Sleep quality predicts future mood symptoms in adolescents with bipolar disorder

Abstract

Background: Poor sleep is prevalent in adolescents with bipolar disorder, precedes illness onset, and is associated with worse mood symptoms. We examined interrelationships between sleep quality and mood symptoms in adolescents with bipolar disorder, particularly effects of sleep quality on emergent mood symptoms.

Methods: Adolescents with bipolar disorder participated in a two-year longitudinal treatment study. Sleep quality (Pittsburgh Sleep Quality Index, PSQI) was assessed quarterly during treatment (baseline, 3-, 6-, 9-, 12-month visits) and twice during follow-up (18-, 24-month visits). Mood symptoms (ALIFE Psychiatric Status Ratings) were retrospectively rated weekly by an independent clinician. Lag models tested whether sleep quality predicted next month's mood symptoms and whether mood symptoms predicted future sleep quality.

Results: Adolescents with bipolar disorder had poor sleep quality. Sleep quality initially improved but remained stable thereafter. Worse sleep quality at 6-months predicted worse depression, hypomania, and suicidal ideation the following month. Sleep quality was worse for adolescents who had a suicide attempt during the study compared to those who did not and was worse preceding months with a suicide attempt compared to months without attempts. Alternatively, worse depression predicted worse future sleep quality at baseline, 3-, and 18-months and worse suicidal ideation predicted worse future sleep quality at baseline, 12-, and 18-months.

Limitations: Mood symptoms were rated retrospectively and the PSQI may not capture all dimensions of sleep important for mood symptoms.

Conclusions: Targeted evidence-based sleep treatment in adolescents with bipolar disorder may alleviate sleep problems and have additional benefits on mood symptoms and suicidality risk.

Keywords: Adolescents; Bipolar disorder; Pittsburgh sleep quality index; Sleep quality; Suicidal ideation; Suicide attempts.

Fig. 1.

Example of lagged model structure. Diagonal arrows represent lagged relationships in which each predictor (boxes to the left; e.g., Pittsburgh Sleep Quality Index [PSQI]) is a separate variable entered in the model used to predict future scores of a particular outcome (boxes to the right; e.g., the next month’s average mood [depression, suicidal ideation, hypomania] score). Downward arrows between predictors and outcomes control for autoregressive effects. Circles with e represent error terms.

Fig. 2.

Changes in sleep quality across treatment (baseline – 12-month visits) and post-treatment follow-up (18- and 24-month visits), with higher scores representing worse sleep quality. Sleep quality is collapsed across treatment group (Dialectical Behavioral Therapy vs Standard of Care), since scores did not differ by treatment group. Error bars represent standard error. *p < .05, **p < .01, ***p < .001.

Fig. 3.

A: Changes in average depression, suicidal ideation, and hypomania symptom ratings the month after each assessment visit. B: Unstandardized beta coefficients for sleep quality (i.e., Pittsburgh Sleep Quality Index global score) predicting the following month of mood symptom ratings. Diamond symbol represents timepoints in which sleep quality significantly (p < .05) predicted mood symptoms. Circle symbol represents timepoints in which sleep quality marginally (p < .10) predicted mood symptom ratings. Error bars represent standard error.

Fig. 4.

A: Changes in average depression, suicidal ideation, and hypomania symptom ratings for the month timeframe two months before each assessment visit. B: Unstandardized beta coefficients for Psychiatric Status Ratings predicting future sleep quality (i.e., Pittsburgh Sleep Quality global score). Diamond symbol represents timepoints in which mood symptoms significantly (p < .05) predicted sleep quality. Error bars represent standard error.