Community Orientation Scale among Community Health Nurses in Fiji: Scale development and psychometric evaluation

Abstract Aim To develop and evaluate the reliability and validity of the COSCHN, a scale that aims to measure community orientation among community health nurses in Fiji. Design Descriptive cross‐sectional design. Methods A self‐administered questionnaire that included the 51 items in the preliminary COSCHN was distributed to community health nurses in Fiji from April–July 2016. Results Exploratory factor analysis of 226 responses (77.4% response rate) to the COSCHN revealed 30 items loading on four factors: Community Initiative Promotion, Consensus Building for Community Needs and Activity Goal, Commitment towards Work and Community Members and Mutually Trusting Relationships with Community Members towards Empowerment. Confirmatory factor analysis with high‐order factor modelling revealed a reasonable fit to the data. Cronbach's α values for the COSCHN and the four factors ranged from 0.78–0.94. Weak correlations were noted for concurrent validity, while known‐groups validity and time stability were generally satisfactory.

CHNs, who are required to make sound decisions about health (Institute of Medicine Committee for the Study of the Future of Public, 1988). Community assessment is a core nursing process that requires a logical and systematic approach to identify community health needs, strengths and resources for community health services (Shuster, 2010). Various models for management of community health needs have been developed and used to enable effective community assessment and its use for community health services (Green & Kreuter, 2005;Kreuter, 1992). One of the models aims specifically to achieve community orientation (CO), defined as the organization-wide generation, dissemination and response to community intelligence to address present and future community health needs (Proenca, 1998). The concept is derived from market orientation, which assumes that the best way to understand and satisfy consumer needs is to develop a closeness to the market (Kohli & Jaworski, 1990). Proenca (1998) noted that CO must be carried out in collaboration with other community systems and organizations, since community health needs are both affected by this and beyond the control of the healthcare system. Successful collaboration requires comprehensive competency that should fit a nation's healthcare system and culture and the responsibilities of CHNs to deliver these services (Lin, Hsu, Li, Mathers, & Huang, 2010). Although a measurement tool to assess CO was developed and widely in use for health service organizations in the United States (Ginn, Shen, & Moseley, 2009), no research has been published about CO among CHNs in Fiji. Several factors affect CHN task performance. With the recent health policy reform, clinical services have been decentralized to health centres (MOHMS, 2015). This shift has required CHNs to spend more time providing outpatient care at their facilities. The working environment can impede their work in that CHNs usually work alone in their assigned communities, with few opportunities to mimic and receive timely advice from senior CHNs, except during special events such as outreach team visits. Senior CHNs have developed their own competency through meaningful yet challenging experiences that have sharpened the effective management of community health activities. However, these experiences are intangible and thus difficult to convey verbally to junior CHNs, especially given the working conditions of CHNs in Fiji. Therefore, identifying detailed characteristics of CO for CHNs in Fiji could help them to gauge how much attention must be paid to understand community health needs and manage community activities.

