Introduction

Background

Heart failure (HF) is one of the most common syndromes in adults, with a prevalence of 1–2% [1]. Exacerbations affect the natural history of HF. They often require urgent therapy and hospitalisation, and in-hospital mortality during the acute heart failure (AHF) episode remains high. In addition, despite most patients having symptomatic improvement in response to therapy received during hospitalisation [2], AHF patients have been identified as a group with extremely high post-discharge death and rehospitalisation rates, especially during the first 60–90 days after discharge. During this early post-discharge period with increased risk of adverse outcomes, termed “the vulnerable phase” [3], readmission for HF occurs in 15–30% of cases, and the mortality rate ranges from 7 to 11% [2,3,4,5].

Importance

The first episode of AHF (de novo AHF) usually marks a change in the natural history of this syndrome, as it usually represents the first step towards progressive patient deterioration and functional decline. Recently, it has been reported that this first decompensation is followed by a rate of 0.87 of subsequent hospitalisations per year, half of which are caused by new AHF episodes [6, 7]. In this scenario, the de novo AHF episode constitutes an essential step for the identification of patient- and episode-related risk factors associated with poor outcomes, as active actions on modifiable risk factors at this early moment could, in turn, improve patient prognosis [8,9,10,11,12,13]. Since patients seek medical care at the emergency department (ED) in most AHF episodes, EDs constitute a key setting for investigating AHF-related aspects. However, to the best of our knowledge, very few studies have analysed the natural history of patients after being diagnosed with de novo AHF, especially in unbiased patient samples from EDs.

The NOVICA (apocopate from the Spanish words “de NOVO Insuficiencia Cardiaca Aguda; in English, de novo AHF) project was designed to cover this gap in the literature. In two previously published studies, we described the characteristics and outcome of patients presenting their first AHF episode in the ED in comparison with patients with previous history of AHF (NOVICA-1) [14], as well as the natural history of a cohort of patients followed for a mean of 2.4 years after the de novo AHF episode (NOVICA-3) [7].

Goals of this investigation

In the present study (NOVICA-2), our objective was to identify risk factors associated with in-hospital mortality and poor outcomes (consistent with a combined endpoint formed by need for rehospitalisation or death) during the early post-discharge period (vulnerable phase) after the de novo AHF episode.

Methods

Setting

The EAHFE (Epidemiology of Acute Heart Failure in Emergency Department) Registry was initiated in 2007 and every 2–3 years carries out a 1- to 2-month recruitment period of all consecutive patients diagnosed with AHF in Spanish EDs participating in the project. To date, six recruitment phases (2007, 2009, 2011, 2014, 2016 and 2018) have been performed with the participation of 45 EDs from community and university hospitals across Spain (representing about 15% of the Spanish public health care system hospitals), enrolling a total of 18,370 AHF patients. The NOVICA-2 study only included patients from the EAHFE registries 4 and 5 since the previous registries did not have the necessary information for the present analysis (rehospitalisation due to AHF was added from EAHFE-4 and is one of the outcomes assessed in the present study), and the EAHFE-6 follow-up was not completed at the time of the study design.

Design

Details of patient inclusion dynamics have been reported previously [15, 16]. In brief, any attending emergency physician in the participating EDs, who received specific study protocol instructions during a weekly ED meeting preceding patient recruitment, can enrol patients. These physicians are responsible for the detection of potential cases of patients with AHF. All suspected cases are confirmed by the principal investigator of each centre to ensure the patients meet the diagnostic criteria of AHF based on the Framingham clinical criteria [17]. If possible, the diagnosis is also confirmed by the measurement of plasma natriuretic peptide and/or echocardiography during ED or hospital stay following the current recommendations of the European Society of Cardiology (ESC) guidelines, and this was available in about 92% of cases. The principal investigator of each centre is responsible for the final diagnostic adjudication of the cases. All principal investigators are provided with a common dictionary of terms to have standard definitions at all the centres (Supplemental Table 1).

Selection of participants

We recruited patients in the ED by a face-to-face interview with them and their caregivers at admission to the emergency department. For the present study, we included patients with a first episode of AHF irrespective of whether they did or did not have a previous diagnosis or treatment of HF made by general practitioners or ambulatory specialists. The main condition was that the patient had never consulted to an ED or been hospitalised because of AHF. All patients with a final diagnosis of first episode of AHF after ED care were recorded and used to evaluate one of the co-primary outcomes (all-cause in-hospital mortality), and those discharged alive after the index AHF episode were followed up to 90 days to be included in the evaluation of the other co-primary outcome (the combined endpoint formed by rehospitalisation or death).

