Design
This study was designed as a questionnaire-based cross-sectional survey.
Ethics clearance
It was approved by the XXX Research Ethics Committee (No. 201905XXXXX), and official written consent from each facility was obtained. Prior to the study, the research process was explained in detail to the participants.
Settings and sample population
During the period from April to June 2019, random sampling was performed using the multi-stage sampling method. For factor analysis, the sample size for items must be 5-10 according to previous studies [14,15]. The Chinese PaArticular Scales have a total of 35 items; therefore, a sample size of 175-350 participants was appropriate for factor analysis in this study [16]. The inclusion criteria of the participants were as follows: (1) ≥ 65 years old; (2) more than 6 months living at the facility; (3) having the language skills to fill out or answer the questionnaire; (4) severe joint contractures in any one important joint (knee, hip, ankle, shoulder, elbow and hand) with confirmation from a doctor, a nurse, or a therapist. Severe joint contractures were defined as 3 on a 4-point scale (loss > 2/3 of joint range of motion) [17,18]. Those with cognitive impairment and major mental illness diagnosed by physicians were excluded.
Study instrument
Disease-related and socio-demographic data
Nursing care dependency can be classified as mild, moderate, severe, and extremely severe, which is determined based on the evaluation report of medical service experts of the national statutory LTC insurance system. To further describe the study population, the minimum data set (MDS) tool recommended by InterRAI Country Websites was used to record socio-demographic data (such as gender and age) and the location of joint contractures (based on medical records and the MDS) [19].
Cognitive status
The Mini-Mental Status Examination (MMSE) was used to evaluate the cognitive status of participants [20]. The MMSE has a total of 13 items, with a total score of 33, and only takes 5-10 minutes to complete. It is a simple quantitative assessment scale that is widely used in clinics and research to evaluate cognitive function and screen cognitive impairment. The higher the score is, the better the cognitive function. The test-retest reliability is good, and the correlation coefficient of inter-rater reliability is .8 [20]. An MMSE score below 25 is defined as cognitive dysfunction.
Chinese version of the PaArticular Scales
The PaArticular Scales consist of 35 items: The Activity subscale has 24 items, and the Participation subscale has 11 items. As an organized face-to-face questionnaire to evaluate activity limitations and participation restrictions, before the interview, the interviewer emphasized that the participants must consider their current environment, not hypothetical environments or their former home environment. In terms of reliability, the Cronbach’s α values of the internal consistency of the Activity subscale and the Participation subscale were .96 and .92, respectively, and the McDonald’s ω total were .98 and .95, respectively, indicating the high internal consistency. In terms of validity, the Pearson correlation coefficients of the 2 subscales (the Activity subscale and the Participation subscale) using the criterion validity of the visual analogue scale of the EQ-5D, which is one of the most frequently used generic health status measurement tools, demonstrated good validity and reliability at -.40 (p > .001) and -.30 (p > .001), respectively [12].
The World Health Organization Quality of Life (WHOQOL)-BREF
To evaluate the criterion validity of the Chinese version of the Scales, we used the Chinese version of the WHOQOL-BREF developed the WHO's WHOQOL group, which contains 26 items. Questionnaires with over 20% of missing data should be discarded. Missing values are replaced by the average domain value. If there are more than 2 missing values in a domain, the domain score is not calculated (except for domain 3, whose score is only calculated if the missing value <1). The Cronbach’s α value of the internal consistency of the overall questionnaire is .90, and the test-retest reliability of each category reaches .75 or above. For the Pearson correlation, the correlation between each item and its category ranges from .45 to .82 (p < .01), and the correlation between different categories ranges from .48 to .63 (p <.01). For the confirmatory factor analysis (CFA) of the construct validity, the structural equation model of the 4 factors echoes the potential structure designed by the questionnaire, and the comparative fitness indices (CFI) of these 2 analyses are all .886, which is equivalent to that of the Hong Kong version of the questionnaire (CFI = .894) and similar to that of the questionnaire using global data (CFI = .903) [21].
