Development of the young spine questionnaire

The YSQ was designed as a self-report measure primarily to be used in cross-sectional cohort studies in early adolescence (9 to 11 years) [20]. It is based on the conceptual model of Wilson and Cleary (Figure 1) [21] and items were chosen to fit the symptom and functional status levels on the health continuum in addition to characteristics of the individual and environment (Figure 2). At the symptom status level, we included spinal prevalence estimates (daily, weekly and lifetime prevalence), frequency of pain episodes and pain intensity. At the functional status level, questions regarding restrictions in school and sports activities were chosen. Back pain in family members can relate to both characteristics of the individual and the environment. Parents’ behaviour during an episode of back pain has the potential to influence an individual psychologically (environment) whereas parents’ opinions about back pain may have an effect on symptom experience, personality and motivation, and finally there is a genetic component influencing the characteristics of the individual. For these reasons we linked family influence to both aspects of the model. Finally, an item mapping care seeking behaviour was included even though it did not fit the model. This was incorporated because it is considered an important consequence of back pain with the potential to be a proxy of severity.

Separate items are analysed on an individual basis and no summary scores are generated. Sections 1-3 are core questions of the YSQ whereas section 4 and 5 are optional and can be included or excluded according to the purpose of the study.

Initially, we constructed an item bank consisting of published questions and drawings fitting the conceptual model [10, 18, 22‐25]. The back drawings were adapted from the Standardised Nordic Questionnaire [22] while the prevalence estimates questions came from a questionnaire used in a study of school children 8 to 16 years old [10, 24, 25]. Items regarding activity restrictions, care seeking behaviour and the influence of parental back trouble were adapted from a questionnaire used in the Funen Back Pain study [18].

To measure pain intensity we included a translated version of the “Revised Faces Pain Scale” (rFPS) which has been well validated in children aged 5 – 12 years [23]. We decided that the rFPS was the most suitable for several reasons. Firstly, the rFPS is a commonly used instrument in many trials (used in > 22 trials) which allows for comparing results [26]. Secondly, it is one of six well documented instruments in the targeted age group. In a review by Stinson et al. it was found to have good reliability, content validity, construct validity and responsiveness [27]. Thirdly, contrary to the original FPS, the revised version has enhanced scoring compatibility with other self-rating scales as it uses a common metric (0-5 or 0-10) [23, 28]. Lastly, it has been shown that 85-90% of all children aged 9 or older reliably can provide self-report of pain given an age-appropriate scale under optimal conditions [29]. Permission to use the scale was obtained from the original developers.

As some of the items were developed for use with an adult population each item was carefully scrutinised and adapted to the reading capacity of children aged 9-12 in an expert group consisting of the authors and a professor in pedagogy and compulsory schooling. This iterative process resulted in the first draft version of the YSQ which consisted of 6 sections (YSQ-1).

Section 1-3 covered prevalence estimates and a pain scale relating to the severity of the cervical, thoracic and lumbar spine. At the beginning of each section a picture demarcating the location of the spinal area was included.

Section 4 included three questions measuring the consequences in terms of activity restrictions during sports, school absenteeism, and care seeking behaviour. Because self-reported pain is completely subjective, consequences of pain is often considered to be a proxy measure for severity of the pain, and as the consequences are more important than the pain itself from a societal point of view, we wanted to include this in the questionnaire. As reduced physical activity, seeking health care and staying at home a few days are the most common consequences reported in the literature [10, 30], these became part of the questionnaire.

Section 5 covered the child’s perception of parental back trouble and its consequences. It is important to understand that this is not a measure of the parents’ actual spinal pain, but the children’s perception of it. This was included as several studies have reported a significant relationship between symptom reporting of family members and back pain [11, 30‐32].

Section 6 was two pain drawings (front and back) asking the child to locate areas of pain at other sites.

We devised an iterative method to test and adapt the draft version of the YSQ for item understanding which consisted of two pilot tests (Figure 3). The method was a simplified version of the content validity method for assessing respondent understanding as proposed by Patrick et al. [33].

The study was not reported to the local ethics committee as this is not required according to the rules and regulations of the Danish scientific ethical committee [34].

In the testing phase we calculated descriptive statistics for the children in pilot test I (YSQ-1) and II (YSQ-4). In addition, percentage agreement was used between the prevalence estimates from the questionnaire (YSQ-1) and the interviews for two reasons. Firstly, it highlighted problematic items which subsequently could be adapted. Secondly, it shows how well the question wording corresponded with the children’s own explanation of their problem. Finally, median, inter-quartile range and Spearman’s correlations were calculated between the 11-box numeric rating scale and the rFPS (YSQ-1).

The children participating in pilot test I had a mean age of 10 with slightly less girls (47.2%) compared to boys (Table 1). Based on YSQ-1 a high proportion of the children reported that they had experienced low back pain (47.2%) or neck pain (41.5%) while slightly less had experienced thoracic pain (28.3%) at least once in their life. The 18 children participating in pilot test II also had a mean age of 10 and there were 11 girls (61%). The focus of pilot test II was on the pain drawings for which reason no spinal pain data is available.

The drawings of the spinal areas included in the YSQ were adapted from the Standardised Nordic Questionnaires which originally were developed for the adult population [22]. Our results demonstrate that children at the age of 10 have a different understanding of the boundaries of especially the thoracic and lumbar spine compared to adults. Many found it difficult to localise the transition from the thoracic to the lumbar spine which might be due to immaturity of the children’s anatomical knowledge. Also, most adults have an understanding of the concept “low back pain” which most of the children in this age group have not yet developed. Changing the drawings to include more bony landmarks helped to avoid these problems.

