Development and Psychometric Properties of a Compassionate Care Questionnaire for Nurses

Background: Compassionate care is emphasized within professional ethics codes for nursing and is a key indicator of care quality. The purpose of this study was to design and assess the psychometric properties of a compassionate care instrument for nurses. Methods: This methodological study was carried out in two qualitative and quantitative phases from February 2016 to October 2018. The rst phase was composed of a qualitative study using a content analysis approach to explain the concept of compassionate care through interviewing nurses, patients and family caregivers. The original draft of the questionnaire was developed using qualitative ndings and a subsequent review of literature. In the second phase, the psychometric properties of the questionnaire were assessed for validity and reliability. Data analysis was performed using descriptive and inferential statistics via SPSS v.16 software. Results: According to the results of the qualitative phase and review of literature, 80 items were extracted. In the quantitative phase, face and content validity led to retaining 40 items. In construct validity, 28 items with a factor load above 0.4 were retained. Convergent validity showed a moderate correlation coecient between the questionnaire and the nurses' caring behaviors tool (r = 0.67, P = 0.01). Reliability of the 28-item questionnaire using the calculation of the Cronbach's alpha coecient and the intra-class correlation coecient were reported as 0.91 and 0.94 for the whole questionnaire, respectively. Conclusion: The questionnaire had an appropriate validity and reliability for measuring nurses' compassionate care. Therefore, the instrument can be used to measure and report the quality of nursing care.


Background
Compassion is the main focus of health care policy and the essential characteristic of person-centered nursing care [1][2][3]. It is widely considered the rst principle of health care ethics [4]. as well as the basis of high-quality care delivered by healthcare professionals [5]. suffering that requires personal awareness of others' suffering and moral response". From Dewar's perspective, compassionate care is the relief of individuals' suffering [6]. Several studies have reported the positive clinical and health outcomes of compassionate care in both patients and nurses. For example, compassionate care can increase patient satisfaction and increase nursing job satisfaction [7].
On the other hand, lack of compassionate care leads to lower standards of care [8].
Compassionate care has been considered the patient's right [9] and is one part of the professional performance standards that health care providers need to be educated about and healthcare systems should measure and report it [10,11].
Although, it is always emphasized on ethical principles such as compassion in the educational context, the real problems arise when nurses face organizational realities [12]. The providing of compassionate care depends on not only the therapist, but the members of the healthcare team and the organizational context [11].
One of the important barriers to improving the quality of patient care and satisfaction with care is the lack of a compassionate clinical scale with strong psychometric properties [13]. Currently, there is no a standard instrument for measuring compassionate care in the health care system. Compassion is one aspect of the quality of care that should be continuously evaluated [14].
Assessment of compassionate care is essential for evaluating and enhancing clinical performance [11].
Empirical evidence of compassionate care in health care systems is scare, since the empirical perception of the nature of compassion has not been well developed. In addition, most studies have been based on prede ned theoretical de nitions that lack speci city, clinical applicability, conceptual validity, and are not patient-orientated [10]. Also, studies have not been conducted using appropriate compassionate care tools in nursing.
One of the challenges for the measurement of compassion care in nursing is that its meaning varies depending on people's perspectives. The concept of compassion is complex and its measurement needs to re ect the concept from the perspective of patients, family members and clinical staff [15].
There are some instruments available to measure compassionate care for physicians [16]. and the public, and on compassionate competence [17,18]. or non-verbal compassionate communication [19]. In addition, in most available tools, the de nition of compassion has been borrowed from the dictionary or a review of literature. Therefore, they do not cover all aspects of compassionate care as delivered by nurses. The present study aimed to design and analyze the psychometric properties of an instrument for measuring nurses' compassionate care.

