M. Méndez-Bailón; F.J. Martín-Sánchez; M. Varas-Mayoral; M. Asenjo-Martínez; M. Yebra-Yebra; B. Sánchez-Sauce; B. Herreros-Ruiz; A. Quesada-Simón; I. Vives-Beltrán; J. Álvarez-Troncoso; L.A.Martínez-Marín; P. GilMartínez; E. Díaz deMayorga;M.A.Moreno-Palanco; L. Soler-Rangel; J. Abellán-Martínez; A.M. Colás-Herrera; G.T. López-Castellanos; R. Ruíz-Ortega; E. Ruiz-Barraza; M.L. Martín-Jiménez; E. Montero-Hernández; J.C. Arévalo-Lorido; J. Carretero-Gómez; P. Calderón-Jiménez; A. Herrero-Domingo; S. Martín-Barba; J.C. Blázquez-Encinar; C. Jiménez-Guardiola; J.M. Cepeda-Rodrigo; D. Quiles-García; S. Carrascosa-García; P. Llacer-Iborra; M.C. Moreno-García; L.F. Díez-García; P. Sánchez-López; M.J. Martínez-Soriano; E. Menor; M. Montero-Pérez-Barquero; M.P. Anguita-Sánchez; M. Sánchez-Moruno; M. Fuentes-Espínola; J.L. Zambrana-García; E. Guisado-Espartero; I. Mejías-Real; J.N. Alcalá-Pedrajas; F.J. Carrasco-Sánchez; C. Díaz-Pérez; M. Guzmán-García; S. Domingo-Roa; B. Cortés-Rodríguez; C. García-Redecillas; R. Martín-Navarro; R. Quirós-López; P. Macías-Ávila; I. Antequera-Martín-Portugués; M. Blanco-Soto; F.J. Flores-Álvarez; R. Aparicio-Santos; M.D. Nieto-Martín; R. García-Serrano; C. Jiménez-de-Juan; J. Ternero-Vega; M. Villalonga-Comas; M. Díaz-Cañestro; J. Asensio-Rodríguez; A. Gil-Díaz; I. Marrero-Medina; A. Puente-Fernández; D. Gudiño-Aguirre; M.F. Dávila-Ramos; E. Calderón.

      References

        • Kavalieratos D
        • Mitchell EM
        • Carey TS
        • Dev S
        • Biddle AK
        • Reeve BB
        • Abernethy AP
        • Weinberger M
        Not the 'grim reaper service'": an assessment of provider knowledge, attitudes, and perceptions regarding palliative care referral barriers in heart failure.
        J Am Heart Assoc. 2014; 3e000544
        • Ponikowski P
        • Voors AA
        • Anker SD
        • Bueno H
        • Cleland JG
        • Coats AJ
        • Falk V
        • González-Juanatey JR
        • Harjola VP
        • Jankowska EA
        • Jessup M
        • Linde C
        • Nihoyannopoulos P
        • Parissis JT
        • Pieske B
        • Riley JP
        • Rosano GM
        • Ruilope LM
        • Ruschitzka F
        • Rutten FH
        • van der Meer P
        Authors/Task Force Members; Document Reviewers. 2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: The Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.
        Eur J Heart Fail. 2016; 18: 891-975
        • Levy WC
        • Mozaffarian D
        • Linker DT
        • Sutradhar SC
        • Anker SD
        • Cropp AB
        • Anand I
        • Maggioni A
        • Burton P
        • Sullivan MD
        • Pitt B
        • Poole-Wilson PA
        • Mann DL
        • Packer M
        The Seattle Heart Failure Model: prediction of survival in heart failure.
        Circulation. 2006; 113: 1424-1433
        • O'Connor CM
        • Hasselblad V
        • Mehta RH
        • Tasissa G
        • Califf RM
        • Fiuzat M
        • Rogers JG
        • Leier CV
        • Stevenson LW
        Triage after hospitalization with advanced heart failure: the ESCAPE (Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness) risk model and discharge score.
        J Am Coll Cardiol. 2010; 55: 872-878
        • Lee DS
        • Austin PC
        • Rouleau JL
        • Liu PP
        • Naimark D
        • Tu JV
        Predicting mortality among patients hospitalized for heart failure: derivation and validation of a clinical model.
