Re-assessing the validity of the Moral Sensitivity Questionnaire (MSQ): Two new scales for moral deliberation and paternalism

RATIONALE, AIMS, AND OBJECTIVES
The current study and previous research have called the six-component model of Lützen's 30-item Moral Sensitivity Questionnaire (MSQ) into question. For this reason, we re-examined the construct validity of this instrument.


METHODS
In this cross-sectional study, which was based on a convenience sample of Dutch nurse practitioners (NPs) and physician assistants (PAs), we tested the validity of MSQ items using exploratory and confirmatory factor analyses (EFA and CFA, respectively).


RESULTS
The EFA revealed a two-component model, which was then tested as a target model with CFA and was found to have good model fit. Some items were correlated with two uncorrelated latent constructs, which we labelled as "paternalistic" and "deliberate" attitudes towards patients.


CONCLUSIONS
As in previous studies, the analyses in the current study, which was conducted among PAs and NPs, did not reveal six dimensions for the 30 items. Two new latent dimensions of moral sensitivity were psychometrically tested and confirmed. These two components relate to studies investigating ethical behaviour, and they can be used to describe the moral climate in healthcare organizations. The scales are indicators of the extent to which health professionals behave in a deliberate (sensitive) or paternalistic (insensitive) manner towards the opinions of patients within the context of medical decision-making.

(sensitive) or paternalistic (insensitive) manner towards the opinions of patients within the context of medical decision-making.

K E Y W O R D S
deliberation, moral sensitivity, nurse practitioner, paternalism, physician assistant

| INTRODUCTION
In Western health systems, two interesting shifts with regard to professional and patient responsibility have taken place over the last few decades. First, the professional responsibility of making medicoethical decisions that exclusively belonged to the realm of medical doctors (MDs) has been extended to other health professions, including nurse practitioners (NPs) and physician assistants (PAs). 34,36 Second, in the past, MDs guided their patients through the medical treatment process according to a strong paternalistic attitude. In current practice, the perspective has shifted towards emphasizing the central role of patients in healthcare. 50 Within the models of shared decision-making (SDM) that are now prevalent, assigning a central role to the patient is regarded as an ethical imperative. Such models of SDM are consistent with the four principles of ethics in care: respecting autonomy, propagating beneficence, avoiding harm, and achieving justice. 7 Medical decisions established through SDM have been shown to be associated with improved medication compliance, health-related quality of life, an increase in patients' perceived control over their choices with regard to treatment options, and a decrease in healthcare utilization. 16 In the past, clinicians were accustomed to employing protocols and guidelines that were accepted as the gold standard for treatment. In contrast, computer-literate and empowered patients are adding a new dimension to the treatment relationship, thus potentially increasing the risk of tension and conflict. 25 In light of such changes in the treatment relationship, tension is likely to arise between what a clinician regards as the best treatment option (or even what rules and regulations dictate that they propose) and the treatment that is perceived as the best in the eyes of the patient. Such tension could create a moral dilemma, which could be described as a situation in which for example there are conflicting opinions (between health professional and patient) regarding what is the best treatment option. 15 Health professionals may employ one of essentially two decisionmaking strategies or coping mechanisms to reduce dilemma-related stress: (i) a predominantly patient-centred, deliberate attitude focused on patient autonomy 42,47 or (ii) a more dominant, clinical view, known as the "paternalistic approach." 40,50 Health professionals adopting a paternalistic attitude are less likely to engage in dialogue regarding treatment options or the health beliefs of patients. They are more likely to decide what is best for the patient based on their own selfpresumed professional knowledge and evidence-based practice.
Health professionals who have adopted a deliberate attitude that takes the opinions and wishes of patients into account must reflect on their decisions in the light of the patient's views. 1

