This study consisted of two phases: translation of the original DRS into the simplified Chinese language; and a descriptive survey study was conducted to examine psychometric properties of the translated Chinese versions.
Phase one: translation process of the DRS, testing of face and content validity and semantic equivalence
The Brislin’s model of translation (and back-translation) was adopted to translate the original English DRS scales into simplified Chinese [16]. The translation process included the following steps: (1) Forward translation. Two bilingual Doctor of Philosophy (Ph.D.) students who were native Chinese speakers and registered nurses, translated the instrument from the source (English) language into the target (simplified Chinese) language. By discussion and consensus between the two translators, amendments were made before the revised simplified Chinese version was ready for back-translation process; (2) Back-translation. A bilingual nursing teacher with a Ph.D. degree blindly (without access to the original language version) back translated the revised Chinese version into English. Two translators then compared and discussed about any inconsistencies in meanings of the translated items. If disagreements on translated items and terminologies not resolved between two translators, the third and fourth translator would translate and back-translate the disagreed items, respectively, to seek agreements and finalized the items among the four translators.
The DRS-C scales were then tested with 10 family dyads of people with hypertension conveniently recruited from one village clinic to examine their face validity. Their comments on the items’ relevance to the study topic and context were collected through face-to-face interviews.
A panel of five experts, consisting of three academic experts on chronic illness management and two clinical nurses with more than 2 years of clinical experience in chronic illness care in the Chinese communities, was invited to rate the items of the DRS-C scales for their relevance to the construct of family dyadic relationship. A four-point Likert scale (1 = not relevant to 4 = highly relevant) was adopted to the rating of the level of relevance [17]. Content validity index (CVI) was used to evaluate the content validity of DRS-C at the item and scale levels [18]. The values of I-CVI ≥ 0.78 and S-CVI ≥ 0.90 were considered acceptable values for content validity of the DRS-C scales [17].
Furthermore, semantic equivalence of the DRS-C scales with the original scales was established by using the cross-language testing method [19]. Before using an instrument in a new language, cross-cultural validation should be performed. Family values in Chinese and Western cultures (e.g., Americans) are different [20]. These cultural specificities can influence family members’ or carers’ responses to the needs of their patients, self-efficacy of an ill relative and relationship of the family dyad in daily care [20]. The positive results of semantic equivalence computed by a group of bilingual participants can support the cultural relevance of the DRS-C scales to the Chinese population or culture. According to this method, the original and translated versions of the scales were administered to a group of bilingual participants. Semantic equivalence could be achieved when a high correlation by items was observed between the item scores of the two versions. However, the family carers and people with hypertension in the rural communities of Mainland China were unable to read and understand the English version of the DRS.
In this study, a panel of 22 members (six bilingual nursing researchers, 14 Ph.D. students in nursing, and two clinical nurses with master’s degree in nursing) were invited to rate the semantic equivalence of the translated DRS-C scales. Individual items in the DRS scales were rated on a four-point Likert scale in an ascending level of equivalence (“1 = Not Appropriate” to “4 = Most Appropriate”). An item was considered not equivalent if more than 20% of the panel members (i.e., at least four of the 22 panel members) rated the item as < 3 point according to the 4-point Likert scale used [17]. The non-equivalent item(s) would be revised by re-run of the translation and back-translation process described above.
Phase two: psychometric testing of the two DRS-C scales
A cross-sectional descriptive study with correlational design, together with a re-test in 40 randomly selected participants over a two-week interval, was carried out to test the reliability and validity of the two DRS-C scales. A sample of 132 people with hypertension and their family caregivers living in a village in China were recruited in the current study.
Participants and study setting
People with hypertension receiving home visit and care at a village clinic at Liuyang City Hunan Province China were the potential eligible participants of this study. The village clinic is a public clinic, providing primary health care to all people in the village. The researcher obtained permission from the clinic manager to review the patients’ medical records in the village clinic under study and created a list of potential participants after screening.
The patients on the list were approached by the researcher during home visits to confirm the eligibility of participants and collect data. Included patients were: diagnosed as an essential hypertension (SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg), with or without adequate BP control [2]; ≥ 18 years old; and living together with one or more family members. The patients were excluded if they were: with terminal illness; with mental illness (e.g., dementia and schizophrenia); having one or more comorbidities of severe cardiovascular cerebral and respiratory diseases (e.g., stroke, COPD and myocardial infarction); or having needs for assistance with daily activities, like toileting, feeding, dressing, grooming, physical ambulation or bathing.
