A four-domain approach of frailty explored in the Doetinchem Cohort Study

With ageing, changes occur in physical, psychological, cognitive, and social functioning. Accumulation of problems in one or more of these domains of functioning is characteristic for frail people. Originally, frailty was mainly focused on the physical problems that older people encounter, such as in Fried’s popular ‘phenotype of frailty’ [1]. Broader definitions of frailty, looking beyond physical functioning, have now been put forward [2‐4] one of which is the definition by Gobbens et al. [5]. According to them, frailty is ‘a dynamic state affecting an individual who experiences losses in one or more domains of human functioning (physical, psychological, social) caused by the influence of a range of variables and which increases the risk of adverse outcomes’. A multidimensional approach to frailty is coherent with the interdisciplinary diagnostic process used in the Comprehensive Geriatric Assessment for frail older people, which also examines physical, mental (including both psychological and cognitive functioning), and social functioning [6, 7].

Frailty often leads to restrictions in mobility and reduced self-reliance, and a greater risk of clinically significant adverse outcomes such as hospitalization, institutionalization and mortality [1, 4, 8‐10]. In several European countries it is government policy to stimulate older people to participate in society and to live at home for as long as possible [11, 12]. Primary prevention of frailty is therefore needed, directed at both delaying the onset of frailty and slowing down the frailty process as prevention of frailty might eventually lead to prevention or postponement of hospitalization and institutionalization of elderly people. Insight in factors that are associated with the presence of frailty is a first step to assist the identification of potentially vulnerable groups. For the physically frail, a series of socio-demographic, lifestyle, and health-related factors have been shown to be associated with frailty, such as age, female sex, cardiovascular diseases, multimorbidity, BMI, and smoking [13]. As part of a broad frailty definition, little is known about factors associated with the psychological and social domains of frailty. Especially, the association between lifestyle factors and frailty has rarely been studied [13].

Recently, the concept of cognitive frailty has been proposed [14]. Since there is increasing support for the idea of cognitive frailty being a separate frailty domain [15, 16], a four-domain approach of frailty was adopted for the current study including the physical, cognitive, psychological and social domains of frailty. It was recently shown that the overlap between these frailty domains was limited, which implicates that the domains largely entail distinct populations and frailty prevention may target multiple frailty domains [17].

In this study, we explored how sociodemographic factors, lifestyle factors, life events, biological risk factors and chronic diseases were associated with physical, psychological, cognitive, and social frailty in a population-based study of men and women aged 40–81 years.

In our population aged 40–81 years, 2.7% was physically frail, 6.3% was psychologically frail, 7.7% was cognitively frail, and 4.1% was socially frail (Table 2). Women were more frequently physically and psychologically frail (63.6% and 68.7%) than men (36.4% and 31.3% respectively), whereas men were more often cognitively frail than women (68.8% vs. 31.2%). The mean age of those with physical, cognitive and social frailty was higher compared to that of the total study population. Each of the domains of frailty showed a higher percentage of respondents with a low educational level compared to the total study population.

17.1% of the population was frail on one or more of the domains. Of the frail population, 81.5% was frail on one of the domains, 15.2% was frail on two domains, 2.9% was frail on three domains, and 0.4% was frail on all four domains. The greatest overlap was observed for physical and psychological frailty and for social and psychological frailty (one third of the physically and socially frail being also psychologically frail) (Table 2).

This study suggests that each of the different frailty domains all had a specific combination of associated factors. Most socio-demographic factors and lifestyle were associated with being frail on each of the domains. Being physically active was consistently associated with a lower risk of being frail on each of the four domains. A short or long sleep duration was associated with a higher risk of being physically, psychologically, and socially frail. Other factors associated with one or more domains of frailty were female sex, high age, a low educational level, being married, a paid job, current smoking, a healthy diet, and multimorbidity.