| BACKGROUND
Regardless of whether health professionals cope with moral dilemmas through either a deliberate or paternalistic attitude, moral dilemmas arising within interactions must necessarily be resolved through an ethical decision-making process. For example, James Rest captures this ethical decision-making process in the "four-component model of moral behavior" (FCM). The FCM states that moral decision-making is influenced by moral sensitivity, moral reasoning, moral motivation, and moral character. In this model, Rest conceptualizes moral sensitivity as the first and essential precursor in ethical decision-making, defining it as "a combination of one's recognition of moral issues, and how one reacts and processes these issues from an affective perspective within a social context." 45 Lützén and colleagues 32 defined the concept of moral sensitivity (MS) in theoretical terms as "a personal attribute involving the ability to recognize a moral conflict, a contextual and intuitive understanding of a person's vulnerable situation and insight into the ethical consequences of decisions made on behalf of another person." They operationalized this concept of moral sensitivity using the Moral Sensitivity Questionnaire (MSQ) in study populations consisting of psychiatrists 30 and psychiatric nurses. 29,31 Based on their results, they reported six dimensions (ie, latent variables). With reference to exploratory analysis, Lützén and colleagues label these dimensions as follows: 1) interpersonal orientation, 2) structuring moral meaning, 3) expressing benevolence, 4) modifying autonomy, 5) experiencing moral conflict, and 6) having confidence in medical knowledge. 29 In a methodological and statistical appraisal of the results as published, however, a weak structure emerges as a result of three observations. First, factor loadings (correlations between items and the underlying construct) were too low, as items should be sufficiently correlated (factor loading ≥.40) with the target dimension in the data. Second, some correlations were biased such that the target construct could not be interpreted, as the full matrix of factor loadings was not presented. And third, several items were correlated with more than one construct of moral sensitivity, thereby violating the necessary condition that each item should exclusively tap an aspect of only one underlying construct or dimension. As a consequence of these problems, the indices of reliability or internal consistency (Cronbach's alpha) for these six scales ranged from 0.36 to 0.61, thus indicating poor intercorrelations between the items.
Other MSQ studies conducted in many different countries 10,14,23,54 have also evaluated the content and psychometric quality of a 30-item MSQ. These combinations of items proposed in these studies deviate from the latent constructs proposed by Lützén.
As was the case with the instrument-testing performed by Lützén and colleagues, the aforementioned studies consisted exclusively of exploratory factor analysis (EFA). This method is not the most suitable for arriving at conclusive results about the factor structure of a scale, given that EFA based solely on the Kaiser criterion could potentially generate an excessively inclusive result. 18 Thus, as the mixed results of the above studies suggest, there are still some unclarities about what the MSQ measures and how it should be used. On top of this, the 30-item MSQ has so far only been validated among psychiatrists and among nurses. 29,30,33 However, it will be particularly interesting and relevant to develop and validate the MSQ among PAs and NPs. This is because these healthcare professionals have a special role that distinguishes them from nurses and doctors. As their responsibility lies in between that of MDs and nurses, their role is largely characterized by having medical-decisional responsibilities. In this role, both deliberate and paternalistic attitudes may take a prominent place in their professional identities. Indeed, in a feasibility study that we first performed, we found some indication that the MSQ administered among this specific group particularly distinguishes paternalistic and deliberative attitudes. We conducted this feasibility study among Master's-level PA students (N = 32). By employing the method developed by Ruscio and Roche, 48 we found a two-factor structure. One of these factors apparently reflects a paternalistic attitude in decision-making, while the other reflects a deliberate attitude.

| Study objective
To test the reproducibility of this two-component factor structure, we performed another study based on a larger sample consisting of the particular group of Dutch PAs and NPs. The objectives of this study are as follows: a) evaluate how many factors of the MSQ should be retained for further factor analysis, and b) apply both exploratory and confirmatory factor analysis (CFA) to test the dimensionality, scalability, and construct validity of the items remaining from the MSQ.

| Research Questions
To meet the study objectives, the following research questions will be addressed: Which items of the MSQ are unequivocally correlated with latent constructs when using the criteria of EFA after having determined how many factors should be retained, and how strong is the model fit, based on CFA?
Are the items of these components scalable? And do these scales have sufficient internal consistency (reliability)?
Do the retained scales confirm discriminant or convergent validity as hypothesized when correlated with the following scales, which are known to tap moral aspects of the decision-making process:

