Examining the Directional Effects of Insomnia, Anxiety, and Depression Symptoms: Results From a Longitudinal Study Conducted During the COVID‐19 Pandemic
Jamie Walker, Ana J. Bridges, Alexandria Muench, Michael L. Perlis, Ivan VargasABSTRACT
Objectives
While some research suggests that increases in anxiety symptoms drive changes in insomnia and depression symptoms, other studies support that these relations may be bidirectional. The present study therefore aimed to further examine how anxiety, insomnia and depression symptoms predict changes in these symptoms over time.
Methods
Participants were recruited via an online participant database, social media advertisements, and other online sites, resulting in a large sample ( N = 2,666) of adults from across the United States. Well validated instruments were used to measure symptoms of depression (CES‐D), anxiety (STAI), and insomnia (ISI). Data were collected via a series of Qualtrics surveys administered online at two time points: March–June 2020 (baseline) and January–March 2021 (follow‐up).
Results
Insomnia symptoms ( β = 0.142, p < 0.001) but not anxiety symptoms ( β = 0.019, p = 0.445) significantly predicted depression symptoms. Depression symptoms ( β = 0.077, p = 0.001) but not anxiety symptoms ( β = 0.039, p = 0.075) significantly predicted insomnia symptoms. Both insomnia ( β = 0.060, p = 0.001) and depression symptoms ( β = 0.281, p < 0.001) significantly predicted anxiety symptoms. Although multiple cross‐lagged paths were statistically significant, effect sizes were generally small, and no single pathway was meaningfully stronger than others. Additionally, secondary exploratory mediation models revealed significant indirect effects between anxiety and depression through insomnia (model 1 indirect β = 0.027, p = 0.007; model 2 indirect β = 0.008, p = 0.036).
Conclusions
Findings suggest that insomnia symptoms were prospectively associated with subsequent depressive and anxiety symptoms. Depression symptoms also predicted later anxiety symptoms, highlighting the complex longitudinal associations among these symptom domains. However, results did not support a temporal pathway in which anxiety precedes the development of comorbid insomnia. Future studies should assess temporal order more closely, using high density longitudinal sampling strategies (i.e., more than two data collection points) to further clarify the temporal ordering of these symptom domains.