| BACKG ROU N D
To identify detailed characteristics and develop a framework for CO among CHNs in Fiji, we conducted semi-structured individual interviews with 20 participants, nine CHN supervisors/managers/lecturers at a nursing school, three novice CHNs, three policymakers in the MOHMS and three community members (Tanabe, Yanagisawa, Waqa-Ledua, & Tukana, 2019). Data were analysed using descriptive qualitative methods. Each occupation comprised a unit for analysis.
As most codes were extracted from expert CHNs, this group was set as the main group for analysis. None of the codes from expert CHNs contradicted those of other groups. Extracted sentences were merged into 57 final codes, which were then consolidated to form 12 subcategories. Finally, these were combined into three categories: Trusting Relationships, Commitment and Activity Management.
Trusting Relationships and Commitment were inter-related and serve as a foundation to promote Activity Management. Reflection and a sense of self-accomplishment during Activity Management further strengthened Commitment and Trusting Relationships. Therefore, each of the three main characteristics is linked to others and circulated as experienced.
Competency is an underlying characteristic of an individual that is causally related to criterion-referenced effective and/or superior performance in a job (Spencer & Spencer, 1993). When evaluating competency, it is not the performance and action that are important, but rather why and how one performs (Streiner, Norman, & Cairney, 2015). Collecting reasons and thoughts underlying the actions and identifying a thought pattern can maximize the generalizability to various outcomes of actions rather than simply listing superficial actions (McClelland, 1973). Thus, development of a self-assessment scale specific to CHNs in Fiji to measure thought patterns such as one's attitude, value and self-image towards CO could allow CHNs to understand these features within themselves as they relate to community health activities, allowing them to be strengthened and educated. To this end, the purpose of the present study was to develop the Community Orientation Scale among Community Health Nurses (COCHN) and assess its reliability and validity.
The present study was part of a larger research project, entitled "Community Orientation among Community Health Nurses in Fiji: Scale Development and Influencing Factors." Data for our analysis were extracted from a cross-sectional survey conducted from April-July 2016 in Fiji.

| DE S I G N
This study was a descriptive cross-sectional design.

| Participants
The survey was administered to all CHNs in Fiji. During the study period, the total number of CHNs in Fiji was 269. Although the whole CHN population was included in this study, we assessed the number of participants since the factor pattern that emerges from a large-sample factor analysis will be more stable than that from a smaller sample (DeVellis, 2016). When determining sample size for factor analysis, one must consider the research questions, type of model, the number of factors or items and more . In general, however, as a checklist for assessing methodological quality of studies of measurement properties has proposed that a ratio of at least 5 participants per item and more than 100 participants would represent an adequate sample size , we followed these guidelines. As described later, the preliminary COSCHN consisted of 51 items; therefore, the necessary sample size was set to 255 participants. Collection rate was not anticipated to be high, due to geographical challenges and limited means of transportation. Therefore, in addition to all 269 active and practising CHNs during the study period, we also administered the survey to 31 former CHNs who had transferred from CHN work to other positions within 1 year prior to the survey administration (hereafter, "former CHNs"). Contact with former CHNs was attempted after obtaining information from practising CHNs when the questionnaires were distributed. In total, the target number of research participants was set to 300 CHNs (269 practising CHNs and 31 former CHNs).
In this study, CHNs in Fiji comprised zone nurses and district nurses who provide both basic clinical treatment and health-promoting services to community members and settings in their respective environments in the assigned area. Zone nurses work at health centres with other health workers, the latter of which includes CHN colleagues and allied health personnel such as physicians or nurse practitioners, nutritionists and health inspectors. District nurses work alone in nursing stations in areas without easy access to health centres.

| Questionnaire components
The self-administered questionnaire comprised four parts: 1. 51 items of the preliminary COSCHN 2. Sociodemographic background: age, position (zone/district nurse) and educational achievements 3. Items to assess known-groups validity: self-reported supervisor's competency assessment measured on a 10-point scale where 1 is "very bad" and 10 is "excellent." 4. Items to assess concurrent validity: six items about normative commitment taken from the academic version of the Three

Component Model Employee Commitment Survey Normative
Commitment Scale (TCM-NCS, Meyer & Allen, 2004). This measures the sense of obligation to the organization (such that employees with strong normative commitment remain because they feel they ought to do so) along with a 7-point Likert scale for which 1 is "strongly disagree" and 7 is "strongly agree." For one item, this is reversed, such that 1 is "strongly agree" and 7 is "strongly disagree." Evidence for reliability and validity of the scale accumulated through years of research (Allen & Meyer, 1996). The normative commitment scale was chosen for concurrent validity because normative commitment was one of the main components of the CO framework developed in the previous study. Permission to use the TCM-NCS was obtained from the authors' organization.

