Setting and sample
The study population was comprised of family members of patients with dementia who visited the specialist memory and geriatric psychiatric clinics of four hospitals in Beijing of China from October 2017 to March 2018. Family members were considered eligible for this study if their relatives with dementia were taken care of at home for at least six months. According to the equations of the minimum sample size (\({\text{N}}_{{\epsilon } }=\frac{{{\delta }}_{1-{\beta }}}{ {{\epsilon }}^{2}\text{d}\text{f}}+1\)), the minimum sample size of our study was 108, when the df = 218, ε = 0.05, the power (1-β) = 0.80, and the \({{\delta }}_{1-{\beta }}=58.182\)(δ is the non-centrality parameter ) [16]. We initially recruited 250 patients who were diagnosed with dementia at least six months prior. Through the physician's referral, we invited one to two family members of each patient, who were the primary caregivers of each patient to fulfil our scales on behalf of the family. In total, 250 family members were invited to participate in our survey. Of those, 25 refused to participate in this study because they had no interest or no time to answer the questions. Thus, a total of 225 family members agreed to participate in this study. For this analysis, we excluded 10 participants who had data missing for more than 10% of the survey. Therefore, the final sample included 215 participants (Fig. 1). The Ethical Review Committee of School of Nursing, Peking Union Medical College approved the study protocol in April 15 (Approval no. [2015] 03) and the study was conducted in accordance with the principles of the Declaration of Helsinki. All of the participants were informed about the aim and significance of this study, their right to quit at any time, and how to fill in the scales. All participants provided written informed consent.
Measurements
Characteristics of patients and families
Information on the following variates was collected through the questionnaire, including demographic and disease-related characteristics of patients, and demographic characteristics of family members as well as the whole family. The patient information included age, gender, subtypes of dementia (Alzheimer's disease, vascular dementia, and others or unknown), dementia severity based on Clinical Dementia Rating evaluated by the physician (moderate, severe), and time or duration since diagnosis. The information on family members and the whole family included age, educational level (junior middle school and below, high school, college degree, and graduate degree), whether they have religious faith (yes or no), and family income per month per person.
Family Crisis Oriented Personal Evaluation Scales (F-COPES)
The F-COPES was the most commonly used instrument to measure family coping via self-report. It was comprised of 30 item and five subscales. The five subscales included a subscale of acquiring social support (the family’s behaviours of actively engaging in acquiring support from extended family, friends, and neighbours), subscale of mobilizing the family to acquire and accept help (the family’s behaviours of seeking out and using community resources to cope with problems), subscale of seeking spiritual support (the family’s behaviours or attitudes of acquiring spiritual support), subscale of reframing (the family’s behaviours or attitudes of redefining stresses/situations to make them more manageable), and subscale of passive appraisal (passive/inactive coping behaviours or attitudes in managing problems). Respondents were asked to choose the frequency of using different coping methods on a scale of 1 (almost never) to 5 (almost always). Items of Nos. 12, 17, 26, and 28 should be reverse scored when obtaining the total score of the F-COPES by summing the responses of all items. This is due to the fact that the coping behaviours or attitudes described in the items of Nos. 12, 17, 26, and 28 were considered negative. Higher total scores of the F-COPES represent better family coping with stressful situations. The F-COPES has strong internal consistency with Cronbach’s alpha coefficient ranging from 0.61 to 0.87 [17, 18], and good factorial and concurrent validities [19, 20].
Multidimensional Scale of Perceived Social Support (MSPSS)
The MSPSS was selected to evaluate the convergent validity of F-COPES in this study. The MSPSS measured the perceived social support from three informal sources: family, friends, and significant other. Participants were rated on a seven-point Likert response format (1 = “very strongly disagree” to 7 = “very strongly agree”). The total score is summed by the scores of items, with higher scores indicative of greater perceived social support. Zimet et al. [21] tested the MSPSS and reported high internal consistency of 0.88. The test-retest reliability was 0.85 over a 2- to 3-months period after completing the questionnaire [21]. The MSPSS had a Cronbach’s alpha coefficient of 0.90 in the current study.
