The autonomy preference index scale (API) has been designed to measure patient preference for 2 dimensions of autonomy: Their desire to take part in making medical decisions (decision making, [DM]) and their desire to be informed about their illness and the treatment (information seeking; [IS]). The DM dimension is measured by 6 general items together with 9 items related to 3 clinical vignettes (3 × 3 items). The IS dimension is measured by 8 items. While the API is widely used, a review of literature has identified several inconsistencies in the way it is scored. The first aim of this study was to determine the best scoring structure of the API on the basis of validity and reliability evidence. The second aim was to investigate the long-term stability of API scores. Two-hundred and 85 patients with a diagnosis of psychosis were assessed as they were about to be discharged from involuntary psychiatric hospitalization and they were reassessed after 6 and 12 months. Confirmatory factor analysis (CFA) revealed that a 3-factor solution was most adequate and that 2 distinct DM subscales should be preferred to 1 total DM score. While internal consistency estimates of the 3 subscales were good, the long-term stability of API scores was only modest. Multigroup-CFA revealed scalar invariance indicating API scores kept the same meaning longitudinally. In conclusion, a 3-factor structure seemed to be most adequate for the API scale. Long-term stability estimates suggested that clinicians should regularly assess patients' preferences for autonomy because API scores fluctuate over time. (PsycINFO Database Record