1. Introduction
Population aging is occurring all over the world and is accompanied by an increasing number of people with disability; after all, the prevalence of disability is associated with aging. In the USA, the prevalence of disability is 10.6% among people aged 18–64 years and 35.2% for those aged 65 years or older [
1]. There are different definitions of disability. According to the World Health Organization (WHO), disability has three dimensions referring to impairment in a person’s body structure or function (e.g., loss of a limb), or mental functioning (e.g., loss of memory), activity limitation (e.g., difficulty in walking), and participation restrictions in performing daily activities (e.g., engaging in recreational and social activities) [
2]. In research focusing on older people, disability is often defined as having difficulty conducting activities of daily living (ADL) and/or instrumental activities of daily living (IADL) [
3,
4]; the second dimension was established by the WHO (activity limitation). Examples of ADL include getting on and off the toilet and standing up from sitting in a chair; doing the shopping, washing, and ironing clothes refer to IADL [
5,
6]. IADL disability reflects a less severe form of disability than ADL disability [
7,
8]. In a sample of Polish people aged ≥60 years, it was observed that 35.75% and 17.13% reported at least one problem with IADL and ADL, respectively [
9]. In a sample of Dutch people aged ≥75 years, these percentages were 54.6% for IADL disability and 67.4% for ADL disability [
10].
Disability is associated with adverse outcomes in older people, such as increased use of health care [
11,
12], lower quality of life [
10,
13], and premature death [
14,
15,
16]. Early identification, prevention, and intervention targeting older people living with a disability is, therefore, of utmost importance. To achieve this, knowledge of factors influencing disability is essential. In Poland, it was shown that the occurrence of disability is influenced by a lack of social contacts, multimorbidity, pain, and the presence of barriers in the environment for older people [
9]. In the Netherlands, risk groups include those living alone, women, people who are widowed or divorced, and those with a low educational level [
17]. In addition to these risk factors and at-risk groups, frailty can be considered a major predictor of disability [
18,
19,
20]. A systematic review using 28 studies showed that physical frailty indicators could predict ADL disability; the most powerful predictors were slow gait speed and low physical activity [
20]. A more recent systematic review and meta-analysis including 20 studies quantitatively showed that frail older people were more likely to develop or have more severe ADL and IADL disability [
19]. Finally, a systematic review and meta-analysis of prospective cohort studies including 32,998 people aged 60 years or older also concluded that those who are frail have the highest relative risk of disability [
18].
As with disability, there are also different definitions of frailty [
21]. There are definitions that view frailty exclusively as a concept that refers only to physical limitations that older people may have. One measurement tool that goes along with these definitions is the phenotype of frailty frequently used in studies [
22]. This tool contains five physical criteria by which a healthcare professional can determine if a person is frail: weakness, unintentional weight loss, poor endurance, slowness, and low physical activity [
22]. As a counterpart to physical frailty definitions, there are also definitions that emphasize the multidimensional nature of frailty. The measurement tools that align with these definitions not only pay attention to physical limitations but also to the psychological and social limitations of older people. Examples of such tools are the Frailty Index [
23], based on the Canadian Study of Health and Aging (CSHA) Cumulative Deficit Model, and the Tilburg Frailty Indicator (TFI) [
24].
As mentioned, many studies have been performed that aimed at examining the association between frailty and disability and, in particular, predicting disability by frailty [
18,
19,
20]. In general, in these studies, frailty was assessed using a tool containing physical items (e.g., the phenotype of frailty). Studies using a multidimensional tool are rarer. In the present study, we used the TFI items belonging to physical, psychological, and social frailty as predictors of disability. We have assigned one point per item if there was a deficit. Linear and logistic regressions are frequently used as techniques to develop a prediction model; however, the relationship between TFI items and disability may differ. In this study, we aimed to develop nomograms with the TFI items as predictors by using cross-sectional data and longitudinal data, focusing on the prediction of disability, where we distinguish between total disability, ADL disability, and IADL disability. We will show that the nomograms differ per outcome and per time point. The different time points refer to the longitudinal aspect of the study. The knowledge we gather and present in the nomograms can support healthcare professionals in determining the risk of disability in community-dwelling older people.
4. Discussion
Disability is associated with increased healthcare utilization [
11,
12], lower quality of life [
10,
13], and premature death [
14,
15,
16] among older people. Frailty, which is also common in older people, can be considered a determinant of disability [
18,
19,
20]. Thus, to prevent disability, it is useful to understand the contribution of individual frailty components in predicting disability. This insight is important for healthcare professionals to be able to intervene early so that disability is prevented or at least delayed. In this study, we aimed to develop nomograms with items of the Tilburg Frailty Indicator (TFI) [
24] for predicting disability, assessed with the Groningen Activity Restriction Scale (GARS) [
27], where we distinguish between total disability, ADL disability, and IADL disability. We used both cross-sectional data and longitudinal data, with a follow-up of five and nine years of a sample of Dutch community-dwelling people aged ≥75 years. Our premise was that the nomograms for predicting disability would differ by outcome variable and by time point. In this section, we will discuss only the main findings.
