Psychometric properties of the Caring Efficacy Scale among personal care attendants working in residential aged care settings

This study assessed the psychometric properties of the Caring Efficacy Scale (CES) among personal care attendants providing care to older residents in residential aged care settings.


| INTRODUCTION
Relationships based on caring are central to achieving healing and positive health outcomes. 1,2Care workers' understanding of caring includes engaging in reciprocal relationships, embracing the essence of caring, engendering instances of caring and embodying caring in practice. 38][9] One of the popular tools to measure caring self-efficacy is the Caring Efficacy Scale (CES).The 30-item self-report CES was developed by Coates 9 in the United States to measure the confidence of nurses and other caregivers in developing caring relationships with their patients, and expressing caring skills, attitudes and behaviours when providing care. 9,10The conceptual and theoretical basis of this scale relies on Watson's transpersonal caring theory and Bandura's selfefficacy theory. 9][17] However, scales validated in certain population groups, settings or countries may not translate fully to other contexts. 18 1.1 | Caring self-efficacy in aged care and

Caring Efficacy Scale
A scoping review by Shrestha et al. 7 found that caring self-efficacy in aged care was described by care workers as their ability to value their work, to understand older residents, to deal appropriately in difficult circumstances when caring for them, to accept feedback about their caring behaviours, and to understand whether their actions have resulted in high quality care and if not, what their weaknesses could be.These notions are reflected in the CES, which should therefore be appropriate to use in aged care settings.
Psychometric properties of this scale have been tested in Australia among registered nurses working in diverse sectors including aged care, 16 however, not separately in residential aged care settings.Separate validation of the scale is required in aged care settings because of differences in the characteristics of aged care and acute care, such as the aims of care, health conditions of patients/clients and their level of dependency on the care workers.Moreover, the CES has not yet undergone psychometric property testing among personal care attendants.Personal care attendants are an integral part of the care team in any care setting.In aged care settings, personal care attendants provide the most assistance to older residents, and therefore the quality of aged care largely depends on their ability to care for them. 19Although several other measures to assess the selfefficacy of aged care workers have been identified in the literature, most were developed specifically to measure self-efficacy in dementia care. 7None of these measures emphasises measuring the self-efficacy of residential aged care workers to provide overall care to older residents.
Hence, we aimed at testing the psychometric properties of the Caring Efficacy Scale among personal care attendants providing assistance to older residents in residential aged care facilities in Australia.

| Study design and setting
A cross-sectional study was conducted in residential aged care settings across Australia in 2020-21.This study was approved by the Human Ethics Committee (HEC) at La Trobe University, Australia (HEC20039).

Policy Impact
The 22-item Caring Efficacy Scale is a valid and reliable tool to measure the perceived caring-self efficacy of direct care workers working in residential aged care settings in Australia.In the context of severe shortages in the aged care workforce, it can be used to help policymakers assess strategies to improve the confidence of care workers.

Practice Impact
This scale can assist aged care providers to identify areas of relative strength and weakness in the caring self-efficacy of their staff and use this information to design targeted interventions within the facility.Aged care providers and researchers can also use this scale as an outcome measure, for example, to examine the impacts of in-service training or other interventions for direct care staff in residential aged care settings.

| Study participants and sample size
The study population comprised personal care attendants working in Australian residential aged care facilities (RACFs).In Australia, personal care attendants are also referred to as 'Personal Care Workers', 'Personal Care Assistants' and 'Assistants in Nursing', depending on which state/jurisdiction they work in.Internationally, popular alternative terms include 'Care Worker', 'Nursing Aide', 'Care Assistant' and 'Care Aide'.Personal care attendants were eligible for the study if they had been providing personal care services to older residents for at least a month following completing Certificate III or IV training in Individual Support (ageing), which is the minimum training required to work in residential aged care in Australia.
Study participants were recruited through convenience sampling.A total of 280 participants were included in the final analysis, which was considered enough for both models based on the ratios of the number of variables to the number of factors (p/f). 20 2.1.2 | Participant recruitment and data collection An online survey link was developed using REDCap to collect the data. T recruit study participants, 25 residential aged care facilities were first randomly selected among the list of residential aged care facilities listed by the Australian Government.21 However, only two facilities accepted the request to forward our survey link to their eligible staff amid pressure associated with increasing COVID cases within facilities and in the general population.So we used social media platforms, emails and text messages to distribute the survey links to potential participants.Considering the significant presence of overseas-born personal care attendants in the Australian aged care workforce, we reached out to communities of different cultural backgrounds to ensure that personal care attendants from diverse cultural backgrounds were represented in this study.Data were collected between August 2020 and March 2021.
The online link consisted of the participant information statement and the structured questionnaire.A participant information statement-which explained the research purpose, procedures, potential benefits and risks, sources of support and researchers' contact information-was included at the beginning of the survey.Participants then had the opportunity to consent and commence the survey.They could also withdraw participation if they chose to.No potential identifying questions were included in the survey to maintain confidentiality.

