Data were collected from the National Health Insurance Service of Korea (NHIS)-Senior database, which included data of 312,736 populations collected via a 10% simple random sampling method from a total of 5,500,000 populations 65 years or older in the National Health Information Database [14, 15]. The NHIS-Senior database included socioeconomic and insurance status, medical history, and check-up test results.

This study included the data of 8982 older adults (age $\ge$65 years) with a history of HF from the Korean NHIS-Senior database, for whom health check-up data were available between 2009 and 2013 and who were followed up until December 2015 (Fig. 1). Criteria of HF in this study was based on ICD-10 codes (I11.0, I50, and I97.1) for one inpatient or two outpatient visit in the database (Supplementary Table 1).

Fig. 1.

Flow diagram of the study population. A total of 8982 heart failure patients aged 65 years or older in Korea were classified according to exercise intensity.

The level of physical activity during waking time was evaluated through a self-report questionnaire using the 7-day recall method [16, 17, 18]. The questionnaire asked about frequency (days/week) of the following three intensity levels: (i) low-intensity physical activity (LPA) for at least 30 min, (ii) MPA for at least 30 min, and (iii) VPA for at least 20 min. LPA include slow speed of walking, MPA include like normal speed of walking and cycling and VPA include high speed of running and cycling.

Using the Ainsworth et al. [19] compendium, 3.3 METs for LPA, 4.0 METs for MPA, and 8.0 METs for VPA were assigned. And the total physical activity-related energy expenditure (MET-min/week) was the sum of duration frequency and intensity of exercise [19]. The participants were stratified according to their total weekly levels of physical activity as follows: (1) highly active group: physical activity exceeding the guideline target range ($\ge$1000 MET-min/week); (2) active group: physical activity within the guideline target range (500–999 MET-min/week); (3) insufficiently active group: physical activity less than the guideline target range (1–499 MET-min/week); and (4) inactive group: no physical activity other than minimal movement in their spare time, considering the guidelines and previous studies [6, 7, 20].

Baseline comorbidities were assessed with the ICD-10 codes and drug prescription data (Supplementary Table 1). To ensure diagnostic accuracy, as in previous studies, if the diagnosis was registered at the time of discharge or the same diagnosis was registered twice or more in an outpatient clinic, the case was considered a corresponding comorbidity [15, 21, 22]. We performed analysis of drug history for heart failure. Among ACE (angiotensin-converting enzyme) inhibitor or ARB (angiotensin II receptor blocker), BB (beta blocker), loop diuretics and mineralocorticoid receptor antagonist, HF medication group I was assumed if no medications were taken or only ACE inhibitors or ARBs were taken. HF medication group II was assumed if BB or loop diuretics were taken but not mineralocorticoid receptor antagonist, and HF medication group III for patients taking mineralocorticoid receptor antagonist [23].

The primary outcome was all-cause mortality. Information on death (cause and time) was checked through a unique identification number in the demographics of the National Statistical Office [15, 22]. Since the NHIS is a national institution for all Koreans, this approach was used to ensure event-related information. And the mortality was analyzed by dividing them into CVD-related death and non-CVD-related death.

The secondary outcome was all-cause hospitalization. The hospitalization cause was based on the diagnosis at the time of discharge. Eligible causes of hospitalization such as exacerbation of HF, CVD, chronic ischemic heart disease, coronary vascular disease, and ischemic stroke were included. The accuracy of diagnosis based on NHIS claims data has already been verified [22]. The details of the clinical outcomes are described in Supplementary Table 1. Follow-ups of mortality and hospitalization were conducted until December 2015.

Descriptive statistics were used to analyze the baseline characteristics of participants. Categorical variables are expressed as ratios (percentages), and continuous variables are expressed as median and interquartile range. Fisher’s test or chi-square test was used for comparison between categorical variables, and Student’s t-test was used for comparison between continuous variables. The incidence of each clinical outcome was expressed as number of events/100 person-years.

Competing risk regression for all-cause death events was carried out with the Fine-Gray method. Multivariable regressions were carried out with adjustment for age, gender, hypertension, diabetes, dyslipidemia, vascular disease, stroke, renal disease, lung disease, cancer, body mass index, smoking and drinking. Additionally, investigate the effects of physical activity as continuous value (0 MET-min/week as reference) on mortality via cubic spline curve.

All test with values of p 0.05 was considered statistically significant. Statistical analyses were performed using R programming version 4.0.3 (The R Foundation for Statistical Computing, Vienna, Austria).

