Background
Frailty has been recognized as a biological syndrome [
1]. It is theoretically defined as a clinically recognizable state of increased vulnerability to stressors, characterized by a decreased reserve capacity to maintain homeostasis resulting from an age-related cumulative decline across multiple physiologic systems [
1,
2]. Frailty confers an increased risk of adverse health outcomes, including falls, delirium, disability, hospitalization, long-term care, and mortality [
3,
4]. The incidence and prevalence of frailty are expected to increase with population aging, which consequently poses a great challenge to public healthcare and social care systems as demands for medical and care resources increase [
5]. Therefore, early screening for frailty in routine clinical practice, especially in primary care settings, is of great significance considering its high prevalence, reversibility, and prognostic value [
6,
7].
The best evidence-based process to detect frailty and grade its severity is comprehensive geriatric assessment, but this is a resource-intensive process [
1,
7]. Although there is no universal consensus regarding specific operational criteria in different practice settings, two main operational definitions receiving broad acceptance are the frailty phenotype proposed and validated by Fried and colleagues in the Cardiovascular Health Study (CHS) [
3] and the Frailty Index proposed and validated by Rockwood and colleagues in the Canadian Study of Health and Aging [
8]. The Fried frailty phenotype is the most commonly used definition in community settings worldwide [
9]. Compared to the Frailty Index, the frailty phenotype has been deemed more suitable for the immediate identification of non-disabled elders who are at increased risk for negative events, such as non-institutionalized community-dwellers [
10,
11].
The CHS frailty phenotype defines the presence of frailty and pre-frailty using five core components of the frailty cycle: unintentional weight loss, low grip strength, exhaustion, low gait speed, and low physical activity. In this measure, the presence of three or more components indicates frailty, one to two components designates pre-frailty, and zero components specify that the individual is not frail. Of these five components, low physical activity has been assessed in previous studies using questionnaires, which seemingly are feasible for routine practice, but prone to possible recall bias and a lack of diagnostic accuracy and comparability between different questionnaires. Specifically, for the frailty phenotype, the physical activity energy expenditure was assessed with the Minnesota Leisure Time Activity questionnaire, which does not capture physical activities in contexts other than specific leisure physical activities included in the questionnaire [
3]. Furthermore, many studies have used various questionnaires containing different kinds of leisure physical activities from those of the CHS [
12-
15]. The measurement of the low physical activity domain has not been standardized, which to some extent hinders the widespread application of the frailty phenotype in primary care practice. Thus, we addressed this issue in our study. We defined the low physical activity domain of the frailty phenotype using accelerometer-based measurement to detect frailty and evaluated whether our measures could statistically aggregate into a syndrome on their own, among older community-dwellers in a suburban area in Japan. Secondly, we examined correlates of frailty across a constellation of social, psychological, environmental, and health-related factors.
Discussion
In this report, we defined the low physical activity domain of the frailty phenotype with accelerometer-based measurement and confirmed the statistical aggregation of the five components of the frailty phenotype into a syndrome using LCA models. The accelerometer-based measurement of the low physical activity domain could potentially be beneficial in improving the diagnostic accuracy of the frailty phenotype and increasing its feasibility in primary care practice. We observed that frailty affected approximately one out of ten elderly adults aged 65 and over in this community-dwelling population, in which the care burden of frailty is the focus of exponentially rising demands for public healthcare resources. We also found significant associations between frailty status and age, living alone, self-perceived health, depressive and anxiety symptoms, current alcohol consumption, engagement in social activities, and cognitive impairment, independent of co-morbidities.
