Testing Bidirectional, Longitudinal Associations Between Disturbed Sleep and Depressive Symptoms in Children and Adolescents Using Cross-Lagged Models

This cohort study evaluates the bidirectional, longitudinal associations between disturbed sleep and depressive symptoms in children and adolescents using cross-lagged models.

1. Indicate how long in total he/she sleeps during the NIGHT (on average). Do not count the hours that your child is awake 2. Indicate how long in total he/she is awake during the NIGHT (on average). 3. In general, how many hours does your child sleep during the DAY (total of all naps)? 4. Does your child TALK in his/her sleep? 5. Does your child WALK in his/her sleep? 6. Does your child have NIGHTMARES? 7. Does your child have NIGHTTERRORS (wakes up suddenly, crying, sometimes drenched in sweat and confused)?
Each item was rated on a 4-point scale (1=never, 2=sometimes, 3=often, 4=always), except for sleep duration and time awake in bed, which were reported in number of hours and minutes, and were re-scaled to vary between 1 and 4. Sleep duration was reverse-coded to have the same direction of the other items and in the assumption that shorter sleep duration is a risk factor for depression 49 .

eMethods 3
Eight-item parent-reported questionnaire to assess depression in children 1. Seemed to be unhappy or sad? 2. Was not as happy as other children? 3. Was unable to concentrate, could not pay attention for long? 4. Was too fearful or anxious? 5. Was worried? 6. Cried a lot? 7. Was nervous, highstrung or tense? 8. Had trouble enjoying HIM/HER self?
Each item was rated on a 3-point scale, (1=never, 2=sometimes, 3=often). Cronbach's alphas at 5, 7 and 8 years were 0.70, 0.72 and 0.73, respectively. A mean score was computed, and the score re-scaled to range 0-10. The higher the score, the more depressive symptoms.
Ten-item self-reported questionnaire to assess depression in adolescents 1. I am unhappy or sad 2. I am not as happy as other people of my age 3. I can't concentrate, I can't pay attention 4. I am too fearful or nervous 5. I worry a lot 6. I cry a lot 7. I am nervous, high strung or tense 8. I have trouble enjoying myself 9. I have temper tantrums or easily get angry 10. I am not interested in doing activities with other children Each item was rated on a 3-point scale (1=never, 2=sometimes, 3=often). Cronbach's alphas at 10, 12, 13, 15 and 17 years were 0.74, 0.78, 0.82, 0.88 and 0.89, respectively.

eMethods 5
Measurements' details for covariates/moderators, i.e. socio-economic stauts (SES), maternal depression and pubertal status SES was the standardized composite of 5 items enquiring about parental education, occupational status, and household income (range, −3 to 3, centered at 0, higher scores indicating higher SES) 63 .
For maternal depression, mothers reported how often they experienced depressive symptoms on a 4-point scale (0=never, 1=sometimes, 2=often, 3=always) using a shortened version of the Center for Epidemiologic Studies Depression Scale 64. Scores ranged from 0 to 10, with higher scores indicating higher depression.
For pubertal status, 12 and 13-year-old adolescents were provided with sex-appropriate drawings of secondary sex characteristics based on Tanner stages of pubertal development. Adolescents were asked to choose which of the drawings were most like them 65, and Tanner scores were assigned from I, prepubertal, to V, full pubertal, accordingly (eTable 1 in the Supplement). Self-rated pubertal status has been shown to be sufficiently accurate for large epidemiologic studies 66. The average scores were used as an overall index of pubertal status at each time point (12 years: M=2.76, SD=0.85; 13 years: M=3.50, SD=0.82). Given the high correlation of pubertal status between ages 12 and 13 years (r=0.72, p<0.001), the average score was used in further analyses (M=3.08, SD=0.81). eResults 1: Missing data pattern for outcome variables in childhood and adolescence.

Missing data pattern for outcome variables in childhood.
Rate of missingness for depressive symptoms and disturbed sleep at each time point are provided in eTable 2 in the Supplement. eTable 3 in the Supplement reports the characteristics of the study variables at each time point in the complete versus with-missing data subgroups. At 6 years of age, children with complete data were significantly less depressed compared to children with missing data. Furthermore, SES was lower and maternal depression was higher in those with missing data compared to those with complete data. The missing data pattern across the 3 time points was MCAR for depressive symptoms (χ2 = 14.90, df = 9, p = .09) and missing at random (MAR) for sleep (χ2 = 19.18, df = 9, p = .02). As for depressive symptoms, there was not systematic differences in missing data related to time since enrollment. As for disturbed sleep, there was only a slight difference in missing data related to time since enrollment Missing data pattern for outcome variables in adolescence. Rates of missingness for depressive symptoms and disturbed sleep at each time point are provided in eTable 5 in the Supplement. eTable 6 in the Supplement reports the characteristics of the study variables at each time point in the complete versus with-missing data subgroups. At 10 years of age, adolescents with complete data were significantly less depressed compared to adolescents with missing data. At 15 years of age, adolescents with complete data had significantly less disturbed sleep compared to adolescents with missing data. Furthermore, SES was lower and maternal depression was higher in those with missing data compared to those with complete data. The missing data pattern across the 5 time points was MCAR for depressive symptoms (χ2 = 42.45, df = 50, p =.77) and for sleep (χ2 = 56.84, df = 50, p =.24). As for disturbed sleep and depressive symptoms, there were not systematic differences in missing data related to time since enrollment.

Cross-lagged paths between the insomnia index and depressive symptoms in childhood.
Findings showed that depressive symptoms were significantly associated with later insomnia at each time point, whereas insomnia was not associated with later depressive symptoms at any time points (eFigure 1 in the Supplement).

Cross-lagged paths between disturbed sleep and depressive symptoms in childhood including time points 5, 7, 8 and 10 years.
Findings showed that depressive symptoms were significantly associated with later disturbed sleep, and disturbed sleep was associated with later depressive symptoms between 5 and 8 years (eFigure 2 in the Supplement).

years.
Findings showed longitudinal, reciprocal associations between depressive symptoms and disturbed sleep between 5 and 8 years, and unidirectional associations between depressive symptoms at 8 years and disturbed sleep at 10 years, and between depressive symptoms at 10 years and disturbed sleep at 12 years (eFigure 5 in the Supplement).