| Study design, participants, and data collection
In this cross-sectional study, five PA and one NP degree programs were selected as sources for approaching alumni. Based on the Dutch Personal Data Protection Act, the researchers were not granted permission to use the databases of the programs in order to retrieve alumni email addresses. The information letter concerning the current study was therefore sent to 470 NP alumni and to 426 PA alumni by the programs' administrators. By activating a hyperlink to a private web-based system included in this letter, individual alumni were free to reveal their contact details to the researchers. When respondents returned permission to use their email addresses, this was regarded as informed consent. Upon receipt of their permission, these alumni were sent the access key to the web-based set of questionnaires.
In all, 294 subjects were willing to participate: 176 PAs and 118 NPs, meaning a response rate of 52.7% (ie, 155/294). Upon closure of the online survey (between January and March 2015), 155 respondents had completed all of the questionnaires. In all, 139 alumni, who initially consented to participate, eventually did not reply to the survey. Therefore, no information about this group was available that could be used to test for selection bias. All questions in the Qualtrics online survey environment were forced choice, so there were no missing data. The "Strengthening the Reporting of Observational Studies in Epidemiology" (STROBE) checklist was employed (see Supporting Information S1).

| Ethical approval and consent to participate
According to the statement by the Central Committee on Research Involving Human Subjects (www.ccmo.nl), no Institutional Review Board (IRB) approval was warranted for this type of survey study among volunteer professionals. This study was performed in accordance with the tenets of the Declaration of Helsinki. 21 Only the first author (LK) had access to the online survey data.

| Academic integrity statement
The dataset in the current study was the same as the one in Kuilman et al (2019). 26 However, from that pool different, variables were used, focusing on different research questions. Only the MDS and the indicator for moral reasoning (DIT-N2) were used in both studies, albeit with different hypotheses and functionality (independent vs dependent variable).

| Statistical analyses
All data were analysed using SPSS Statistics for Windows, Version 25.0. CFA was performed using SPSS AMOS, Version 23.0.

| Bivariate analysis
For categorical data, we used the chi-square test (Fisher's exact tests for 2 × 2 contingency tables) and the difference-between-proportions test. 38 For continuous variables, we used the Student's t-test for independent samples.

| Multivariate analysis
To assess the structural validity of the MSQ, we performed factor analyses and calculations of reliability estimates, as explained below.

| Model fit through Confirmatory Factor Analysis
We used the following goodness-of-fit indices to determine model fit using CFA: a) chi-square/degrees of freedom (χ 2 /df ), b) root mean square error of approximation (RMSEA), c) standardized root mean square residual (SRMR), d) comparative fit index (CFI), and e) a goodness-of-fit index (GFI). The χ 2 /df with a ratio between 0 and 2 is indicative of a good fit. 49,52 For the RMSEA, a cut-off value less than or close to 0.06 was assumed to be appropriate. 24 The lower limit of the confidence interval (CI) should be close to 0, and the upper limit should not exceed 0.08. 35 We also report the SRMR, as its standardized nature makes it easier to interpret. Values for the SRMR ranged from zero to 1.0, with good-fitting models having an acceptable threshold of less than 0.08. 24 For the CFI, values equal to or greater than 0.95 are deemed indicative of a good model fit. 24 For the GFI, cut-off values greater than or equal to 0.95 are recommended for relatively low factor loadings and sample sizes. 37 The Akaike information criterion (AIC) was used to compare different models.
This criterion is a descriptive measurement, in which the preferred model is the one with the lowest value. 3

| Internal consistency
Cronbach's alpha values were calculated to examine the reliability of all scales. In general, values equal to or greater than ≥0.70 are considered sufficient. 9 3.9 | Convergent and divergent validity In this study, convergent validity was imputed according to statistically significant associations (linear associations between measurements of moral behaviour), while divergent validity was assumed when there was no correlation (ie, P > 0.05). The degree of overlap between constructs was estimated by calculating the nonparametric effect size of Rho (given the sample size). The statistically significant small effect size, with Rho in the range of ≥0.10 to <0.30 and Rho ≥0.30 to < 0.50, indicates a medium effect that is comparable to relevant effect sizes in terms of differences between two means. 11 Divergent validity was analysed according to correlations between measurements of moral behaviour that were expected to be unrelated (ie, no statistically significant correlation).

| Sociodemographic characteristics
In this study, the following sociodemographic characteristics were self-reported: age, gender, working environment, and religion.

