OBJECTIVES: To describe and compare symptom burden and healthcare resource utilization among patients diagnosed with major depressive disorder (MDD) with and without comorbid substance use disorder (SUD) in a real-world data (RWD) setting.

METHODS: A retrospective cohort of adults (≥18 years old) with ≥2 records of ICD-9/10 MDD diagnoses and no lifetime diagnosis of schizophrenia were analyzed using NeuroBlu, a RWD platform using US electronic health records. Two sub-cohorts were defined based on the presence of ≥1 ICD-9/10 SUD diagnoses within 3 months before the date of the first recorded MDD diagnosis (index date) or absence of SUD. NLP techniques were used to identify psychiatric symptoms from semi-structured Mental State Examination recorded within 30 days of index date. The prevalence of these symptoms and post-index psychiatric hospitalizations within one year were compared between patients with and without SUD using chi-square test.

RESULTS: Out of 81,518 patients with MDD, 18,218 (22.3%) had comorbid SUD (M+S sub-cohort) while 50,426 (61.9%) had no recorded diagnosis of SUD (M-S sub-cohort). Baseline symptoms related to mood, judgment, affect, attitude, memory, and suicidality were more prevalent in M+S sub-cohort than M-S. Notable differences include judgement issues (57.6% vs 33.2%; p <.001), suicidal ideation (22.7% vs 18.9%; p <.001) and memory issues/impaired memory (24.4% vs 10.0%; p <.001). No significant difference in insight-related issues was observed (77.2% vs 77.7%; p = 0.17). Furthermore, post-index hospitalizations were more frequent in the M+S group (27.3% vs 12.1%; p <.001) compared with the M-S group.

CONCLUSIONS: These findings suggest that patients with MDD and comorbid SUD had a higher symptom burden and were hospitalized more frequently. NLP-derived data provide an opportunity to derive insights from unstructured clinical notes which are otherwise difficult to summarize in large-scale RWD, leading to a more holistic view of the overall disease burden and economic impact.