Discussion
This is a 1-year follow-up study of a cohort of real-world AHF patients who were discharged after an index hospitalisation from an internal medicine department of a tertiary care hospital, in Portugal. Considering that there may be different risk factors for death and rehospitalisation in AHF patients after hospital discharge, the main goal of this study was to identify the independent predictors of rehospitalisation for AHF, considering death as a competing event. In addition, it was sought to determine the risk factors for 1-year overall mortality, independently of the occurrence of a rehospitalisation before death. Of note, our study reports data from a real-world geriatric HF population, predominantly having a LVEF >40%, which is largely underrepresented in clinical trials and cardiology-based registries.
After performing the competing risks analyses, we were able to identify which factors independently contribute to the risk of rehospitalisation 1 year after hospital discharge from AHF in our study population: AF, renal dysfunction or beta-blocker prescription at discharge. In addition, we have identified predictors of death without rehospitalisation: being a male patient, aged ≥80 years, with dementia, and RDW at Q4 on index admission were considered as high-risk factors, while taking beta-blockers at discharge and having higher PDW levels on the index admission were found to be low-risk factors. Interestingly, we have previously reported the predictors of 1-year overall mortality after AHF hospitalisation,17 and identified those having ≥80 years of age, with active cancer, dementia, higher urea, RDW at Q4 and lower PDW levels as being at higher risk of mortality. Consistent with this, we are now reporting that, after adjusting for gender, age and LVEF values, having dementia, RDW at Q4 and lower PDW levels remain as independent risk factors for overall mortality after hospital discharge. Altogether, this knowledge will allow us to stratify patients by risk factors for each event (rehospitalisation or death) separately, at discharge of the index hospitalisation.
Hospital readmission is a key marker of healthcare quality, particularly when assessing AHF patient outcomes. However, despite its widespread use, there is some controversy regarding the statistical methods employed to analyse the risk of rehospitalisation, as these analyses mostly do not consider patients who die before experiencing a readmission event within the period of interest.18 Significantly, it has been estimated that 46% of studies in high-impact medical journals are susceptible to competing for risk bias,19 highlighting the importance of considering death and rehospitalisation separately in the risk analysis of HF patients.
Many pivotal clinical trials included in the pharmaceutical and device development programmes for HF management have tested the benefit of the intervention on the composite primary endpoint of cardiovascular death and HF hospitalisation. However, several researchers have noted that the emphasis on the reduction of 30-day hospitalisation rates may have adverse consequences, such as an increased risk in the 30-day mortality,20 and that risk-adjustment models do not account for the competing risk of mortality.21 In fact, there is one report of increased 30-day and 1-year mortality after the successful implementation of a Hospital Readmissions Reduction Programme when competing risks analyses were used.22
These findings are particularly important considering the characteristics of our study population, a real-world cohort of geriatric patients with multiple comorbidities (having an average of three comorbidities per patient) and with the majority of patients having LVEF >40%.16 We should not neglect that HF is regarded as the cardiovascular syndrome of ageing, resulting from the combination of cardiovascular conditions of older adults with age-related changes in cardiovascular structure and function23, and that the mortality, hospitalisation and rehospitalisation rates are still high in this population.2 24 Therefore, it is of utmost importance to find risk factors for these events in such patients.
AF has been associated with a worse prognosis for patients with chronic HF and is a classical trigger for HF decompensation. Herein, the more advanced age and HF severity of this group of patients justify the finding that AF is associated with a higher risk of rehospitalisation, in line with previous reports.25 In the ESC-HF-LT Registry, the proportion of patients with AF varied widely among clinical profiles, with the highest prevalence of AF (34.7%) documented in patients with right HF. Curiously, this was also the group with the highest 1-year rehospitalisation rate (31.2%),26 in accordance with what was observed in this study.
Other authors have also reported that risk predictors for rehospitalisation are different from those for mortality. For instance, SBP <120 mm Hg at the index hospitalisation predicts mortality, whereas an increase in body weight predicts rehospitalisation.21 Although we included SBP in our risk analysis, it was not identified as an independent predictor of rehospitalisation or death without rehospitalisation.