The CHS frailty phenotype has been shown to have satisfactory internal validity in the Canadian Study of Health and Aging [
10]. Prior LCAs of the frailty phenotype have been performed in the Women’s Health and Aging Study (WHAS) [
26] and the Survey of Health, Ageing and Retirement in Europe [
27] and have demonstrated satisfactory internal validity. Likewise, using LCA, we found that the five frailty criteria aggregated statistically into a syndrome. We further observed that the estimated probabilities of low physical activity exhibiting within latent frail classes were similar to those of two other objectively-measured components, regardless of whether subjects were stratified into two or three latent classes. These results suggest that our measures have satisfactory construct validity and could be used in this population. Additionally, our results did not conclude that classifying subjects into three subgroups (or classes) was better than two subgroups in characterizing the population. However, the CHS frailty definition of three phenotypes has been reported to have acceptable criterion validity as it identifies a profile of adverse health outcomes [
3,
10,
13,
26].
Our study established a frailty prevalence of 9.3%, which is in line with previous studies that defined frailty using Fried’s criteria [
5]. Of note is that a large scale population-based survey in Japan reported a frailty prevalence of 11.3% based on the Fried phenotype among community-dwelling adults aged 65 years and older, which was slightly higher than the prevalence in our study [
15]. Perhaps the discrepancy could be explained by the fact that they measured physical activities based on self-reported binary questions or that they did not use the lowest quintile approach for cut-off points. In the present study, we defined low physical activity with objective measurement and the percentages of low physical activity were 19.1% for men and 19.8% for women, which were generally lower than what was reported in the previous study. Many previous studies have reported proportions of low physical activity that range widely from 20% to 30% [
12,
13,
26,
28]. The large discrepancies between previous studies can at least be partly explained by the various methods used to measure physical activity, ranging from validated questionnaires to two simple questions. Limiting comparisons to studies that calculated energy expenditure with validated questionnaires and that used the lowest quintile approach seems likely to ensure that at least the populations identified are similar. However, essentially different questionnaires capture different types of leisure physical activities, for example, only six of the 18 leisure activities considered in the CHS were evaluated in the WHAS [
26], suggesting that different populations may be characterized as having low physical activity, even if the proportion of low physical activity would be similar, with respect to people’s preferences for these activities.
The agreement between accelerometer- and questionnaire-based measurement of physical activity is relatively poor [
29]. The cut-off points of energy expenditure of physical activity in this study were 6.20 kcal/kg/day for men and 7.13 kcal/kg/day for women. Obviously, the estimates of energy expenditure were much higher than those in the CHS, since tri-axial accelerometers are capable of recording energy expenditure derived from a variety of daily physical activities rather than specific physical activities. Although accelerometer-based measurement may be less comparable to existing or historical cohorts, this objective measure of low physical activity may potentially standardize measurement in future cohorts and improve diagnostic accuracy of the frailty phenotype. In addition, the objective measurement of physical activity can be administered by non-professionals, which could possibly raise the feasibility of the Fried frailty criteria in primary care settings. Commercial products of tri-axial accelerometer devices have been increasingly distributed and available in the general population after validation, even in the form of smartphone applications [
30].
It is intriguing that we found that the prevalence of frailty was similar in both genders. This finding was consistent with some reports [
14,
31], while a systematic review showed a greater difference in the weighted prevalence of frailty between men and women (9.6% vs. 5.2%) [
5]. The lack of a gender difference could be partly explained by the differences between North American or European and Japanese populations. Alternatively, there was a higher prevalence of frailty among men in our study than in other international studies. The age distribution for Japanese elderly men is different than for men from other nations: Japanese elderly men (+65 years) live longer and the proportion of the oldest old (+80 year) among elderly men is higher than those in many other nations (
https://www.cia.gov/library/publications/resources/the-world-factbook/geos/ja.html). The age distribution for elderly men in our study was similar to the national age distribution, as shown in Additional file
1. Moreover, Japanese elderly men have reported a similar prevalence of sarcopenia with that of women [
32], which plays a central role in the pathogenesis of frailty [
3].
