Study design
The prevalence estimates of BP and care seeking are presented by age and gender based on three cross-sectional studies of a cohort of children studied at the ages of 9, 13 and 15 years. The tracking of BP and care seeking is based on the individual trajectories over time. The three studies are referred to as T1, T2, and T3, where "T" stands for "time".
The flow of the study
The original cohort was sampled for the European Youth Heart Study (EYHS) in 1997 among 3rd grade schoolchildren in the municipality of Odense, Denmark. The main purpose of that study was to investigate risk factors for cardiovascular disease and diabetes in a mixed longitudinal study design. As a part of that study, an interview was conducted to determine whether back problems were present. All interviews at T1 were performed at the schools of the children [8] but due to logistic problems, 110 of the sampled children were not then offered a back interview.
Four years later, in 2001/2, those participating in the study at T1 and still living in the same area were invited to a second investigation regarding BP. At this second time point the children had an MRI of the lumbar spine, were questioned about BP, and measurements of body composition were taken [12]. All children were picked up at their schools by taxi and brought to the Spine Centre of Southern Denmark, Ringe.
The third data collection took place in 2003/2004. At this time, all teenagers sampled for the original cohort were invited to an interview about BP and its consequences. They also had an MRI of the lumbar spine. Furthermore, a number of objective measurements were taken for back performance, body composition, and aerobic capacity. In addition, physical activity was measured objectively over a one-week period. Also this data collection took place at the Spine Centre of Southern Denmark. The children were brought to the centre by taxi or by train, if they had moved to other parts of the country. The flow of participants is shown in Figure 1.
Study population
The 38 relevant state schools in the municipality of Odense (180,000 inhabitants) were stratified according to their location (urban, suburban, rural) and the socio-economic character of their uptake area [8]. From each stratum, a proportional, two-stage cluster sample of children was selected. The primary units (clusters) were the schools. The sampling frame was all the schools in the town, from which schools were selected using probability, proportional to school size. Each school on the list was allocated a weighting equivalent to the number of children enrolled, who were eligible to be selected for the study. The secondary units were the children in the schools. Equal numbers of children in the third and ninth grades (a maximum of 30 individuals from each grade and school) were sampled from each school. Children in the appropriate age bands (8 to 10 years and 14 to 16 years) were allocated code numbers and then randomly selected using random number tables. For this report, only the young cohort is used.
Experience from previous studies of children using a similar method suggested that a likely response rate from schools would be 90%, and that a 75% to 80% response rate from the children or their parents could be expected [13, 14]. Estimates of power were based on the cardiovascular aspect of the study, which is described elsewhere, and a maximum non-response rate of 25% was predicted [11]. This required a minimum of 4 × 250 participants (i.e. 250 children in each age and gender subgroup).
Generalizability
It has been shown that the income of the parents of the children participating at T1 was similar to the rest of the Danish general population, whereas the parents' educational level was slightly higher [15]. We have not further addressed the characteristics of non-responders.
Ethics
The study was approved by the local ethics committee (ref. no. 20000042) and the database was approved the Danish Data Protection Agency (ref. no. 2000-53-0037). The children as well as their parents gave their consent to participate in the study
Data collection
Data on LBP were collected using an interview developed on the basis of previously used nation-wide surveys and tested for feasibility in relation to the first data collection [8, 16]. The children were asked if they had any spinal pain (LBP, MBP or NP) at the moment, within the past week, or within the past month in order to establish the one-month period prevalence. Those who responded positively to any of these questions were asked to show the location of the pain. The lumbo-pelvic, thoracic, and cervical spine and the corresponding posterior aspects of the body surface were defined as low back, middle back, and neck, respectively. If the child had problems showing the area, the interviewer put one hand subsequently on the neck, thoracic and lumbar area while asking: "Was it in this area?"
If the child reported back pain, the following questions were asked: "Did you because of back pain a) stop participation in physical activity such as sports or play? b) stay home from school up to 3 days, c) stay home from school for more than 3 days? d) see a physician once? e) see a physician more than once? f) see a physical therapist or participate in special gymnastics? g) go to a hospital?" The question f) "see a physical therapist" was included in the second and third surveys only [12].
At T1 the interview took place at the schools and was conducted by NW. All questions were explained to ensure the children's understanding of the content of the questions. The same procedure was used at T2 and T3. At T2 the interview took place in relation to the MRI scanning and was conducted by either of two radiographers and at T3 a research nurse conducted the interview.
Validity
At T2 the children also filled in a questionnaire on back pain that had previously been used on Danish schoolchildren [17]. Therefore, the possibility of comparing answers from interviews with answers from questionnaire existed. Analyses of the different responses lead to the conclusion that data from the interview were the most credible [18].
Data manipulation
From the interview, "BP" was recorded if the child reported pain in any of the spinal areas on the day of examination, within the past week, or within the past month. "LBP" was defined as positive if the child answered yes to pain related to the lumbar area, "MBP" for pain related to the thoracic spine, and "NP" for pain related to the cervical area. "Seeking care" was defined as such if "yes" was the answer to any of the questions d-g about consequences of back pain.
Quality of data
Data from the interviews were collected on paper and entered into data files using the software Microsoft Access for the first study. All computerised data were double checked against the original data on paper, and corrected if necessary. Data from the second round were entered into the software database EpiData and were checked randomly with an extremely low error rate. Information from the third data collection was entered directly into EpiData, leaving no possibilities for missing data. If the child could not provide a lucid answer, the response was nevertheless entered in a log-file for future decision/classification.
Statistical analyses
Prevalence data were reported for each variable in each cross-sectional study. Exact 95% confidence intervals were constructed and differences in proportions between genders tested with Fisher's exact test. Test for trend over time was performed using logistic regression accounting for repeated measures by Stata's cluster option in order to define statistically significant differences in prevalence rates between the surveys. The tracking patterns of BP, LBP, MBP, NP, and care seeking were studied by investigating how study subjects stayed in or left their respective group when surveyed at the next time point. The patterns of reporting at the second and third surveys were also established for previous non-responders in order to see if they were biased towards more BP than the responders.