Participants
The participating students came from randomly selected (by The Swedish Bureau of Statistics), schools in Sweden, representing different geographical and socioeconomic regions. They were originally part of a wider, multidisciplinary base study conducted during the spring of 2001, where 1975 students, from grades 3, 6 and 9 participated [16, 17].
The present study was carried out a year and a half later, during four weeks in October 2002, with a sub sample of the original 6th grade classes from eleven schools. The students had entered 8th grade and were age 14 at the onset of the year. Only those students present at school on the test days, were included in the investigation. Parents or legal guardians of the students received written information outlining the study and they gave their signed informed consent before the study commenced. The parents were also invited to participate in the study through answering similar questions as the students in a mailed-out separate questionnaire with a prepaid envelope to return.
In total 232 students completed their questionnaire and 200 parental responses were collected. The majority of the parental questionnaires were completed by the mother/stepmother (86%) and thereafter by the father/stepfather (13%). One questionnaire was completed by another female adult family member (0.5%) and there was one missing answer (0.5%). One hundred and eighty-six (84 girls: 45% and 102 boys: 55%) corresponding student-same parent questionnaires were registered for which comparisons of answers could be made and analysis conducted. This gives a corresponding same child same parent response rate of 82%.
The identity of the students and parents was coded to ensure anonymity. The base study, including permission for the follow-up study was approved by the Ethical committee at the Karolinska Institutet, Stockholm, Sweden (Ref. no. 00–416).
The questionnaire
The principal investigator (G.B.S.) visited each selected school, administering and assisting the students while responding to a specially designed self-complete questionnaire. The questionnaire was, with minor revision, identical to the questionnaire answered by the same students in the base study (2001). A more detailed description of the questionnaire and the base study has been presented earlier [16, 17]. Questions addressed their a) medical background, i.e. handicaps, chronic or prolonged diseases, and if any recent surgeries or fractures, requiring a cast, had occurred since the onset of the fall term; b) injuries and accidents during the recall period and since the base study, including information of site of, type of injury and setting; and c) perceived health. All students reporting an injury orally clarified their injury with the principle investigator so it complied with the definition and recall period. As a measure of the students' subjective well-being they were asked to recall their perceived health "since the onset of the fall term", i.e. mid August until the testing date, in October. Thus the recall period was 7–11 weeks. The students answered by grading on a five-point Likert scale [(1) never or almost never, (2) now and then, (3) often (every week), (4) very often, and (5) always] how often they suffered from headaches, abdominal-, back-, and/or musculoskeletal pains. Furthermore, they were asked if they frequently felt stressed, sad, and lonely and if they had problems sleeping or often felt tired.
The parents' questions were derived from the students' and addressed the same three parts, the students' medical background, their injuries and accidents and questions of perceived health, i.e. headaches, sense of fatigue, pain symptoms not attributed to an injury or disease in addition to back or knee-pain, as well as questions of perceived general health status.
Reliability
Reliability is associated with the accuracy, consistency as well as the repeatability of a test e.g. questionnaire [18, 19]. A reliability coefficient differentiates between the ratios of measured variance that is a true score from a random error. To test for reliability the same subject must answer the questionnaire at least once within no longer time than four weeks [18]. The health questionnaire used in this study was tested in a test-retest procedure. For ordinal variables, a comparison was made using the statistical procedure of Spearman Correlation and Intra class coefficient (ICC Alpha). The strength of agreement was good to very good (Cohen 1988 cited in [20, 21]) with values above 0.8 (ICC: 0.9) for pain variables and above 0.9 (ICC: 0.9) for sleeping problems and tiredness. In spite of statistical tests, a low test-retest score may reflect actual changes in feelings or opinions, and on the other hand, a high score can be due to recollection of answers earlier given. In the test-retest study eight students reported that an injury had occurred during the recall period. One student failed to complete the questionnaire at the second occasion. The other seven students gave an identical answer on 99% of the 54 questions/items given.
Validity
Steps to secure validity includes initial review from experts in the field, pilot testing with subjects, resembling the target group, and assuring the test subjects anonymity and confidentiality [18]. For content validity the present questionnaire was constructed in collaboration with a pediatrician and orthopedic surgeons trained in sports medicine. The survey, at all stages coded for anonymity, was first pre-tested for relevance and comprehension by school students of the corresponding age groups, and thereafter in a pilot study in November 2000 with 103 students from grade 3, 6 and 9.
Statistics
The strength of agreement between responses from student and their matching parent was studied by means of absolute agreement. Absolute agreement is the shared positive and negative answers from both students and parents divided by total number of responses presented in percent. Agreement was also analyzed with the Kappa coefficient. Kappa corrects for chance and takes into account both the observed and expected value on the diagonal of a cross tabulation. For ordinal variables weighted Kappa was calculated. Weighted kappa includes weights given to values according to their distance from the diagonal so to account for the magnitude of disagreement [22]. Descriptive statistics, with frequencies of answers was used in those cases where the students' and parents' questions did not share the same format.
For all analyses, the statistical significance was set at p < 0.05.
The questionnaires were converted into a database using the SPSS (Statistical Package for the Social Sciences (SPSS 11.0, Chicago, IL, USA) computer software. For quality control, both the students' and parents' questionnaires were re-read and compared to the database to help establish the highest possible level of accuracy.