The researcher guided the patients to nominate one main caregiver from each family to be the participants in this validation. The family carer should be the one who had provided more assistance with the patient’s daily health care and stayed a longer time with the patient in daily life among family members. Additional inclusion criteria of family carers were aged 18 years or above; and with kinship, marital or co-residence relationship with the patient. The carers were excluded if they were diagnosed with mental illness (e.g., schizophrenia and depression) or learning disorder; or taking care of two or more patients in the family at recruitment. Illiterates could be included since the questions of the questionnaire would be read out by the researcher for completion unless they were unable to understand the questionnaire items.
Sample size estimation
For a confirmatory factor analysis, the sample size would be at least 10 subjects per item of the scale(s) to be tested [21]. The DRS-C-caregiver consisted of the highest number of items (i.e., 10 items), and thus about 120 family dyads were required after taking account of a potential non-response and/or incompletion rates of 17% [22]. This sample size could allow the achievement of a study power of 0.80 at 5% significance level, with a moderate correlation between dyadic relationship and self-efficacy in caregiving [23].
Data collection
After obtained the research ethics approval from the Survey and Behavioural Research Ethics Committee of the Chinese University of Hong Kong (Reference No. SBRE-18-677), the researcher confirmed the eligibility of the participants during home visits. Informed written consent was obtained from all individual participants, with assurance of confidentiality and anonymity of data and right to withdrawal, before collecting any data. All data collected were anonymous, kept confidential, and used for research purposes only. Personal information or identities of the participants were not in any way identifiable in the papers. The research data were stored safely in a locked cabinet. The personal data were kept for six years after the study, after which the researcher would destroy it.
Dyadic Relationship Scales (Chinese version, DRS-C)
The quality of dyadic relationship between family dyads was measured by the translated Chinese version of the DRS-C scales (DRS-C-patient and DRS-C-caregiver), which was self-rated on a four-point Likert-type scale ranging from 0 (strongly agree) to 3 (strongly disagree). A higher scale score indicated a worse dyad relationships [15] (Additional file 1).
Hypertension self-care profile (HBP SCP)
The Chinese version of the HBP SCP-self-efficacy scale was used to measure hypertensive patients’ self-efficacy in hypertension management. The self-efficacy scale consisted of 20 items; each rated on a four-point Likert scale ranging from 4 (very confident) to 1 (not confident). A higher score indicates a higher level of perceived self-efficacy in hypertension management [24]. The Chinese version of HBP SCP revealed good psychometric properties in Chinese with hypertension with a Cronbach’s alpha coefficient of 0.93, and a moderate correlation between the HBP SCP and Treatment Adherence Questionnaire for Hypertension (TAQPH), r = 0.45 and 0.65 respectively, all p < 0.001 [25].
Zarit burden interview (ZBI)
ZBI was used to measure the family carers’ caregiving burden [26]. The 22-item ZBI was a self-reporting instrument with a five-point Likert-type scale ranging from 0 (nerve) to 4 (nearly always); and a higher score indicated a greater perceived burden. It was valid and useful in Chinese dementia population [27], with a good intraclass correlation coefficient of 0.89 and split-half correlation coefficient of 0.87, as well as significant correlations with the Geriatric Depression Scale (r = 0.57, p < 0.001) and Hamilton Anxiety Scale (r = 0.44, p = 0.003).
Measurement of blood pressure (BP)
A well-validated electronic upper-arm sphygmomanometer (OMRON HEM-752), the results of the validation have been published [28], was used to measure one’s BP. The method or procedure for measuring BP followed the international and national hypertension management guidelines [1, 2]. BP measurements (on upper arm) were repeated at an interval of 1 to 2 min; and the mean value of the two readings was recorded. If the difference between these two readings of SBP/DBP were more than 5 mmHg, the measurements would be repeated once; and an average value of the three readings would be recorded [2].
