Background
Since the World Health Organization (WHO) published the International Classification of Functioning, Disability and Health (ICF), interest in better understanding and assessing the “participation” construct has been a frequent topic in rehabilitation literature. In the ICF, participation is defined as “involvement in life situations”, or the result of a complex relationship between a person, his or her health condition, and the environment [
1,
2]. A person’s participation is acknowledged as an important outcome of the rehabilitation process, even though several researchers have identified the need for a better conceptualization of the participation construct in order to build better measurement tools [
3‐
6].
The rehabilitation process encompasses a set of procedures aiming to assist individuals who experience or are likely to experience disability to achieve and maintain optimal functioning in interacting with their environments [
7]. Thus, information regarding a person’s functioning, including participation, is necessary to guide rehabilitation planning and assess the impact of intervention. Recent systematic reviews have identified a variety of participation measures available in the literature [
8,
9]. However, most of these participation tools were developed in high-income countries and may not be suitable to represent the experiences of those living in less developed societies [
10,
11].
In an effort to better describe the participation of people living in middle-income and low-income countries, van Brakel et al. (2006) proposed the Participation Scale (P-Scale) – an instrument that was designed to assess the participation of individuals with a health condition or disability, especially conditions associated with stigma and discrimination [
10]. The P-Scale aims to quantify the restrictions perceived by the individual in eight of the nine major areas of life defined by the ICF: learning and applying knowledge; communication; personal care; mobility; domestic life; interpersonal interactions and relationships; major areas of life; and community, social and civic life [
10]. An innovative characteristic of the scale is that the individuals are asked to compare themselves with a real or hypothetical “peer” – that is, someone who is similar to them in all respects, except for illness or disability. This comparison was proposed to allow the representation of the roles and expectations for participation in different social and cultural contexts [
10]. These special features indicate that the P-scale might be useful to assess clients’ participation restrictions in diverse life situations.
In Brazil, the public rehabilitation services are structured in an integrated network, organized with multidisciplinary teams across three levels of care (basic, specialized, and hospital) [
12]. In addition to the diversity of professionals and services, the rehabilitation networks have to address diverse patient profiles because the services provide assistance to people of a wide variety of health conditions, socio-demographic backgrounds, and functional needs [
12‐
14]. In this clinical context, information about patients’ function, including participation restrictions, may help service planning and better direct investments in rehabilitation [
15,
16]. Therefore, it would be helpful to investigate the P-Scale properties and its suitability to support rehabilitation services in the country.
In previous validation studies, the P-Scale showed good psychometrics properties and was found to be valid for use in several different health conditions and cultural environments [
10,
11,
17,
18]. All these studies, however, used Classical Test Theory (CTT) procedures, and, to the best of our knowledge, there is no study that assessed the P-Scale properties using Rasch analysis. Classical Test Theory has a number of limitations, including sample dependency (the item and scale statistics apply only to a specific group of respondents) and the assumption of item equivalence (individual items are treated as being equally difficult) [
19,
20].
On the other hand, Rasch analysis transforms ordinal data (i.e., ratings with non-equal intervals) into linear measures with equal-interval units called logits, which are used to describe the measures of both individuals and items [
21]. The transformation of raw scores into an abstract linear continuum of ability (for individuals) and difficulty (for items) allows one to predict the likelihood of a person choosing, for example, “yes or no” on a specific functional item [
22]. Thus, one is able to identify the location of each item on a continuum of ability and compare where the person’s level of ability is located on the same continuum.
Once the person and item measures are described using the same “logit” unit, Rasch analysis allows for the comparison of a person to other individuals, one item to other items, and individuals to items [
21]. Furthermore, the Rasch model can be used to build new scales, to suggest improvements to existing scales and to estimate the stability of item difficulty estimates among different groups, thus allowing for comparisons of homogeneous measures [
22]. The aims of this study were to use Rasch analysis for the following:
-
to assess the P-Scale items in terms of their item and person fit, dimensionality, item difficulty, reliability, and Differential Item Functioning (based on gender and duration of the present symptoms);
-
and to examine whether the Brazilian-Portuguese version of P-Scale is suitable to assess the perceived ability to take part in participation situations by patients in a rehabilitation services network who have diverse levels of function.
Discussion
This was an initial study that investigated the psychometric properties of the P-Scale using Rasch analysis to examine whether the P-Scale may be a suitable tool to collect data regarding participation among the patients from a Brazilian public rehabilitation services network.
