Introduction

Hand injuries are the second most common work-related musculoskeletal injury that leads to severe impairment [1]. Severe impairments cause disability and drastically affect the participation levels of a patient. Participation has been defined by the International Classification of Functioning, Disability and Health (ICF) [2] as an individual’s involvement in life situations. Based on the ICF, activity represents disability at the individual level and participation at the societal level [3]. Participation is using an activity to interact with others or with the environment. For instance, if we consider eating as an activity, eating with friends in a restaurant is participation. Any health condition can hinder involvement in life situations and thus restrict participation. Participation restriction is distinct from the outcome of activity limitation and impairment level and should be measured and targeted separately by the clinicians [4].

Participation contains different domains such as learning and applying knowledge; mental functions in general tasks and demands; communication; mobility; self-care; domestic life; interpersonal interactions and relationships; employment and economical life; and community, social, and civic life. The restriction of the participation domains is not the same in different populations with different disabilities. Different musculoskeletal conditions can cause different activity limitations, which will impact participation. For example, a patient with median nerve injuries will experience difficulties in picking up a coin or using a key (activity limitation) and restrict the participation level. Alternatively, one might expect patients with lower extremity impairments to have more mobility problems than those with upper extremity conditions. In contrast, patients with upper extremity issues might have more challenges manipulating, carrying, or interacting with technology. Different activity limitations profiles will also impact participation restrictions.

A consensus about measuring participation restriction as a different construct than activity limitation [5] has inspired the development of some generic questionnaires to measure participation. These generic questionnaires mainly measured interpersonal relationships and social participation [6–8]. Some participation questionnaires were also developed for specific conditions, such as brain injuries [9], distress [10], and hand injuries [11], which used the ICF framework and integrated patients’ reported restrictions in developing the questionnaires.

The Participation Behaviour Questionnaire (PBQ) [12] was developed to evaluate participation for patients with hand injuries using the ICF conceptual framework and considering patients’ reports of limitations in life situations and IRT methodology. The PBQ showed good measurement criteria regarding reliability, validity and unidimensionality [11, 13, 14]. The IRT/Rasch analysis confirmed the construct validity of the PBQ and reported it as a unidimensional measure. However, in measuring the construct validity, factor analysis is another method performed on all items of a PROM to evaluate the number of subscales of the PROM and the clustering of items within subscales (i.e., structural validity) [15]. In this study, we aim to explore the subscales underlying the board construct of participation with classic test theory via factor analysis [16]. Therefore, the purposes of this study were to evaluate the structural validity of the PBQ utilizing CTT and to confirm dimensionality as evaluated through IRT approaches.

Materials and methods

Results

Participants

A total of 404 participants completed the PBQ. The mean age of participants was 37 (16.0) years old, 32% were female, and 57% had an injury on their right hand. Demographics can be found in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Patients’ demographic characteristics (n = 404).

https://doi.org/10.1371/journal.pone.0267872.t001

Factor structure

The Kaiser–Meyer–Olkin test of sampling appropriateness was satisfactory (0.95), and Bartlett’s test of sphericity was statistically acceptable with [X2 = 264.092, p < .001]. The scree plot in Fig 1 shows that more than one component can be merged from the EFA. The eigenvalues explained variance and factor loadings for orthogonally and obliquely rotated two, three and four factors to explore the best factor structure. Table 2 displays the factor structure of the 30-item PBQ with all factor solutions after rotation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. In this scree plot, the point at which the curve first begins to level off indicates that there is one dominant latent variable which is obviously dominant over the remaining latent variables.

https://doi.org/10.1371/journal.pone.0267872.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. EFA of the PBQ; comparing three different factor solutions.

https://doi.org/10.1371/journal.pone.0267872.t002

All the models were significant, but the variances accounted for data were higher in the four factors solution (64.40%).

We intended to use parallel analysis to provide further evidence or a basis for deciding the number of factors more easily. The results showed that the eigenvalue of the first factor in the actual data was 14.52, while it was 1.91 in the simulative data set. The second factor’s eigenvalue in the actual data was 2.06, whereas it was 1.78 in the simulative data. The eigenvalue of the third factor in the actual data was 1.47, whereas it was 1.66 in the simulative data. The fourth factor’s eigenvalue in the actual data was 1.27, while it is 1.58 in the simulative data. When we shifted from the fourth factor to the fifth, the case was different, and thus the number of the scale factors is determinedly restricted to 4 because the eigenvalue of the simulative data (1.52) of the fifth factor was higher than the actual data (0.94).

