The Participation Assessment with Recombined Tools-Objective (PART-O): measurement properties of the Norwegian version after traumatic brain injury

ABSTRACT Objective To translate and evaluate the validity of the Participation Assessment with Recombined Tools-Objective (PART-O) in a Norwegian context. Methods One hundred and twenty persons with TBI with verified intracranial lesions and persistent symptoms lasting more than 2 years, included in a randomized controlled trial, rated their participation using the PART-O at baseline. The PART-O with its three subscales (Productivity, Out and About, and Social Relations) was translated to Norwegian. Descriptive statistics, Cronbach's alpha, Rasch analysis, and correlation analysis were applied. Results The Rasch analysis indicated a unidimensional construct of PART-O and its subscales (χ2 < 12.69, p > 0.28). The internal consistency was moderate (Cronbach's alpha 0.48–0.52) and there was a need to reduce scaling options for most of the items. The Out and About and Productivity subscales had considerable floor effects. PART-O showed moderate positive correlation to TBI-related quality of life and global functioning. Conclusions PART-O and its subscales reflect unidimensional aspects of participation. In the present Norwegian TBI population the original scaling of PART-O was too detailed for all subscales. The floor effects and suboptimal targeting between items and subjects participation level of the Out and About subscale is a matter of concern.


Introduction
Traumatic brain injury (TBI) represents a huge global health problem and is considered a chronic disease process with dynamic and life-long impact on health and well-being (1).The long-term consequences of TBI involve a wide range of symptoms that relate to physical, emotional, cognitive, behavioral, and psychosocial functioning resulting in challenges with participation in daily life and society at large (2,3).
Persons with a TBI consistently report the importance participation with respect to the perception of being valuable and feeling like an included and contributing member in their communities and society at large (4,5).
In the International Classification of Functioning, Disability and Health (ICF) (6), participation is conceptualized in major life areas including work and education, interpersonal interactions and relationships, as well as social and civil life.These areas are also reflected in most of the measurements developed to capture participation (7).Hence, participation is an important rehabilitation outcome, but operationalization of the different aspects of participation remains challenging.
One tool developed to measure participation outcome in the TBI population is the 17-item Participation Assessment with Recombined Tools-Objective (PART-O).This tool is developed based on the ICF and targets challenges specific to the TBI population (8).The PART-O is included in the Common Data Elements by the National Institute of Neurological Disorders and Stroke for assessing outcome in Social Role Participation and Social Competence in the TBI population (9).Its measurement properties have been validated and refined (10).As a recommended measure of the common data elements, the PART-O allows comparison across studies.Participation is assumed to be a distinct construct although related to both functioning and health-related quality of life (7).The participatory challenges may be heterogeneous and highly individual.The present study was part of a randomized controlled trial evaluating individualized goal oriented rehabilitation (11) applying Target Outcomes to capture their individual problem areas (12).Participation was one of four main domains nominated as Target Outcomes (12) and render a possibility to associate the individual problem areas to PART-O.Furthermore, PART-O has up to now been most widely used in US populations, and participation is assumed to be influenced by cultural factors (13).
Thus, the aim of this study was to translate PART-O to Norwegian and evaluate its measurement properties in a Norwegian context.

Design and setting
Cross-cultural adaptation and psychometric testing of the PART-O and its three subscales was conducted as part of a larger randomized controlled trial evaluating the effectiveness of home-based rehabilitation in chronic TBI, with patient recruitment being conducted at Oslo University Hospital (OUH), the Regional Trauma Referral Center for Health South East, from 2018 throughout 2020.The baseline assessment was conducted at OUH in an outpatient setting.The trial was prospectively registered with Clinicaltrials.gov(NCT03545594).Written informed consent was provided by all participants and the study was evaluated by the Regional Committees for Medical and Health Research Ethics 2017/1081 and approved by the Data Protection Office at OUH registration number 2017/ 10390.

