The aim of this study was to (1) calibrate an item bank to measure patients' experience of drug therapy for adult patients with SMIs and (2) develop computerized adaptive testing (CAT) to improve its use in routine practice. This is a cross-sectional, multicentric study involving 541 patients with schizophrenia, bipolar disorder, and major depressive disorder. Analyses based on classical test and item response theories were performed. After 7 highly inter-correlated items and 4 items with low factor loadings were removed, the remaining 26 items were sufficiently unidimensional (RMSEA = 0.069, CFI = 0.969, TLI = 0.963) and showed adequate fit to the generalized partial credit model. There was no differential item functioning by gender, age, care setting, or diagnosis from moderate- to large-magnitude. The mean score was 46.0 ± 16.9 and was significantly higher for patients reporting good medication adherence. The resulting PREMIUM-DT item bank has strong psychometric properties, and CAT facilitates widespread use in clinical settings (an average of 8 items administered, corresponding to a reliability of >0.90). Our results suggest that practical information and information about the side effects of psychotropic treatments and how to cope with them should be targeted as a priority to improve patients' experience of drug therapy.