BACKGROUND: Subjective cognitive complaint is a sensitive marker of decline.OBJECTIVE: This study aimed to (1) examine reliability of subjective cognitive complaint using the Cognitive Function Instrument (CFI), and (2) assess the utility of the CFI to detect cognitive decline in non-demented elders.METHODS: Data from a four-year longitudinal study at multiple Alzheimer's Disease Cooperative Study (ADCS) sites were extracted (n = 644). Of these, 497 had Clinical Dementia Rating (CDR) global scores of 0 and 147 had a CDR of 0.5. Mean age and education were 79.5±3.6 and 15.0±3.1 years, respectively. All participants and their study partners completed the subject and study partner CFI yearly. Modified Mini-Mental State Exam (mMMSE) and Free and Cued Selective Reminding Test (FCSRT) were administered. Scores below the predetermined cut-off scores on either measure at annual visit were triggers for a full diagnostic evaluation. Cognitive decline was defined by the absence/presence of the trigger.RESULTS: Three-month test retest reliability showed that inter-class coefficients for subject and study partner CFI were 0.76 and 0.78, respectively. Generalized estimating equation method revealed that both subject and study partner CFI change scores and scores from previous year were sensitive to cognitive decline in the CDR 0 group (p < 0.05). In the CDR 0.5 group, only the study partner CFI change score predicted cognitive decline (p < 0.05).CONCLUSION: Cognitive decline was predicted differentially by CDR level with subject CFI scores providing the best prediction for those with CDR 0 while study partner CFI predicted best for those at CDR 0.5.