Participants
Subjects included in a randomized controlled trial (11) with the following inclusion criteria participated; having a TBI diagnosis with radiologically verified intracranial abnormalities, aged 18-72 years, time since injury ≥2 years with presence of TBI-related difficulties, and living at home in the Eastern part of the Norwegian Health Region South-East.Exclusion criteria were severe neurological or psychiatric illness that would confound outcome measurement, inability to cooperate in the intervention (e.g., severely reduced awareness), inability to provide informed consent, insufficient fluency in Norwegian, or ongoing violent tendencies or substance abuse that would put study personnel at risk.See Figure 1 for the inclusion process.

Procedure
The baseline assessments were conducted at an outpatient clinic at OUH throughout the study period, i.e., the assessment procedures were similar before and during the COVID-19 pandemic.The baseline assessment was conducted before the pandemic for 82% of participants.
Determining persistence of TBI-related symptoms was based on the interview and the standardized questionnaires administered at baseline (11).

Demographic and injury related characteristics
Time since injury and mechanism (transport/fall/sport/violence/other), and lowest unsedated acute Glasgow Coma Scale (GCS) score, was collected retrospectively from the medical records.
Age, gender, relationship, years of education and if the subject was partipating in paid work was recorded.Level of functioning at inclusion was assessed by the Glasgow Outcome Scale-Extended (GOSE), and persistence of TBI-related problems was also determined according to the GOSE interview and the nomination of their main ongoing TBI-related problems (Target Outcomes).Overview of the demographic and injury-related characteristics of the participants is provided in Table 1.A total of 120 subjects, 71% males (36% mild, 16% moderate and 48% severe based on the lowest GCS within the first 24 hours after injury) and median Glasgow Outcome Scale Extended score of 6 (IQR 5-7) were included.

PART-O
The revised version of PART-O with 17 questions with its three subscales Productivity (3 items related to paid employment, academic activity, and homemaking), Social Relations (7 items related to social activity and intimate relationships) and Out and About (7 items related to leisure and civic activities in the community) was translated to Norwegian (Appendices 1 and 2).The initial translation was conducted by two Norwegian neuropsychologists with extensive experience in neurorehabilitation (authors SLH and ML), where disagreements were resolved by consensus.An authorized translation service then completed the back translation without access to the English original version.A consensus meeting was thereafter held between the Norwegian group and the translator, thus establishing a consensus-based back translation.The final Norwegian version was sent to the authors of the English original version (Whiteneck, G.G.), along with the backtranslation, allowing them to critically review and approve the translation.Minor cultural adaptations were performed in the Out and About subscale, item 9, where going to café was added to visiting a restaurant to better fit the Norwegian social culture ('In a typical month, how many times do you go to a café or eat in a restaurant?').Norwegians might dine out less frequently but are much more likely to go out socializing at a café.
In accordance with the original version (10) (Appendix 1), the participants scored the items from 0 to 5 based on the frequency option given for the items, with higher scores reflective of greater levels of participation.The exceptions were item 8 (Out and About subscale) which was scored 0, 1.25, 2.5, 3.75, and 5, based on how many times the participants went out of their house in a typical week, and items 15, 16, and 17 (Social Relations subscale) which were scored 0 (No) or 5 (Yes).For analytical purposes, item 16 was set to score 5 for participants reporting to live with their spouses.The mean score for each subscale along with a total score based on the mean score of the 3 subscales were calculated.

Additional questionnaires
Target Outcomes was adopted from a study by Winter and colleagues (12,14).Participants were asked at baseline to identify their three main ongoing problems related to their TBI.After nominating the problem areas, they rated the current severity by defining the degree of difficulty in handling each of the problems on a Likert scale from 0 to 4 (0 = not difficult at all, 4 = extremely difficult).The mean severity score per participant was calculated by dividing the sum of the severity scores by three.
The Quality of Life after Brain Injury Overall Scale (QOLIBRI-OS) consists of 6 items scored on a Likert scale from 1 (not at all satisfied) to 5 (very satisfied) (15).Total score is calculated by summing all the responses, and then dividing by the actual number of responses and subsequently converted to a 0 (worst possible quality of life) to 100 (best possible quality of life) scale.The six items comprise physical condition, cognition, emotions, function in daily life, personal and social life, current situation, and future prospects.

Statistical analysis
Descriptive statistics were applied for assessment of the level of participation and Spearman Rank Order correlation to assess the association between Part-O and whether the subjects were included before or during the COVID-19 pandemic.

Internal consistency
The internal consistency of the PART-O was evaluated with Cronbach's alpha and based on raw data and rescored and transformed Rasch based values.
All items were analyzed regarding their response-level thresholds.If the thresholds were disordered, i.e., the score levels did not separate the level of the underlying construct, the responses were rescored (17).
Local dependency of the items was evaluated using a correlation analysis of the residuals of the items.A coefficient of 0.3 was chosen as the threshold value to indicate that the responses of two different items were dependent upon each other (17).
Fit to the Rasch model was investigated for the items and the participants.A final summary fit for all 17 items in PART-O and for each of the three subscales was also provided.The fit of the persons and items were reported on a logit scale with mean and SD, Mean logit of 0 and an SD of 1 represent an optimal fit to the Rasch model.The fit of the items was statistically evaluated using standardized residuals and Chi-squared statistics.Item residuals below −2.5 or above 2.5 and a nonsignificant Chi-squared probability value were considered to indicate an adequate fit to the Rasch Model (18).The overall summary fit of PART-O and its subscales were also evaluated by Chi-square statistics, where a non-significant probability value indicates a fit to the Rasch model (18).Differential Item Functioning analyses (DIF) was assessed by analysis of variance for each item, comparing scores across age, above and below 45 years, and gender (19).
Verification of unidimensionality was undertaken by creating two subsets of items representing the items with the most positive and most negative residuals according to a Principal Component Analysis.Person estimates for each of the two subsets were calculated, and compared by paired sample t-tests (20).Similar estimates indicated the unidimensionality of the underlying construct.The percentage of t-tests with p-values below 0.05 and the corresponding Confidence Intervals (CIs) were reported.The recommendation for a unidimensional construct is that the CI should include 5% (18).The targeting, i.e., the match of participation level between subjects and items of the PART-O, was evaluated by examining the hierarchical distribution of the items/statements and their response levels and was compared to the distribution of the patients along the same metric scale.
Concurrent validity was explored by the association to the health-related quality of life (QOLIBRI-OS), to the severity of Target Outcomes assumed to partly reflect individual participatory challenges and to functional level (GOSE) reflecting social and productivity challenges.A correlation coefficient below 0.4 was considered low, 0.4 to 0.6 moderate and above 0.6 high (21).
The statistical analyses were performed with IMB SPSS v 28 and RUMM 2030 (RUMM laboratory, Perth, Australia).A significance level of p < 0.05 was applied.

Participation level
Mean score on the PART-O total was 1.86 (± 0.54).Highest degree of participation was found on the Social Relation subscale 2.63 (± 0.91), followed by Productivity subscale 1.48 (SD 0.88) and the Out and About subscale 1.47 (± 0.51).There were no significant associations between scores on the PART-O and its subscales and assessment before or after the COVID-19 pandemic (rho < 0.1, p > 0.18).At the item level, only use of internet for communication (rho = 0.54, p < 0.001) was associated with baseline assessment completed after the COVID-19 pandemic.

Score distribution
Distribution of scores were highly skewed with floor effects for several items, see Table 2. Particularly, many participants reported low level of participation in the Productivity and the Out and About subscales, with going out of the house (item 8), shopping (item 10) and sport activities (item 11) as exceptions.

Internal consistency
Cronbach's alpha for the internal consistency of PART-O was 0.54 (CI 0.47 to 0.69), thus not reaching the recommended 0.7 level (22).Cronbach's alpha for the Productivity subscale was 0.06 (CI −0.27 to 0.32), for the Social Relations subscale 0.52 (CI 0.38 to 0.64), and for the Out and About subscale 0.48 (CI 0.32 to 0.61).

Rasch analysis
Rasch analysis revealed that the original 6-point scoring was too detailed to discriminate between the scores for 11 of the items and thus had to be rescored with fewer scoring options (Table 3).
After rescoring of the items, PART-O fit the Rasch model (χ 2 = 34.53,p = 0.44).There were no misfitting items.The scoring intimate relationship (item 16) is linked to having a spouse (item 15), thus the high residual correlation (0.65) between these two items was expected.No other items revealed residual correlation above 0.3.Cronbach's alpha was 0.60 for the total PART-O score, indicating that rescoring of the items improved internal consistency, yet, remaining below the recommended value of 0.7.Paired t-test revealed that 7.5% of the tests had p-values below 0.05 with a CI of 3.3% to 14.5%, supporting a unidimensional construct of participation.Mean person location was 0.38 (SD 0.70), indicating that the items are constructed for higher level of participation than the level revealed in the present population (floor effect) (Figure 1).No invariance across age or gender was revealed.
Fit to the Rasch model of the Productivity subscale was indicated (χ 2 = 7.42, p = 0.28), but due to only three items being included in this subscale, the power of the analysis was very low and Cronbachs alpha 0.11.Mean location of persons was −0.52 (SD 1.08), indicating suboptimal targeting with participants revealing lower level of participation than its items (Figure 2).
The Social Participation subscale fit the Rasch model (χ 2 = 12.69, p = 0.55).Paired t-test revealed that 2.5% of the t-tests had p-values below 0.05 with a CI of 1.0% to 8.5%.Cronbach's alpha was 0.58, and location of persons 0.45 (SD 1.03) indicating higher level of participation of persons compared to items for this subscale (Figure 2).
The Out and About subscale also fit the Rasch model (χ 2 = 11.80,p = 0.62).Paired t-test revealed that 0.8% of the tests had p-values below 0.05 with a CI of 0.2% to 5.5%.

Concurrent validity
The PART-O total score showed moderate correlation with QOLIBRI-OS and GOSE and, and low negative correlation with mean severity of Target Outcomes (Table 4).The Productivity subscale showed high correlation with GOSE, whereas the Out and About subscale revealed low correlation with all measurement.