        JAMA. 2003; 290: 2581-2587
        • Lehrke M
        • Marx N
        Diabetes Mellitus and Heart Failure.
        Am J Cardiol. 2017; 120: S37-S47
        • Dei-Cas A
        • Fonarow GC
        • Gheorghiade M
        • Butler J
        Concomitant diabetes mellitus and heart failure.
        Curr Probl Cardiol. 2015; 40: 7-43
        • Fernández-Martinez J
        • Romero-Correa M
        • Salamanca-Bautista P
        • Aramburu-Bodas Ó
        • Formiga F
        • Vázquez-Rodríguez P
        • Conde-Martel A
        • García-García JA
        • Páez-Rubio I
        • López-Reboiro M
        • Sánchez-Sánchez C
        • Arias-Jiménez JL
        Prevalence of advanced heart failure and use of palliative care in admitted patients: Findings from the EPICTER study.
        Int J Cardiol. 2021; 327: 125-131
        • Medical guidelines for determining prognosis in selected non-cancer diseases
        The National Hospice Organization.
        Hosp J. 1996; 11: 47-63
        • Bernabeu-Wittel M
        • Ruiz-Cantero A
        • Murcia-Zaragoza J
        • Hernández-Quiles C
        • Barón-Franco B
        • Ramos-Cantos C
        • Nieto-Martín MD
        • Fernández-López A
        • Fernández-Moyano A
        • Moreno-Gaviño L
        • Ollero-Baturone M
        Precisión de los criterios definitorios de pacientes con enfermedades médicas no neoplásicas en fase terminal. Proyecto PALIAR [Reliability of different criteria in identifying end-of-life trajectory of patients with chronic medical diseases. PALIAR Project].
        Rev Esp Geriatr Gerontol. 2010; 45: 203-212
        • Bernabeu-Wittel M
        • Murcia-Zaragoza J
        • Hernández-Quiles C
        • Escolano-Fernández B
        • Jarava-Rol G
        • Oliver M
        • Díez-Manglano J
        • Ruiz-Cantero A
        • Ollero-Baturone M
        • Researchers PALIAR
        Development of a six-month prognostic index in patients with advanced chronic medical conditions: the PALIAR score.
        J Pain Symptom Manage. 2014; 47: 551-565
        • Gómez-Aguirre N
        • Fuertes-Ruiz D
        • Gracia-Tello B
        • Clemente-Sarasa C
        • Artajona-Rodrigo E
        • Cabrerizo-García JL
        • de Escalante-Yangüela B
        • Bueno-Castel MC
        • Velilla-Marco J
        • Díez-Manglano J
        External validation of the PALIAR index for patients with advanced, nononcologic chronic diseases.
        Aging Clin Exp Res. 2019; 31: 393-402
        • Breiman L
        Random forests.
        Machine Learning. 2001; 45: 5-32https://doi.org/10.1023/A:1010933404324
        • Liaw A
        • Wiener M
        Classification and Regression by randomForest.
        R News. 2002; 2: 18-22
        • Strobl C
        • Boulesteix AL
        • Zeileis A
        • et al.
        Bias in random forest variable importance measures: illustrations, sources, and a solution.
        BMC Bioinformatics. 2007; 8: 25https://doi.org/10.1186/1471-2105-8-25
        • Iglay K
        • Hannachi H
        • Joseph-Howie P
        • Xu J
        • Li X
        • Engel SS
        • Moore LM
        • Rajpathak S
        Prevalence and co-prevalence of comorbidities among patients with type 2 diabetes mellitus.
        Curr Med Res Opin. 2016; 32: 1243-1252
        • Cittadini A
        • Salzano A
        • Iacoviello M
        • Triggiani V
        • Rengo G
        • Cacciatore F
        • Maiello C
        • Limongelli G
        • Masarone D
        • Perticone F
        • Cimellaro A
        • Perrone-Filardi P
        • Paolillo S
        • Mancini A
        • Volterrani M
        • Vriz O
        • Castello R
        • Passantino A
        • Campo M
        • Modesti PA
        • De-Giorgi A
        • Monte IP
        • Puzzo A
        • Ballotta A
        • D'Assante R
        • Arcopinto M
        • Gargiulo P
        • Sciacqua A
        • Bruzzese D
        • Colao A
        • Napoli R
        • Suzuki T
        • Eagle KA
        • Ventura HO
        • Marra AM
        • Bossone E
        • T.O.S.CA. Investigators
        Multiple hormonal and metabolic deficiency syndrome predicts outcome in heart failure: the T.O.S.CA.
