Study Uses Machine Learning to Differentiate Between Depressive Disorder and Insomnia Disorder Across Age Groups

A recent study aimed to differentiate between depressive disorder and insomnia disorder characterized by depressive symptoms among various age groups. Conducted at Hangzhou Seventh People’s Hospital from January 2016 to October 2024, the research involved 1,873 patients classified into two groups: depressive disorder and insomnia disorder with depressive symptoms, based on DSM-5 criteria. The sample included 190 adolescents, 1,253 adults, and 430 older adults, who were assessed using the “Good Sleep 365” platform.

Researchers employed machine learning techniques, particularly logistic regression models, to analyze symptom features, utilizing inputs from the Pittsburgh Sleep Quality Index (PSQI) and the Patient Health Questionnaires (PHQ-9 and PHQ-15). The findings revealed significant demographic and symptom differences between the two groups, although many clinical presentations overlapped. Symptoms such as depressed mood, fatigue, impaired concentration, and sleep-related issues were found to be stable indicators across age groups, while specific patterns varied by age.

The performance of the models varied, with adults showing the highest ability to distinguish between the disorders (AUC of 0.8216), followed by older adults (0.7446) and adolescents (0.6229). This suggests that the symptom-level evaluation might provide valuable insights for clinical differentiation between these conditions.

The study indicated that depressive disorder is highly prevalent, significantly affecting global mental health, while insomnia disorder also poses a serious issue, with a notable comorbidity with various mental disorders, especially depression. The results suggest that age-specific symptom assessment could enhance clinical diagnosis and treatment interventions.

Overall, the research emphasizes the importance of symptom-level differences in the differential diagnosis of depressive disorder and insomnia disorder with depressive symptoms and highlights the potential use of machine learning in refining mental health assessments.

Future studies are encouraged to explore these findings further across broader and more balanced populations, taking into account the complexities of age-related symptom presentations.