Relationship between senile HF and frailty
During the progression of HF, a series of complications often arise, such as renal dysfunction, frailty, pulmonary congestion, and edema. Frailty, in particular, diminishes the heart's ability to regulate stress events, not only increasing the incidence of cardiovascular diseases but also exacerbating the condition of existing cardiovascular patients, leading to adverse outcomes. Research has indicated that frailty can be used to assess the prognosis of cardiovascular diseases [14]. HF patients already suffer from impaired physical function, making them prone to frailty symptoms and further contributing to the development of frailty in the body, ultimately worsening the HF condition. Therefore, HF and frailty interact with each other, creating a vicious cycle that leads to poor prognosis in elderly HF patients with concurrent frailty, significantly impacting their overall physical and mental health. A systematic analysis study involving 6,896 HF patients revealed that the prevalence of HF patients with concurrent frailty ranged from 36.2% to 52.8% [15]. Gastelurrutia et al. (2013) conducted a study with 621 HF patients and clinically assessed that 44.2% of patients experienced frailty at least once, with significant impacts on their daily lives [16]. In another study in 2014, among 316 HF patients, 70.2% were found to meet the criteria for frailty after evaluation [17]. In 2015, Uchmanowicz et al. (2015) [18] found that the incidence of frailty was high among HF patients. Furthermore, frailty had a detrimental effect on patients’ self-care ability. In their 2018 study, Uchmanowicz et al. [19] also proposed that frailty symptoms are highly prevalent among elderly HF patients, and the severity of frailty is directly associated with higher readmission rates. In the years 2016-2017, Vidán et al. (2016) [20] and Díaz-Toro et al. (2017) [21] independently reported frailty occurrence rates of 76% and 50.6% respectively in elderly HF patients, suggesting that the incidence of frailty in this population is relatively high. The relevant articles published between 2018 and 2021 also demonstrated that frailty was very common in patients with senile HF [22,23]. Due to the fact that the incidence of frailty in HF increases with age, as the population continues to age, the occurrence rate of frailty is also expected to rise. As the aging population intensifies, the incidence of frailty is generally on an upward trend. The level of population ageing and the incidence of frailty from 2010 to 2021 were summarized in Table 1.
Table 1 Summary of population ageing and the incidence of frailty
Year
|
Population ageing (over 60 years old)
|
Incidence of frailty (%/n)
|
Number of people (hundred million)
|
Proportion (%)
|
Number of patients with frailty (cases)
|
Proportion (%)
|
2010
|
1.19
|
8.87
|
--
|
--
|
2011
|
1.23
|
9.1
|
--
|
--
|
2012
|
1.94
|
14.3
|
--
|
--
|
2013
|
2.02
|
14.8
|
273
|
44.2
|
2014
|
2.1
|
15.5
|
227
|
70.2
|
2015
|
2.22
|
16.15
|
55
|
50
|
2016
|
2.31
|
16.7
|
342
|
76
|
2017
|
2.41
|
17.3
|
40
|
50.6
|
2018
|
2.49
|
17.9
|
214
|
64.8
|
2019
|
2.53
|
18.1
|
101
|
50
|
2020
|
2.64
|
18.7
|
662
|
56.1
|
2021
|
2.67
|
18.9
|
825
|
66.5
|
According to Figure 3, the number of elderly people over 60 increased year by year from 2010 to 2021. In contrast, the incidence frailty showed no apparent growing trend as population ageing did, which might be related to the differences in regional scope, age range, number of patients, and HF type.
The aforementioned studies highlight the close relationship between HF and frailty status. Firstly, there is a higher probability of frailty occurrence in HF patients. Secondly, frailty impacts the prognosis, mortality rate, and readmission rate of HF patients, posing a significant threat to their overall health and well-being. Thus, timely improvement of frailty status holds crucial importance for the life and health of HF patients. Currently, there are various methods for screening and assessing frailty in clinical practice (as shown in Table 2). Some studies have suggested that the Clinical Frailty Scale (CFS) has high sensitivity (87%) and specificity (89%), with the lowest misclassification rate (12%). It is straightforward to implement and yields good application results, making it the most valuable among the screening methods. However, among the assessment methods, no specific standout exists. The Fried standard, Edmonton Frail Scale (EFS), and Deficit Index (DI) have comparable misclassification rates, with advantages in sensitivity and specificity for each. As a result, their overall application effects are similar [24].