| Preliminary Community Orientation Scale among Community Health Nurses in Fiji
The preliminary COSCHN was created based on the conceptualized framework in the previous qualitative descriptive study (Tanabe et al., 2019), as described above. Final codes were used as items for the preliminary COSCHN. Six items were excluded from the 57 final codes, as they contained similar content to other items. In the end, 51 items were created. Four items were reversed meaning of the statements to reduce response bias. The items were assessed along a 7-point Likert scale to reflect the degree of practice, whereby 1 is "not at all" and 7 is "extremely." While a higher number reflected a higher level of practice, this was reversed in the case of four reversed items for which a higher number reflected a lower level of practice.
To verify content validity, an expert review and a pilot test were conducted. The questionnaire was checked by eight experts: four medical and nursing managers of the MOHMS, one senior nursing lecturer and three public health researchers. A pilot test was also conducted to examine about its clarity of explanation, response time and questions with a low response rate. The questionnaire was distributed to 29 CHNs selected by convenience sampling; 28 forms were returned. Some text of the 51 items was modified to reflect the comments from the experts and the pilot test participants. The changes in items were made after discussion among co-researchers.

| Data collection
Questionnaires with return envelopes were distributed to each individual CHN by either the chief researcher or nursing managers. The questionnaire forms were collected 1 week after distribution from CHNs who did not complete them immediately. The chief researcher also visited former CHNs when introduced to them by practising CHNs. Orientation and instruction was given at the time of distribution. Forms were distributed and collected by nursing managers when the chief researchers were not available. Nursing managers were instructed that they were to neither insist on submission nor confirm responses. Participants of the survey received F$6 (approximately US$3.0) phone vouchers. Nursing managers who cooperated with distribution and collection received F$30 (approximately US$15.0) phone vouchers.
The retest survey was conducted 1 month after CHNs returned the questionnaire forms. The target population comprised CHNs in the central division who submitted a consent form for the retest, which followed the same procedure as the survey.

| Data analysis
Of the collected questionnaires, those with no responses or multiple answers in the preliminary COSCHN were eliminated. All items with reversed scores (four in the preliminary COSCHN and one in the TCM-NCS) were analysed after rescoring to reflect the reversed points.
Initially, sociodemographic background was analysed.
Subsequently, item analysis was conducted. Skew, ceiling effect and floor effect were analysed to examine score distribution and normality.
Skew of each item was determined using the exclusion criteria of either >1.0 or <−1.0. Ceiling effect was calculated by the sum of the mean and standard deviation (SD), and any items exceeding the maximum score (7) were excluded. Floor effect was calculated by subtracting SD from the mean, and any items scoring below the minimum score (1) were excluded. The item-total correlation analysis was conducted to eliminate items that did not correlate with a given component by determining Pearson's correlation coefficients (Pearson's r) between an individual item and the total score without that item; any items for which Pearson's r < .2 were eliminated. Good-poor analysis was conducted to verify the discriminative power of each item by Welch's t test and items that showed no statistically significant differences (p < .05) were eliminated. Correlations within items were assessed to eliminate items that were too similar by calculating Pearson's r (one of the two items was eliminated if Pearson's r was >.7 between them).
Subsequently, exploratory factor analysis (EFA) was conducted to assess construct validity with the exclusion point for factor loading set to 0.4. Items with factor loading that was just slightly lower than the exclusion point were reconsidered and carefully examined.
Extracted factors were named. Next, confirmatory factor analysis (CFA) was conducted to determine the factor structure. The fitness of models was assessed using the following: p value of CMIN chisquare (χ 2 ), goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI) and root mean square error of approximation (RMSEA).
Confidence coefficients, Kaiser-Meyer-Olkin (KMO) and Cronbach's α were calculated to measure suitability of items. Known-groups validity was examined by Welch's t test. Concurrent validity was examined by Pearson's r between TCM-NCS and a factor indicating commitment.
For reliability assessment, internal consistency was assessed using Cronbach's α for COSCHN and factors. Test-retest reliability was computed as a measure of stability reliability. Analyses were performed by SPSS 24.0 for Windows and Amos 24.0.

