Quantifying the diagnostic contribution of neuropsychological evaluation using natural language processing.

Abstract

Objective: Diagnostic clarification is regarded as a core function of neuropsychological evaluation, yet its impact has rarely been quantified because diagnoses are typically embedded within unstructured narrative reports. This pilot study quantified the diagnostic contribution of neuropsychological evaluation by examining diagnostic reclassification and diagnostic clarification using a novel natural language processing (NLP) pipeline. Method: A retrospective observational study was conducted using 220 adult outpatient neuropsychological evaluation reports from the National Neuropsychology Network. A rule-based NLP pipeline extracted and harmonized pre-evaluation referral diagnoses and post-evaluation clinician diagnoses from reports in the UCLA electronic health records into structured diagnostic profiles. Diagnostic reclassification, diagnostic specificity gain, and changes in diagnostic category distributions were quantified. Exploratory analyses examined demographic correlates of diagnostic clarification. Results: Diagnostic reclassification was common following neuropsychological evaluation. Overall, 87.7% of patients received at least one new diagnosis, 55.0% had at least one preexisting diagnosis not retained, and 84.5% demonstrated increased diagnostic specificity. Among patients with nonspecific cognitive symptom codes, 90.1% transitioned to formal psychiatric, neurological, or neurocognitive diagnoses. Collectively, these changes shifted diagnostic profiles away from nonspecific cognitive diagnoses toward more specific formal psychiatric and neurocognitive diagnoses (all p < .001). Exploratory demographic associations were modest. Conclusions: Neuropsychological evaluation provides substantial incremental diagnostic information beyond referral documentation by increasing diagnostic specificity and refining diagnostic formulation. These pilot findings establish diagnostic clarification as a measurable clinical outcome and demonstrate the potential of NLP to transform routine neuropsychological reports into scalable metrics for evaluating the clinical impact of neuropsychological services.