Validation of the ELAN-HF Score and self-care behaviour on the nurse-led heart failure clinic after admission for heart failure

The concept of HF nurses working in an outpatient clinic was for first time described in 1983 [6]. This was followed by the first nurse‐led HF clinic started in Sweden in 1990 after which they spread out to many Swedish hospitals. Nurse‐led HF clinics reduce the need for hospital care since titration of drugs can be rapidly achieved. Furthermore, studies indicate that early follow‐up after hospitalization may prevent readmissions [6].

Similar to these successes on HF clinics, several studies on other nurse-led clinics also indicated positive effects [7]. Rich et al. investigated the effect of a multidisciplinary, nurse‐directed intervention and found that the intervention improved the patients’ compliance, quality of life and decreased the rate of readmission and the healthcare costs [8]. Nurses independently perform anamnesis and physical examination, and are responsible for the diagnostic processes.

Nurse-led HF clinics provide education on self-care and psychosocial support to patients and their family. Programs employing multidisciplinary teams and in-person communication led to fewer HF hospital readmissions [9]. High-risk HF patients (advanced stage, low self-care skills, elderly, and those with frequent readmissions) may be expected to benefit the most from improvements in HF self-care knowledge and home care behaviour skills [9].

Prognostication of patients is useful in triaging patients during and after hospitalization [10, 11]. For this purpose, specific predictors for readmission of patients with HF have emerged [12]. By combining different clinical and laboratory parameters in a clinical prediction model, patients can be triaged just before discharge. In patients with HF plasma biomarkers Brain Natriuretic Peptide (BNP) and N-terminal pro-B-type Natriuretic Peptide (NT-proBNP) are commonly used. They indicate the severity of congestion and cardiac dysfunction and predict morbidity and mortality [13].

Various risk models for readmissions and mortality in HF have already been developed [11, 14‐17]. They incorporated the natriuretic peptide levels, measured either at admission or discharge, while some models also use their change during hospitalization. The PRIMA II trial, that investigated the influence of changing of NT-proBNP with guided therapy and intensified HF care pre-discharge, did not demonstrate to improve prognosis in these patients [18]. This study may have been hampered by a relatively overestimated expected benefit in the third of patients who were guided after having not reached the expected drop in NT-proBNP levels [18, 19].

The ELAN-HF (European collaboration on Acute decompensated Heart Failure) score is a model which is different from other risk models because it incorporates absolute discharge NT-proBNP levels, but also the percentage change in NT-proBNP, along with clinical risk markers [14]. Although it has been already validated retrospectively in a cohort of 325 patients, multiple external validations are needed to generalize the ELAN-HF score as a prediction model before it can be implemented in the clinical practice of the nurse-led HF clinic [20, 21].

Self-care behaviour is defined as the behaviour that consists of the decisions and strategies that a person undertakes for the sake of livelihood, healthy functioning and well-being [22, 23], as shown in Table 1. Previous studies indicated that optimal self-care behaviour can lead to fewer hospital admissions for HF [24]. Furthermore, is has been prospectively demonstrated that the use of information to improve self-care in HF led to a 30% decrease in readmission and outpatient visits within 30 days of discharge [24].

Self-care behaviour can be scored by using the European Heart Failure Self-care Behavioural Scale (EHFScBS-9), containing 9 items grouped around consulting behaviours and adherence with the regimen. Each of the items is graded with a 5-point Likert scale. This questionnaire was validated in several countries and was improved in 2014 [5].

The first aim of this study was to validate the predictive value of the ELAN-HF score on readmission and/or mortality in a prospective study of a hospitalized HF population.

The second aim was to assess the effect of self-care behaviour on readmission and mortality in these patients.

We conducted a quantitative, prospective, single centre cohort study. The primary endpoint is a composite endpoint of re-admission and /or all-cause mortality at 180 days.

The secondary endpoint is all-cause mortality at 180 days. Patients hospitalized for ADHF were included for three months (October—December 2017, Fig. 1).

A readmission was defined as an urgent clinical admission with a duration of at least 24 h after a previous discharge from the hospital.

Two sources were used for data collection. The data for both the outcome variables (readmission and / or mortality within 180 days), as well as the baseline characteristics and biomarkers (independent variables), were extracted from the electronic patient file, whereas missing data on mortality were retrieved from the general physician. The ELAN-HF score was calculated with this data (Table 2).