The only exclusion criterion for inclusion in the EAHFE Registry is a primary diagnosis of ST-elevation myocardial infarction (STEMI) while concurrently developing AHF (which occurs in 3% of AHF cases). The EAHFE Registry does not include any planned intervention, and the management of patients is entirely based on the attending ED physician decisions.

Variables recorded

Trained clinical researchers extracted data from the initial interview with patients and their caregivers regarding functional class before admission according to the New York Heart Association [NYHA] classification and functional status (Barthel Index) in the 2 weeks before admission and was checked their medical history. We completed the data obtained from the initial interview by reviewing the hospital and primary care clinical report. We recorded data directly for each patient on a pre-established data collection form. We recorded the following variables: demography (sex, age and place of residence); cardiac history (HF, ischaemic cardiopathy, valvulopathy, LVEF) and preexisting comorbidities (hypertension, atrial fibrilation, diabetes mellitus, dyslipidemia, renal failure defined as creatinine ≥ 2 mg/dl, chronic obstructive pulmonary disease, stroke, peripheral artery disease, dementia) were assessed (yes/no) if they were recorded in the patient's medical records. HFpEF was defined as an ejection fraction > 50%. We assessed the severity of the de novo AHF episode by (1) calculation of the MEESSI (Multiple Estimation of risk based on the Spanish Emergency department Score In patients with AHF) risk score [18], and (2) the need for hospitalisation during the index episode. We defined the index episode as “patients requiring ED care due to an episode of AHF”.

Outcomes

The two co-primary outcomes in the present study were in-hospital all-cause mortality and 90-day post-discharge combined endpoint. The latter was considered if rehospitalisation due to AHF or all-cause death occurred during follow-up. These two components of the 90-day post-discharge combined endpoint were individually considered as the secondary outcomes. The follow-up was performed through telephone contact and consultation of medical records. We completed the data obtained in the initial interview and 90 days after discharge by reviewing the hospital and primary care clinical report. If we did not find information, patients, relatives, or caregivers were contacted by phone 90 days after ED discharge and data concerning mortality were recorded. We chose the 90-day time span for the vulnerable period based on the previous definition by Greene et al. [19].

Statistical analysis

Continuous variables are expressed as mean and standard deviation (SD) or median [interquartile range (IQR)], and discrete variables as absolute values and percentages. Comparison among groups was carried out using the Student’s t test, after checking with the Kolmogorov–Smirnov test to ensure that they fit a normal distribution, or the Mann–Whitney non-parametric test for non-normally distributed variables. The chi-square test (for trend, when needed) was used for comparing discrete variables.

The effect size of the independent variables on the risk of in-hospital all-cause mortality (co-primary outcome) was expressed by odds ratio (OR) calculated by logistic regression, while for the other co-primary outcome (90-day post-discharge combined endpoint) and for the secondary outcomes (90-day post-discharge readmission or death) it was estimated by hazard ratio (HR) calculated by Cox regression. All effect size estimations were calculated in adjusted multivariate models by entering all the variables showing a significant association in the univariate analyses. For calculations of adjusted OR and HR, missing values in the independent variables were replaced using the multiple imputation technique provided by SPSS software, generating five datasets in which there are no misses among all the variables included in the adjustment.

We also investigated if there was a significant different size effect according to patient age and the left ventricular ejection fraction (LVEF) for the risk factors that finally resulted independently associated with the primary outcomes. This was assessed by measuring the first-order interaction of age (dichotomised as ≤ 80 years or > 80 years) and LVEF (dichotomised as < 50%—heart failure with reduced or mid-range ejection fraction or ≥ 50%—heart failure with preserved ejection fraction) with each risk factor in the multivariate adjusted logistic models.

Statistical significance was accepted if the 95% confidence interval (CI) of the OR or the HR excluded the value 1, or if the p value was less than 0.05. Since this was an exploratory study, a pre-hoc sample size calculation was not made. We used the SPSS, version 24.0 (SPSS Inc Chicago, USA) for all analyses.