Statistical analysis
The IBM SPSS version 25.0 and AMOS (IBM Crop. Armonk, NY) was employed for data analyses. Frequencies and percentages were calculated for categorical variables; and mean and standard deviations were for continuous variables. The missing values, which could not be replenished by reviewing the raw data, were replaced by the means imputation strategy. Since parametric tests (e.g., Pearson’s product-moment correlation and independent-sample t-test) were calculated to determine convergent and discriminant validity of the two DRS-C scales. The total scores of the scales should meet the statistical assumptions of normality, linearity and homoscedasticity [17]. Q-Q plot, skewness and kurtosis statistics were performed to examine the normality of the scores of DRS-C-patient, DRS-C-caregiver and self-efficacy subscale of the HBP SCP and ZBI. To those that were originally skewed, logarithm transformation was performed [29]. In this study, the scores of the DRS-C-patient and self-efficacy of HBP SCP were originally normally distributed, whereas the normality of the DRS-C-caregiver and ZBI scores was met with logarithm transformation. All statistical analyses were two-sided, and any p value of < 0.05 was considered statistically significant.
Validity
Confirmatory factor analysis (CFA)
This study employed the CFA to confirm whether the constructs of DRS-C were similar to the original English versions in American communities in which two-factor solutions (positive dyadic interaction and dyadic strain) of the two DRS scales were found in both exploratory (EFA) and confirmatory factor analysis (CFA) [15]. Therefore, the two-factor solution in the two DRS-C scales was tested to see whether they were similar to the original DRS versions. The Maximum Likelihood Estimation for the CFA was performed to test the model fit. The model-fit indices and criteria [30, 31], including root-mean-square error of approximation (RMSEA), comparative fit index (CFI), goodness-of-fit index (GFI), Tucker and Lewis Index (TLI), and standardized root-mean-square residual (SRMR), were adopted. RMSEA values at 0.05 or lower would indicate a good model fit, and those ranging from 0.05 to 0.08 would represent a moderate fit. The GFI, CFI and TLI values at 0.95 or above could indicate a well-fitting model, and those between 0.90 and 0.95 would indicate an acceptable fit. In addition, a SRMR value at 0.08 or lower could be indicative of a good model-fit.
Convergent validity
Convergent validity was examined by hypothesis-testing approach. The efficacy scale in Hypertension Self-Care Profile (HBP SCP) was employed in the evaluation of convergent validity of the DRS-C-patient. The Shared Care Model posited that family dyadic relationship in daily care could influence patients’ self-care efficacy [32]. Previous studies have also demonstrated the improved dyadic relationship might lead to positive effects on self-efficacy of patients with different heart diseases [33,34,35]. As such, a positive correlation between the mean scores of the DRS-C-patient and self-efficacy scale of the HBP SCP would be tested to provide evidence on the convergent validity of the DRS-C-patient.
The convergent validity of the DRS-C-caregiver was assessed by correlation test between the mean scores of the scale (DRS-C-caregiver) and Zarit Burden Interview (ZBI) scale. The ZBI has been widely used to assess the subjective burden of caregivers. The Pearlin Stress Process Model proposes that stress factors, including dyad strain, caregiver’s role captivity, role strain, and perceived stress, can determine caregiver burden [36]. Therefore, a correlation between the mean scores of the DRS-C-caregiver and the ZBI with similar construct or conceptualisation can provide evidence for the convergent validity of the DRS-C-caregiver.
Known-groups validity
The known-groups validity of the DRS-C was examined using known-groups comparison test. Based on the available empirical evidence [12], patients with a better dyadic relationship would exhibit better levels of self-care and self-efficacy of lifestyle behaviours and medication adherence, which could determine the people with hypertension’ BP values [2]. Therefore, the patients with well-controlled BP in normal level (SBP < 140 mmHg and DBP < 90 mmHg) were hypothesized to have a significantly better dyadic relationship than those without controlled normal BP as tested by independent-sample t-test.
Reliability
Internal consistency
The internal consistencies of DRS-C scales were examined using Cronbach’s alpha statistics and item correlation analysis. Cronbach’s alpha of 0.7 or higher indicates an acceptable reliability [37]. Item correlation analysis of DRS-C scales was performed using corrected item-to-total correlation and Cronbach’s alpha statistics with item deletion(s). A corrected item-to-total correlation coefficient of at least 0.3 is the acceptable level of homogeneity of individual items to the scale/subscale [38]. The inconsistent item could be considered to be deleted if its corrected item-total correlation was less than 0.3; and its deletion would not cause a decrease of 0.1 or more in the Cronbach’s α coefficient of the scale [39].
Stability
The weighted Kappa statistic was computed to examine the test–retest reliability of the DRS-C scales at a two-week retest interval [40]; and the intraclass correlation coefficient (ICC) ≥ 0.7 can indicate a satisfactory stability [41].