Drawing on the integral conceptual model of frailty, we observed a prevalence of 17.1% among men and women of 40 to 81 years who lived independently. The proportion of frail persons in a population is dependent on the definition of frailty used [34] and on characteristics of the study population. The prevalence of frailty that we observed is relatively low compared to recent other studies based on the integral conceptual model of frailty [21, 22, 35], which could be explained by the large age range of our population and the absence of persons over 81 years of age. The majority of studies directed to frailty have focused solely on people over the age of 65, despite emerging evidence suggesting that frailty begins much earlier than that [36, 37]. Our findings show that frailty may already exist at a relative young age and therefore extend the findings of previous studies.

The relationship between socio-demographic factors and physical frailty has been described in the literature [13, 38]. Like age, sex also contributed differently to each of the frailty domains: being female was associated with a higher risk of being psychologically frail and a lower risk of being cognitively and socially frail. A recent review showed socioeconomic status to be inversely associated with physical frailty [13]. Our findings support a higher risk of being psychologically and socially frail for people with low education. Briefly, socio-demographic factors are important for frailty but their impact varied for each of the domains of frailty.

Lifestyle factors in relation to multiple domains of frailty have not (yet) been studied extensively. An unhealthy lifestyle was previously found to be associated with a higher risk of being physically and socially frail [39‐41], and psychologically frail (including cognitive frailty) [39, 40]. However, in these studies lifestyle was assessed by a single item in a self-report questionnaire. Such assessment of lifestyle precludes unambiguous interpretation, because it remains unclear which lifestyle factors participants had in mind when answering the question and what aspect they considered to be unhealthy [40]. Our findings provide novel insight into the specific lifestyle factors (physical activity, smoking, diet, alcohol consumption, sleep) associated with the different domains of frailty. Physical activity was significantly associated with all domains of frailty in our study. A previous study of Strawbridge also considered a broad range of risk factors including lifestyle, in relation to a multidimensional definition of frailty [42]. Being physically inactive, either at one instant or at several measurements over a period of 29 years, was associated with a higher risk of being frail. Other studies confirmed the associations of physical activity with physical frailty and cognitive decline [43, 44], as far as we know no studies included the domains of psychological and social frailty. Our results are in line with earlier studies showing that current smoking was associated with a higher risk of being physically frail [45], and a healthy diet was associated with a lower risk of being physically and cognitively frail [46, 47]. A new insight based on our findings is that short sleep duration was associated with a higher risk to be physically, psychologically and socially frail, and long sleep duration was associated with a higher risk to be psychologically frail. Sleep deprivation contributes to a number of molecular, immune and neural changes that play a role in the development of health problems [48]. Previously, sleep quality and sleep disturbances, but not sleep duration, were reported to be associated with physical frailty [49]. More detailed studies are needed to understand the relation between sleep and each of the frailty domains [50].

In addition to sociodemographic factors and lifestyle, we studied life events, multimorbidity and overweight. Life events were not associated with any of the frailty domains in our study. Other studies reported life events to be associated with a higher risk of being psychologically frail [39‐41]. Overweight and obesity were not associated with frailty in our multivariate models, which is comparable to the findings of Strawbridge [42]. Some studies that adjusted for socio-demographic variables and smoking, but not for other lifestyle factors, did report an association between obesity and (physical) frailty [51, 52]. These inconsistent results regarding the relationship between life events and overweight with frailty can be explained by methodological differences such as the definition of life events. Finally, multimorbidity was associated with a higher risk to be physically and psychologically frail, which is in line with other studies [39‐42]. Fried illustrated that frailty is distinct but overlapping with comorbidity, with almost 70% of the frail persons also having two or more diseases [8].