Table 2 Methods for screening and evaluating frailty
|
Name
|
Presenter
|
Time of presentation
|
Evaluation effect
|
Sensitivity
|
Specificity
|
Error classification rate
|
Screening methods
|
CFS
|
Rockwood
|
2005
|
87%
|
89%
|
12%
|
Derby frailty index
|
Woodard
|
2013
|
76%
|
73%
|
26%
|
Acute frailty network standard
|
Unknown
|
2018
|
79%
|
78%
|
22%
|
Evaluation methods
|
Fried standard
|
Fried
|
2001
|
93%
|
76%
|
17%
|
EFS
|
Rolfson
|
2006
|
62%
|
98%
|
19%
|
DI
|
Mitnitski
|
2002
|
75%
|
92%
|
17%
|
Effects of frailty on the prognosis for HF patients
The above research findings confirm that frailty is a common clinical syndrome in elderly HF patients. Furthermore, studies have demonstrated that frailty impacts the prognosis, mortality rate, readmission rate, and acute events of HF patients. For instance, Gastelurrutia et al. (2013, 2014) [25] confirmed through research that frailty affects the quality of life of HF patients of all ages and is a crucial and determining factor in the survival of outpatient HF patients across all age groups. Sanders et al. (2018) [26], analyzing experimental data and baseline characteristics of 1,767 preserved ejection fraction HF patients in the TOPCAT trial, found that an increase in frailty index (FI) leads to an increase in body mass index, systolic blood pressure, and pulse pressure. Moreover, with a higher degree of frailty, there is an increased risk of cardiovascular disease outcomes and mortality. Rodríguez-Pascual et al. (2017) [27] studied 497 HF patients to explore the correlation between frailty and mortality, readmission rate, and reduced heart function in outpatient HF patients. The results showed that frailty increased the likelihood of 1-year mortality, readmission, and declining heart function in elderly HF patients, also serving as an independent risk factor for HF patient readmissions. Uchmanowicz et al. (2018) [19], in their analysis of 330 HF patients with frailty syndrome, found a positive correlation between frailty syndrome and the number of hospitalized patients. Higher levels of frailty were identified as a determining factor for increased HF patient readmission rates. Vidán et al. (2016) [20] included 450 non-dependent HF inpatients aged 70 or older and analyzed the impact of frailty on mortality, functional decline, and readmission risk. The study indicated that frailty is an independent predictor of early disability, long-term mortality, and readmission rates. Jujo et al. (2021) [23] conducted a 1-year follow-up observation on 1,240 HF patients with frailty and found that the combined endpoint and all-cause mortality rate were higher in HF patients with frailty compared to those without frailty. Wang et al. (2018) [28] also demonstrated in their research that frailty holds prognostic significance for HF patients, with patients experiencing frailty having worse outcomes and higher mortality rates.
Based on the above research results, it is evident that frailty is closely linked to the prognosis of HF patients. Therefore, it was indispensable to further explore the influencing factors for the prognosis for HF patients with frailty and the corresponding action mechanism in the background of population ageing in China. The effects of frailty on readmission rate of HF patients were presented in Figure 4 below.
The analysis results of all-cause death rate among HF patients with frailty in the above research were illustrated in Table 3.
Table 3: All-cause death rate among HF patients with frailty
References
|
Hazard ratio (HR)
|
95% confidence interval (CI)
|
P
|
[20]
|
2.13
|
1.07~4.23
|
0.005
|
[23]
|
1.53
|
1.01~2.30
|
0.044
|
[25]
|
1.38
|
1.15~1.66
|
<0.001
|
[26]
|
Level 2
|
1.32
|
1.00~1.73
|
0.05
|
Level 3
|
1.42
|
1.06~1.89
|
0.02
|
Level 4
|
1.81
|
1.31~2.50
|
<0.001
|
[27]
|
Low-level physical activity
|
1.64
|
1.10~2.45
|
<0.001
|
Fatigue
|
1.83
|
1.21~2.77
|
<0.001
|
[28]
|
1.70
|
1.41~2.04
|
<0.001
|
Application of AI in the diagnosis of HF
With the increasing incidence of HF, the quality of life and health of humans are greatly threatened, especially those of the patients with HF with preserved ejection fraction (accounting for nearly 50% in all HF patients) [29]. However, there is no single clinical diagnosis and treatment standard for HF and the prognosis for most patients is poor. Indeed, despite the significance of heart function assessment in the diagnosis and treatment of HF patients, the abundance of assessment methods can lead to inconsistent diagnostic outcomes, significant individual differences, and complexity in indicators, making it challenging for clinicians to conduct comprehensive evaluations. However, with the emergence of AI, heart function assessment methods have seen some improvements, leading to more accurate results in aiding clinical diagnosis by healthcare professionals. As the subsets of AI, ML and deep learning have been widely applied in medical field.