| Item analysis
In total, 14 items were discarded from the preliminary COSCHN: seven due to skew, eight due to the ceiling effect and four due to I-T correlation analysis, yielding a final total of 37 items included in the preliminary COSCHN.

| Exploratory factor analysis
EFA was conducted using the maximum likelihood technique and promax rotation. The eigenvalue rule (DeVellis, 2016) and scree plot of an initial EFA revealed four factors. As a result of the first EFA, seven items were discarded, and as a result of the second factor analysis with 30 items, one item was discarded. Factor loading for the third factor analysis with 29 items exceeded 0.4 for all items. All discarded items were examined before finalizing the item list. In so doing, Item 25 ("I pay attention to vulnerable people and minority groups when collecting and analysing information (hereafter, Vulnerable Sensitivity)") was reconsidered because of its significance and because the factor loading was barely below 0.4 (0.399) in the first analysis. Therefore, the fourth factor analy-   indicators is presented in Table 3. Model 2 showed the better GFI (>0.8) and AGFI, while Model 1 showed the better RMR and RMSEA; however, the differences were sufficiently small. Therefore, this comparison revealed no remarkable differences between the two models.

| Known-group validity assessment
To test the known-groups validity, Welch's t test was conducted.
Self-reported supervisor assessments were divided by a median value (8). Significantly higher scores for COSCHN (p = .010) were observed in the group with higher scores for self-reported supervisor competency assessment (Table 5).

| Concurrent validity assessment
To test concurrent validity, Pearson's r values were calculated for TCM-NCS and Commitment. Of the 226 valid respondents, data from 10 were discarded due to missing values in the TCM-NCS; data from the remaining 219 were used in the analysis. The correlation coefficient was r = .263 (p < .001) between TCM-NCS and Commitment. Given the possibility that the reversed item in the TCM-NCS might be misinterpreted, additional calculations were made using five items of the TCM-NCS after discarding the one reversed item. This yielded a slightly higher correlation coefficient of r = 0.284 (p < .001).
To analyse time stability, a retest was conducted among 74 participants who submitted retest consent forms. Of the 74, 54 returned questionnaires (73.0% response rate), 46 of which were valid forms (62.2% valid response rate). Mean COSCHN for the retest was significantly higher (p < .001) than that of the original test, with a Pearson's r of 0.519.

| D ISCUSS I ON
In this study, we developed the COSCHN and examined its reliability and validity. This scale measures CO among CHNs in Fiji, with the aim to promote effective management of community health activities.
The EFA identified a four-factor model from the COSCHN ( Figure 1), even though the conceptual framework of CO among CHNs in Fiji in the previous study consists of three main characteristics. This discrepancy is primarily due to the division of Activity Management in the previous study into two factors: (a) information collection and planning and (b) preparation, implementation and evaluation. This division is also in line with the two distinctive capabilities needed to achieve a higher degree of CO as proposed by Proenca (1998): community sensing and community linking. Community sensing is the ability to learn more about community through the structured, ongoing process of tracking community events and trends and is necessary in the information collection process. Community linking indicates the ability to create and manage close relationships with communities. This is particularly important when preparing and implementing activities together with community members and collect honest feedback from them.
Factor validity was thereby confirmed. Next, COSCHN was assessed by CFA with high-order modelling. CMIN chi-squared test of the path diagram of COSCHN yielded a p < .001, revealing that the data were not suitably fitted to the model. However, multiple correlation coefficients of the four factors ranged from 0.81-0.54 in Model 1, which was Our analysis concluded that Vulnerable Sensitivity was included in the COSCHN. Several studies pointed out persistent issues with gender inequality, poverty and ethnicity in Fiji (Chattier, 2005;Wiseman et al., 2017). The MOHMS highlights equity as a main value (MOHMS, 2015). The WHO recommends health professionals in Fiji to increase awareness of issues pertaining to gender norms and inequality in disease perpetuation (MOHMS, 2015;WPRO, 2011). Therefore, CHNs are unable to improve health situations without exercising principles of fairness to all population; it is essential that CHNs examine health situations and environments to identify disparities among community members.
Known-groups validity was tested by self-reporting supervisor competency assessments. As CHNs with higher supervisor assessment had significantly higher COSCHN scores than those with lower supervisor assessment, known-groups validity was confirmed to be reasonable. Internal consistency was evaluated using Cronbach's α values.
Our analysis revealed α values of 0.94 for COSCHN and that ranging between 0.79-0.89 for the four factors. A Cronbach's α exceeding 0.7 indicates high reliability (Oshio, 2004). Therefore, the internal reliability was confirmed.
Time stability was tested among 46 patients. The necessary sample size was found to be 29 for a correlation coefficient of 0.5, an α level of 0.05 (two-tailed) and a β level of 0.2 (Browner, Newman, & Hulley, 2007 Finally, we would recommend that the MOHMS takes steps to improve the accessibility of epidemiological information that could promote CO among CHNs in Fiji. As the MOHMS has recently introduced an online information management system, CHNs should be able to access this database for their division and nationwide. Easy access to the system will allow CHNs to obtain the latest statistics for their areas to analyse trends and compare them with those in other areas, allowing them to assess better the community health needs.