In general, there is no consensus about a definition of frailty [34, 53] as shown by the different approaches described in the literature. Besides the well-known frailty phenotype approach [1], the frailty index is another dominant approach in frailty research [54]. The frailty index involves the accumulation of diseases, symptoms, signs, disabilities or any deficiency in health with age [55]. Although different domains of human functioning are incorporated in the frailty index, it differs from our approach to frailty because it considers frailty as much broader than functioning alone. The incorporation of social functioning in the concept of frailty is an area of discussion. During the development of the integral conceptual model of frailty a group of frailty experts agreed upon including social functioning in this model [19]. The social domain cannot be left out because it is relevant to an integrated view of human beings [3, 56], the relationship with adverse outcomes is demonstrated [57, 58], and ‘social relationships’ and ‘social support’ are viewed as determinants of frailty [42, 59, 60]. Two other reasons to consider social functioning as part of a multidimensional definition of frailty are (1) social functioning is regarded as separate health domain in the Comprehensive Geriatric Assessment and therefore viewed as relevant in clinical practice [7], and (2) in a qualitative study the majority of interviewed elderly persons reported to consider reduced social functioning as an important component of frailty [61].

The Doetinchem Cohort Study is a unique cohort for studying frailty because of the relatively wide age range of the participants and the ability to define multiple frailty domains due to the wide array of collected variables. The original TFI scale is based on self-report, but we were able to combine self-reported and objectively measured variables to define frailty. Since the population has a wide age range and includes middle-aged adults, we used similar or more stringent cut-off points than applied earlier by van Campen [22]. A sensitivity analysis for cognitive frailty with a lower cut-off (<7.5%) confirmed the findings of cognitive frailty defined by 10th percentile, except for alcohol consumption, obesity and multimorbidity. Since data required to define social frailty (as well as some of the indicators for physical and psychological frailty) were measured for the first time in the most recent completed round of the Doetinchem Cohort Study, longitudinal analyses were not possible. As such, all analyses were cross-sectional and causal inferences cannot be made. To illustrate this, we take the example of having a paid job. We found that having a paid job was strongly associated with a lower risk to be physically frail. This may suggest a protective effect of having a paid job for being physically frail. However, it may also imply that frail persons have stopped working because of limitations in work functioning. Future studies are needed to address the prospective association between a various range of factors and the development of physical, psychological, cognitive, and social frailty. Another limitation is the measurement of life events. Other studies have shown that especially death of a loved one was associated with frailty [39]. The questionnaire used for the Doetinchem Cohort study did not assess death of a loved one. We therefore decided to use the variable ‘being widowed’ as a proxy. We should however, acknowledge that the life events being widowed and death of a loved one may partly overlap but are not the same. This may explain the lack of association with psychological frailty. Although the response rates for the examination rounds varies between 75 and 80%, we cannot exclude selective response that may have caused an underrepresentation of severely frail participants, in particular of physically and cognitively frail individuals. To minimize a healthy cohort effect, home visits were offered if participants were not able to get to the municipal health services where the examinations were carried out. This way, frail participants could still participate in the study. However, due to potential selective response the observed associations may be underestimated.

The present study suggests that frailty, in particular psychological frailty, may already be present at a relatively young age. Sociodemographic factors, lifestyle and multimorbidity contributed differently to each of the frailty domains. This highlights the relevance of a multidimensional approach to frailty as operationalized in the integral conceptual model of frailty. Understanding which groups of older adults are at risk of being frail on each domain may help to prevent frailty and to identify frail individuals in an early stage. Identification of frail individuals is an important step to preclude the development of undesirable outcomes, to provide adequate healthcare and support, and to effectively prevent and delay the development of frailty by health professionals. Since we know little about factors associated with the psychological, cognitive and social domains of frailty, our findings add to available literature and are relevant for clinical practice [34]. We found that lower educated adults were at higher risk of being frail, and therefore preventive strategies should be directed at this group. Furthermore, our findings suggest that lifestyle factors, specifically physical activity and sleep, are associated with the presence of frailty. As lifestyle is potentially modifiable, interventions directed to improve lifestyle may provide new opportunities for the prevention of frailty in the future. But first longitudinal research should be conducted to better examine how lifestyle affects the development of frailty and its dynamic course.