Objective: Although condition-specific measures are commonly used in dementia, they cannot be used in analyses of cost per quality-adjusted life-year because they do not incorporate preferences. We addressed this gap by estimating two preference-based single index measures: the DEMQOL-U from the self-report DEMQOL (mild-to-moderate dementia severity) and the DEMQOL-Proxy-U from the carer-report DEMQOL-Proxy (all levels of dementia severity). Methods: We conducted valuation studies on 593 members of the general population (306 for the DEMQOL-U, 287 for the DEMQOL-Proxy-U) using the time trade-off elicitation technique. We then fitted a range of mean and individual-level multivariate regression models to the valuation data to derive preference weights for each measure. We applied the estimated weights to a large, clinically representative sample. Results: Mean observed time trade-off values ranged from 0.18 to 0.95 for DEMQOL-U and from 0.33 to 0.96 for DEMQOL-Proxy-U. The best performing models for each measure were main effects models estimated using individual-level data. DEMQOL-Proxy-U had inconsistent but insignificant coefficient estimates for one dimension. Models were estimated to remove these inconsistencies. Conclusion: Preference-based single index measures from DEMQOL and DEMQOL-Proxy for use in economic evaluation will enable economic evaluation using quality-adjusted life-years to be undertaken for people across the full range of dementia severity. Future research will examine how the utilities from each measure can be used and combined to populate cost-effectiveness models.