| LI M ITATI O N S
Our analysis included data from 21 former CHNs as valid respondents who had transferred within the year prior to the survey administration. Their current tasks could have influenced their responses, yielding responses that may have differed from those they may have produced as CHNs. However, the questionnaires were distributed to all CHNs except one CHN who were stationed in the outer island, with most (90.71%) comprising active CHNs (Table 1). Therefore, our analysis was fairly reflective of the views of CHNs. Second, all reversed items in the preliminary COSCHN were discarded due to the negative I-T correlation. The wording used for reversed items may not be suitable for CHNs in Fiji, as it may cause misinterpretation. We surmise that the wording must be adapted to be more culturally appropriate, so that CHNs are better able to comprehend fully the content and think deeply about the meaning and context. Third, concurrent validity was not supported. This could be because TCM-NCS was to measure different characteristics of commitment from those measured by the COSCHN, as discussed above. Forth, time stability was not fully supported due to seminars and workshops for CHNs during the retest period.
Further investigation and modification are needed to refine the COSCHN to increase its potential for generalizability. Also, the COSCHN is applicable only to zone and district nurses in Fiji.
Future studies should examine the application of the COSCHN to other nursing positions in Fiji, and those in other Pacific island countries.

| CON CLUS IONS
The present study developed the COSCHN, scale to measure thought patterns such as one's attitude, value and self-image towards management of community health activities, and examined its reliability and validity. Construct validity, known-groups validity and internal consistency were supported. This scale can be used for self-assessment and discussion between CHNs and their supervisors. Further research is necessary to refine the scale and increase the potential for generalization.

ACK N OWLED G EM ENTS
We express our sincere thanks to CHNs, CHN supervisors and managers who supported this study. We also express our gratitude to MOHMS Fiji and Fiji National University School of Nursing for their kind support. This study was performed as part of a doctoral dissertation for the School of Nursing and Health, Aichi Prefectural University. The survey was supported by grant-in-aid YAMAJI FUMIKO NURSING RESEARCH FUND (Heisei28-3 Shiba-079).
Modification and publication was supported by JSPS KAKENHI Grant Number JP17H07223.

CO N FLI C T O F I NTE R E S T
The authors declare no financial or other conflicts of interest.

AUTH O R CO NTR I B UTI O N S
ST and SY were responsible for the study conception, design and drafting of the manuscript. ST performed the data collection. All authors performed the data analysis. SY made critical revisions to the manuscript and supervised the study. All authors read and approved the final manuscript.