The participants were adults who were in treatment or seeking rehabilitation services, with diverse health conditions, socio-demographic backgrounds and functional needs, and from different services of the rehabilitation network. These sample features corresponded to the usual daily variability found in the services.
The findings are encouraging, especially when we recognize the challenges defining the construct of participation, as well as developing good measurement tools to capture it. There is consensus in the field that participation is a multifactorial construct that is related to an individual’s physical and social environments, personal factors, and health conditions [
1,
3‐
5]. Such complexity makes participation a difficult construct to assess during a rehabilitation process, although involvement in real life activities is the final goal of most people receiving rehabilitation care. The P-Scale was found to be helpful in this particular rehabilitation context, despite some issues that should be addressed for further improvement of the scale.
Rasch analysis was performed using a modified rating scale, collapsing the original categories into a dichotomous scale since some items did not have a sufficient number of answers in all response categories [
31]. This low number did not seem to be an issue related to the sample size, as the item separation index confirmed that the number of participants in our study was sufficient to test the scale items [
26]. A low rate of response to some categories can occur for reasons besides sample size, such as when the respondents have difficulty distinguishing between similar categories. Further research with the P-Scale is needed to confirm whether some categories on the original rating scale are actually underused by the respondents.
After collapsing the rating scale to a dichotomous scale, we were able to perform the next steps of the analysis. In general, the P-Scale showed a good fit to the Rasch model with only one item (
N15 – Start or maintain a relationship) showing misfit, which was thus removed in the final analysis. The low rate of misfitting items is a good indicator of unidimensionality [
28]. Additionally, the dimensionality analysis performed using the Winsteps also showed that the scale could be considered unidimensional for the purposes of the Rasch analysis. Although the variance explained by the measures reached a low percentage (approximately 35%), it was closely matched to the variance expected in the model. The variance explained by the residuals in the first contrast was below 2.0 Eigenvalues, indicating the low likelihood of a second dimension [
26].
Even using a modified rating scale format, the P-Scale can be useful as a screening tool for participation problems reported by patients in order to lead rehabilitation professionals in addressing such problems in the patients’ recovery process. In the analysis, the items were well spread along the continuum of difficulty, with just four items overlapping at the same measure of difficulty (N3 – Contribute economically to household and N9 – Opportunity to take care of yourself; and N13 – Visit public places in village neighbourhood and N17 – Comfortable meeting new people). Thus, the items were able to show where most of the patients were located on this difficulty continuum, according to their ability to participate. However, as seen in the person-item map, there are some gaps between the items. In the same way, there were no items covering the top of the continuum, in which the best performing patients were located (approximately 37% of the sample).
The gaps and lack of items to represent the patients with higher participation abilities may explain the person separation index found in this study (1.51). This index is influenced by factors such as the length of the scale, number of categories per item, and the match between the items and the ability of the respondents (i.e., sample-item targeting) [
26]. Although the person separation index found in this study indicates that the items were enough to discriminate the sample in two groups (low and high ability to participate), a higher index value is desirable for more sensitivity and improved identification of different levels of ability among the respondents [
21,
26]. Because the P-Scale was first developed to assess the participation of people with disabilities related to stigma [
10], some items (e.g.,
N8 – Same respect in the community) seem to be more useful for individuals with severe restrictions than for those with low or moderate restrictions. In this sense, it might be useful to expand the scale by adding some items that could better identify participation restrictions among individuals with better functioning.
In an effort to make the P-Scale more suitable to people with different health conditions and disabilities, the scale authors removed one item from the former version [
10] –
N16 – In home, are the eating utensils you use kept with those used by the rest of the household? – because this item was considered more appropriate for individuals with infectious health conditions, such as Leprosy. It is expected that the new item added in the latest version [
24]:
N10 – Do you have the same opportunities as your peers to start or maintain a long-term relationship with a life partner? will be more relevant for diverse health conditions. However, because only a portion of the patients in the present study sample answered the latest P-Scale version, we were not able to include this new item on the analysis. Thus, it would be valuable if future research include this item as it potentially can contribute to fill some of the gaps found on the continuum.
The analysis of DIF demonstrated that item difficulty varied between men and women in two items (N2 – Work as hard your peers do and N14 – Household work), while two other items (N6 – Take part in casual recreational/social activities and N12 – Move around inside/outside house/village/neighbourhood) showed differences between people with acute symptoms and people with chronic symptoms. These variations may be attributed to different engagements of each group in the situations described by the items due to cultural factors. However, before suggesting any changes on these items, further investigation should be carried out to clarify why the respondents answered differently to those situations and whether DIF has a significant impact on the overall score.