Objective: Many people with epilepsy report subjective cognitive impairment (SCI), i.e., problems with memory, attention, or executive functions, reducing quality of life. Nevertheless, overlap with objective cognitive impairment (OCI) is often weak. One reason may be a domain-specific mismatch between subjective reports and objective tests. We aimed to evaluate relations between SCI and OCI of corresponding domains and to assess whether these differ between persons who over-or underestimate their performance.
Methods:In this prospective, cross-sectional sample of 104 adult inpatients with epilepsy, we performed multiple regression analyses predicting SCI in the domains attention, memory, and executive functions. We tested relationships with measures of psychomotor speed, short-term memory, verbal learning, verbal delayed recall, and word fluency while controlling for age, sex, seizure frequency, structural lesions, mono-versus polytherapy and adverse events of antiseizure medication (ASM), depressive and anxiety symptoms, level of education, and employment status. Furthermore, we tested whether these relationships differed between realistic raters and over-and underestimators.
Results:We found domain-specific relations for attention and executive functions for the full sample, explaining a small proportion of variance of SCI (general dominance index = .03 and .004), whereas ASM adverse events and psychological variables were more important predictors. When dividing the sample according to the concordance of SCI and OCI, we found high frequencies of both over-(23%-46%) and underestimation (31%-35%) depending on the domain. The explanatory power of OCI for SCI was stronger within the subgroups compared to the full sample, suggesting nonlinear relationships and different underlying mechanisms for realistic raters, underestimators, and overestimators.Significance: Domain-specific SCI and OCI are related, and both should be assessed with standardized instruments. These relationships differ between overand underestimators as well as realistic raters. Based on the concordance of