The nomograms derived from linear regression analyses showed the points that must be given to the 15 frailty components assessed with the TFI. The three monograms belonging to total, ADL, and IADL disability showed that the frailty item ‘difficulty in walking’ is the most important predictor at baseline (T0); ‘difficulty in walking’ scored 100 points for all three disability outcomes. The relative importance of this TFI item was less at T5 (total, IADL) and T9; regarding IADL disability, ‘difficulty in walking’ scored 0 points at T9. A systematic review including 28 studies showed that slow walking speed was one of the most powerful predictors of ADL disability in community-dwelling older people aged 65 years or older [
20]. Another systematic review also found an association between walking speed and the probability of disability [
32]. A recently developed prediction model demonstrated that age, walking speed, and cognitive function were the strongest predictors of disability-free survival in healthy older people [
33]. We only found two studies, both longitudinal, carried out in the Netherlands aimed at predicting disability assessed with the GARS by frailty items [
34,
35]. The first used the physical subscale of the TFI [
35]; this study was conducted in a sample of 429 Dutch people aged ≥ 65 years and showed, based on linear regression analyses, that slowness predicted both total and IADL using a follow-up of two and a half years. The second study aimed to predict ADL and IADL disability using an objective measure of walking speed, the Timed Up and Go (TUG) test, with a follow-up period of one year [
34]. After controlling for previous disability and other predictors (background characteristics, body mass index, physical activity, handgrip strength, fatigue, balance), walking speed was predictive for total, ADL, and IADL disability.
Another physical frailty component, ‘unexplained weight loss,’ was the most important predictor for total, ADL, and IADL disability at T9; this item achieved the highest number of points for all three outcome variables (100 points). Unexplained weight loss is common in frail older people, with prevalence figures rising to 27% [
36]. Based on this finding, prevention of unexplained weight loss in older people seems to be very important, especially since unexplained weight loss in this target group is associated with increased morbidity as well as increased mortality [
37]. Additionally, after controlling for health, functional status, and social network, this frailty item is an important predictor of early institutionalization [
36]. According to Alibhai et al. [
38], for managing unexplained weight loss, identifying and treating the underlying causes (e.g., malignant disease, psychiatric disorder, gastrointestinal disease) should be the first priority.
Lack of strength in the hands and problems with memory were important predictors for IADL disability, in particular at T9, with 93 and 84 points, respectively. This was not the case with ADL disability; at T9, ‘lack of strength in the hands’ scored 36 points and ‘problems with memory’ scored 0 points. In addition to the two items mentioned earlier (difficulty in walking, unintentional weight loss), ‘lack of strength in the hands’ belongs to the five criteria of the phenotype of frailty by Fried et al. [
22]. Our finding is confirmed by an umbrella review of systematic reviews using meta-analyses of observational studies [
39]. In this review, handgrip strength was not only considered a useful indicator for disability but also for general health status and mortality. It should be noted, however, that the operational definition of disability was different from our definition, IADL disability assessed with the GARS. With regard to the frailty item ‘problems with memory’, referring to cognitive impairment, many previous studies showed that cognition predicts disability, e.g., Shimada et al. [
40] and St John et al. [
41]. In a community-dwelling baseline sample of 1715 people aged ≥65 years, cognition determined with the mini-mental state examination (MMSE) predicted disability 5 years later [
41]. In addition, in a Japanese sample consisting of 4290 community-dwelling older people aged 65 years or older, cognitive impairment showed a significant association with disability [
40].
The TFI assesses physical, psychological, and social frailty. The findings of the present study showed that the eight items belonging to physical frailty predicted total, ADL and IADL disability to a greater extent, in particular, difficulty in walking and unintentional weight loss, or, to a lesser extent, especially poor hearing, poor vision, and physical tiredness. The items referring to psychological and social frailty received far fewer points. The social frailty item ‘lack of social support’ was the only frailty item that scored zero points. For predicting disability, psychological and social frailty are obviously less important than physical frailty. It is assumed that in other adverse outcomes of frailty (e.g., increased healthcare utilization, lower quality of life), the nomograms will strongly differ, and more points will be given to psychological and social frailty items. If we consider the prediction of ADL and IADL disability using the TFI, then we can observe that the Rsq-values were nearly the same; the main difference in Rsq-values existed between ADL and IADL at T5 (0.52 versus 0.44). In addition, for both T0 and T5, there were more predictors of disability than for T9. At T9, there were only nine, seven, and eight predictors for total, ADL, and IADL disability, respectively. This finding is not so surprising; after all, making a long-term prediction is trickier than making a short-term prediction.
Our study has several limitations. Firstly, the generalisability of the findings may be questioned because the response rate at baseline was only 42%. Secondly, for the prediction of total, ADL, and IADL disability, we only used frailty items and did not control for background characteristics of the participants, e.g., individual diseases or multimorbidity and age. However, in a previous Dutch study using the TFI and GARS, it was observed after sequential linear regression analyses that only previous disability and age significantly contributed to the prediction of disability; multimorbidity and other background characteristics (e.g., sex, education, income) did not [
35]. Third, the sample size can be considered small at T9. Because of the high age at baseline (mean 80.3 years; SD 3.8), many older people were unable to participate in the follow-up of this study, especially for the measurement of disability at T9. In a previous study using the same sample, it was observed that 162 died between 2008 and 2015 [
42]. Finally, we would like to indicate that discussion of our findings in light of findings from previous studies has its limitations because both frailty and disability were usually measured with instruments other than the TFI and the GARS, such as the phenotype of frailty [
22], the FI [
23] for assessing frailty, and the Katz Scale [
43] and the Lawton and Brody Scale [
44] for assessing ADL and IADL disability, respectively.