Settings and participants
Two hundred and ninety-nine participants were recruited from five community elderly centres and nine day-care centres for the elderly in Hong Kong between May 2015 and January 2016. A convenience and snowball sampling method was used to recruit community-dwelling older people aged ≥ 65 who could communicate in Cantonese and who had been assessed as being in a pre-frail or frail state according to the Fried Frailty Index (FFI) [
20]. The items in the index included: i) unintentional weight loss: a self-reported unintentional loss of 10% of body weight in the past year; ii) exhaustion: by answering “Yes” to either “I felt that everything I did was an effort” or “I could not get going in the last week”; iii) slowness: a 4.5-m walk with an average walking speed in the lowest quintile stratified by median body height; iv) weakness: with a maximal grip strength, as measured by hand dynamometers, in the lowest quintile stratified by the body mass index quartile; and v) low activity: a Physical Activity Scale for the Elderly-Chinese (PASE-C) score in the lowest quintile (i.e., < 30 for men and < 27.5 for women). The presence of ≧ 3 items indicated that the elderly person was in a state of frailty, while 1–2 items indicated pre-frailty [
20].
Participants were excluded if they were cognitively impaired (with an abbreviated mental test score of < 6), had been admitted to hospital in the past 6 months, or were confined to bed or restricted by the permanent use of a wheelchair.
Measurement
Participation restriction was measured by the C-RNLI, which was translated from the RNLI. The translation process followed standard procedures involving translation and back-translation [
21]. A professional translator first produced a provisional translated version of the RNLI. This provisional version was then back-translated by LJWY (the first author). A comparison was made between the back-translated version and the RNLI, and the discrepancies between them were discussed. The required modifications were made and a pre-final version of the C-RNLI was drawn up. Then, five older people were invited to comment on the pre-final version of the C-RNLI in terms of its difficulty, quality, clarity, and language use. The pre-final version of the C-RNLI was reviewed by the research team according to the feedback from the older people, and the final C-RNLI was produced.
Although Pang et al. [
9] had developed the Chinese version of the RNLI based on the extent to which 75 patients with chronic stroke had reintegrated to normal living [
9], in this study it was decided that a new version of the C-RNLI should be developed. This was because the target participants of this study were frail older people, most of whom had received little or no education when they were young because of the disruption of the war years, and hence had a low level of literacy [
22]. Thus, a version of the C-RNLI using simple words and structures was needed to ensure that they would be able to understand it. This was achieved by soliciting the opinions of a gerontologist and a group of frail older people during the process of developing the C-RNLI in this study. The C-RNLI consists of 11 declarative statements rated on an 11-point numerical rating scale (with 0 indicating the least agreement and 10 the greatest agreement with the statements). Item scores were summed and proportionally converted to 100 through dividing the score by 1.1 to provide a total score, with a lower score indicating a higher level of participation restriction [
5].
The original RNLI was validated among 109 patients from hospitals and rehabilitation centres [
4,
5]. The Cronbach’s alpha of the RNLI ranged from 0.87 to 0.97 in studies involving patients with stroke [
11], spinal cord injuries [
8], and limited mobility [
7], which supported its internal consistency with a measure construct of RNL. The two-factor structure of the RNLI was first proposed by Wood-Dauphinee et al. [
5]. The scores from items 1 to 8 were summed to give the total score of factor 1, which was called “daily functioning”; and the scores from items 9 to 11 were summed to give the total score of factor 2, which was called “perception of self” [
5]. This factor structure was also found in the Chinese version of the RNLI developed by Pang et al. [
9]. However, the exploratory factor analysis conducted by Stark et al. [
7] showed another two-factor structure for the RNLI when the RNLI was validated with community-dwelling people with limited mobility [
7] and comorbidities [
6]. For this factor structure, the scores from items 1 to 5 were summed to give the total score of factor 1, which was called “physical reintegration”; and the scores from items 6 to 11 were summed to give the total of factor 2, which was called “social reintegration”.