Procedure
Translation and adaptation of F-COPES
We conducted a rigorous ‘‘forward-backward’’ translation following the guidelines to develop the initial Chinese version of F-COPES [22]. Then, during the adaptation process, we invited six clinical nurse supervisors, nursing researchers, or psychological researchers who were professors or associate professors and familiar with both dementia caregiving and psychometrics to form an expert panel to judge whether each item of the initial Chinese version of F-COPES should be deleted or modified and whether we should add other items.
In the adaptation process, we deleted the items of Nos. 9, 29, and 30 from the Chinese version of the F-COPES. Item 9 (“Seeking information and advice from the family doctor”) was deleted because the system of family doctors, which is commonly understood internationally, does not exist in China [23]. Item 29 (“Sharing problems with neighbours”) was deleted because families of patients with dementia are reluctant to share problems with neighbours when facing the discrimination and prejudice against patients with dementia and their families in China [24, 25]. Item 30 (“Having faith in God”) was deleted, because only 10.4% of Chinese participants identified themselves as religious [26]. Only 6.5% of family members in the current study had religious faith, although they participate in religious practices. We modified the items (items of Nos. 14, 23, 27) that described religious practices as coping methods in the subscale of seeking spiritual support. We broadened the Christian terms in these items to comprehensively suit other religions, which was similar to the Turkish version of the F-COPES [12] since the Chinese participants were engaged in some form of religious practice across the various religions [26]. Finally, all experts agreed there was no need to add any new items.
Statistical analysis
After translation and adaptation of the F-COPES, we did a content validity analysis, item homogeneity analysis, factor structure analysis, convergent validity analysis, and internal consistency reliability of the F-COPES. We used AMOS 20.0 software for factor structure analysis and SPSS 17.0 software for other analyses.
Content validity analysis was explored by the item content validity index (I-CVI) and the scale content validity (S-CVI). Six experts were invited to quantify the relevance of items to family coping by using a rating scale (not relevant or somewhat relevant = 0, quite relevant, and very relevant = 1). The I-CVI is the average of all experts rating the results of each item, and S-CVI is the average of I-CVIs. The I-CVI is recommended to be no lower than 0.78 and an S-CVI of 0.80 or higher is acceptable [27, 28]. Item homogeneity analysis was tested using a corrected item-subscale correlation, item-total correlation, and Cronbach’s alpha coefficient of the subscale after deleting each item. The corresponding item is acceptable if the corrected item-subscale correlation is higher than 0.20, and the value of Cronbach’s alpha coefficient of the scale after deleting each item is lower than before the deletion [29].
The factor structure of the F-COPES was analysed by confirmatory factor analysis. The criteria used to determine whether the models of factor structure fit the data were 3.00 or lower on the chi-square of model fit/degree of freedom (χ2/df) [30], 0.90 or greater on the comparative goodness-of-fit index (CFI), goodness-of-fit index (GFI), and the Tucker-Lewis Index (TLI); 0.50 or greater on parsimony goodness-of-fit index (PGFI) [31-34]; and 0.08 or lower on a root-mean-square error of approximation (RMSEA) [35]. In the confirmatory factor analysis in the current study, the original model established based on the factor structure of the English version of the F-COPES was tested first. The competing model was then tested if the original model did not meet the above-mentioned criteria [36]. The data fitness of the original model and competing models of F-COPES were compared. If the competing model—which meets the above fitness criteria and the value of CFI of this model—is at least 0.01 higher than the original model, the competing model is then considered significantly better than the original one [37]. Convergent validity of the F-COPES was tested by estimating its correlation with social supports as measured by the MSPSS, since social support has been shown be to be correlated to caregiver coping [38]. The correlation coefficient for convergent validity between 0.10 and 0.29 was considered weak, 0.30 to 0.49 was considered moderate, and 0.50 to 1.0 was considered strong [39]. The internal consistency reliability of the scale and subscales were evaluated by Cronbach’s alpha coefficients. A Cronbach’s alpha coefficient between 0.80 and 0.90 was considered perfect, and 0.65 was the lowest acceptable cut point [40].