Prediction of Dementia and MCI by Subjective and Objective Complaints

Parallel Session 13 - Risk Reduction and Risk Factors - no. OC074Introduction: Subjective cognitive complaints serve as an important indicator of progression to dementia or mild cognitive impairment (MCI). However, there are limited research in comparing subjective (by help-seeker) and objective (by carer) complaints associating with cognitive impairment. The current study aims to explore the relationship between cognitive impairment and the subjective and objective complaints so as to identify the predictors of suspected dementia and MCI. Method: A total of 1038 participants (age 42 to 96, mean age = 75.7±8.36, 378 males) were recruited from a community-based dementia early detection program in Hong Kong. Both of the helper-seekers and their carers were asked to report the reasons of help seeking. Contents of the complaints were classified into different themes by a trained research assistant, namely memory, executive function, time orientation, language, neuropsychiatric, personality change and mood. Demographic characteristics (i.e., age, gender, educational level) were obtained. The Clinical Dementia Rating (CDR) was conducted by an occupational therapist or a social worker as an indicator of suspected MCI (cutoff 0.5) or Dementia (cutoff 1.0). Cognitive function (MMSE) and functioning (Lawton IADL) were also assessed. Logistic Regression Models (LRM) were computed to identify the predictors of dementia and MCI respectively with the categories of subjective and objective complaints as independent variables after controlling for demographics, cognitive function and functioning (significance level at p<0.01). Result: Using CDR cutoffs, 552 help-seekers were suspected as MCI while 327 as dementia. LRM showed that carers’ complaints of memory (p<0.01) and language impairment (p<0.01) significantly predicted the risk of dementia (p<0.01, Nagelkerke R2=66%, percentage correct 86.3%). In terms of the prediction of MCI, only carers’ memory complaint was significant (p<0.01) with the model significant at p