Ethics

The EAHFE Registry 4 and 5 protocols were approved by a central Ethics Committee at the Hospital Universitario Central de Asturias (Oviedo, Spain) with the reference numbers 166/13 and 160/15. Due to the non-interventional design of the registry, Spanish legislation allows central Ethical Committee approval, accompanied by notification to the local Ethical Committees. At admission, all participating patients gave informed consent (writing consent) to be included in the registry and to be contacted for follow-up. If a patient does not have the capacity to consent their participation in the study, to determine what his/her wishes would have been, we consult his/her close family or caregiver. The NOVICA-2 study was carried out in strict compliance with the Declaration of Helsinki principles. The authors designed the study, gathered, and analysed the data, vouch for the data and analysis, wrote the paper, and decided to publish.

Results

EAHFE-4 and 5 included 7946 patients with AHF, and 3422 patients met the inclusion criteria of the NOVICA-2 study (Fig. 1). The mean age of the patients was 80 (SD 11) years, and 52.1% were women. The main comorbidities were hypertension (79%), diabetes (37%) and atrial fibrillation (38%). Most patients had preserved left ventricular ejection fraction (72%), and nearly half had some degree of dependence in basic activities of daily living (Barthel Index of 90 or less; 48%). The most frequent precipitants of the index episode were infection (35%) and rapid atrial fibrillation (19%). With respect to the severity of the acute episode, 48% were classified as low risk by the MEESSI scale, 36% as intermediate risk and 16% as high/very high risk (distribution of patients in the 13 variables included in the MEESSI scale is presented in the Supplemental Table 1), and 2501 patients (73%) needed to be hospitalised during the index episode. Patient distribution according to the NYHA class recorded at patient arrival to the ED differed from that observed at baseline (Table 1), as few patients were in class I and II (1.7% and 14.1%, respectively) and most were in class III and IV (46.8% and 37.4%). The demographic, baseline and episode characteristics of the patients are summarised in Table 1. There were three variables (Barthel Index, left ventricular ejection fraction and classification of the severity of index episode by MEESSI scale that had more than 10% of missing values (comparison between patients with missing values for these three variables with the rest of patients in presented in Supplemental Table 2).

Fig. 1
figure 1

Flowchart for patient inclusion

Full size image
Table 1 Baseline and acute heart failure episode characteristics corresponding to the 3422 patients included in the present study
Full size table

Two hundred and thirty-five patients died during hospitalisation (in-hospital mortality 6.9%). Table 2 shows the differences between survivor and non-survivor patients during the index AHF event. Patients who died were older, and more commonly had chronic kidney insufficiency, dementia and active neoplasia as comorbidities compared to survivor patients. In addition, they had a worse functional and NYHA baseline status and were less frequently on chronic beta-blocker treatment. The acute episode was more frequently triggered by infection and less frequently by hypertensive episodes, and the severity of decompensation was higher (Table 2). However, only five of these risk factors remained statistically significant in the adjusted analysis: two comorbidities (OR = 2.25, 1.62–3.14, for dementia and OR = 1.97, 1.41–2.76 for active neoplasia), the functional dependence at baseline (OR = 1.58, 1.02–2.43), being on chronic treatment with beta-blocker (OR = 0.62, 0.44–0.86), and the severity of decompensation (OR = 6.38, 2.85–14.26 for the high-/very high-risk patients compared to low-risk patients) (Fig. 2) (see Supplemental tables 3–6: Unadjusted analysis of risk factors related to outcomes during the first episode of acute heart failure).

Table 2 Comparison of the patients’ baseline and acute heart failure episode characteristics depending on the presence or absence of the outcomes evaluated in the present study
Full size table
Fig. 2
figure 2

Adjusted analysis of risk factors related to all-cause in-hospital mortality in patients during the first episode of acute heart failure. Adjustment was performed by including in the multivariate logistic regression all variables with a significant different distribution in the univariate analysis (see Table 2). Bold numbers denote statistical significance. AHF acute heart failure.

Full size image

The median length of stay for all patients discharged alive was 5 days (IQR = 1–9) and 7 days (IQR = 5–11) in the subset of patients that were hospitalised (i.e., excluding patients directly discharged from the ED without hospitalisation). The median length of stay for patients dying during the index event was 5 days (IQR = 2–11).

