δ Scores Identify Subsets of "Mild Cognitive Impairment" with Variable Conversion Risks

Background: The latent variable 'δ' (for 'dementia) is a transdiagnostic measure of dementia severity. δ can be reified and applied to individuals as a composite 'd-score'. Like Spearman's general intelligence factor 'g', δ can be constructed from almost any cognitive battery. So many are available that we must further distinguish each composite as a δ 'homolog'. Fourteen have been validated. All are strongly associated with dementia severity and potentially with mild cognitive impairment (MCI) conversion. Objectives: To assess δ's impact on MCI conversion risk. Methods: A new δ homolog (dDx) was constructed in 1,230 Mexican-American (MA) and 2,215 non-Hispanic White (NHW) participants in the Texas Alzheimer's Research and Care Consortium (TARCC). 1,445 normal controls (NC) and 723 MCI were followed annually for up to 6 years. Results: Each SD decrease in the dDx score increased the risk of conversion sixteen-fold [OR=16.39 (CI: 5.0-52.6)]. Cases below the optimal diagnostic threshold for Alzheimer's disease (AD) versus NC were labeled as having a functionally salient cognitive impairment (FSCI). Such cases were at a 73-fold increase risk of a diagnosis of AD [OR=73.19 (95% CI: 58.3-92.0)]. However, 25.6% of MCI cases were also FSCI(+). They accounted disproportionately for prospective conversions. Age <80 years, the absence of an ϵ4 allele, <12 years of education, and MA ethnicity independently increased the risk of diagnosing FSCI as MCI. Conclusion: A sizable minority of MCI cases may be misdiagnosed and they account disproportionately for AD conversions.