A total of 8982 individuals 65 years or older (mean age, 75.5 $\pm$ 6.3 years) with HF were included in the analysis. The comparison of baseline characteristics among groups classified by activity level are summarized in Table 1. The participants were divided into inactive, insufficiently active, active, and highly active groups, accounting for 15.6%, 13.0%, 11.4%, and 10.2% of the study population, respectively.

During the mean follow-up period of 92.8 $\pm$ 36.9 months, 1359 (15.1%) participants had all-cause mortality with an overall incidence of 5.0/100 person-years. The all-cause mortality rates were 6.7, 4.2, 3.8, and 2.4/100 person-years in the inactive, insufficiently active, active, and highly active groups, respectively (Table 2). In the Fine-Gray competing risk multivariable regression analysis, the insufficiently active (HR, 0.82; 95% CI, 0.71–0.94), active (HR, 0.81; 95% CI, 0.70–0.95), and highly active (HR, 0.52; 95% CI, 0.41–0.67) groups were showed lower risks of mortality compared to the inactive group. Fig. 2 presents the risk of hospitalization associated with continuous values of level of physical activity with a cubic spline curve. The risk of mortality associated with continuous values of level of physical activity with a cubic spline curve. There was a linear relationship between physical activity and risk of mortality. As the level of physical activity increased, the mortality risk was further reduced.

Fig. 2.

Weekly level of physical activity and adjusted hazard ratio for all-cause hospitalization in older adults with heart failure. MET, metabolic equivalent task. Non-linear cubic spline curve of weekly physical activity level against hospitalization. Black line represents the fitted line of the association between weekly physical activity level and adjusted hazard ratio of hospitalization, whereas the shaded region represents the 95% confidence interval.

The incidence of non-CVD-related mortality was 4.5, 2.8, 2.5, and 1.5/100 person-years in the inactive, insufficiently active, active, and highly active groups, respectively. Compared to the inactive group, the adjusted risk of non-CVD-related mortality was reduced in the insufficiently active (HR, 0.78; 95% CI, 0.65–0.93), active (HR, 0.75; 95% CI, 0.62–0.91), and highly active (HR, 0.46; 95% CI, 0.34–0.62) groups. However, the adjusted risk of cardiovascular-related mortality was not reduced by physical activity (Table 2).

During the follow-up period, 5788 patients (64.4%) were hospitalized. The overall incidence of all-cause hospitalizations during follow-up was 35.8/100 person-years. When divide the patients according to their level of intensity of physical activity, the incidence rates were 39.8, 35.9, 33.1, and 27.9 cases/100 person-years in the inactive, insufficiently active, active, and highly active groups, respectively (Table 3). In the Fine-Gray competing risk multivariable regression models, compared with the inactive group, the insufficiently active (HR, 0.98; 95% CI, 0.91–1.04), active (HR, 0.94; 95% CI, 0.88–1.01), and highly active (HR, 0.81; 95% CI, 0.74–0.89) groups were showed a lower risk of hospitalization. Fig. 2 presents the risk of hospitalization associated with continuous values of level of physical activity with a cubic spline curve. A non-linear relationship between physical activity and the risk of hospitalization was evident. Although the risk of hospitalization was further reduced with increased level of physical activity, the benefit of increased physical activity did not increase beyond 1500 MET-min/week.

Compared to the inactive group, the incidence and risk of stroke were reduced in the highly active group (2.2 vs. 1.5/100 person-years; HR, 0.70; 95% CI, 0.51–0.96; p = 0.030) (Table 3). However, the risks of HF admission and CVD were not reduced by increased physical activity.

The risk of mortality was assessed in different patient subgroups. A risk reduction of mortality was observed regardless of the patient’s age, sex, hypertension, diabetes, chronic kidney disease, and previous history of stroke (Fig. 3). In subgroup analysis according to gender, the adjusted risk of all-cause mortality was reduced in the insufficiently active (in female: HR, 0.81; 95% CI, 0.67–0.97 and in male: HR, 0.81; 95% CI, 0.64–1.03), active (in female: HR, 0.79; 95% CI, 0.64–0.99 and in male: HR, 0.85; 95% CI, 0.68–1.07), and highly active (in female: HR, 0.58; 95% CI, 0.40–0.84 and in male: HR, 0.50; 95% CI, 0.36–0.68) groups (Supplementary Table 2). The adjusted risk of all-cause hospitalization was presented at Supplementary Table 3.

Fig. 3.

