A psychometric analysis of the Caring Assessment Tool version V

Abstract Aim The aim of this study was to examine the factor structure and construct validity of the Caring Assessment Tool version V (CAT‐V) for patients in Australian hospitals. Design Secondary analysis of CAT‐V surveys from the Australian Nursing Outcomes Collaborative (AUSNOC) data set was used. The CAT was originally developed in the United States of America. Methods The 27‐item CAT‐V was administered to patients prior to discharge from eight wards in three Australian hospitals in 2016. The psychometric properties of the CAT were evaluated using item analysis and exploratory factor analyses. Results Item analysis of surveys from 476 participants showed high levels of perceived caring behaviours and actions. Exploratory factor analysis revealed a two‐factor structure consisting of: Nurse–patient communication; and Feeling cared for. The CAT‐V is a reliable and valid instrument for measuring patients’ perceptions of the attitudes and actions of nurses in Australia.

There is a strong global commitment to improving health care and ensuring that the care provided by nurses is of the highest possible standard (McCance, Wilson, & Kornman, 2016). Recent reports into health system failures have highlighted how fragile the healthcare system can be and made recommendations for nurses to improve patient outcomes through focusing on the culture of caring and development of person-centred approaches to care delivery (Francis, 2013;Garling, 2008). National regulation bodies and indus- There is only limited empirical research that examines links between improved patient outcomes and the presence of caring cultures (Feo & Kitson, 2016). Research that examines this phenomenon is usually related to person-centred care. This is seen in the positive associations between person-centred care and patient outcomes for people who have experienced an acute myocardial infarction (Meterko, Wright, Lin, Lowy, & Cleary, 2010) and haematology-oncology patients (Radwin, Cabral, & Wilkes, 2009). The patient-nurse relationship is less frequently studied, but seen as pivotal in examining the effectiveness of person-centred cultures (Duffy et al., 2014).
There are several approaches used to examine patient-nurse relationships and the caring attitudes and actions of nurses from a pa- A discussion of the theoretical foundations of these instruments is beyond the scope of this paper. The most frequently used instruments for assessing caring behaviours and action of nurses from the patients' perspective in acute care hospitals are the CBI and the CAT (Kuis, Hesselink, & Goossensen, 2014).
The CBI was originally developed by Wolf and colleagues in 1994 and assesses patient and nurse perceptions' of caring using identical self-report surveys with a six-point Likert scale (Wolf, Giardino, Osborne, & Ambrose, 1994). The CBI was revised in 2006 to a 24item scale (CBI-24) for both patient and nurse surveys (Wu, 2006).
The CAT was originally developed by Duffy in 1990 as a 100-item survey to assess patients' perceptions of nurse caring behaviours (Duffy, 1990). The CAT has been iteratively revised (Duffy et al., 2014;Duffy, Hoskins, & Seifert, 2007) and is currently (CAT-V) a unidimensional 27-item survey. The CAT is supported by the Quality Caring Model© (Duffy & Hoskins, 2003) which combines multiple theories from multiple disciplines to help explore the nurse's relationship with the patient and the contribution that nursing attitudes and actions have on patient outcomes (Kim, 2016). The CAT is completed by patients using either a paper-and-pencil approach (Duffy & Brewer, 2011) or via electronic survey (Duffy, Kooken, Wolverton, & Weaver, 2012). Iterative versions of the CAT have had different numbers of items (100, 36 and 27) and different factor structures (between 8 and 1), and each version has reported appropriate reliability and validity (Duffy et al., 2014(Duffy et al., , 2007O'Nan, Jenkins, Morgan, Adams, & Davis, 2014). However, all of the studies using the CAT have been undertaken in different population groups in the USA.
The CAT was chosen as the data collection instrument in this study because of its conceptual link with the Quality Caring Model© and the use of the Quality Caring Model© as the foundational model for evaluating nursing practice in over 40 hospitals in the USA (Duffy et al., 2012). In addition, the CAT had previously been used in an electronic format and this was an important factor in this study (Duffy et al., 2012). Once the decision to use the CAT in the Australian Nursing Outcomes (AUSNOC) data registry had been made, it became appropriate, given the differences between the healthcare systems in the USA and Australia, to test the construct validity of the CAT-V in the Australian healthcare context. Therefore, the purpose of this study was to examine the factor structure and construct validity of the CAT-V using exploratory factor analysis (EFA).

| Aim
The aim of this study was to examine the factor structure, reliability and construct validity of the CAT version V (CAT-V) in the Australian healthcare setting using survey data collected in the AUSNOC data registry.

| Design
The AUSNOC data registry is a multi-site repository of structure, process and outcome measures that explore the quality and safety of nursing practice (Sim, Crookes, Walsh, & Halcomb, 2018). This study used cross-sectional data from patients at the time of discharge in three hospitals who were participating in the feasibility testing of the AUSNOC data registry. The feasibility testing of the AUSNOC data registry is described elsewhere (Sim, Joyce-McCoach, Gordon, & Kobel, 2019). Hospitals were chosen based on convenience and willingness to participate in the AUSNOC project. The data from the CAT-V are focused on measuring patients' perceptions of the caring attitudes and actions of nurses and the nurse-patient relationship.

