INTRODUCTION: We evaluated short versions of a 16-item odor identification (OID) test, with regard to their ability to identify individuals at high dementia risk.

METHODS: Participants from the population-based SNAC-K study (n = 2418) were followed across 12 years. We formed 13 abbreviated clusters based on the identifiability and perceptual characteristics of the Sniffin’ Sticks Test (SST) items, and pre-existing test versions. Dementia hazard was estimated with Cox regressions.

RESULTS: Lower OID scores were associated with an increased dementia hazard across all odor clusters. Lower performance in the high identifiability cluster showed the strongest association with dementia (hazard ratio = 1.39, 95% confidence interval [1.28–1.51]). Moreover, the high-intensity odor cluster showed a stronger association with dementia than the low-intensity cluster (P = 0.02).

DISCUSSION: The findings suggest that the SST items differ with regard to their association with dementia and support using a reduced set size for clinical practice.