        Registry. Eur J Prev Cardiol. 2021; 28: 1691-1700
        • Bauters C
        • Lamblin N
        • Mc Fadden EP
        • Van Belle E
        • Millaire A
        • de Groote P
        Influence of diabetes mellitus on heart failure risk and outcome.
        Cardiovasc Diabetol. 2003; 2: 1
        • Swan JW
        • Anker SD
        • Walton C
        • Godsland IF
        • Clark AL
        • Leyva F
        • Stevenson JC
        • Coats AJ
        Insulin resistance in chronic heart failure: relation to severity and etiology of heart failure.
        J Am Coll Cardiol. 1997; 30: 527-532
        • Dean J.
        Organising care for people with diabetes and renal disease.
        J Ren Care. 2012; 8: 23-29
        • Bailey RA
        • Wang Y
        • Zhu V
        • et al.
        Chronic kidney disease in US adults with type 2 diabetes: an updated national estimate of prevalence based on Kidney Disease: Improving Global Outcomes (KDIGO) staging.
        BMC Res Notes. 2014; 7: 415
        • Karnib HH
        • Ziyadeh FN
        The cardiorenal syndrome in diabetes mellitus.
        Diabetes Res Clin Pract. 2010; 89: 201-208
        • Low PA
        • Benrud-Larson LM
        • Sletten DM
        • Opfer-Gehrking TL
        • Weigand SD
        • O'Brien PC
        • Suarez GA
        • Dyck PJ
        Autonomic symptoms and diabetic neuropathy: a population-based study.
        Diabetes Care. 2004; 27: 2942-2947
        • Dries DL
        • Sweitzer NK
        • Drazner MH
        • Stevenson LW
        • Gersh BJ
        Prognostic impact of diabetes mellitus in patients with heart failure according to the etiology of left ventricular systolic dysfunction.
        J Am Coll Cardiol. 2001; 38: 421-428
        • Stratton IM
        • Adler AI
        • Neil HA
        • Matthews DR
        • Manley SE
        • Cull CA
        • Hadden D
        • Turner RC
        • Holman RR
        Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study.
        BMJ. 2000; 321: 405-412
        • Fischer S
        • Gozansky WS
        • Sauaia A
        • Min SJ
        • Kutner JS
        • Kramer A
        A practical tool to identify patients who may benefit from a palliative approach: the CARING Criteria.
        J Pain Symptom Manage. 2006; 31: 285-292
        • Crespo-Leiro MG
        • Metra M
        • Lund LH
        • Milicic D
        • Costanzo MR
        • Filippatos G
        • Gustafsson F
        • Tsui S
        • Barge-Caballero E
        • De Jonge N
        • Frigerio M
        • Hamdan R
        • Hasin T
        • Hülsmann M
        • Nalbantgil S
        • Potena L
        • Bauersachs J
        • Gkouziouta A
        • Ruhparwar A
        • Ristic AD
        • Straburzynska-Migaj E
        • McDonagh T
        • Seferovic P
        • Ruschitzka F
        Advanced heart failure: a position statement of the Heart Failure Association of the European Society of Cardiology.
        Eur J Heart Fail. 2018; 20: 1505-1535
        • Brophy JM
        • Dagenais GR
        • McSherry F
        • Williford W
        • Yusuf S
        A multivariate model for predicting mortality in patients with heart failure and systolic dysfunction.
        Am J Med. 2004; 116: 300-304
        • Levy WC
        • Mozaffarian D
        • Linker DT
        • Sutradhar SC
        • Anker SD
        • Cropp AB
        • et al.
        The Seattle Heart Failure Model. Prediction of Survival in Heart Failure.
        Circulation. 2006; 113: 1424-1433
        • Sacco A
        • Morici N
        • Villanova L
        • Viola G
        • Lissoni B
        • Forni L
        • Mazza U
        • Oliva F
        Withdrawal of active treatments in terminally ill heart failure patients.
        Int J Cardiol. 2021; (S0167-5273(21)00815-9)https://doi.org/10.1016/j.ijcard.2021.05.003