The original drawings also included the buttocks and thus the coccyx. Therefore, the frequent falls (e.g. from play and sports) on the buttocks causing coccygeal pain and/or bruised buttocks would be classified as low back pain. In the original drawing, the buttocks are included to capture pain radiating from the lumbar spine to the buttocks, but due to a high rate of non-spinal buttock/coccyx pain in this age group, it was a necessary trade-off to exclude the buttocks from the drawing.

In all the test phases, the sequence of sections was the same (cervical, thoracic, lumbar). Since only very few of the children had problems identifying the neck area, this may have been used as a reference point improving the identification of the other areas. It is not known, how the drawings of the thoracic and lumbar areas will be perceived if they are shown in another sequence or used individually.

The YSQ contain a frequency question (which can be dichotomised to represent lifetime prevalence) as well as point and 1-week prevalence estimates for all the spinal regions. Pilot test I revealed prevalence estimates of 41.5% and 18.8% (lifetime and 1-week) for the cervical spine, 28.3% and 3.8% for the thoracic spine and 47.2% and 11.3% for the lumbar spine. These estimates agreed well with findings from the interviews which ranged between 85.7-97.9% and we therefore believe that the prevalence questions accurately reflect the children’s perception of pain prevalence. In addition, our estimates were similar to reported lifetime and 1-week estimates in the same age group which ranged from 18.8-51.0% [38] and 23-44.3% [31, 39] in the cervical spine, 9.5-72% [40, 41] and 3.4-51.4% [31, 42] in the thoracic spine, and 7.0-72.0% [6, 41] and 9.5-20.0% [39, 43] in the lumbar spine. To our knowledge no point estimates have been reported in the literature for comparison. On the basis of this, we believe that by asking for specific answer categories and subsequently collapse them for analyses, a more precise estimate can be obtained compared to simply asking “Have you ever had pain?”. However, because the prevalence estimates are based on the first version of the questionnaire and due to the small sample sizes these estimates should be interpreted with caution.

When developing the YSQ we decided not to include a focused lifetime prevalence estimate question due to the risk of “memory decay” where children are more likely to forget ‘episodes’ of spinal problems with the passage of time [16]. Secondly, “forward telescoping of events”, a tendency to recollect episodes of e.g. spinal problems as having happened more recently, are more likely to bias estimates with long recall periods [15]. Instead we included a frequency of spinal problems question which does not focus on a set time-frame. If needed for comparison to other studies, this can be converted into lifetime prevalence by collapsing the three last answer categories. However, we believe there should be more focus on frequency and duration, as a single short period of back pain is less likely to have long term consequences than more long-lasting or recurrent back trouble [17].

For each of the three spinal regions we added the rFPS as measure of pain intensity. To our knowledge the rFPS has not been used in spinal pain studies of children or adolescents, and it is therefore unclear how well the pain scale performs for this condition. However, our results showed acceptable correlation between the rFPS and the NRS (r = [0.67-0.79]) supporting the validity of the rFPS also for spinal pain. Similar results were found in a study of post-operative pain ratings comparing the same scales (r = 0.89) [44].

In the Danish translation of the rFPS we used “rigtig meget ondt” (very much pain) rather than the commonly used adult version “værst tænkelige smerte” (worst pain imaginable) as the latter is considered difficult for children to understand since they have a limited experience of pain intensity [29].

Section 4 of the questionnaire contains questions related to activity restrictions (physical activity and school absenteeism) and care seeking behaviour (treatment received). We included these questions as proxy measures for severity of pain, and in combination with the rFPS they may have the potential to distinguish trivial from significant pain. Our experience from the pilot-tests is that children’s reporting of pain is more spontaneous and immediate compared to adults resulting in a higher frequency of trivial pain being recorded when answering the prevalence questions. This supported our decision to include these measures, so it is possible to some degree to qualify the answers in order to target primary intervention programmes to children with high risk of developing spinal pain in their adult life.

The last section measures the child’s perception and consequences of the parents back trouble. These questions had to be completely rephrased to be very specific, illustrating the need for testing questions in the target population. We believe children to a certain extent will imitate their parents’ behaviour reinforcing the relevance of the information when interpreting the children’s answers to the participation questions in section 4.

After the completion of final version of the YSQ it was decided to include section 4 and 5 as optional since they may not be relevant in studies which primarily focus on prevalence.

The study has several strengths. Firstly, the questionnaire includes all 3 spinal areas and considers each as a distinct region [10]. Secondly, the emphasis on relevant content and content understanding of the target population during the development phase resulted in high content validity of the questionnaire. This was achieved using an iterative process combining ordinary questionnaire data with interview data.

Several drawbacks need to be mentioned. First of all, the questionnaire needs to be tested for reliability. Secondly, the questionnaire has only been tested on school-children in the age range of 9-11 years and cannot be generalised to other age groups. Thirdly, the iterative process and testing of the YSQ was mainly performed on preliminary versions (YSQ-1 to YSQ-3). It resulted in changes to all included items (either the question or the answer categories) and drawings for which reason caution is advised when applying these results to the final version. We recommend that future validation studies include a thorough evaluation of the items in the final version of the YSQ. Finally, we reinforce that the questionnaire has been developed and tested for use in cross-sectional cohort studies. At this point in time it is not recommended to use it in a longitudinal setting measuring change for several reasons: 1) the questions and response options have not been designed for the purpose of being responsive to change, and 2) the questionnaire does not have a summary score and change scores are therefore reliant on the validity of scores on single questions. Future studies need to ascertain whether or not the included questions can be used in a longitudinal design, measuring change.