The longitudinal dataset comes from the “SALVe- Cohort” study, which aims at following youngsters from two cohorts, born in 1997 and in 1999. The youngsters were contacted when they were 12-13 and 14-15 years-old (T1), 15-16 and 17-18 years-old (T2), and 18-19 and 20-21 years-old (T3). At T1, participants were contacted by regular mail and invited to participate in the longitudinal study. They were informed that their participation was voluntary and that they could interrupt it at any time. They returned a self-reported questionnaire at T1, T2 and T3. The study was approved by the Ethical Review Board in Uppsala (Dnr. 2012/187).
The original eligible adolescents were N=4712, of whom N=1868 (38.46%), responded at wave 1 (T1). At wave 2 (T2), the adolescents were contacted again, and 1575 of the original sample filled in the questionnaire (see 20). Finally, 1174 of the adolescents who participated at T1 also returned the questionnaire at T3.
In sum, the final sample that participated at T1 and T2 consisted of 1575 young people, of whom 58% (N=664) were female; 20% (N=319) had non-Scandinavian parents; 49% (N=804) were born in 1999 and 51% (N=774) in 1997. The final sample that participated at T1 and T3 consisted of 1174 young people, of whom 61% (N=722) were female; 20% (N=229) had non-Scandinavian parents; 51% (N=601) were born in 1999 and 49% (N=573) in 1997.
Psychosomatic symptoms were measured using eight items from the WHO scale assessing the frequency of symptoms in the last three months (18). The answers range from never (0) to always (4). A Confirmatory Factor Analysis (CFA) showed that the model best fitting these items was from a second-order confirmatory factor analysis with a latent dimension of psychosomatic symptoms that comprised three subdimensions of symptoms, namely psychological (i.e., feeling nervous, feeling irritable, feeling sleepy, 3 items), somatic (headache, stomachache, 2 items) and musculoskeletal (pain in the shoulders/neck, pain in the back/hips, pain in the hands/knees/legs/feet, 3 items) (c2=77.03, p>.01, df=17, RMSEA=.04, CFI=.98, SRMR=.034). Accordingly, in the analysis we used both the total index of Psychosomatic symptoms (a=.71) and its three dimensions, namely Psychological, Somatic, and Musculoskeletal symptoms. Since our interest was in high levels of psychosomatic symptoms, the total index and the three subdimensions were dichotomized into high (top 25th percentile) and low.
Depressive symptoms. The Depression Self-Rating Scale Adolescent version, DSRS-A (21) was used. The scale comprises 15 items based on the DSM-IV criteria for a major depressive disorder. The adolescents were asked about their depressive feelings in the last two weeks, with a yes/no response alternative. In accordance with the DSRS scale, the index used in the analysis was calculated by adding reported symptoms, where each set of symptoms was counted only once (0-9 points). Cronbach´s alphas were 0.81, 0.77 and 0.87, for T1, T2 and T3 respectively. Moreover, a dichotomous variable was created, in which adolescent boys and girls fulfilling the DSM-IV A-criterion were classified as having high symptoms of depression.
Anxiety. Anxiety symptoms were measured using the Spence Children’s Anxiety Scale (SCAS, 22) at waves 1 and 2, while a short version of the same scale for adults was used at wave 3. The SCAS consists of 44 items, of which 38 cover all the six categories of anxiety disorder highlighted in the Diagnostic and Statistical Manual of Mental Disorders (23), while 6 are used as filler items to reduce negative bias. Alternative responses go from 0 (never) to 3 (always). The score of 33 has been identified as the cut-off for a diagnosis of generalized anxiety disorder (24) for children and adolescents. At wave 3, an adapted version with 15 items of the scale was used to assess the anxiety symptoms of the young adults. In the Swedish population, the cut-off for total anxiety disorder was >18. Cronbach alphas were 0.87, 0.89 and 0.91 and respectively. We used the above-mentioned cut-offs to classify the adolescents with high vs low symptoms.
Diagnoses of depressive and anxiety disorders. At T3, the young adults were asked whether they had received a diagnosis of depression and/or or some anxiety disorder, with yes/no as response alternatives.
Covariates. Sex (male, female), age (born in 1997 or 1999) and parents´ country of birth (Scandinavian parents vs at least one of the parents born outside Scandinavia) were used as covariates in the analyses.
To investigate the effects of psychosomatic symptoms at T1 on levels of depression and anxiety at T2 and T3 and on diagnoses of depressive and/or anxiety disorder at T3, multivariate logistic models were constructed, controlling for initial levels of depression and anxiety respectively, and also for sex, age and SES. First, we ran models assessing the effect of the dichotomized total index of psychosomatic symptoms (low vs high, based on the top 25th percentile at T1) on depressive or anxiety symptoms (low vs high, based on DSRS and RCMAs cutoffs) at T2 and T3, and on diagnoses of depression or anxiety disorder at T3. Then, we ran models using the subscales of psychosomatic symptoms, i.e., high somatic, high musculoskeletal, and high psychological symptoms (high=top 25th percentile). Finally, to test whether sex was a moderator of the effects of psychosomatic symptoms, all the models were re-run adding an interaction term, i.e., psychosomatic symptoms * sex.