The vast majority of patients dying during the index event (221 out of 235, 94%) had been hospitalised, as only 14 died shortly after ED arrival before hospitalisation. By limiting the calculation of in-hospital mortality to only hospitalised patients, the percentage would increase from 6.9% (235 out of 3422 of index episodes) to 8.8% (221 out of 2501 hospital admissions).

Of the 3187 patients discharged alive, we obtained follow-up data on 3138 patients for the 90-day post-discharge analyses (Fig. 1). The 90-day combined endpoint was observed in 607 patients (19.3%) and six risk factors were independently associated with this outcome in the adjusted analysis (Fig. 3): four comorbidities (HR = 1.40, 1.11–1.76 for hypertension, HR = 1.23, 1.01–1.49 for chronic renal insufficiency, HR = 1.24, 1.01–1.51 for heart valve disease, and HR = 1.22, 1.01–1.48 for chronic obstructive pulmonary disease), the NYHA class 3 or 4 at baseline (HR = 1.40, 1.12–1.74) and the severity of decompensation (HR = 1.23, 1.01–1.50; and HR = 1.64, 1.20–2.25; for intermediate and high-/very high-risk patients, respectively). The secondary outcomes were observed in 546 patients (17.4%) for 90-day post-discharge rehospitalisation and in 240 patients (7.6%) for 90-day post-discharge death. Although many of the risk factors associated with the secondary outcomes coincided with the risk factors for the combined endpoint, different risk factors were observed for 90-day post-discharge rehospitalisation (hypertension, peripheral arteriopathy, chronic obstructive valve disease, NYHA 3–4, dietetic/therapeutic transgression as AHF trigger and severity of the de novo AHF episode) and death (chronic kidney insufficiency, NYHA 3–4, and severity of the de novo AHF episode) (Fig. 3).

Fig. 3
figure 3

Adjusted analysis of risk factors related to adverse events observed during the vulnerable post-discharge phase (90 days following the discharge after the index de novo acute heart failure event): a combined endpoint (panel A), rehospitalisation (panel B) and death (panel C). Adjustment of hazard ratios was performed by including in the multivariate Cox regression all variables with a significant different distribution in the univariate analysis (see Table 2). Bold numbers denote statistical significance. AHF acute heart failure.

Full size image

The effect size of the risk factors found to be significantly associated with the two co-primary adverse outcomes in the adjusted models did not significantly differ according to age, with the exception of the relationship between active neoplasia and in-hospital mortality, which was significantly higher in patients ≤ 80 years of age (OR = 3.28, 1.80–5.96) than in patients with > 80 years (OR = 1.58, 0.62–4.05; p-for-interaction = 0.049). On the other hand, no interaction was found in the relationship between any of the risk factors and either of the two co-primary endpoints when the analysis was repeated and stratified by the LVEF.

Discussion

Our study provides further evidence on prognostic factors in patients presenting their first AHF episode and seen at the emergency department. We identified four factors which increased the risk of in-hospital mortality in patients with de novo AHF (dementia, neoplasia, functional dependence in basic activities of daily living, and severity of the de novo AHF index episode). During the vulnerable phase, six independent risk factors for combined endpoint were related to poor outcomes (hypertension, chronic kidney insufficiency, heart valve disease, chronic obstructive pulmonary disease, NYHA functional class III–IV, and severity of the index decompensation).

The main baseline characteristics of patients with de novo AHF were similar to those of other studies conducted in EDs, but differed from those reported for hospitalised patients (especially in cardiology departments) because our patients were older, fewer had previous history of coronary artery disease, and the most predominant type of HF according to left ventricular function was a preserved ejection fraction [11, 13]. In line with previous studies, we found an in-hospital mortality of 6.9% as well as 7.6% of mortality and 17.4% of readmission for AHF during the vulnerable post-discharge period (90 days post-discharge) [4, 11, 20, 21].