| Study tool
A structured self-administered questionnaire was used to collect the data.The 30-item CES by Coates 9 was used to assess caring self-efficacy.This scale comprises 15 positively worded statements expressing confidence and 15 negatively worded statements expressing doubts about one's ability to deal with certain situations while caring. 9Response options were along a 6-point Likert-type scale, ranging from −3 (strongly disagree) to +3 (strongly agree).We replaced the term 'clients/patients' with 'older residents' to align with the work of our study population.The author of the CES approved its use for our study before it commenced.
Sociodemographic data, including age, gender, marital status, education, migration history, employment status, working hours and type of facility, were also collected.

| Data analysis
Data were analysed using IBM SPSS version 27.0 and AMOS version 27.0.Descriptive analyses were carried out to identify the characteristics of study participants.Descriptive characteristics of the original scale, such as mean, skewness and kurtosis, were explored before factor analysis.
Likert-type scale values of the 30-items CES were converted from '−3 to +3' to '1 to 6'.Negatively worded items were reverse-scored to maintain consistency among items 22 where 6 represented the highest self-efficacy for each item.The total score ranged from 30 (lowest) to 180 (highest).
Confirmatory factor analysis (CFA) was used to assess how well the prespecified structure of the CES fit in our study population.A one-factor model (Model 1) developed by Coates 9 and a two-factor model (Model 2) proposed by Reid et a.l 16 were assessed.The goodnessof-fit of these two models was examined using the chisquared value, the Comparative Fit Index (CFI), the Tucker Lewis Index (TLI), the Standardised Root Mean Square Residual (SRMR) and the Root Mean Square of Approximation (RMSEA).An acceptable model fit would have resulted in CFI and TLI values close to 0.95 or greater, SRMR close to 0.08 or below and a RMSEA value close to 0.06 or below. 23We planned to conduct exploratory factor analysis with our data if the CFA found both models to have poor fit.
As both models, 1 and 2, had unsatisfactory fit indices in the preliminary analysis, we conducted exploratory factor analysis (EFA) to examine the dimensionality of the CES and underlying structure in our study population.The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (MSA) and Bartlett's test of sphericity were assessed to determine the suitability of the data for factor analysis.The data set was considered suitable for factoring if it had a KMO > 0.5 and p < 0.05 on Bartlett's test of sphericity. 24The anti-image correlation matrix was calculated to identify item-specific MSA values.Any item with an MSA value below the threshold level of 0.50 was excluded from the factor analysis. 25rincipal component analysis was followed by a parallel analysis.The number of factors was determined based on the comparison of eigenvalues from the principal component analysis with randomly generated eigenvalues created from the parallel analysis. 26Finally, principal axis factoring with promax rotation was carried out to determine which items belonged to each subscale.An a priori decision was made to consider only items with loadings ≥0.40 in interpreting subscales. 26nternal consistency of the final scale and subscales identified was examined using item-total correlations and Cronbach's alpha coefficients.The minimum thresholds of 0.3 for item-total correlations and 0.7 for Cronbach's alpha coefficient were considered satisfactory. 25

| Sample characteristics
Of the 280 personal care attendants included in our study, 95% were women.Participants were aged between 18 and 67 years, with a mean age of 40.0 (SD 12.2) years.Four in 10 participants were born overseas.Nearly 24% of participants primarily used language other than English to speak at home.Forty-one per cent of participants had more than the minimum required qualification.While 78% of participants had a permanent job, three-quarters were working as part-time employees.The mean number of years of working as a personal care attendant was 7.7 (SD 7.8; Table 1).

Scale before factor analysis
Descriptive statistics for the original scale among study participants are shown in Table 2.The total score of the scale was normally distributed based on values for skewness and kurtosis, which were between +1 and −1.

| Confirmatory factor analysis
According to the fit indices, Models 1 and 2 had poor fit in our study population (Table 3).The corresponding fit indices for Model 2 were better than for Model 1, but still unsatisfactory.This outcome of the confirmatory factor analyses prompted exploratory factor analysis of the data.

| Exploratory factor analysis
The value of KMO MSA was greater than 0.5, and Bartlett's test of sphericity was significant (p < 0.001), indicating that the scale was appropriate for factor analysis.All items on the scale had high MSA values.Parallel analysis, following principal component analysis, showed that first two eigenvalues based on the

| Internal reliability analysis
Internal reliability analysis of the scale and subscales following factor analysis and removal of items with low factor loadings indicated that the modified versions were internally consistent.Item-total correlations amongst items in the scale and subscales following EFA were found to be satisfactory, with values equal to or greater than 0.3 (see Table 4).
Similarly, Cronbach's alpha coefficients of the modified scale and subscales were in acceptable level-0.85 for the modified scale, with 0.83 for Confidence to Care and 0.79 for Doubts and Concerns.

