Browsing by Author "Smith, Glenn E."

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    A harmonized memory composite score for cross‐cohort Alzheimer’s disease and related dementia research: development and validation
    (Wiley, 2025-01-03) Sanderson-Cimino, Mark E.; Gross, Alden L.; Gaynor, Leslie S.; Paolillo, Emily W.; Casaletto, Kaitlin B.; Chatterjee, Ankita; Albert, Marilyn S.; Apostolova, Liana G.; Boersema, Brooke; Boxer, Adam L.; Boeve, Brad F.; Clark, Lindsay R.; La Joie, Renaud; Eloyan, Ani; Tomaszewski Farias, Sarah; Gonzales, Mitzi M.; Hammers, Dustin B.; Wise, Amy B.; Cobigo, Yann; Yballa, Claire; Schonhaut, Daniel R.; Hampstead, Benjamin M.; Mechanic-Hamilton, Dawn; Miller, Bruce L.; Rabinovici, Gil D.; Rascovsky, Katya; Ringman, John M.; Rosen, Howard J.; Ryman, Sephira; Salmon, David P.; Smith, Glenn E.; Decarli, Charles; Kramer, Joel H.; Staffaroni, Adam M.; Neurology, School of Medicine
    Background: The Uniform Data Set (UDS) neuropsychological battery, administered across Alzheimer’s Disease Centers (ADC), includes memory tests but lacks a list‐learning paradigm. ADCs often supplement the UDS with their own preferred list‐learning task. Given the importance of list‐learning for characterizing memory, we aimed to develop a harmonized memory score that incorporates UDS memory tests while allowing centers to contribute differing list‐learning tasks. Method: We applied item‐banking confirmatory factor analysis to develop a composite memory score in 5,287 participants (mean age 67.1; SD = 12.2) recruited through 18 ADCs and four consortia (DiverseVCID, MarkVCID, ALLFTD, LEADS) who completed UDS memory tasks (used as linking‐items) and one of five list‐learning tasks. All analyses used linear regression. We tested whether memory scores were affected by which list‐learning task was administered. To assess construct validity, we tested associations of memory scores with demographics, disease severity (CDR Box Score), an independent memory task (TabCAT Favorites, n = 675), and hippocampal volume (n = 811). We compared performances between cognitively unimpaired (n = 279), AD‐biomarker+ MCI (n = 26), and AD‐biomarker+ dementia (n = 98). In a subsample with amyloid‐ and tau‐PET (n = 49), we compared memory scores from participants with positive vs negative scans determined using established quantitative cutoffs. Result: Model fit indices were excellent (e.g., CFI = 0.998) and factor loadings were strong (0.43‐0.93). Differences in list‐learning task had a negligible effect on scores (average Cohen’s d = 0.11). Higher memory scores were significantly (p’s<.001) correlated with younger age (β = ‐0.18), lower CDR Box Scores (β = ‐0.63), female sex (β = 0.12), higher education (β = 0.19), larger hippocampal volume (β = 0.42), and an independent memory task (β = 0.71, p<0.001). The memory composite declined in a stepwise fashion by diagnosis (cognitively unimpaired>MCI>AD dementia, p<0.001). On average, amyloid‐PET positivity was associated with lower composite scores, but was not statistically significant (β = ‐0.34; p = 0.25; d = 0.40). Tau‐PET positivity was associated with worse performance, demonstrating a large effect size (β = ‐0.75; p<0.002; d = 0.91). Conclusion: The harmonized memory score developed in a large national sample was stable regardless of contributing list‐learning task and its validity for cross‐cohort ADRD research is supported by expected associations with demographics, clinical measures, and Alzheimer’s biomarkers. A processing script will be made available to enhance cross‐cohort ADRD research.