The WHO Disability Assessment Schedule 2.0 (WHODAS 2.0) - 36 items
To evaluate the criterion validity of the Chinese version of the Scales, we used the Chinese version of the WHODAS 2.0- 36 items. A 5-point Likert scale is used by participants to answer questions related to difficulties performing activities. The score ranges from 0 (lowest difficulty) to 100 (maximum difficulty) and is calculated as the sum of each domain score [22]. The higher the score is, the higher the degree of disability and the more severe the restricted situation. Restriction severity refers to the difficulty level classification method of the ICF and WHODAS 2.0. The classification of impairment severity is as follows: below 4% is none; 5-24% is mild; 25-49% is moderate; 50-95% is severe; and more than 96% is extremely severe [23]. In this study, only 32 items were calculated because all the participants were retired and unemployed. Among the reliability indices of the Chinese version of the WHODAS 2.0, Cronbach’s α for internal consistency is between .70 and .99, and the intra-class correlation coefficient is between .80 and .89 [24,25]. Among the validity indices, the content and the concurrent validity have some correlation, and based on exploratory factor analyses (EFAs), 5-7 factors have an explanatory power higher than 55%. The factor loadings of the CFA are all higher than .56 [24]. It has excellent reliability and validity and is consistent with the item response theory (IRT).
Data collection procedure
By communicating with the long-term care facilities, the resident list was obtained after approximately one week, and institutional residents who met the criteria were selected from the list. The study participants generally completed the questionnaire independently. However, if a participant was unable to complete the questionnaire independently because of vision, hearing, reading, or writing limitations, then the researchers provided assistance (e.g., explaining some sentences to make the meaning clearer). Care was taken that the method of assistance provided by the researchers in answering the questions was consistent for all of the participants (e.g., the examples provided for answering the questions were the same), which followed the data collection procedure of Chen et al [10].
Data analysis
In the process of completing the survey, due to refusal of the respondents, negligence of the investigators, or issues with the questionnaire itself, missing data occurred but was resolved by linear interpolation. Descriptive statistics were used to characterize the study population. Absolute and relative frequencies are used for categorical variables, while continuous variables are expressed as the mean and standard deviation (SD).
Test-retest reliability
The test-retest reliability of the Chinese version of the questionnaire was evaluated using residents from 2 LTC facilities. Participants were revisited 3 days later by a different interviewer (i.e., not the first interviewer) and asked to fill out the questionnaire again.
Cohen’s kappa statistics and 95% confidence intervals (weighted and unweighted) were used to evaluate agreement (above chance level) between the administrations’ person rating and the personal items under continuous testing. Kappa ranges from 0 to 1, where 0 indicates no agreement, and 1 indicates complete agreement. Kappa ≥ .8 indicates almost perfect agreement, between .8 and .6 indicates substantial agreement, and between .6 and .4 indicates moderate agreement [26].
Internal consistency reliability
The internal consistency reliability was evaluated based on different tests. Cronbach’s α [27], McDonald’s ω hierarchical, and McDonald’s ω total were used [28]. All of these tests range from 0 to 1. The higher the value is, the higher the reliability. Two types of item analysis were used, that is, (1) the relevance within-item and (2) the correlation between item-to-total, to analyse the homogeneity of the research tool. Finally, the correlation between the subscale and the total scale was analysed. Cronbach’s α coefficient was used to measure the internal consistency reliability between the Chinese version of the PaArticular Scales and its subscales.
Construct validity
According to the principle of the varimax rotation, EFA was used to assess the validity of the Chinese version of the PaArticular Scales. The original English version of the PaArticular Scales has good criterion validity and internal consistency reliability [12]. EFA is used to find the essential structure of multivariate observations. The factors are first selected based on a screening index of the eigenvalue > 1.0 [29]. Based on a screen plot, clinical experience and original factor structure of the Scales, the factors are selected again [30-33]. Finally, the items are selected as long as the minimum variance in each factor is 5%.
Criterion validity
Criterion validity was tested by the convergent construct validity of the tool. To evaluate the convergent structure validity of the new scores, we calculated the Pearson correlation coefficients of the scores obtained from the new questionnaires and the WHOQOL-BREF and WHODAS 2.0 ̵36 items and compared the correlation coefficients. This comparison is meaningful because changes in activity limitations and participation restrictions are accompanied by changes in health-related QOL [34,35]. The point-biserial correlation coefficient was used to calculate the correlations between the Chinese version of the WHOQOL-BREF, the Chinese version of the WHODAS 2.0 ̵36 items and the Chinese version of PaArticular Scales total score to establish concurrent validity. All the data were statistically analysed using the SPSS 22.0 software package (IBM, Armonk, NY, USA).