Defining the landscape of sleep problems in young adults using machine learning on nationwide register data from 2 million individuals
Abstract
Background: Sleep problems among young adults pose a major public health concern. To understand the patterns underlying the complexity of sleep in this population, we report trends in sleep problems and clusters of life-course factors using Machine Learning on nationwide surveys and registries.
Methods: We explored trends in the last decade using data from the Danish National Health Survey, MEDSTAT and The Danish National Patient Registry in self-reported sleep problems, medications such as melatonin and diagnoses for organic and non-organic sleep disorders. We used Natural Language Processing to learn life-course constellations based on registry data from the DANLIFE cohort of 2 million individuals. We then explored clusters of childhood adversity, diagnoses, medications and medical procedures centered around sleep-related medical terms to identify related factors.
Results: In the last decade, self-reported sleep problems and sleep medications have been increasing while diagnoses have remained steady. When looking at life-course constellations based on known sleep-related diagnoses, medications and medical procedures we found multiple clusters. Organic clusters involved factors related to respiratory issues, surgical interventions and fatigue from various causes while non-organic clusters were populated by mood and neurodevelopmental disorders. Interventions also differed among clusters where organic clusters had a higher prevalence of pharmacological and medical procedures while non-organic clusters reflected an enrichment in parental counselling and individual psychoeducation.
Conclusion: Sleep problems in young adults are increasing. By zooming out to identify clusters and life-course constellations of sleep problems, we provide a basis for zooming into the mechanisms and targeted interventions for young adults.
