BACKGROUND: Non-adherence to medication is associated with increased risk of relapse in patients with bipolar disorder (BD). OBJECTIVES: To (1) validate patient-evaluated adherence to medication measured via smartphones against validated adherence questionnaire; and (2) investigate characteristics for adherence to medication measured via smartphones. METHODS: Patients with BD (n=117) evaluated adherence to medication daily for 6-9 months via smartphones. The Medication Adherence Rating Scale (MARS) and the Rogers' Empowerment questionnaires were filled out. The 17-item Hamilton Depression Rating Scale, the Young Mania Rating Scale and the Functional Assessment Short Test were clinically rated. Data were collected multiple times per patient. The present study represents exploratory pooled reanalyses of data collected as part of two randomised controlled trials. FINDINGS: During the study 90.50% of the days were evaluated as 'medication taken', 6.91% as 'medication taken with changes' and 2.59% as 'medication not taken'. Adherence to medication measured via smartphones was valid compared with the MARS (B: -0.049, 95% CI -0.095 to -0.003, p=0.033). Younger age and longer illness duration were significant predictors for non-adherence to medication (model concerning age: B: 0.0039, 95% CI 0.00019 to 0.0076, p=0.040). Decreased affective symptoms measured with smartphone-based patient-reported mood and clinical ratings as well as decreased empowerment were associated with non-adherence. CONCLUSIONS: Smartphone-based monitoring of adherence to medication was valid compared with validated adherence questionnaire. Younger age and longer illness duration were predictors for non-adherence. Increased empowerment was associated with adherence. CLINICAL IMPLICATIONS: Using smartphones for empowerment of adherence using patient-reported measures may be helpful in everyday clinical settings. TRIAL REGISTRATION NUMBER: NCT01446406 and NCT02221336.