The derived EFA factor model was entered into a CFA using the second half of the sample. The CFA supported four subscales that fit the four described subscales of the PBQ. A non-significant item trait interaction (chi-square, 3759.49, P<0.001, df = 406), relative chi-square) (chi-square index divided by the degree of freedom more than 2 (9.25) and RMSEA less than 0.08 (0.070) empirically confirmed verification of the four-factor structure solution. GFI (0.96), CFI (0.99), and TLI (0.96) indices more than 0.9 in the four-factor structure model indicated a good fit to the data (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Four factor structure of the PBQ based on the results of CFA.

Four factors were retrieved from the FA and can be seen as 1 to 4, which were named as follow: Social Participation and Interpersonal Relationships, Autonomy and Role, Subjective Satisfaction with Participation, Recreational, Sport, and Leisure Time. Each subscale was formed with a group of items from the original questionnaires, which can be seen in gray rectangles. All the subscales are intercorrelated, as can be seen with arrows among subscales. The correlation between each item and the subscale can be followed with the arrows.

https://doi.org/10.1371/journal.pone.0267872.g002

Based on the four-factor solution, names were assigned to the factors based on the recommendations from the experts, and internal consistency was tested for each factor. We emailed the final four-factor structure questionnaire to the invited experts and asked them to name each factor. The research team finalized the factors’ names and concluded on the suggestions. Factor one included ten items with loading values between 0.56 and 0.74 and reported Cronbach’s alpha of 0.91. The first factor, which explains 48.39% of the common variance of PBQ, with ten items, contained items that measured interpersonal relationships and social participation; it was entitled based on the recommendations from experts as "social participation." The second factor included ten items with loading values between 0.43 and 0.44 and a Cronbach’s alpha of 0.90. Factor two evaluates the individual’s assessment of their functional ability to participate, and so it was entitled "Autonomy and role." The third factor included six items with loading values between 0.40 and 0.76 and Cronbach’s alpha of 0.85. Factor 3 assessed the degree to which the individuals feel their participation is accepted by others, are satisfied with their participation and entitled to "Subjective satisfaction with participation." Factor 4 included four items with loading values between 0.47 and 0.63 and Cronbach’s alpha of 0.85. Factor 4 assessed the individual’s assessment of their ability to participate in recreational activities and was entitled, "Recreational, sport, and leisure time." Cronbach’s alpha was 0.96 for the total scale of 30 items (Table 2).

Discussion

This study provides additional information regarding the structural validity of the PBQ using factor analysis to extract the factor structure of the PBQ derived from Classic Test theory (CTT) approaches. CTT and IRT differ significantly in their modelling processes, making fundamentally different assumptions about the nature of the measured construct [24, 25].

Item response theory with Standard Rasch analysis is based on unidimensional models where a single latent trait is assumed to determine individuals’ performances on the test. The questionnaire could be analyzed for unidimensionality as a whole or separately for each subscale. The first approach cannot retrieve the subscale structure of the test, and the second approach ignores the potential intercorrelations between related but not identical latent traits [25, 26]. In the development of PBQ, the unidimensionality was tested on the whole questionnaire. The result confirmed that the PBQ measures the construct of participation, and no latent trait exists. The CFA demonstrated that a four-factor solution provided the best fit. We asked invited experts to label each factor considering the content of the items in each factor.

The research team then finalized the labelled: Social Participation and Interpersonal Relationships, Autonomy and Role, Subjective Satisfaction with Participation, and Recreational, Sport, and Leisure Time.

The first factor contained items that evaluate social participation. Based on definitions, the core central theme of participation is the social aspect of participation and interpersonal relationships. Social participation is an essential domain of participation which can discriminate participation from activity. Based on the ICF definition, participation is involvement in life situations [27]. However, in the ICF classification system, activity and participation are defined separately but included under the umbrella term of disability, with "d" codes used to specify the specific areas of disability, which leaves some conceptual ambiguity between activity and participation (28). Multiple authors have tried to further refine definitions of participation in ways that discriminate participation from activity. Operational definitions of participation have alluded to: fulfilling social roles and performance at the societal level [29], active engagement in intrinsically social activities [30], purposeful things that people do and defined by a social role (societal involvement) [28], societal involvement [29–31], being with others, interacting with others, doing an activity with others, helping others [32]. Some authors indicated the specific chapters of the ICF from d chapters; including assisting others (d660), particular interpersonal relationships (d730–d779), education (d810–d839), work and employment (d840–d859), economic life (d860–d879), community life (d910), recreational and leisure (d920), religion and spirituality (d930), and political life and citizenship (d950) can be considered as participation [29].