Patients with ADHF were included at admission. Excluded were patients with cognitive limitations, inability to speak the Dutch language and patients who were unable of follow-up.

The local ethics committee of the Catharina Hospital approved the study. All investigators adhered to the principles of the Declaration of Helsinki. The measurements performed during this study were part of routine care. In addition, all included patients were informed orally and in writing by the investigator and gave written consent.

The ELAN-HF score, as shown in Table 2, is categorized by Salah et al. into four risk categories (low, intermediate, high and very high) which corresponds to increasing 6-month mortality rates (3.6%, 9.2%, 23.5% and 51.1%).

To assess self-care behaviour, the Dutch version of the European Heart Failure Self-care Behavioural Scale (EHFScBS-9), was used [25]. The questionnaire was handed out by the investigator during admission and was completed by the patient without the presence of the investigator. The EHFScBS-9 score ranges from 9 – 45, but has been standardised to a scale of 0 – 100 with higher scores indicating better self-care [5]. The standardised score can be obtained by using the SPSS syntax provided by the study of Vellone et al. and also for the interpretation of which a cut-off point < 50 means sub-optimal score of self-care, a score 50–100 can be seen as an optimal/ good score of self-care [5].

The statistical analyses consisted of two parts: (1) external validation of the ELAN-HF score in our study cohort (2) and survival analysis of the ELAN-HF score, EHFScBS-9 score and other clinically relevant variables.

External validation is the process of evaluating model performance in a sample independent of that used to develop the model. The outcome used for external validation was 6-month all-cause mortality, analogous to Salah et al. [14]. The 6-month mortality rates for the four risk groups as reported by Salah et al. were compared to those of our study cohort. The external validation steps performed in this study are described in more detail by Royston et. al 2013 [26]. First, the ELAN-HF linear predictor was calculated by using the regression coefficients in Table 2. The ELAN-HF linear predictor was then used as a covariate in a Cox PH model. A likelihood-ratio (LR) test was performed to test whether the slope of the ELAN-HF linear predictor was equal to 1. Secondly, the model misspecification was tested formally by running a Cox PH model on all the ELAN-HF covariates and constraining the coefficient of the ELAN-HF linear predictor to 1. Thirdly, the discriminative ability of the ELAN-HF score was evaluated using Harrell’s c-index. Finally, calibration was evaluated for predicting all-cause mortality at 6-months. The baseline-hazard at 6 months was obtained through personal correspondence with the authors of the ELAN-HF paper [14]. Patients were grouped based on expected/predicted probabilities and observed probabilities were calculated. Plotting expected versus observed probabilities yielded a calibration plot.

Survival curves were analysed for the ELAN-HF and EHFScBS-9 scores and compared with log-rank tests. For the ELAN-HF the score categories described by Salah et.al were used as reference. In case of the EHFScBS-9 score, patients with an EHFScBS-9 normalized score lower than or equal to the median EHFScBS-9 normalized score were categorized as “low”, and patients above the median as “high”. Kaplan–Meier (KM) curves were analysed using the log rank test to assess if there were significant differences between groups in cumulative incidence of events. An event was defined as time to first readmission or time till death from any cause.

Secondly, survival was analysed by Cox proportional hazards (PH) models by using time to readmission or all-cause mortality within 6 months as the outcome.

Univariate Cox PH models were fit to a subset of clinically relevant variables that were not in the ELAN-HF score. The variables that were tested significant in univariate analysis were then included in a multivariate Cox PH model to assess whether the ELAN-HF score could be improved by the EHFScBS-9 score or other variables. Statistical analyses were performed in R version 4.0.3 and a p-value < 0.05 was considered statistically significant.

The ELAN-HF score consists of clinical variables (age, peripheral oedema on admission, SBP and NYHA class) and biomarkers (NT-proBNP, sodium and urea). All variables were collected from the electronic patient file. Biomarkers were measured in the clinical laboratory on a Cobas 8000 Pro (Roche Dx, Basel, Switzerland) instrument.

The reliability and validity of the EHFScBS-9 to measure self-care has been extensively researched in multiple studies [16, 27, 28]. The studies show that the psychometric properties of the EHFScBS-9 are satisfactory.