Subgroup analyses of all-cause mortality in older adults with heart failure. AF, atrial fibrillation; CI, confidence interval; DM, diabetes mellitus; HF, heart failure; HTN, hypertension; PYR, person-years at risk. Forest plot showing the adjusted hazard ratios of all-cause death in older (age $\ge$65 years) patients with heart failure. In all cases of HF, regardless of the presence or absence of DM, HTN, AF, and stroke, exercise reduced all-cause mortality. Even minimal exercise reduced all-cause death compared to no exercise, and the degree of mortality reduction increased as the level of exercise increased.

In a subgroup analysis of the HF medication group, for all-cause mortality and non-CVD-related mortality, there was no difference in outcome with increasing exercise in HF medication group I—none or ACE inhibitor/ARB, but all-cause mortality and non-CVD-related mortality were decrease with increasing exercise in HF medication group II—beta blocker or diuretics, III—spironolactone (Supplementary Table 4). On the other hand, in the case of hospitalization due to heart failure and cardiovascular disease, exercise had a positive effect on outcome in HF medication group I—none or ACE inhibitor/ARB, but in HF medication group II—beta blocker or diuretics, III—spironolactone, there was no significant correlation between exercise change and outcome (Supplementary Table 5).

The recent Cochrane review of exercise training [12], including 33 trials in 4740 patients with HF (primarily HF with a reduced ejection fraction), showed that exercise tends to reduce mortality in trials using 1 year of follow-up. Compared to the control group, the exercise-trained group had lower hospitalization rates (both all-cause and HF-related hospitalizations) and improved the quality of life of patients. Practical recommendations for exercise training have been published by the Heart Failure Association [24]. Also, a single large-scale randomized controlled trial showed a modest decrease in the primary composite outcome of all-cause mortality or all-cause hospitalization [25]. However, this study did not show a significant reduction in mortality [25, 26]. Therefore, although the importance of exercise in HF has been proven, there is insufficient information on the results of exercise intensity, with limited number of related studies, especially on Asians.

As the life expectancy of HF peoples increases, comprehensive medical support for them, including lifestyle modifications, such as exercise, has become increasingly important. In this study, the dose-response relationship between physical activity and mortality was evaluated by calculating the total duration physical activity as a continuous variable. The positive benefits of increased physical activity on mortality started to appear at lower levels of activity (Fig. 4). These results showed that little activity is better than no activity at all.

Fig. 4.

Adjusted hazard ratio for all-cause mortality in older adults with heart failure by weekly physical activity levels. MET, metabolic equivalent task. Non-linear cubic spline curve of physical activity level against all-cause mortality. Black line represents the fitted line of the association between weekly physical activity level and adjusted hazard ratio of mortality, whereas the shaded region represents the 95% confidence interval.

Also, the same results were confirmed in the analysis by gender difference. Although many studies demonstrated that physical activity and physical exercise may prevent negative cardiovascular adaptations associated with the postmenopausal period, which in turn could lead to the development of heart failure [27, 28], this study showed that physical activity reduced all-cause mortality and all-cause hospitalization even in elderly females with HF.

In the past, exercise was performed for the prevention of diseases including heart failure and it is already known that exercise has therapeutic implications [29, 30]. This is also specified in current treatment guidelines [7]. This study showed that the therapeutic effect of such exercise was also applied to Asian, older adults, and heart failure patients, and that even lower-intensity exercise than the guideline was meaningful.

This study had limitations. First, it did not include echocardiographic data, such as left ventricular ejection fraction, or assessment of symptoms, such as the New York Heart Association Functional Classification, since the diagnostic criteria of HF were based on the ICD codes. Although we presented the outcome according to the different degree of medication for heart failure, the severity of HF could not be evaluated. This can be an important limitation of this study, as severe HF condition can reduce physical activity. Further studies based on the severity of HF are required. Second, such studies using administrative databases could be susceptible to errors arising from coding inaccuracies. To minimize this problem, we applied the definitions that were already validated in previous studies using the Korean NHIS cohort [15, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31]. Third, CV-related information was relatively well known in NHIS data, but it was difficult to identify all other causes of death. Therefore, we only divided total deaths into two categories: CVD-related and non-CVD-related deaths. Fourth, this study relied on self-reported data at a single point in time. Therefore, questionnaire answers may not reflect the overall patient’s condition. Fifth, although physical activity plays an important role in reducing CVD-related incidents, recent data showed that activity alone is not enough to reduce the risk of CVD in older adults [32, 33]. In this study, other factors such as sedentary behavior were not considered; therefore, additional research on these factors in the elderly Asian population is required. Despite these limitations, this study was still significant in that it was a large-sized study using physical activity data in older Asian population with HF. Herein, we assessed the correlation between exercise intensity and clinical outcomes in HF patients. Overall, our analysis may serve as the first step in determining the optimal exercise intensity for older adults with HF.