| Moral Sensitivity Questionnaire
In order to adjust the psychiatry oriented MSQ 29 for use in research populations of NPs and PAs, it was necessary to rephrase nine items.
For example, references to "psychiatrist" were replaced with references to either "NP" or "PA" in two items, and the terms "psychiatric care" and "psychiatric practice" were rephrased as "care" or "practice," respectively, in five items. Furthermore, two items referring to "treatment under the Mental Health Act" were rephrased to refer to "care provided to incapacitated patients." Respondents were asked to use a 7-point Likert scale (1 = fully disagree to 7 = fully agree) to indicate how they perceived their own manner of decision-making in moral dilemmas. Each of the items reflected either a paternalistic or deliberate attitude, as assumed in a previous feasibility study conducted among PA students.
For each scale, item scores were coded, summed, and transformed into a scale ranging from 0 to 100 (with higher scores reflecting greater sensitivity or insensitivity) and calculated by subtracting the lowest possible scale score from the raw summed scale score, divided by the range of scores on the scale and multiplied by 100.
The instruments used for testing the convergent and divergent validity of the hypothesized latent MSQ constructs (as found in the feasibility study), as described in Appendix S1, include the following: a) the BCPH scale; b) the EAS; c) the MDS, and d) the DIT-N2. Appendix S1 contains the following references: 2, 4, 5, 43, 44, 46. All of the scales used in the current study were transformed towards normality through a two-step transformation process, conducted prior to the analyses. 53

| Translation of measurement instruments
Questionnaires were translated into Dutch following the procedure proposed by Guillemin and colleagues. 22

| Hypotheses regarding convergent and divergent validity
We examined the strength of the correlation coefficients as indicators of conceptual overlap between paternalistic and deliberate attitudes according to four concurrent self-report measurements. The following hypotheses were formulated:

| Divergent validity
Although paternalistic and moral deliberate attitudes are usually pictured as two opposites, the traits are nevertheless expected to be independent of each other. This is because the features of both traits are not incompatible with each other. For example, an important feature of a deliberate attitude is valuing to have a relationship with patients. This is not necessarily in contradiction with one's inclination to follow rules and regulations and base one's decision on medical practice (which is a feature of a paternalistic attitude).
So, even though a healthcare professional may be aimed at having a relationship with a patient and treat the patient with respect (ie, deliberate attitude), still the healthcare professional can decide to base his/her decision on medical knowledge or regulations, even if that is against the will of a patient (ie, paternalistic attitude), if he/she really thinks this is in the best interest of the patient. We therefore hypothesize that H1 :There is no correlation between the two scales measuring a deliberate attitude and paternalistic attitude respectively.
We further assume that paternalistic and the deliberate attitude are different from moral reasoning. After all, moral reasoning reflects a cognitive, intra-personal process, in which a person engages in a deliberation on what is the moral thing to do. The paternalistic and deliberative attitudes refer more to a person's general preferences for how they relate to patients. This is more an inter-personal issue and reflect one's tendencies of how to behave in a patient-professional relationship. We therefore hypothesize that H2 :Neither the paternalistic nor the deliberate attitude scale is expected to have any significant overlap with the level of moral reasoning (DIT-N2).

| Convergent validity
We assume that deliberative attitude is related to several ethical tendencies of people. After all, the main characteristic of a deliberative attitude to value a respectful relationship with patients. Hence, deliberate patient-centred decision-making is expected to more strongly possess preferences and traits that are ethics-related. For example, they may be more likely to adhere to the fundamental principle of "First, do not harm"; they may be more likely to value ethical considerations in care and are less inclined to use mechanisms of moral disengagement. We therefore hypothesize that  Despite the expectation that a paternalistic attitude is negatively related to the need to advocate for ethics and positively to moral disengagement, this does not necessarily mean that they do not care about the "First, do not harm" principle. In fact, people who score high on paternalism may also adhere to this principle, although they try to achieve this in different ways than people who score high of moral deliberation (namely, through the adherence to regulations and expert information rather than through building a relationship with the patient). Therefore, we hypothesize

| Sociodemographic characteristics
An overview of the sociodemographic characteristics broken down by professional group (ie, NPs and PAs) is provided in Table 1. The mean age of the PAs (42.5 years) was lower than that of the NPs (48.8 years). The two groups did not differ in terms of gender, religion, or work setting. Moreover, no significant differences were found between the two professions with regard to the prevalence of politically conservative or liberal orientations. Based on these results, we considered it acceptable to merge the samples for analyses.