We found many similarities between the results in our studies and those reported in previous research [
5], we found that the rate of frailty increased dependent on age after adjusting for co-morbidities. Frail individuals reported poorer self-perceived health [
6]. A possible reason for is that as the level of frailty increases, so does the tendency to rate their health poorly [
31,
33]. Although we used a psychological distress scale to measure depressive and anxiety symptoms in our study, our result, which indicated a significant association with frailty, is in agreement with findings reported in previous studies using other depression measures [
14,
34,
35]. Living alone, another factor that has been found to be related to frailty, is related to poorer nutrition, which is a cardinal component of frailty [
3]. The diversity of social ties might exert a beneficial effect on frailty [
34,
36], while living alone may indicate poorer social ties [
21].
In agreement with some previous studies [
6], we observed that frailty was unrelated to socioeconomic status, such as education, income, employment status, or house tenure. This may be explained by the universal health coverage of social health insurance in Japan, especially for the oldest elders, and by the equity in social economic conditions adequate to health maintenance [
37]. The observed associations between cognitive impairment and frailty could be explained by several mechanisms, such as Alzheimer’s disease pathology, hormone dysregulation, and impaired nutrition [
38]. Concordant with previous studies [
12,
36], we also observed that current alcohol consumption was associated with lower odds of frailty. The association may be explained by an avoidance of alcohol [
39] or a decrease of alcohol-related socialization among those who were frail, in line with Japanese culture that people commonly believe drinking alcohol facilitates socialization and mutual understanding between individuals [
40]. Older adults who engaged in social activities were less likely to become frail. Frequent engagement in social activities could help to maintain physical and mental fitness [
41] and then compensate for age-related decline in reserve and function. Another explanation is that withdrawal from social activities could be a behavioral precursor of frailty. These findings favored the notion that an overt state of frailty may be preceded by behavioral adaptation, such as withdrawal from social activities, made in response to declining physiologic reserve and capacity [
2]. Early detection of frailty and pre-frailty before decreased reserves become more pronounced helps to shift towards more appropriate goal-directed and individualized care provision. The potential role of the correlates of frailty and pre-frailty in prevention and intervention merits further studies to explore their clinical application, since frailty is a reversible process.
Strengths and limitations
To our knowledge, this is the first attempt to date using a tri-axial accelerometer to define energy expenditure of physical activity for the frailty phenotype. We examined a wide range of potential correlates of frailty covering social, psychological, environmental, and health-related factors. This study also has several important limitations. The sample was not nationally representative. The cross-sectional design prevents conclusions of directional relationships. There might have been selection bias due to the relatively low participation rate. However, given the similar prevalence findings in our study and previous studies, we may extrapolate that potential response or selection biases would not tend to lead to underestimation or over-estimation in the prevalence of frailty in this study.
Acknowledgements
We would like to thank Dr. Nofuji Yu and Ms. Matsuo Eri’s contributions to the data collection. We also would like give our thanks to the municipal staffs in primary care-giving division of Sasaguri town, especially Ms. Gunjima Kumiko who provided us with official data about the inhabitants without placement in long-term care insurance and helped us coordinate the survey in the community. None of them had role in the study design, data analysis, data interpretation, writing of the manuscript, nor in the decision to submit the manuscript. This study was supported in part by Health and Labour Sciences Research Grant of the Japan Ministry of Health, Labour and Welfare (H25-Ninchisho-Ippan-004) and a research grant from Sasaguri town, Fukuoka, Japan. As a financial sponsor, they had no role in the study design, data analysis, data interpretation, writing of the manuscript, or in the decision to submit the manuscript. CSM is supported by China Scholarship Council (CSC).
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
CSM carried out the conception and design, analysis, drafting and revising manuscript; HT was involved in the design, analysis, and revising manuscript; CT participated in the design, and revising manuscript. NK was involved in the design and revising manuscript. HY participated in the design and revising manuscript. SA participated in the design and revising manuscript. KS contributed to the design, planning, coordination, and revision. All authors have read and approved the final manuscript.