Our study has two major findings. The first was related to in-hospital mortality. We identified four factors which increased the risk of death (dementia, neoplasia, functional dependence in basic activities of daily living, and severity of the de novo AHF index episode) and one conferring protection against in-hospital death (being on chronic treatment with beta-blockers). Our findings highlight the importance of the initial clinical assessment in the ED including functional and cognitive evaluation of the patient. Dementia and activities of daily living have been proposed as the most representative prognostic indicators for survival in elderly patients in many chronic diseases [22, 23]. Multiple studies have highlighted their importance as factors of poor short-term prognosis in elderly patients who present at the ED with an episode of AHF [22, 24, 25]. It seems that beta-blocker treatment exerted a protective effect against adverse short-term outcomes during the de novo AHF episode [26]. We do not know the exact reason for the treatment with beta-blockers in our patients at the time of the debut of AHF (perhaps due to previous HF diagnosis without any previous decompensation, but also to other comorbidities such as hypertension or atrial fibrillation). An LVEF-stratified mortality analysis was not performed, and our study was not designed for this. Finally, the severity of the first decompensation, assessed by the MEESSI score designed to predict 30-day mortality, was associated with the risk of in-hospital mortality in de novo AHF patients.

The second major finding concerned factors related to poor outcomes during the vulnerable phase. We identified six independent risk factors for the combined endpoint (hypertension, chronic kidney insufficiency, heart valve disease, chronic obstructive pulmonary disease, NYHA functional class III–IV, and severity of the index decompensation). Many of these factors have been reported in previous studies [27,28,29,30], and underline the importance of comorbidity and functional status for identifying patients with the first episode of AHF with a poor short-term prognosis. Strikingly, dementia and functional dependence were not found to be markers of poor outcomes in our patients. We believe the differences found in our post-discharge model with respect to in-hospital mortality model may have been influenced by the organisation of the local health system and by the patients’ clinical characteristics. Thus, it is likely that since the patients had a worse functional status and dementia, some of the consultations could have occurred in primary care (consultation with the on-call primary care physician) or other emergency health care providers for fragile patients or highly dependent patients (i.e., nursing home).

It is also of note that some variables classically related to short- and long-term prognosis in patients with HF were not related to the short-term prognosis of the de novo AHF episode. In this sense, the LVEF was not associated with any outcomes, while the presence of limited functional capacity assessed by the Barthel Index and, especially, advanced NYHA classes at baseline were. It is possible that this indicates that there is a subset of patients with a previous long history of well-compensated HF, although this aspect was not investigated in the present study. On the other hand, in the present study, hospitalisation during the de novo AHF episode was not related to outcomes, and this contrasts with the 2016 ESC Guidelines that recommend to admitting every patient in the first episode of AHF (26). Perhaps, if outpatient multidisciplinary pathways were well developed, all the study and therapeutic approaches needed after de novo AHF could be carried out in a subset of patients without hospitalisation. Finally, the lack of influence of age on any of the short-term outcomes of patients with de novo AHF is also highly remarkable. This finding seems to be consistent since age did not modify the effect on outcomes of the significant risk factors (with the exception of actual neoplasia on in-hospital mortality during the index AHF event, which was higher in younger patients).

The NOVICA-2 has a number of limitations. First, AHF was defined using Framingham criteria. These criteria were validated for chronic HF, but they have also shown good diagnostic accuracy for the diagnosis of AHF, with similar specificity for systolic and diastolic HF (89%) and higher sensitivity for systolic (97%) than for diastolic HF (89%) [31, 32]. Second, because of the observational nature of this study, not all patients had variables of interest, such as the LVEF or NT-pro-BNP, and some variables were not recorded (i.e., frailty). This could introduce some bias in our findings. However, the main clinical variables were systematically collected. Nevertheless, it should be kept in mind that our intention was to perform the study under conditions of routine clinical practice, mainly using the variables available to ED physicians in their daily decision-making processes. Third, by design, the study cannot distinguish patients with de novo AHF with known previous heart disease from patients without prior heart disease. Fourth, we excluded AHF in patients with STEMI. Fifth, we do not have the specific causes of death. Sixth, since there was no sample size calculation due to the exploratory nature of the study, a type II error cannot be excluded in some of the estimations made. The strength of the study is that, in contrast to most previous studies based on hospital patients, NOVICA-2 recruitment was done in the ED, providing a more global picture due to the inclusion of patients with different AHF profiles.

In conclusion, our findings highlight the importance of stratifying patients according to risk severity and assessing basal status with NYHA functional class and Barthel Index. In addition, it is important to assess some comorbidities and basic activities of daily living as well as the cognitive status of patients in the ED, as they are also related to in-hospital and vulnerable post-discharge phase adverse outcomes in patients with de novo AHF. Finally, the age of debut of AHF does not seem to influence in-hospital mortality or adverse outcomes during the vulnerable post-discharge phase.