| Confirmation of the latent MSQ dimensions
The Velicer's Minimum Average Partial (MAP) test yielded three factors to be retained (MAP squared: 0.017) for consecutive analysis. 13 This test was followed by EFA, which was also based on a polychoric correlation matrix 39 using principal axis factoring and oblimin-quartimin-Q rotation. 6 A three-factor model converged well, detecting 16 items with loadings exceeding 0.40 and having no cross-loadings exceeding 0.20 on any other factors. 12 CFA was used to assess the model fit of the three-factor solution retained from the MSQ item pool. At first glance, the model fit parameters for the CFA appeared to indicate an acceptable model fit.
Upon closer examination, however, five items (all of which were clus- Once all of the items with factor loadings less than 0.40 were excluded from the analysis, the model became unidentifiable. For that reason, these three items were not included in Factor 1. As in the feasibility study, the results of this study indicate that the MSQ item pool represents two dimensions. In light of these findings, we conducted a meticulous recheck of the content of the factors retained from the MSQ items. This led to the conclusion that the findings were congruent with the content of the moral deliberation and paternalism of the physician-patient relationship models hypothesized by Emanuel and Emanuel 17 and empirically tested by Falkum and Førde. 19 Factor 1 thus represents a construct that we have labelled the "Paternalistic Attitude Scale" (MSQ-PATER), as the items reflect the thinking that one is acting in the patient's best interest while disregarding the patient's will in the matter. In this study, we defined the concept of paternalism as "a tendency to avoid empathizing with the patient's dilemmas and taking decisions with a strong emphasis on rules and regulations, as well as on medical knowledge and practice, and based on professional opinions about the best treatment options". Factor 2 represents a construct that T A B L E 1 Sociodemographic characteristics of participants, stratified by profession (NPs and PAs) we describe as the "Deliberate Attitude Scale" (MSQ-DELIB). All of the items in this scale centre on the dimension of a professional relationship between the clinician and the patient, as indicated by such socio-cognitive, affective themes as "autonomy," "relationship," "giving respect," and "providing patients with insight." The concept of moral deliberation thus implies that NPs and PAs engage in careful and serious deliberation before making any important medical decisions. This finding is based on the independent content analysis of MSQ items. We defined the concept of moral deliberation, measured by the MSQ-DELIB as "medical decision-making aimed at helping patients to determine the best healthrelated values that can be realized in the clinical situation after considerable deliberation." The new scales and their assigned items are presented in

| Convergent validity
Our analyses revealed several statistically significant correlations, which could be used to establish convergent validity, as hypothe-  Table 2.

