Abstract
Multivariable modeling in dementia risk factor studies is limited by the number of factors that can be analyzed practicably. Index variables, which integrate exposures, can efficiently reduce dimensionality. The Consortium to Investigate Vascular Impairment of Cognition study, a Canadian memory-clinic-based 30-month cohort study of 1,347 patients, used a vascular risk factor index (from 20 exposures) and a vascular clinical profile index (17 items). Patients with vascular cognitive impairment had higher index counts compared to those without cognitive impairment (0.16 ± 0.11 vs. 0.07 ± 0.07 for the risk factor index and 0.21 ± 0.16 vs. 0.09 ± 0.07 for the clinical profile index; p < 0.05). Both the death rate and the rate of cognitive impairment increased exponentially with the index variable (r > 0.90 for each index). The risk ratio for death was 1.12 (95% CI 1.09–1.15) for each increment of the risk factor index and was 1.23 (95% CI 1.1–1.28) for each increment of the clinical profile index. With each index, the areas under the receiver operating characteristic curves for predicting death and institutionalization ranged from 0.73 ± 0.01 to 0.75 ± 0.01. Construction of index variables that integrate multidimensional factors is a promising approach to assessing risk in multidetermined states.