Methods
This methodological study was conducted from February 2016 to October 2018 in one of the southeastern cities of Iran. In the rst phase, dimensions of the concept of compassionate nursing care were developed. In this regard, individual interviews were completed with 20 nursing staff (including 18 nurses and 2 nursing instructors working in different in-patient wards of hospitals a liated with the university. Clinical specialisms included surgical, emergency, burn, hemodialysis, CCU, and ICU), 8 patients and 6 family caregivers. In addition, two focus group interviews were conducted with one including 6 nurses from internal, emergency, CCU and ICU specialisms, and the other involving 6 nursing instructors.
At the end of the qualitative phase, items for the questionnaire were extracted so that, in accordance with the qualitative study goal, the de nition of the concept of compassionate nursing care and its constituent dimensions was determined. Next, using this template, the pool of items based on the domains and sub-domains of the concept of compassionate nursing care was created (inductive approach). Also, a review of literature and relevant questionnaires (deductive approach) were used. The research team then merged overlapping items and the initial 80 item version of the questionnaire was considered for psychometric analysis. The second phase of the study was the psychometric assessment of the instrument. Face and content validities were performed using qualitative and quantitative methods. Also, structure validity and reliability were performed.
For qualitative face validity, the perspectives of 12 nurses working in different specialisms, as well as the perspectives of 3 nursing tutors and 2 language experts were sought in terms of the di culty level, ambiguity, and grammar in face-to-face interviews. Their comments resulted in amendments to the questionnaires' contents, but no items were deleted. After reviewing the items, the quantitative method using the impact scores of items helped evaluate validity of the questionnaire. For this purpose, 10 nurses working in the hospital were asked to comment on the importance of each of the items based on the 5-point Likert scale (5 = very important, 4 = important, 3 = relatively important, 2 = not very important, 1 = It does not matter at all). The impact score of each item was calculated and a score more than 1.5 was considered appropriate [20].
For content validity using a qualitative method, 15 nurses who had signi cant knowledge and experience in the eld of instrument development and nursing ethics participated. They were asked to comment on each item in terms of the grammar, use of appropriate words, the placement of the items and scoring. For content validity, the content validity ratio (CVR) was calculated to determine the necessity of each item.
Also, the content validity index (CVI) was used to examine the relevance of each item regarding the concept of compassionate care [21]. The Kappa coe cient for the agreement between evaluators was calculated using the total content validity index (S-CVI) [22]. The initial form for determining the Content Validity Ratio (CVR) was on the basis of the 3-point Likert Scale (necessary, useful but not necessary, not necessary). According to the Lawshe table, items with a score equal to or greater than 0.49 were retained [23].
The Content Validity Index (CVI) was calculated through the Waltz & Bausell (2010) approach. This involved the evaluation of the perspectives of 15 specialists based on the criteria of relevance, simplicity and clarity on a 4-point Likert scale. The cutoff point for the CVI was considered 0.78 and higher [23] Similarly, the Kappa statistics were calculated for an agreement between evaluators [22]. The mean of the content validity index (S-CVI) was also used to calculate the total content validity index (S-CVI) [24] Before exploratory factor analysis, item analysis was conducted with a sample size of 40 people. Item analysis was performed to assess the Cronbach's alpha coe cient for initial reliability and to identify items that affected reliability [25,26] In this study, exploratory factor analysis was used to assess the construct validity of the questionnaire.
To determine the required sample size for factor analysis, 5-10 people per item has been recommended though a higher sample size has also been suggested [27]. Three individuals for each item has been suggested if the factor load for each item is above 0.80 [28]. In the present study more than 10 nurses were sampled for each item (n = 420). The exploratory factor analysis was performed using the Kaiser-Meyer-Olkin Index (KMO) and the Bartlett's test of sphericity, main component analysis, Scree plot and varimax Rotation with a sample size of 420 nurses.
To determine the number of constructs, initial Eigenvalues and scree plot were used [29]. In the next step, the exploratory factor analysis method was performed using the varimax rotation. The factor load of each item in the factor matrix and the rotated matrix should be at least 0.4 [30]. To assess convergent validity, this questionnaire and the Wolf et al.'s questionnaire (CBI-42) (1998) were given simultaneously to 100 nurses and the correlation between scores was examined.
The internal consistency and stability were used to assess reliability of the questionnaire. Internal consistency was assessed with a sample size of 420 nurses. The Cronbach's alpha of 0.7 to 0.8 indicated a favorable internal consistency [26]. The test-retest method was used to assess consistency of the questionnaire by 50 nurses within a two-week interval. The scores of the two tests were calculated using the calculation of the intra-class correlation coe cient for each of sub domains and the whole questionnaire. Burns and Grow (2014) recommended the stability of the questionnaire to be assessed over a period of 2 weeks in a month. 31 The index rating above 0.8 was a sign of the instrument's stability [32]. The ease of use of the questionnaire, as well as the effect of ceiling and oor were studied.
Inclusion criteria for nurses were having at least a bachelor's degree, willingness to participate in this research, having manageable workload, appropriate physical and mental status. Data analysis was performed using the SPSS software v.16.

