OBJECTIVES: To extract severity diagnosis and neuropsychological test result from Veterans with Alzheimer's disease (AD) in the Department of Veterans Affairs healthcare system.

METHODS: Clinical notes from the electronic health record of Veterans with AD containing both a severity diagnosis and mini-mental state examination (MMSE) and/or Montreal cognitive assessment (MoCA) test result were included in the study. A proprietary Python algorithm was applied to extract MMSE and/or MoCA test result(s) as the objective assessment(s). AD severity keywords such as "mild," "moderate," and "severe" were extracted as the subjective assessment. Objective and subjective assessments of AD severity were compared to evaluate concordance.

RESULTS: A total of 7,514 notes from 4,469 Veterans met the study requirements. Overall concordance of objective and subjective assessments was 53%. Concordance of subjective assessment with either MMSE or MoCA test result was also 53%. Concordance of severity assessments from dementia clinics (61%) was higher than non-dementia clinics (53%). Concordance increased when Veterans had symptoms such as wandering (73%) and aberrant motor behavior (67%) as well as comorbidities such as type I diabetes (61%).

CONCLUSIONS: The reported concordance of subjective and objective assessments highlights the complexity of AD diagnosis. Various factors such as level of education, cultural influences, and socioeconomic background may affect the objective assessment but are often not reported. As interventions designed to delay disease progression are developed, diagnosis of AD along with severity is critical to identify candidates for early-stage interventions.