Currently, the commonly used methods for heart function assessment are echocardiography and magnetic resonance imaging (MRI) examinations. They are typically utilized for the quantitative evaluation of left ventricular function. Left ventricular ejection fraction (LVEF) is the most common indicator used to predict the prognosis of HF. The specific calculation methods for left ventricle and LVEF are displayed in Figure 5 and equation (1).
Due to various factors that can introduce errors in the estimation of ejection fraction from ultrasound images, it presents an opportunity for the application of AI techniques. Asch et al. (2019) [30] utilized over 50,000 echocardiograms to establish a deep learning model for automated LVEF assessment, based on the recognition of endocardial borders in echocardiography. The model was then clinically evaluated on 99 patients, and the automated LVEF results were found to have high concordance, sensitivity (93%), and specificity (87%) when compared to reference values obtained using traditional evaluation methods by three experts. This indicates that the automatically obtained ejection fraction values from the model are reliable. Attia et al. (2019) [31] improved the convolutional neural network (CNN) using TensorFlow and Keras, and applied it for automatic recognition of LVEF from electrocardiograms. The model predicted LVEF<35% with an accuracy of 86.5%, specificity of 86.8%, sensitivity of 82.5%, and an area under the curve (AUC) of 0.918, suggesting its good performance for LVEF recognition.In MRI examinations, accurate identification of the endocardial borders is crucial. Glading et al. (2021) [32] applied ML to advanced electrocardiogram (AECG) and AI-enhanced echocardiography (Echo AI), and their study showed the potential of ML for HF assessment. Overall, AI has shown promising applications in improving the accuracy and efficiency of heart function assessment methods, benefiting the clinical evaluation of HF patients. Winther et al. (2018) [33] applied v-net deep learning algorithm in the automatic analysis of left and right ventricular endocardium and pericardium to improve the precision and efficiency of segmentation. Eventually, excellent achievements were made. Avendi et al. (2016) [34] realized the automatic segmentation of cardiac 4D blood flow MRI through AI technology to perform the quantitative assessment of hemodynamics and function of cardiac valve (Figure 6). To sum up, AI had a good application prospect in assisting cardiac function assessment.
Application of AI in HF treatment equipment
Although angiotensin converting enzyme inhibitor, angiotensin receptor antagonist, and β-receptor blocker could improve the prognosis for patients during the treatment of chronic HF, most of prognostic effects were not satisfactory and 5-year survival after the treatment reached only 50% [35]. Therefore, auxiliary therapeutic devices have been developed to improve clinical symptoms and quality of life, extend survival, and reduce mortality in HF patients. These devices include triple-chamber pacemakers, ventricular assist devices (VADs), implantable cardioverter-defibrillators (ICDs), cardiac resynchronization therapy defibrillators (CRT-Ds), and more (Figure 7). These devices have been widely utilized in clinical practice.
The mentioned cardiac assist devices are essential for the treatment of many HF patients. However, during device installation, issues may arise, and subsequent adjustments require knowledge of the device model and structure. With a wide variety of cardiac assist device models available, searching for the corresponding model can be labor-intensive and time-consuming, potentially delaying treatment. To address this challenge, experts have developed intelligent identification programs using AI algorithms to recognize device models and structures from chest X-ray images. The results indicate that the AI algorithm's accuracy in device recognition is generally above 86% [36]. This advancement in AI technology offers a more efficient and accurate way to identify the cardiac assist device, streamlining the treatment process for HF patients. In addition, acute attack of HF was related to atrial fibrillation (AF) and ischemic stroke. European Society of Cardiology (ESC) suggested that the comprehensive assessment of cardiac parameters detected by the implanted equipment was conducive to the prediction of high-frequency attack of heart disease. To improve discrimination, Kim et al. (2022) [37] applied ML algorithm, random forest (RF) algorithm, support vector machine (SVM) algorithm, and eXtreme Gradient Boosting (XGB) algorithm to establish clinical heart disease-related predicting model. It was concluded that the above 3 algorithm models all improved the discrimination of the predicting model, which demonstrated that AI could predict disease attack through pacemaker parameters during the treatment.