Objective: To develop a classification system (CS) for a diabetes-specific preference-based measure of health (PBMH). Methods: Plausible attributes for the PBMH were identified by Classical Test Theory, using Factor Analysis of responses from Type 2 Diabetes patients (n=385) to the 18-item Audit of Diabetes-Dependent Quality of Life (ADDQoL). A seven-member expert panel then provided qualitative input for content. Three pilot rounds in outpatient and community settings produced data from people with Type 1 and Type 2 diabetes (n1=52, n2=65, n3=111) that were analyzed using Modern Test Theory, based on Rasch Analysis (RA), for 1) fit of selected attributes to the Rasch Model, and 2) scaling of severity levels for attributes. Results: Principal Axis Factoring with Promax rotation identified two plausible attributes from six ADDQoL items. In a structured survey, experts rated the importance of all ADDQoL and additionally important items, and suggested attributes that might be described using sets of related items. A CS was developed consisting of five independent attributes, with each question containing a description based on the item content of the respective attribute and four sentences describing severity levels. Maintaining this format, the wording in the CS was further modified based on additional input from experts and RA after each pilot. The final attributes were: Physical Ability & Energy, Relationships, Mood & Feelings, Enjoyment of Diet, and Satisfaction with Management of diabetes. Results of the third pilot indicated Infit and Outfit MNSQ for the five attributes ranging between 0.88 and 1.10. Person and Item reliabilities were 0.65 and 0.92, while the respective separation ratios were 1.36 and 3.34. Severity levels used were supported by Rating Scale Diagnostics indicated by RA. Conclusion: Results of the statistical analyses indicate that the PBMH has desired psychometric properties. Research on the estimation of a utility scoring algorithm and validation testing of this PBMH is ongoing.