| Sample
Patients being discharged from three hospitals between March-December 2016 were approached to complete the CAT-V survey.
All hospitals included in this study were private hospitals providing acute care services in the state of New South Wales, Australia.
Patients discharged from four surgical wards, three medical wards and one rehabilitation ward participated in the study.

| Survey instrument
The CAT was originally developed in 1990 (Duffy, 1990) and is based on Watson's Theory of Human Caring (Watson, 2008). Several different versions of the CAT have been tested in hospitalized adults (Duffy & Brewer, 2011;Duffy et al., 2012;O'Nan et al., 2014), emergency department settings (Anosike, 2016), settings outside the USA (Melby, 2005), education settings to assess student relationship competency (CAT-Edu) (Duffy, 2005) and among nurses to assess the caring behaviours of their managers (CAT-Adm) (Wolverton, 2016). The most recent version of the CAT is referred to as CAT-V and was validated by Duffy et al. (2014) for use with hospitalized adults. Table 1 provides an overview of the evolution of the CAT.
The CAT-V consists of 27 items and a single factor structure.
Participants rate how often each item occurred in their healthcare experience on a five-point Likert scale where 1 = never, 2 = rarely, 3 = occasionally, 4 = frequently and 5 = always. The CAT-V includes items related to caring, person-centred care and the nurse-patient relationship (Duffy et al., 2014). All items are directly related to the concept of caring which is defined by Duffy (2013) as "a process that involves the person of the nurse relating with the person of the patient" (p.32). No items in the CAT-V are reverse scored. Summed scores for the overall scale range from 27-135, with higher scores indicating higher ratings of caring and person-centred care (Duffy et al., 2014). In this research, pilot testing was undertaken using the CAT-V with a sample of 40 patients from participating hospitals in February 2016. No changes were made to the wording of any items, and data from the pilot testing were not included in the final sample. Permission to use the CAT-V was obtained under licence from QualiCare on 17/9/2015 (Licence #000915).

This study was approved by the Health and Medical Human
Research Ethics Committee at the University of Wollongong and Illawarra Shoalhaven Local Health District (Approval No HE15/425). All participants were given a participant information sheet by a staff member in the ward and had the opportunity to ask questions about the study. Participants were free to choose whether they wanted to participate and provided informed consent prior to completing the survey. No identifiable data were collected from any participant. All data obtained in the survey were stored securely on password-protected computer systems at the University of Wollongong.

| Data collection
Participants completed the survey within 24 hr prior to discharge from the ward. Surveys were completed either by using an online survey tool in RedCap software (Harris et al., 2009) via an iPad™, or using a paper-based form that was subsequently entered into the online survey tool by a nominated staff member in each ward. The survey consisted of demographic questions and the 27 item CAT-V survey. All paper-based forms were given a unique identifier, and data entry accuracy was verified in a random selection of surveys.

| Data analysis
Prior to undertaking the psychometric analysis, missing value imputation and descriptive analyses were undertaken. The expectation-maximization technique was used to impute the missing values as it is reported to be the best method that produces unbiased estimates (Allison, 2012). Descriptive statistics were then used to summarize the demographic data. A two-step approach involving TA B L E 1 Evolution of the Caring Assessment Tool (CAT) in published studies

Psychometric properties
Original CAT (Duffy, 1990)  both confirmatory factor analysis (CFA) and EFA adopted in previous studies (Bhagwat, Kelly, & Lambert, 2012;Servidio, 2017) was then used to examine the psychometric properties of the CAT-V.
The two-step process is more feasible than a study replication in that the two-step process enables researchers to run CFA and EFA independently on both samples to compare and confirm the results (Schumaker & Lomax, 2004). The data (N = 476) were randomly split into two subsamples of approximately 50% of the cases using the for the CAT-V as an index of internal consistency. Generally, an acceptable alpha is >0.75 (Cronbach, 1951). All analyses were conducted using SPSS for Windows version 22 software and AMOS version 22 software (IBM Corp, 2013).

| Descriptive statistics
The means and standard deviations for each item in the CAT-V (N = 476) are displayed in Table 2. The responses were negatively skewed with most participants responding either "Frequently" or "Always" on most items (mean = 4.52, SD: 0.71). The CAT-V inter-item correlation ranged between 0.44-0.81 demonstrating that most selected items measure related phenomena. The subsamples were similar with no significant differences in the mean scores for all the 27 CAT-V items.