| DISCUSSION
The unsatisfactory goodness-of-fit of the original Caring Efficacy Scale prompted exploratory factor analysis of the scale in our study population.Similar to EFA results from previous studies carried out among registered nurses, 16,17 this study suggested a two-factor structure of the scale in the overall sample of personal care attendants.Factor 1-Confidence to Care-comprised items depicting perceived confidence in personal care attendants' ability to relate to and care for older residents.Factor 2-Doubts and Concerns-comprised items that depicted doubts and concerns about this ability.While 22 of the original 30 items were retained, eight items were excluded in this study because factor loadings of these items were below the threshold value of 0.40.
The number of items with factor loadings less than 0.40 in our study is comparable to findings of previous studies. 16,17However, the items retained for use with registered nurses were not necessarily the same as those retained among personal care attendants in the current study.It is not unusual to find a scale validated in one group, which does not fully translate to another, 18 and therefore, revalidation is necessary each time a measure is used with a new population.Moreover, composition of study participants may also have influenced which items were retained.Similar to the national statistics of the Australian aged care workforce, 36% of study participants were from culturally and linguistically diverse (CALD) backgrounds. 27Some items in the original scale, which was tested on native English-speaking care workers in the United States, may make less sense in our study population given the extent to which the backgrounds of personal care attendants differed from those of the older residents.For instance, based on their life experience, personal care attendants from non-English-speaking backgrounds might anticipate that residents may not like them, and therefore item 19-I can usually get residents to like me-may be less appropriate for this group.
To summarise, the Caring Efficacy Scale was shortened following the exploratory factor analysis to assess the caring self-efficacy of personal care attendants working in residential aged care settings.Once items with low factor loadings were omitted, the Caring Efficacy Scale and its two subscales were sufficiently reliable to measure caring self-efficacy.

| Study limitations
This study has some limitations.Ideally, the results of an EFA should be tested using CFA in another sample.We did not examine the construct validity of the revised factor structure obtained from EFA because the sample size was not large enough to split into two groups for separate EFA and CFA.Therefore, further studies are needed to confirm the construct validity of this measure among personal care attendants working in RACFs.Additional validity testing, such as convergent and discriminant validity, is also recommended in future studies.For example, caring selfefficacy should be more strongly correlated with a measure of relationships with residents than with a measure of relationships with co-workers or management.
This study may have been affected by social desirability bias. 28This bias was minimised by assuring respondents that their anonymity would be maintained.The mode of data collection, an online survey, also reduced the risk of underreporting socially undesirable attitudes. 29However, an online survey increases the risk of sample selection bias. 30Participants who had no internet access or limited acquaintance with using the internet would not have participated in this study.
In addition, responses to the CES may be subject to acquiescence bias due to participants' varying ability to comprehend questions on the scale. 31Although an intermediate level of English is required to work as personal care attendant in Australia, it is still possible that personal care attendants from countries where English is not the first language might have had difficulty responding to the survey.
Despite the limitations mentioned above, study findings suggested that the CES is a reliable tool to assess the caring self-efficacy of personal care attendants working in residential aged care.It can be used as an outcome measure, for example, to examine the impacts of in-service training or other interventions used with direct care staff in residential aged care settings.

| CONCLUSIONS
The exploratory factor analysis of this scale identified a two-factor structure.Once items with low factor loadings were removed, the scale and subscales had a high degree of internal consistency.Overall, this study suggested that the 22-item, two-factor Caring Efficacy Scale can be used to assess the caring self-efficacy of personal care attendants working in residential aged care settings, but some caution is required.The subscales-Confidence to care and Doubts and concerns-can also be used individually to assess personal care attendants' confidence or doubts/ concerns in caring for older residents.

ACKNO WLE DGE MENTS
This work is supported by a La Trobe University Postgraduate Research Scholarship (LTUPRS) and La Trobe University Full Fee Research Scholarship (LTUFFRS).We wish to acknowledge Donna Kennett and Caroline Egan from HelloCare, including residential aged care providers, managers and registered nurses for their kind support in recruiting study participants.We would also like to acknowledge all the study participants for their time and invaluable contribution to this study.We thank Dr Xia Li, La Trobe University Statistics Consultant, for her expertise and kind assistance in the statistical analysis for the study.Open access publishing facilitated by La Trobe University, as part of the Wiley -La Trobe University agreement via the Council of Australian University Librarians.
Characteristics of study population.
T A B L E 1 ing with me, I am able to adjust to his/her level; and item 30-I do not use creative or unusual ways to express caring to my residents.
T A B L E 2T A B L E 3 Fit indices for confirmatory factor models.Abbreviations: CFI, Comparative Fit Index; RMSEA, Root Mean Square of Approximation; SRMR, Standardised Root Mean Square Residual; TLI, Tucker Lewis Index.T A B L E 4Note: