Initial psychometric validation of the questionnaire on pain caused by spasticity (QPS)

This was a prospective, observational, one-arm, psychometric validation study of a health outcome measure. The a priori construct validity hypothesis of the study was that patients with greater disability or spasticity would experience a greater level of SRP and a lower QoL.

Children or adolescents with CP and their parents/caregivers were recruited between November 2013 and June 2014 from three hospital sites in China: the Foshan Nanhai Affiliated MCH Hospital of Guangzhou, the Hunan XiangYaBoAi Rehabilitation Hospital, and the Dongguan MCH Hospital. Sites recruited from both their inpatient caseload and their outpatient caseload. The study adhered to the ethical principles for human research and approval was provided by the ethical review board at each participating hospital. Each patient or their legally acceptable representative provided written informed consent to participate in the study.

Male or female patients were included in the study if they were between 2 and 17 years of age with either uni- or bilateral CP, UL and/or LL spasticity, and intermittent SRP in either the ULs or LLs on a weekly basis in one of the following clinical patterns: pes equinus, adducted thigh, flexed knee, flexed elbow, or flexed wrist. Patients were also required to have an Ashworth Scale (AS) score of ≥1 in at least one of the pre-defined clinical patterns and were receiving continuous anti-spastic treatment and/or medication. Key exclusion criteria included fixed contractures, predominant forms of muscle hypertonia other than spasticity, constant SRP over very long time periods, LL and/or UL surgery within the past 12 months, and an indication for orthopedic surgery within the next 2 months.

The study (No.: MRZ99901_0027_4) comprised two clinic visits, a screening/baseline visit (V1/V2) and an end-of-study visit (V3) scheduled for 21 (±3) days later. The QPS was administered at the baseline (V2) and end-of-study visits (V3) at site. Based on the pre-screening of the cognitive, communicative, and motor abilities of a patient, the QPS was self-administered by the patient, or the interviewer-administered module was used by site staff. If no assessment by the patient was possible, due to age or cognitive impairments, only the parent/caregiver module was completed. The parent/caregiver module of the QPS (ObsRO) was completed for each child by the same parent/caregiver throughout the study. The parent/caregiver had regular contact with his/her child/adolescent so that he/she could report reliably on the observed SRP behaviors. Alongside the QPS assessment, the clinic staff responsible for the QPS assessment procedure were asked to complete a questionnaire documenting ease of use of the QPS (e.g. difficulties in selection of modules for children) and any administration issues (e.g. encountered problems and positive experiences).

In addition to the QPS, a number of standard clinical evaluations for the study population and for spasticity were performed during the study. As standard classifications in the CP population, the Gross Motor Classification System – Expanded and Revised [GMFCS] and the Manual Ability Classification System [MACS]) were used [27‐29]. The GMFCS is a classification of general motor independence with five classification levels (Level I: Walks without limitations – Level V: Transported in a manual wheelchair). The MACS is a classification for CP children with five levels on the self-initiated abilities to handle objects and their need for assistance or adaptation to perform manual activities (Level I: Handles objects easily and successfully – Level V: Does not handle objects and severely limited ability to perform even simple actions). For evaluation of motor function, the Gross Motor Function Measure-66 [GMFM]) was used, which utilizes 66 functional tests that span the spectrum from activities in lying and rolling up to walking, running and jumping skills [30]. Cognitive ability by means of cognitive age was based on the standard tools used by sites (e.g. Gesell Developmental Schedule, Sign-Significant Relations Rehabilitation Rating). The level of spasticity in spastic joints in this study was determined with the AS and Modified Tardieu Scale [MTS]) [31‐33]. Spasticity with the AS is assessed during passive range of motion of a joint on a 5-point response scale (‘0 = ‘No increase in tone’ - ‘4 = ‘Limb rigid in flexion or extension’). For the MTS, the spasticity angle was documented, which is the difference between the passive full range of motion angle with slow assessment velocity minus the fast stretch speed angle of a spastic joint. The Pediatric Quality of Life Inventory™ (PedsQL™) was used to capture information on patient QoL [34, 35].There are four sub-scales that refer to physical functioning (8 items), emotional functioning (5 items), social functioning (5 items), and school functioning (5 items). The association of the QPS to overall Quality of Life was of main interest, and hence the total score of the PedsQL™ was the focus for the analyses. Patient demographic data and CP history were also recorded.

For all assessment tools, the official Chinese language versions were used during the study. The Chinese version of the QPS was established using a standard linguistic validation process for developing new language versions for data collection measures [19, 36]. The process included forward and backward translation steps, and cognitive interviews with Chinese children with CP and SRP and their parents/caregivers. Finally, an international harmonization step was used for quality control and finalization of the translations.

The demographic characteristics of the children/adolescents and their parents were summarized using descriptive statistics, controlling for the actual QPS modules used (LL and/or UL) and whether the QPS was self- or interviewer-administered.

To evaluate the individual QPS items, responses were evaluated by means of standard item-reduction statistics (e.g., ceiling/floor effects, missing data, item-to-item correlation, corrected item-total correlation (0.40 threshold) [37]. Factor analyses (principal component analysis with varimax rotation using Kaiser normalization) were conducted to assess scale dimensionality with factor loadings of at least 0.40 as a cut-off [38]. Rasch Measurement Theory (RMT, one-parameter logistic model, RUMM2030 software) [39] models were utilized to review item functioning of tested constructs. Due to small sample sizes, RMT models were only run for the sample of children with LL SRP using combined results from self-reported and interviewer modules. Based on these results, the final QPS items were selected and a scoring algorithm was created. The following validation analyses of the QPS were then performed based on the final scoring approach.

Internal consistency of the QPS was evaluated by Cronbach’s α analysis using 0.70 as the desired threshold [40]. Test–retest reliability between both visits (V2 and V3) using total score and single items, was demonstrated by intraclass-correlation coefficient (ICC 2,1) estimates, using 0.70 as the desired threshold [41].

Convergent/discriminant validity was investigated by Pearson correlations of QPS variables with other measures of interest, such as GMFCS, AS and MTS scores, GMFM, and PedsQL™ using associations of at least 0.30 as desired level of association. Construct validity of the QPS was tested by comparison of reported SRP levels in relation to activities and differences in important demographic groups in the study population such as the GMFCS or MACS categories. Specifically, higher levels of SRP to QoL based on the available studies were expected [6‐8]. Also higher SRP levels were hypothesized to be associated with higher degrees of spasticity (AS, MTS) and disability (GMFM, GMFCS, MACS), but prior studies have indicated not always clear associations in this regard [4, 8, 14]. In addition, ordinal logistic regression analysis models were run with a set of potential predictors for identification of important SRP-related factors. All analyses, with the exception of RMT, were conducted using SPSS [42].

A total of 137 children/adolescents with CP and their parents/caregivers were enrolled and contributed to this study, from which all 137 completed it. None of the children had missing QPS data and only a very small amount of QPS data were missing for the parents/caregivers (maximum 3 cases, 2.4%). Some of the clinical assessments had small amounts of missing data (i.e., AS 21 cases, 15%; MTS 20 cases, 14%).

Most children presented with bilateral LL spasticity (117/137, 85.4%). Although 52 children presented with UL spasticity overall, UL spasticity occurred most often together with LL spasticity and very few (10/137, 7.3%) presented with UL spasticity alone (Fig. 2).

The study population covered the full age spectrum (range 2–17 years) with a mean (standard deviation, SD) age of 6.6 (3.2) years. Most children were 4–8 years of age; 68.6% were male and 93.4% were Chinese (Table 1). Children who could self-administer the QPS had a mean age of 7.9 (3.1) years, those who completed the interviewer-administered QPS had a mean age of 5.3 (2.3) years and those who couldn’t use the QPS had a mean age of 3.8 (1.5) years.

The mean (SD) age of the parents/caregivers was 39.6 (11.0) years (range 24.9–69.1 years) and the majority (83.2%) were female. Most parents were described as not employed outside the home (50.4%) or homemakers (19.0%). The greatest proportion of children (33.6%) were Level II on the GMFCS and 35.8% were Level II on the MACS (Table 1). Overall, the study sample reflected a heterogeneous one, as indicated by the GMFCS classification.

To provide information on the ease of use of the QPS in daily practice and identify any administration issues, clinic staff were asked to complete a questionnaire regarding any difficulties. No difficulties were reported in deciding which module of the QPS to use for most (117/137, 85%) children and difficulties related to uncertain cognitive abilities were reported for only a few children (15/137, 10.9%). Furthermore, clinic staff noted no general administration difficulties for most of the children (80/137, 58.4%), although, some difficulties in assigning pain and stiffness using QPS items 1 and 2 were reported (45/137, 32.8% of children), and when defining SRP in general (11/137, 8.0% of children). Based on this experience, for approximately one third of children (49/137, 35.8%), the clinic staff mainly focused on cognitive ability and ability to communicate/express the location of their SRP when deciding on the appropriate module to use.

The mean QPS score patterns of the items that contribute to the QPS are shown in Table 2. Similar mean score patterns were noted for all the different child/adolescent QPS modules with relatively high SRP levels associated with exercise (physical therapy or stretching exercises) followed by the general item of SRP when tight. Low levels of SRP were expressed while at rest (sitting, watching TV, or trying to sleep). Similar to the child results, the greatest amount of pain observed by the parent occurred when the child was exercising. Almost no pain was noted by parents when the children were at rest.

The low SRP levels in the study population for the ‘SRP at rest’ items and in some instances for the ‘SRP doing usual activities’ items, resulted in floor effects for these items. This was also reflected in the item-to-item and corrected item-total correlations (Table 2 and Table 3). Conversely, the pain levels supported the hypothesis that easier activities should be related to lower SRP levels than more difficult activities. Since the qualitative work for the QPS creation of these items was shown to be of relevance for some children and their parents, and the SRP levels followed the hypothesized pattern, the items were included in the subsequent scoring analyses. No ceiling effects were found for any of the QPS items and the entire spectrum of the scale was seen, with high SRP levels being reported in some patients.

Three scoring algorithms were explored with the data to establish a score for the QPS and included testing to include all QPS items and different combinations of SRP items. In the final scoring system, based on optimal results (IRT and factor analyses), the item on general SRP was removed from the overall score, keeping it as an external overall criterion. The four remaining activity-related SRP items (at rest, usual activities, with exercise and when trying to do a specified hard task) were included to calculate the QPS score. This decision was supported by IRT and factor analyses, in which the general item displayed significant misfit, whereas the other items performed well (50–70% of variance). The other items of the QPS contribute additional information relating to the established overall score. In the children modules they are a useful check of response consistency. In the parent modules, the other items provide further information on the localization of SRP and definition of SRP behaviors [15].

Within the UL modules (self-administered child/adolescent and parent), there were still two 2-dimensional results and not the overall anticipated uni-dimensional score. Here, the exercise item contributed to a second factor, which fitted the assumption that exercise is always reported with a high level of pain. Because all the other factor analyses yielded a uni-dimensional structure it was decided to adopt the single-score approach (the pain level for exercise activity was high here as well), and to revisit these issues when future QPS data become available for analyses.

An overview of the final scoring approach and the conceptual framework of the QPS is presented in Fig. 3. Although not contributing to the score, the general item was reported together with the single activity-related items, to allow reporting of the overall SRP profile of a child with CP and as a general external reference.

Furthermore, separate reporting of the items and QPS score was recommended for the six modules to account for the different CP etiologies and age groups (i.e., a higher proportion of self-administered modules than interviewer-administered modules were expected to be completed in older children).

Using the newly defined QPS score, internal consistency (Cronbach’s α) and test–retest reliability (ICC 2,1) were tested (Table 4). Although not presented in detail in this paper, clinical measures such as the AS remained stable between visits, thus supporting an appropriate setting for testing reliability. While Cronbach’s α was lower than expected for the child self-report group, the coefficients for the child interviewer-administered group were adequate (>0.70, 0.78 for LL and 0.91 for UL) [43]. Parent modules also showed acceptable consistency (0.74 for LL and 0.70 for UL). Test–retest reliability was 0.92 to 0.95 for the QPS score, demonstrating ‘almost perfect’ (>0.81) results [43]. Also for single items, test–retest reliability results were ‘substantial’ (0.61 to 0.80) or ‘almost perfect’ between the 3-week interval of the two study visits (V2 and V3).

With regard to construct validity of the QPS score items, as stated earlier, scores were consistently higher for more difficult activities, with the ‘at rest’ item related to the lowest scores and the ‘exercise’ item to the highest scores (Table 2). For further investigation of convergent/construct validity with other concepts of interest, correlations with the QPS score were performed and QPS scores across groups were explored. According to the 5-point AS rating conducted by the clinicians, SRP was widely reported across the spasticity grades in the main muscle group patterns by children and parents. Thus, associations in the hypothesized direction were not found. In contrast, there was a trend towards some moderate-to-high associations in the hypothesized direction between the QPS score and MTS spasticity angle (Tables 5).

For the association between QPS and motor function, as measured by the GMFCS and GMFM-66, some conflicting (child LL self-report group), but also clear relations were seen (child interviewer group, Table 6 and Additional file 1: Figure S1). This was not the case for the parent questionnaire modules and for the association to MACS, which fitted the general notion that SRP was reported across the disability spectrums (Additional file 1: Fig. S1). Contrary to these findings, clear associations were found with QoL, as measured by moderate-to-high associations with the PedsQL™ (Table 6, Additional file 2: Figure S2), which reached significance in the QPS LL category (p < 0.01). Although it should be noted that a smaller proportion of those in the child interviewer groups completed the PedsQL™, the overall pattern showed that there seemed to be a close association between reported levels of SRP and general QoL.

To explore QPS validity further, an exploratory ordinary least squares (OLS) regression model was run on the QPS baseline scores with the following independent variables being included in the model: age, gender, GMFCS, MACS, health state of child (self-report and parent-report), use of assistive devices (yes/no), and therapy (yes/no) [Additional file 3: Table S1]. For ease of analysis and sample size, the self-report and interviewer-administered QPS modules were combined. Most models showed reasonably high variance. Using the child LL model, age, GMFCS, use of assistive devices and therapy had a significant effect on QPS scores, while in the parent LL model GMFCS, overall health of the child, use of assistive devices and therapy had a significant effect on QPS scores (Additional file 3: Table S1). The sample sizes were too small to run the multiple regressions for UL in both the child and parent samples.