| Exploratory factor analysis (EFA)
The second sample (N = 242) was used to explore the dimensionality of the CAT-V using EFA. Bartlett's test of sphericity revealed statistical significance (χ 2 = 7587.05, df = 351, p < 0.0001) indicating that the data were adequately distributed to allow an evaluation of the potential factor structure. The Kaiser-Meyer-Olkin (KMO) index was 0.961, suggesting that the ratio of the number of participants to CAT-V items was sufficient for factor analysis.
Two factors had eigenvalues greater than one and accounted for 72.44% of the variance of the total factor loading. The inflexion on the scree plot and further analysis suggested a departure from linearity that was consistent with a two-factor solution. Further attempts at different factor structures did not significantly change the number of residuals. Therefore, a two-factor structure was considered best fit for these data. A summary of the EFA for the two subscales of the 27-item CAT-V is presented in Table 4. All items loaded 0.5 or higher on the respective factors. The two-factor model was

| Reliability and criterion-related validity analysis
The Cronbach's alpha (α) reliability coefficient was 0.97 for "Nurse-

| D ISCUSS I ON
The purpose of this study was to evaluate the psychometric properties of the CAT-V in the Australian healthcare setting. The CAT-V was assessed using (a) a pilot study with 40 participants; (b) analysis of data from 476 participants to establish a data set; and (c) a cross-validation study to confirm the factor structure and to ensure reliability of the scale. Using CFA, the hypothesized unidimensional factor of the 27 item CAT-V was rejected. The follow-up EFA suggested a two-factor model. Review of the items that loaded ≥0.50 on factor 1 led to the conceptual label "Nurse-patient communication." Revision of the items that loaded ≥0.60 on factor 2 led to the conceptual label "Feeling cared for."

| Reliability
Internal consistency of the CAT-V was shown because the Cronbach's α confidence coefficient was higher than 0.75 (Cronbach, 1951) across the whole instrument and in each factor. The Cronbach alpha

| Validity
The criterion-related validity of the CAT-V was supported by evidence of a high correlation between the two factors with r = 0.83 (p < 0.001, two-tailed

| Development of the Caring Assessment Tool
Prior research has examined the factor structure of various versions of the CAT using EFA (Duffy et al., 2014(Duffy et al., , 2007. To the best of our knowledge, this is the first study to assess the factor structure of the CAT-V; the first study to assess any version of the CAT in a data registry; and in the Australian context. Previous versions of the CAT have had a range of different subscales. The CAT-IV had eight subscales (mutual problem-solving; attentive reassurance; human respect; encouraging manner; appreciation of unique meanings; healing environment; affiliation needs; and basic human needs) (Duffy et al., 2007). The CAT-V was reported as evaluating a unidimensional construct which was described as an expression of the nurse-patient relationship where the attitudes, skills and TA B L E 4 Rotated loading matrix of the exploratory factor analysis for the two-factor Caring Assessment Tool version V solution (N = 242) behaviours of nurses are assessed in the caring relationships they have with their patients (Duffy et al., 2014). The unidimensional CAT-V described 73% of the variance in the construct and had a high Cronbach's alpha coefficient of 0.97 (Duffy et al., 2014). Our study has produced a two-factor solution with an explained variance of 72.44% and a high Cronbach's alpha (α) coefficient of 0.98.

Item (item number) Communalities
This study builds on prior research and provides a valid instrument to advance the research in the field. This study has evaluated the psychometric properties of the CAT-V and proposes a two-factor solution in the Australian healthcare context. Data obtained for this study were obtained from multiple sites which enables generalizability of the results.

| Study limitations
There are several limitations which must be considered when considering the results of this study. Firstly, a convenience sample from three hospitals in one state in Australia was used. As such, our results may not generalize to other locations. In addition, this study used self-reported data which may limit the validity of findings as participants may have various reasons for over-or underestimating their responses due to social desirability and inaccurate recall. It is also possible that a substantial proportion of patients were not invited to participate in this study at time of discharge due to factors such as unexpected discharge, absence of key staff, busyness of the wards and staff not providing relevant information to potential participants at time of discharge.
Despite these limitations, our findings make meaningful contributions to the body of knowledge and support the ongoing use of the CAT-V to evaluating patients' perceptions of the caring attitudes and actions of nurses at the time of discharge from an acute care hospital. Further evaluation of the CAT-V with different types of patients and various age groups is required.

| CON CLUS ION
The results of this study support the usefulness of the 27-item CAT-V as a brief, reliable and psychometrically sound instrument for measuring patient's perceptions of the caring attitudes and actions of nurses. In evaluating the CAT-V, a two-factor structure was identified which highlights the ability to assess "Nurse-patient communication" and "Feeling cared for." The two-factor, 27-item CAT-V provides important information at unit level about nurse caring that can be used to evaluate and improve the quality of nursing care provided to patients in hospitalized settings.
Assessment of nursing care quality is complex and multi-faceted.
In this study, the CAT-V has been used to evaluate patients' perceptions of the caring attitudes and actions of nurses during hospitalization. The CAT-V provides important information about the quality of the patient-nurse relationship, communication and the perceptions of being cared for. These elements are essential to evaluate the quality and safety of nursing care in a holistic way (Sim et al., 2018). The two subscales of "Nurse-patient communication" and "Feeling cared for" describe meaningful constructs that provide opportunities for hospitals to obtain more precise measures of the quality of nursing care. Additional studies that examine the factor structure of the CAT-V and other measures of quality of nursing care are critically needed.

ACK N OWLED G EM ENTS
The authors gratefully acknowledge participants who completed surveys, and staff in participating hospitals who championed data collection.

CO N FLI C T O F I NTE R E S T
None declared.