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.
Objective: List-learning tests provide learning, recall, and recognition indices that are often interpreted as reflecting partly distinct memory processes. We used principal component analysis (PCA) and network analysis (NA) as complementary approaches to examine the organization of verbal and visual-design list-learning performance, with particular attention to very-long-delay memory assessed after approximately 1 week. Method: Data were drawn from two normative cohorts of healthy adults (N = 805). Analyses of the conventional learning, delayed-recall, and recognition measures were based on 800 complete cases, and 233 participants had complete data for analyses including very-long-delay recall and recognition approximately 1 week later. PCA was used to identify dominant dimensions of performance, whereas NA was used to examine relationships among individual measures. Results: PCA yielded two correlated components organized by stimulus modality, accounting for 73.2% of the variance for the conventional learning, delayed-recall, and recognition measures and 70.6% when very-long-delay measures were included. NA likewise showed clear modality-specific organization but additionally differentiated the relationships among learning, recall, and recognition measures, with recall measures occupying comparatively central and recognition measures more peripheral positions. Very-long-delay measures remained embedded within the modality-specific PCA structure but showed a distinctive pattern of relationships in the network. Conclusions: PCA and NA provided complementary descriptions of the same list-learning data. PCA primarily highlighted organization by stimulus modality, whereas NA differentiated relationships among individual learning, recall, and recognition measures. Very-long-delay assessment added information about the organization of memory performance, but the present findings do not establish a separate very-long-delay cognitive process.