  • Thumbnail Image
    Item
    Development and validation of a harmonized memory score for multicenter Alzheimer's disease and related dementia research
    (medRxiv, 2025-04-03) Sanderson-Cimino, Mark; Gross, Alden L.; Gaynor, Leslie S.; Paolillo, Emily W.; Saloner, Rowan; Albert, Marilyn S.; Apostolova, Liana G.; Boersema, Brooke; Boxer, Adam L.; Boeve, Bradley F.; Casaletto, Kaitlin B.; Hallgarth, Savannah R.; Diaz, Valentina E.; Clark, Lindsay R.; Maillard, Pauline; Eloyan, Ani; Tomaszewski Farias, Sarah; Gonzales, Mitzi M.; Hammers, Dustin B.; La Joie, Renaud; Cobigo, Yann; Wolf, Amy; Hampstead, Benjamin M.; Mechanic-Hamilton, Dawn; Miller, Bruce L.; Rabinovici, Gil D.; Ringman, John M.; Rosen, Howie J.; Ryman, Sephira G.; Prestopnik, Jillian L.; Salmon, David P.; Smith, Glenn E.; DeCarli, Charles; Rajan, Kumar B.; Jin, Lee-Way; Hinman, Jason; Johnson, David K.; Harvey, Danielle; Fornage, Myriam; Kramer, Joel H.; Staffaroni, Adam M.; Neurology, School of Medicine
    Introduction: List-learning tasks are important for characterizing memory in ADRD research, but the Uniform Data Set neuropsychological battery (UDS-NB) lacks a list-learning paradigm; thus, sites administer a range of tests. We developed a harmonized memory composite that incorporates UDS memory tests and multiple list-learning tasks. Methods: Item-banking confirmatory factor analysis was applied to develop a memory composite in a diagnostically heterogenous sample (n=5943) who completed the UDS-NB and one of five list-learning tasks. Construct validity was evaluated through associations with demographics, disease severity, cognitive tasks, brain volume, and plasma phosphorylated tau (p-tau181 and p-tau217). Test-retest reliability was assessed. Analyses were replicated in a racially/ethnically diverse cohort (n=1058). Results: Fit indices, loadings, distributions, and test-retest reliability were adequate. Expected associations with demographics and clinical measures within development and validation cohorts supported validity. Discussion: This composite enables researchers to incorporate multiple list-learning tasks with other UDS measures to create a single metric.
  • Thumbnail Image
    Item
    Federated learning with multi‐cohort real‐world data for predicting the progression from mild cognitive impairment to Alzheimer's disease
    (Wiley, 2025) Pan, Jinqian; Fan, Zhengkang; Smith, Glenn E.; Guo, Yi; Bian, Jiang; Xu, Jie; Biostatistics and Health Data Science, Richard M. Fairbanks School of Public Health
    Introduction: Leveraging routinely collected electronic health records (EHRs) from multiple health-care institutions, this approach aims to assess the feasibility of using federated learning (FL) to predict the progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD). Methods: We analyzed EHR data from the OneFlorida+ consortium, simulating six sites, and used a long short-term memory (LSTM) model with a federated averaging (FedAvg) algorithm. A personalized FL approach was used to address between-site heterogeneity. Model performance was assessed using the area under the receiver operating characteristic curve (AUC) and feature importance techniques. Results: Of 44,899 MCI patients, 6391 progressed to AD. FL models achieved a 6% improvement in AUC compared to local models. Key predictive features included body mass index, vitamin B12, blood pressure, and others. Discussion: FL showed promise in predicting AD progression by integrating heterogeneous data across multiple institutions while preserving privacy. Despite limitations, it offers potential for future clinical applications. Highlights: We applied long short-term memory and federated learning (FL) to predict mild cognitive impairment to Alzheimer's disease progression using electronic health record data from multiple institutions. FL improved prediction performance, with a 6% increase in area under the receiver operating characteristic curve compared to local models. We identified key predictive features, such as body mass index, vitamin B12, and blood pressure. FL shows effectiveness in handling data heterogeneity across multiple sites while ensuring data privacy. Personalized and pooled FL models generally performed better than global and local models.