Results
In the qualitative stage, individual and focus group interviews were conducted to explain the concept of compassionate nursing care. Dimensions of the concept included effective interaction, professionalism, and continuous comprehensive care. De nition of the concept has also been identi ed as follows: Compassionate care is a professional-quality care that takes place through clinical excellence, adherence to ethical values, and openness to needs. Effective interaction through the use of emotional support, building trust and communication skills, along with continuous comprehensive care for the coordination and integrity of the patient's existential dimensions occurs at the moment [33].
Upon completion of the phase, the nurses' compassionate care measurement questionnaire was prepared according to the de nition of the concept of compassionate nursing care and its constituent dimensions.
This consisted of 98 items. Within the review of available literature, 130 possible items were identi ed based on the studies examined. The research team merged a number of overlapping items, so the nal number was reduced to 80.
Qualitative face validity led to some modi cations and revisions of the items. For quantitative face validity, the impact item score was calculated and all items except 5 items which achieved a score more than 1.5. Therefore, the number of items was further reduced from 80 to 75 items.
In qualitative content validity, expert opinions led to merging overlapping items. Next, a questionnaire with 48 items was prepared for quantitative content validity. The CVR and CVI were calculated. For CVR, 40 items achieved scores more than 0.49 and were retained. Given the cutoff point of 0.78 for the content validity index, this index was more than the minimal level in all items and therefore, no items were deleted. Also, the Kappa coe cient score for 40 items was excellent. Mean score of the content validity index (SCVI / Ave) was 0.91, which was considered to be excellent.
Before factor analysis, item analysis was performed with a sample size of 40 nurses. Reliability of the questionnaire was reported as 0.94 based on the calculation of the Cronbach's alpha coe cient. The results of item analysis to assess the correlation coe cient between the items and the total score led to the elimination of 2 items. In other cases, each item had an association with at least one other item reported as 0.2-0.3, and no item was deleted.
To analyze construct validity of the questionnaire, for analyzing the data in the rst stage of factor analysis, the sampling adequacy index by Kaiser-Meyer-Olkin Index test was calculated and reported as 0.928. The results of Bartlett's test of sphericity was signi cant at P < 0.001.
To determine the number of constructs of the questionnaire, the Initial Eigenvalues and the scree plot were used. An initial analysis with a special value more than one was performed, and along with 8 factors accounted for 57.278% of the observed variance. The scree plot showed that the major variance was assigned to the rst factor, and it was at for other 4 factors (Fig. 1). Therefore, through limiting the number of factors to 4 factors, a factor analysis was performed and items were considered. The factor analysis led to 4 factors accounting for 48.05% of the variance.
In the next step, the exploratory factor analysis method was performed using a varimax rotation. In this research, factor load of 0.4 was considered the minimum acceptable degree of correlation between each item and extracted factors. At this point, those items that had a high correlation with each other were included within a category. At this stage, the research team decided to delete ve items that did not reach the minimum load factor of 0.4 or had repetitive concepts. The nal questionnaire had 28 items with four factors of professional performance (9 items), continuous follow up (6 items), Surrogacy (7 items), and empathic communication (6 items) ( Table 1).  34 and the correlation between the scores was examined. The correlation between the scores of these two questionnaires was moderate (p < 0.001 and r = 0.68) For reliability, internal consistency and stability methods were used. For internal consistency, the Cronbach alpha coe cients of the whole questionnaire and each of its domains were calculated ( Table 2). In addition to the Cronbach's alpha coe cient calculation method, the split half technique was used to assess internal consistency. In this method, the correlation coe cient between the rst and second half items of the questionnaire was calculated, and was reported as 0.82, indicating a desirable reliability of the questionnaire (Table 3). To verify stability, a test-retest method was used. The scores of the two tests were calculated using the calculation of the intra-class correlation coe cient for each of the domains and the whole questionnaire (Table 4). To assess the ease of the questionnaire use, the mean time for completing the questionnaire and the percentage of individuals who did not respond to each item were calculated. The mean time of completing the questionnaire was reported as 4 minutes with a range of 3-5 minutes. Also, for all items, the rate of non-response should be 0-5%, which was the same with our questionnaire.
The effect of ceiling and oor was that more than 15% of respondents obtained the highest or lowest achievable scores, respectively. In general, the presence of a ceiling or oor effect indicated that the minimum and maximum severity of the phenomenon were not included in the questionnaire, and showed inadequacy of content validity. The results of the study on the effect of ceiling and oor on the sample size of construct validity (n = 420) showed that the minimum and maximum score in none of the subscales and the whole instrument did not reach 15%. So, the questionnaire had no ceiling and oor effect.