Table 4 presents 6-month all-cause mortality according to subdivisions of the ELAN-HF score risk groups, comparing actual and predicted mortality. Figure 2 presents the calibration plot for predicting 6-month all-cause mortality. Although some mis-calibration appears in low probabilities/risk groups, this is not statistically significant due to small number in each group. The slope of the ELAN-HF linear predictor in the validation cohort was 0.80 (SE = 0.22), the slope is not significantly different from 1 (LR test χ²_{df = 1} = 0.81, p = 0.367), so the discrimination of the ELAN-HF score seems to be preserved in our cohort. There was also no evidence of model misspecification, a joint test of all the predictors was non-significant (χ²_{df = 10} = 14.71, p = 0.143), meaning that the regression coefficients of the ELAN-HF score do not appear biased. The discriminative ability expressed in Harrell’s c-index was 0.719 (SE = 0.056) in our cohort, this is similar to the reported index by Salah et al. of 0.71 [14].

The Figs. 3A and 3B show the relationship between the risk groups derived from both EHFScBS-9 (panel A) and ELAN-HF scores (panel B) and the composite endpoint of readmission and / or all-cause mortality. There was no significant difference in composite endpoint among patients with low EHFScBS-9 score (i.e., below or equal to the median of the normalized EHFScBS-9 score) in comparison to patients with a high score (24%, versus 31% respectively KM-log rank test p = 0.15). Readmission and/or mortality rate was significantly higher in patients with higher ELAN-HF scores in comparison to those with low scores (KM log-rank test p = 0.0071). Due to the smaller sample size, there is an overlap between survival curves of the low and intermediate, and high and very high-risk groups.

Univariate Cox regression analysis for the composite endpoint results is shown in Table 5. Univariate analysis did not show that the normalized EHFScBS-9 score was associated with 6-month readmission and/or mortality. Other than the ELAN-HF score, two additional variables showed a significant association; whether the patient was admitted with ADHF in the previous year, and whether the patient is an outpatient clinic patient. Both factors increased the risk of 6-month readmission and/or mortality. This association remained significant in multivariate analysis. A LR-test revealed that adding these variables to the ELAN-HF score improved the model fit (χ²_{df = 2} = 10.61, p = 0.005) in predicting risk of 6-month readmission and/or mortality.

The ELAN-HF study retrospectively defined a risk score model for all-cause mortality for patients discharged after ADHF [14], which was validated in an independent cohort [21].

In our study, we validated the ELAN-HF score with prospectively collected data and demonstrated that 180-day mortality can be robustly predicted.

For the present analysis, we used the composite endpoint of mortality and readmission for additional analysis of the risk prediction value of the EHFScBS-9 self-care score after admissions for HF to compare with the risk prediction value of the ELAN-HF score for this endpoint. It was notable that the composite endpoint event rate within 180 days in our study was 55%, compared to 43% in the ELAN-HF study. This can be explained by the fact that patients in our study had a relatively higher NYHA-class compared to the ELAN HF study. While the ELAN-HF score was retained in the model as a predictor of the composite endpoint, the EHFScBS-9 self-care score was not.

Based on these results, we think that implementing the ELAN-HF risk model on the nurse-led HF clinic may offer more guidance to follow-up of these patients and we strongly suggest to add this score to the discharge checklist as standard care. Possible consequences may be that the high-risk population can benefit from more aggressive treatment and also from a much closer follow-up by intensive (tele-) monitoring throughout the entire HF care network.

While earlier studies demonstrated a relationship between better self-care and a reduced readmission rate [24], self-care behaviour was non-significantly associated with readmission and/or mortality in our study population. This is most probably caused by an on-average high normalized self-care score (median of 68) within the total cohort, with 20% of them being known patients from our outpatient clinic. These patients already received self-care education and were experienced with adjusting their lifestyle, knowledge of their disease and alarming symptoms.

An optimal self-care score always needs improvement [22, 23]. Therefore, it remains important to invest in improving self-care behaviour and to optimize patient education in HF and self-care activities by nurses, during the discharge and outpatient phases.

Several limitations of our analyses should be acknowledged.

A first limitation of this study is the number of participants. However, the sample size turned out to be large enough to demonstrate the prognostic value of the ELAN-HF model. This does however yield a limitation of possible sub-analyses. A second limitation is that using self-reports as in the EHFScBS-9 may be affected by memory and social desirability biases.