Introduction. This study aims to deepen our understanding of therapeutic adherence in patients with severe mental illnesses, an issue that holds significant public health and economic implications. Worldwide, one in eight individuals is affected by a mental disorder, leading to a substantial cost both direct and indirect, including extended hospital stays and lost productivity. Non-adherence to medication treatments remains a complex issue influenced by various demographic, socioeconomic, and disease-specific factors. The paper seeks to identify variables that contribute to decreased adherence in this population and compare them with existing literature. Method. This study, conducted from September 2022 to December 2022 at the Clinical Hospital of Psychiatry ‘Professor Doctor Alexandru Obregia’ in Bucharest, Romania, involved 177 participants. Data was collected through interviews, patient records, and verified with attending physicians and family members as needed. Exclusions were made for patients not fluent in Romanian or diagnosed with a major neurocognitive disorder per DSM-5. Assessments included the Clinical Global Impression Severity Scale (CGI-S) for disease severity and the modified Overt Aggression Scale (mOAS) for aggressive behavior. Treatment adherence was defined as taking 80-120% of the prescribed psychiatric treatment dose in the 4 weeks prior to current admission and was corroborated by at least two external sources. Statistical analysis employed logistic regression and post hoc Tukey tests to identify significant differences in patient data. Results. In our study, no statistically significant correlation was found between treatment adherence and various sociodemographic or personal factors. However, patients living alone had an Odds Ratio of 4.92, suggesting almost five times higher likelihood of treatment non-adherence. No significant correlations were identified concerning diagnosis, illness duration, or number of hospitalizations. Patients brought in by police had an 11-fold higher risk of non-compliance (OR 11.05), while those brought by ambulance had a 4-fold increased risk (OR 3.923). Non-compliant patients were more likely to experience prolonged involuntary hospitalization (p=0.007) and had a 4-fold greater risk of involuntary admission (OR 3.96). High scores on the Clinical Global Impression Scale (CGI), Admission Experience Survey (AES), and Modified Overt Aggression Scale (mOAS) were statistically significant predictors of patient non-compliance. Conclusion. In our study, we found no significant correlation between sociodemographic factors and medication adherence. However, living alone significantly raised the odds of non-adherence, suggesting the role of social support in treatment compliance. Manner of admission to the emergency room emerged as a key predictor, with involuntarily admitted patients showing heightened risks of non-compliance. Our data supports the linkage between lower adherence and more days of hospitalization, in alignment with the specialty literature. Patients admitted involuntarily were notably four times more likely to be non-adherent (OR 3.96, p=0.001). Additionally, the severity of illness at the time of admission was higher among nonadherent patients, as assessed by CGI, AES, and mOAS scales.