The depressive mood of the participants was measured by the Cantonese version of the Geriatric Depression Scale Short Form (CGDS-SF) [
23]. The participants were required to answer yes or no to 15 statements describing different emotions. The CGDS-SF had good internal consistency, test-retest reliability, criterion-related validity [
24], and good sensitivity and specificity for identifying geriatric depression [
25]. The scores of the items were summed to provide the total CGDS-SF score, where a higher score indicated a higher level of depressive mood. Under the WHO-ICF, depressive mood is a kind of impaired body function [
1], and it has been found to be associated with participation restriction [
16]. Therefore, it was hypothesized in this study that a negative correlation would be found between the C-RNLI and the CGDS-SF to establish the convergent validity of the C-RNLI in relation to participation restriction.
The functional performance in daily activities of all of the participants was measured using the Hong Kong Chinese version of the 9-item Lawton Instrumental Activities of Daily Living Scale (HKC-IADL) [
26]. Each item was rated on a 4-point Likert scale. All item scores were summed to provide the total score, with a lower score indicating a higher level of dependence. The scale showed good internal consistency, test-retest reliability, inter-rater reliability, and construct validity [
27]. The HKC-IADL was used to measure activity limitations, which is a dysfunction level of the WHO-ICF [
1], and it was found that older people who frequently participate in social activities were less likely to display disabilities in the instrumental ADLs [
28]. Thus, it was hypothesized in this study that a positive correlation would be found between the C-RNLI and the HKC-IADL to establish the convergent validity of the C-RNLI in relation to participation restriction.
Data analysis
The data were analysed using the Statistical Package for the Social Sciences 21.0 (SPSS Inc., Chicago, IL), and the confirmatory factor analysis (CFA) was conducted using the Stata Statistical Software 14 (StataCorp., Texas, TX). Descriptive statistics were used to evaluate the demographic data of the participants as well as the other assessments. The reliability of the scale was established using internal consistency and test-retest reliability. Cronbach’s alpha was used to measure the internal consistency of the scale, with a reliable scale having a Cronbach’s alpha of > 0.7 [
29]. Intra-class coefficient (ICC) (2,1) was used to measure test-retest reliability, with an ICC of > 0.75 indicating good reliability and an ICC of between 0.5 and 0.75 indicating moderate reliability [
30]. The item-total correlation and the Cronbach’s alpha were also checked after items were deleted. A reliable item should have an item-total correlation of > 0.3 [
31] and should not cause the Cronbach’s alpha to become larger after it is deleted. A Spearman’s rank correlation coefficient (r
s) was used to establish hypothesis-testing based construct validity, with a correlation coefficient of between 0.1 and 0.29 considered a small effect, that between 0.3 and 0.49 considered a moderate effect, and that ≧0.5 considered a large effect [
32]. It was expected that the C-RNLI would be significantly correlated with the CGDS-SF negatively and with the HKC-IADL positively.
The known-groups method was used to establish construct validity. All of the participants were classified based on their frailty status and whether they had been recruited from community versus day-care centres. The participants from the day-care centres were generally more impaired in their ability to maintain their optimal level of daily activities, and less capable of taking care of themselves than the participants from the community centres. It was hypothesized that participants recruited from day-care centres would have significantly lower C-RNLI scores than participants from community centres. Likewise, participation restriction was found to be more prevalent in frail older people [
16]. Therefore, it was hypothesized that frail participants would have significantly lower C-RNLI scores than pre-frail participants. A Mann-Whitney
U test was used to determine whether there were significant differences between these two groups in C-RNLI scores. A value of
p < 0.05 was considered statistically significant.
The CFA was used to establish the structural validity of the C-RNLI. A CFA with a maximum likelihood estimation was conducted for the respective factor structures proposed by Wood-Dauphinee et al. [
5] and Stark et al. [
7] to identify which one would be the most appropriate factor structure for the C-RNLI in this study. An acceptable model fit would be indicated by: i)
\( \frac{\mathrm{chi}\hbox{-} \mathrm{squared}}{\mathrm{degree}\;\mathrm{of}\;\mathrm{freedom}}\le 3 \); ii) a root mean square error of approximation (RMSEA) score of ≤ 0.08, iii) a comparative fit index (CFI) score of ≥ 0.95; and vi) a Tucker-Lewis Index (TLI) score of ≥ 0.90 [
33‐
37].