| DISCUSSION
Prior to the current study, a feasibility study was conducted among students in a master's program for PAs, in order to test the psychometric properties of several instruments, some of which have also been included in this study. The objectives of the feasibility study also included modifying and validating a revised version of the MSQ, as developed by Lützén and colleagues, 29 for use among NPs and PAs. In that study, however, a simple EFA using Varimax rotation revealed 10 latent components, instead of the six that were theoretically assumed by Lützén and colleagues. A subsequent narrative review of the literature revealed that international scholars building on the work of Lützén and colleagues 10,14,23,54 had also been unable to reproduce the six factors proposed for the original instrument.
In light of these developments, we decided not to re-evaluate the six-component structure, but instead to modify and validate a revised version of the instrument. The outcomes of the current study support the validity and reliability of two new scales: MSQ-DELIB and MSQ-PATER. These findings are obviously preliminary, given that this is the first time that the validity of these new MSQ dimensions have been evaluated among Dutch NPs and PAs. The solid methodology of this study nevertheless contributes to these two new scales, which were established through CFA to produce a two-factor solution with good model fit and satisfactory internal consistency (reliability estimates).
Our findings are in line with work by Emanuel and Emanuel, who identify deliberative and paternalistic attitudes as two of the four parts of the clinician-patient relationship (the other two being informative and interpretive attitudes). 17 The first factor identified in the current study was labelled the MSQ-PATER. Examination of the seven items of these scale reveals that they centre on such themes as "following the rules," "personal opinion about good care," and "best treatment." All of these themes are consistent with the general perception of paternalism, that is, one is acting in the patient's best interest, while disregarding the patient's will in the matter. In this light, we formulated the following operational definition of paternalism was formulated to capture the meaning of the MSQ-PATER: "The tendency toward paternalism in medical decision-making is activated by a clinician's preference for arguments based on rules and regulations. Decisions are established through the interplay between the clinician's own opinion, medical knowledge, and experience, as well as the opinions of others, while ignoring the will of the patient." The second factor identified in the current study relates to the dimension of the professional relationship between the clinician and the patient, as indicated by affective, socio-cognitive considerations (eg, consideration for "autonomy," "relationship," "giving respect," and "providing patients with insight"). We combined these four items to form the MSQ-DELIB, which reflects "the clinician's aim of helping patients to determine the best health-related values that can be realized in the clinical situation". 19   The current study used the same pool of 30 items from the original physician's version of the MSQ, and factor analysis was used in order to assess whether the items correlated with the underlying construct.
Our results indicate a comparable reduction in the number of items.
This suggests that the 9-item MSQ of Lützén and collegues may not actually measure moral sensitivity. 28  son (eg, "The physician expert should decide"). As such, the scales may be more likely to reflect a deliberate predisposition towards a paternalistic and moral stance, rather than any broader, general values concerning moral deliberation and paternalism. We do not wish to make any value judgement concerning whether a deliberate or a paternalistic attitude is better. Even though it may seem that we now regard a paternalistic approach as inappropriate within the clinicianpatient relationship, it might be the case that this is a reflection of our contemporary culture. However, societies change, and it is possible that, in a future era, a paternalistic or a deliberate attitude is differently valued than it is now.

| Strengths and limitations
One of the major strengths of this research lies in its robust study design (employing CFA), including the assessment of the convergent and divergent validity of the scales. Another important strength is that the results are based on a representative sample that reflects the characteristics of the PA and NP workforces in the Netherlands with regard to gender and age. 8 For this reason, the results can be generalized to a certain extent to both the NP and PA workforces at large.
We also expect that the MSQ-DELIB and MSQ-PATER are applicable to other healthcare professionals who share a comparable framework regarding knowledge, skills, and legal boundaries (eg, MDs). One weakness of the study, however, is that the stability of the instrument (ie, its test-retest reliability) was not assessed. The study design did not allow for testing the two scales for longitudinal validity. By definition, cross-sectional studies cannot examine the stability of the attitudes or traits of subjects over time. A follow-up study will investigate longitudinal psychometric research questions focusing on the testretest stability of the instruments.
Another limitation of the current study is that no a priori calculations of sample size were performed. Given the lack of studies assessing moral sensitivity among PAs and NPs, however, the field was open to exploration. Given the actual sample size addressed in the study (155 records), it may not be necessary to assume that the results of our CFA were compromised by the sample size. Although we are aware of the various rules and opinions used to determine the sample size needed for CFA, this study was based on a convenience sample with an N (=155) to P (number of items = 11) ratio of 14.1. We therefore felt confident that the assumption underlying CFA was not violated. 20 Finally, even though our analyses revealed statistically significant correlations for both the MSQ-DELIB and the MSQ-PATER scales based on convergent and divergent instruments, the explained variances were relatively low.

| CONCLUSION
The results of this study provide evidence of two new latent dimen-

| IMPLICATIONS
Healthcare professionals are quite likely to perceive working with patients as a natural calling, prompted by an intrinsic motivation to do good. Such inherent sympathy and empathy, however, which is perceived as beneficent, may become blurred by blind spots with regard to the personal attitudes held by individual clinicians and, consequently, their behaviour. With the exception of prejudice, the majority of complaints and disciplinary cases are based on either miscommunication or a feeling on the part of patients that they have been treated discourteously. With this in mind, both the MSQ-DELIB and MSQ-PATER could be used and applied as self-report tools for clinicians who would like to become more aware of their own underlying attitudes (eg, moral deliberation and paternalism) when communicating with patients. The two scales could also function as a type of "thermometer" with which to assess the moral climate and the workrelated moral stress experienced by health employees. 27

AUTHOR CONTRIBUTIONS
Luppo Kuilman: study conception/design; acquisition (and storage) of data; analysis and interpretation of data; writing and critical revision of the manuscript.