Discussion
In this study, the instrument of nurses' compassionate care measurement was designed and its psychometric properties were evaluated. In the rst stage, individual and focus group interviews were conducted to explain the concept of compassionate nursing care with nurses, nurse instructors, patients and family caregivers. Qualitative ndings were presented in three themes of effective interaction, professionalism and continuous comprehensive care. At the end of the qualitative phase, items were extracted based on the operational de nitions and a review of literature. Next, the questionnaire was assessed for psychometric properties.
One of the important steps of questionnaire's design is the process item generation. In the present study, item generation was made using a combination of inductive and deductive approaches. While in some available instruments, the production of items has been solely generated through a review of literature and based on dictionary de nitions (deductive approach) [19,35,36]. Also, due to the in uence of social and cultural factors around the concept, there is a need to incorporate a deep understanding of related experiences of nurses, patients and family caregivers, assessment of which was performed in this study.
Compared to the present study, the Schwartz center compassionate care scale [16] and the Fogarty Compassion Scale (1999) were designed for physicians. It should be noted that physicians and nurses have a different understanding of the needs of patients due to their different professional roles. Also, the nature of nursing care is different and over a more sustained timeframe than medicine.
Some de nitions of compassion that are based on dictionary classi cations or literature reviews include aspects of empathy or sympathy [10], while compassionate care is conceptually broader than these concepts. In compassionate care is emphasized to interventions for relieving suffering. 2 In some instruments, the word "compassion" itself is used in the scale instead of the use of descriptive variable for compassion [16,35,36].
In the study by Lee and Simon (2017), the concept of compassion competence was developed through analyzing the hybrid concept and indicated speci c nursing behaviors for measuring the effects of compassion. In this study, only nurses in special wards were interviewed [37], but in the present study, experiences of nurses, patients and family caregivers in specialized units (CCU, ICU, hemodialysis, emergency and burn), surgical and internal medicine were used to explain the concept. Therefore, the present study contained richer information about the concept of compassionate care.
Compared to the present study, in most previous studies, face and content validity was assessed using a qualitative method. In the study by Fogarty (1999) [35], face and content validity was not carried out. The calculation of the item impact score, deletion of inappropriate items and determination of the importance of each item were carried out [20]. Also, calculating the content validity ratio in the present study helped identify those items that were necessary for measuring the concept [21]. Calculating the content validity index helped with identifying related concepts based on the opinions of experts [24]. The Kappa score of the questionnaire was excellent, indicating an agreement between evaluators on the relevance of items.
In the present study, before making construct validity, item analysis was performed. The results of the exploratory factor analysis indicated the adequacy of the sample size for performing construct validity assessment. Varimax rotation led to assignment of 28 items to 4 factors as professional performance (9 items), continuous follow up (6 items), surrogacy (7 items) and empathic communication (6 items insight. The convergent validity of the questionnaire showed a high correlation coe cient, but item analysis was not performed for identifying items affecting initial reliability. In the Grimani's study (2017) the way of extracting factors and determining the factor structure in construct validity was not clear [38].
In a study of Burnell & Agan (2013), exploratory factor analysis with a sample size of 250 hospitalized patients was conducted. Twenty-four items were tted into four factors of meaningful relationship, patient expectations, care characteristics, and competent specialist. However, information on the adequacy of sampling was unavailable. Also, the method of extracting factors and determining the factor structure had not been reported [36]. In some of available tools, construct validity has not been conducted [19].
In most existing tools, item analysis has not been performed for identifying items that affect reliability. In this study, reliability was assessed through internal consistency and stability (test-re test method) within a two-week interval. Similar to the current study, reliability of the scale was assessed through an examination of internal consistency and test-retest [37]. In this study, a half split technique was also used to examine reliability. The linear correlation between the rst and second half of questionnaire items indicated appropriate reliability of the questionnaire.
In most existing tools, stability has not been reported [18,19,35,36]. The high stability of this questionnaire showed that the individual score of the test would remain constant over time, so this could not be compared with other questionnaires.
In this study, the effect of ceiling and oor on the sample size of 420 nurses was studied. One of the factors in uencing the reliability of a tool was the effect of ceiling and oor. If there was no such an effect, individuals with the highest and lowest scores were not intelligible and could not be differentiated, and reliability decreased. 6 No information on the effect of ceiling and oor has been reported in any of available tools.
In the present study, a broad spectrum of participants was considered through selecting a sample of nurses, nurse educators, patients and family caregivers with a maximum variability and a comprehensive psychometric assessment.

Limitations
The psychometric properties of the questionnaire were conducted in only one city. Therefore, it is recommended that future studies are carried out in other cities and contexts.

Conclusion
A questionnaire was designed in this study to measure nurses' compassionate care. The results of the study showed that the validity and reliability of the questionnaire was supported. The use of the questionnaire was also easy and only took around 4 minutes. Therefore, this is an appropriate questionnaire to measure nurses' compassionate care. Measuring compassion helps recognize the compassionate performance of clinicians, instructors, and policymakers to adopt a more effective strategy to promote compassionate care as an aspect of holistic care.