Renal dysfunction has been recognised as a risk factor for HF rehospitalisation and mortality,25 and our data support this notion. Indeed, a previous report by Álvarez-García et al performed a competing risk analysis for risk prediction in hospitalised HF patients. They used competing risks analysis to develop the Redin-SCORE to predict the 1-month and 1-year risk of rehospitalisation due to HF acute episodes among ambulatory patients.6 Similar to our findings, they found that renal dysfunction, defined as an eGFR below 60 mL/min/1.73m2, among other clinical and laboratory parameters, were predictors of 1-year rehospitalisation.6
Interestingly, we found that higher levels of PCr at the index HF admission are: (1) a risk factor for overall mortality after discharge in the multivariable Cox model and (2) a risk factor for rehospitalisation, but not for death without rehospitalisation, in the competing risk analysis. Of note, this last finding does not weaken the impact of renal dysfunction on mortality, since patients are known to die during and after AHF rehospitalisations, and rehospitalisation for HF is a known risk factor for death.27 In a systematic review of models for predicting mortality and/or hospitalisation in patients with HF, Ouwerkerk et al found that the strongest predictors were blood urea nitrogen and sodium levels.28 Perkins et al have developed a prediction model for rehospitalisation derived from a cohort of 607 chronic kidney disease patients hospitalised for HF; 19.1% were rehospitalised within 30 days.29
In our study, 52.8% of patients had a prescription for a beta-blocker at hospital discharge,16 and this was associated with a higher risk of rehospitalisation and a lower risk of death without rehospitalisation. These are in line with recent evidence from multiple studies reporting that the uptitration of beta-blockers is associated with more HF hospitalisations in patients with a preserved ejection fraction.30 31 Nevertheless, these results need to be addressed, given that the known mortality benefit of this class of drugs in HF patients seems limited to those with LVEF <40% and in sinus rhythm,25 30 whereas our patients had mostly LVEF >40% (70.5%) and AF (51.7%).16 Notably, 75.2% of patients with LVEF <40% were discharged with a beta-blocker,16 following the guidelines in place at the time the study was conducted, and we adjusted the analysis for LVEF. Therefore, we can presume that the higher risk of rehospitalisation is related to the frequent side effects of beta-blockers—negative inotropism, bradycardia and hypotension—and the contribution of these factors to the decompensation of HF patients.25 Recently, Stolfo et al performed a propensity score-matched analysis from the Swedish HF registry focusing on older patients (>80 years) with LVEF <40%, typically underrepresented in randomised control trials.32 The authors concluded that the use of beta-blockers was high for those patients and was strongly associated with improved all-cause and cardiovascular survival.
HF is not only associated with an increased risk of all-cause dementia, as it may represent a risk factor itself for dementia. In a meta-analysis by Cannon et al, the prevalence of cognitive impairment among HF patients was estimated as 43%.33 Adelborg et al reported in a large, nationwide cohort study in Denmark that HF was associated with an approximately 20% risk increase of all-cause dementia among patients surviving at least 1 year after HF diagnosis.34 The possible mechanisms underlying this risk association are chronic cerebral hypoxia secondary to hypoperfusion or cerebral inflammation, and microvascular dysfunction due to neurohormonal HF activation mechanisms.34 Our findings add new information to this interplay between HF and dementia, showing that dementia is: (1) a risk factor for overall mortality after discharge and (2) a risk factor for death without rehospitalisation, but not for rehospitalisation. These findings reinforce the relationship between dementia and death after AHF hospitalisation and prompt further investigation into the death causes of these patients.
The higher RDW level (at Q4; ie, higher than 16.3%) has been previously reported by us in the PRECIC study as an independent predictor of higher 1-year overall mortality risk in HF patients.28 Here, a higher RDW level is again associated with a higher mortality risk, yet in patients discharged from an index hospitalisation. In addition, this association is reinforced by the finding that patients with higher RDW had an increased risk of death without rehospitalisation. Recently, Melchio et al reported that HF patients discharged for AHF with higher RDW (>14.8%) had more comorbidities and were at higher risk of death (HR for death from any cause=1.73).35 The authors have also demonstrated that RDW adds prognostic information beyond that provided by well-established risk factors or biomarkers (eg, age, renal function, NT-proBNP levels, therapy) and is a powerful marker of worse long-term outcomes in AHF patients. Thus, there is a growing body of evidence that RDW can alert clinicians to the risk of death for individual patients in the first few hours of an AHF hospitalisation.
On the opposite, the prognostic impact of PDW, a specific marker of platelet activation, is still not yet well known. In a prospective observational study, which aimed to evaluate the prognostic impact of PDW among AHF hospitalised patients, Sato et al found an association between PDW levels and the prognosis of HF patients36: patients with higher PDW (>16.9) were at higher risk of all-cause death, cardiac death and a cardiac event (HR 1.72, 1.92 and 1.40, respectively).36 In contrast, our results show that patients with lower PDW levels were at higher risk of overall mortality after discharge and death without rehospitalisation, and had no influence on the rehospitalisation incidence levels.
Study limitations
This study has a few limitations that might introduce bias in our conclusions: (1) the rehospitalisation episodes analysed were only those occurring at the study centre, given the difficulty in accessing patient health data in other public or private health organisations, and, because of this, rehospitalisation rates might have been underestimated; (2) due to the retrospective nature of the study, with secondary use of data, signs of congestion and other clinical and laboratory parameters, like NT-proBNP, could not be evaluated as risk factors for readmission or death; it was also not possible to provide a thorough description of the group, including the aetiology of AHF, as well as the NYHA class on admission (3) the implantation of cardiac devices was not examined, although it was unlikely that a significant number of patients had implanted such devices, considering their age and that only a minority had LVEF ≤40%; (4) other factors that could have contributed to rehospitalisation, such as poor compliance to medication, were also not considered in the risk analysis; (5) the causes of death after discharge were not studied in detail, although we have previously reported that 42.9% of patients from the PRECIC study died of cardiovascular causes.16