The second factor included items that evaluated ’Autonomy and Role’ and explained 6.86% of the variance of PBQ. Furthermore, role performance [33] is an important defining characteristic of participation considered in PBQ. Role performance implies that the individual is executing and experiencing their specific role, such as being a student, mother, etc. [34]. This also can differentiate participation from activity [28, 35].

The third factor contains items that measure the Subjective satisfaction with participation which is the internal aspect of participation, commonly defined as subjective or perceived participation. The ICF does not cover satisfaction with participation and engagement as the definition of ICF focuses on functioning [3, 30], thus may not capture some important perspectives about participation, including satisfaction and engagement. A qualitative study based on an interview with disabled patients identified important aspects of participation such as active and meaningful engagement/being a part of choice and control [3]. Involvement and engagement are subjective subdimensions of participation often neglected in measuring participation [34]. Based on the definition, engagement comprises the individual’s behavioural, cognitive, and affective investment during role performance [34]. However, the PBQ contains items that capture behavioural, cognitive, and affective dimensions of participation and measure satisfaction with participation.

The fourth factor evaluates participation in recreational, sports, and leisure time. Recreational and leisure (d920) is a chapter of ICF recommended as participation and represents the major participation domain of ICF [29].

The content of four extracted subscales indicated that the PBQ could evaluate participation comprehensively in patients with hand injuries. Many standard outcome measures are used in practice and specially designed to evaluate participation; however, their items do not fully capture the construct of participation in patients with hand injuries [36]. There are also some generic measures to evaluate participation [37–39]. The results from a systematic review of the content of existing generic participation measures indicated that no existing participation measure covers the complete set of participation domains identified in the ICF, and different PROMs address different domains of participation. They also reported that many PROMs measure a mixture of concepts (e.g., activity, symptoms) and participation [31, 40]. Some outcome measures evaluate disability in hand injured patients, which contains some items that measure participation; however, there is no specific outcome measure to evaluate participation [13, 41, 42].

The PBQ is a valid PRO that can be used to assess participation across the different domains of ICF, including domestic life, interpersonal interactions and relationships, major life areas, community, social and civic life, role, supporting others, domestic life, and subjective participation. As the PBQ contained items based on the patients’ perspective, it has items related to subjective participation and satisfaction [30, 43–45] and personal preferences in participation [29], which are not considered by ICF in participation coding. The factor analysis of the PBQ showed that it is better to define participation using four separate subscales (social participation and interpersonal relationships, autonomy and role, subjective satisfaction with participation, recreational, sport, and leisure time) rather than as a single summative score.

Additional psychometric analyses using newer techniques, such as those drawn from Rasch or item response theory, to evaluate the unidimensionality of each subscale, targeting and different item functioning, and item difficulty is warranted. Evaluating the longitudinal Validity of the PBQ and its subscales is needed to know the responsiveness of the PBQ in detecting changes over time.

There are some key limitations to interpretation that should be observed. First are the missing items and uncomplete data sets. While most studies would be satisfied with greater than 80% retention, it is worth recognizing that uncompleted items represent a small but important subset of the information. As the data were collected via questionnaire and each participant independently completed the questionnaire, it was impossible to determine the reasons for skipping items. We considered lack of attention, unclear wording, or participants’ discomfort in answering certain items as possible reasons for missing items. Our considerable early work on the first version of the PBQ, which included cognitive debriefing and confirming its content validity, would suggest that unclear wording is an unlikely problem. Another limitation of this study is that PBQ is a multi-dimensional instrument that is scored on an ordinal scale which is difficult to interpret. The content range is unclear, and caution must be considered when interpreting scores.

Supporting information

Acknowledgments

Dr. Joy MacDermid was supported by a Canadian Institutes of Health Research Chair in Gender, Work and Health and the Dr. James Roth Chair in Musculoskeletal Measurement and Knowledge Translation.