Discussion
The present results support the unidimensional construct of PART-O and its subscales.However, major challenges with targeting and floor effects of the Out and About subscale were revealed.The tool behaves similarly across different ages and gender.The moderate correlation with health-related quality of life supports a unique concept of participation.The low correlation with the individual problem areas underscores the discrepancy between individualized and standardized outcome assessment, and the individual profiles not necessarily being related to the same construct of participation.
The gender distribution of the present population is representative for the TBI population (23).Requiring verified intracranial injuries reduces the mild injury predominance generally observed in TBI populations ( 23), yet rather similar to the population included in the development of PART-O (8).In accordance with previous studies, a unidimensional participation construct of the PART-O is supported (10).The constructs of Productivity, Social Participation and Out and About are also supported.One should be aware that the power of the Rasch analysis of the Productivity subscale with only three items is a concern.Participation is important for the quality of life (24) of persons with TBI, but the moderate correlation to QOLIBRI-OS also legitimates that participation should be measured as a separate construct.The lower correlation with the mean Target Outcome severity is not surprising as participants could nominate problem areas unrelated to participation (12).The high correlation between the Productivity subscale and GOSE was expected due to the strong relationship between functional level and participation in working life (25), but functional level does also influence social and overall participation in life situations (26).
The highest degree of participation was found on the Social Relation subscale, followed by the Productivity subscale and the Out and About subscale.The scoring was skewed toward low participation and floor effects (>15% scoring the lowest value) for most of the items in the latter two subscales.For the Productivity and Out and About  subscale, the targeting between items and the subjects was also suboptimal, with the items seeming to be better fitted for participants with higher level of participation than the participants in our sample.This is a challenge from a measurement perspective rendering the scale less able to distinguish participants with low participation levels, as well as capture decline in participation.This challenge is not only restricted to the present population and the PART-O instrument, but may be a more universal challenge in measurement of participation (27).On the other hand, ceiling effects were not detected, and the Part-O subscales thus should be expected to capture improvement in participation.
The observed floor effects may be related to the present population representing subjects with persistent symptoms two years after TBI with intracranial lesions.Persons with severe, moderate, and mild TBI as defined by GCS in the acute phase participated, but the usual dominance of milder TBI was less prominent in the present population which may contribute rather low participation level.However, the measured low participation level may also reflect a cultural challenge with the PART-O.Only 12% of the Norwegian population participates in religious activities (28).The mean annual number of cinema visits per person in Norway is 2.1 (29), and most Norwegians do not go to sport events as spectators.This effectively renders almost half of the items in the Out and about subscale void for most of the Norwegian population.Hence, the floor effects may be related to cultural issues and not TBI severity.However, also in the US population (30) several items in the Out and About subscale were not closely connected to the overall participation over time.
Uneven distribution of item scores was found by Wen et al. (27) in TBI veterans with moderate to severe TBI.Uneven score distribution with disordered thresholds, i.e., lack of discrimination between scores for several items were found also in the present study.The present results support the suggestion by Malec et al. (10) to reduce the scoring options to three levels.Establishing an overall productivity item integrated in the Social Participation subscale may also be a good solution for the Norwegian PART-O.From a Norwegian perspective, slightly changing the contents of the Out and about items may be needed, for example watching sports events on TV with family and friends.Further, the item investigating the frequency of going to the movies could also benefit from including a broader specter of cultural attendance, such as going to the theater, a concert, a show or museum.Overall, in order to accommodate cultural differences in participation, a more inclusive description of activities may be justified.
The present study was undertaken during the Covid pandemic, but 82% of the participants responded to baseline assessments before the pandemic caused a shutdown of society.The only item associated with assessment under the pandemic was use of internet for communication, which increased.Baker-Sparr et al. (31).found significantly lower internet and social media use in TBI patients compared than the general population.However, the majority of internet users with TBI had a profile account on a social networking site (79%).Hence, there may be a potential to facilitate social engagement in the TBI population with increased use of these platforms.

Limitations
One should be aware that only patients with documented intracranial findings, i.e., severe, moderate, and mild complicated TBIs with sustaining symptoms after two years, which may influence the participation level.The sample size is limited and the study was conducted in a Trauma Referral Center in Norway's south-eastern region, where the study setting and the participants' sociodemographic characteristics may not be generalizable to other populations or phases.

Conclusions
The unidimensional construct of PART-O and its subscales was supported in a Norwegian population, and PART-O was robust across age and gender.However, concerns were highlighted regarding the scaling and floor effects on multiple items, particularly the Out and about subscale.There may be potential for improvements, including a reduction of scoring options and changing the content of some some of items to be more robust for cultural differences.To find the optimal adjustments, the validity of PART-O should be examined in several different cultures before concluding on changes to the measurement tool.

Figure 2 .
Figure 2. The distribution of the items and person scores for PART-O total score and subscales along the Rasch calibrated metric scale.

Table 1 .
Demographic and clinical characteristics of the participants (n = 120).

Table 2 .
Distribution of scores (%) for the 120 participants, for each item.

Table 3 .
Original and revised scoring options for the 17 items of PART-O.

Table 4 .
The correlation (Spearmans rho) between PART-O and TBI-related health-related quality of life (QOLIBRI-OS), mean target outcome severity and functional level (Glasgow outcome scale-extended, GOSE).