Medication Complexity and Insomnia Severity in Psychiatric Outpatients: A Moderation Analysis Across Mood and Non-Mood Disorders

Abstract

Background: Insomnia is common and occurs across psychiatric diagnoses. A prior analysis of this outpatient cohort found that psychotropic medication complexity was associated with insomnia severity, but whether this association is stronger in particular diagnostic groups was not examined. We preregistered a moderation analysis to test whether mood disorder status modifies the association between medication complexity and insomnia, and additionally examined modification by sex and by insomnia domain. Methods: A secondary analysis was performed of cross-sectional, routinely collected data from 502 adult outpatients (from an initial cohort of 506) at an Italian community mental health centre. The primary model regressed the Insomnia Severity Index (ISI) total score on medication complexity × mood disorder status (depressive and bipolar vs. other psychiatric disorders), adjusting for age, sex and education; it was preregistered on the Open Science Framework before any interaction model was estimated. Results: Medication complexity was associated with the ISI total (b = 1.54 per class, 95% CI 0.92 to 2.16; attenuated but retained after benzodiazepine adjustment, b = 1.00, p = 0.009). The prespecified medication complexity × mood interaction was not statistically significant (b = −1.02, 95% CI −2.25 to 0.21, p = 0.103). The medication complexity × sex interaction was also non-significant (b = 0.28, p = 0.64), and the moderation did not differ between ISI domains (three-way interaction p = 0.67). Benzodiazepine prescription was independently associated with the ISI (b = 1.87, p = 0.011). Conclusions: The prespecified hypothesis of a stronger medication complexity–insomnia association in mood disorders was not supported. Because no equivalence margin was prespecified, the analysis cannot establish that the association is equivalent across groups; it indicates only that the hypothesised stronger association was not detected. Findings are exploratory with respect to mechanism and should be read in light of the cross-sectional design and probable confounding by indication.