Quantifying the Richness of the Boston Process Approach (BPA) Using a Novel Analytic Approach: the Framingham Heart Study

Abstract Objective Neuropsychological (NP) tests are often interpreted with a one test-one domain analytic approach in research, which misrepresents their clinical utility. The Boston Process Approach (BPA) allows for extraction of multi domain features within and across NP tests although its use is hampered by the data complexity and lack of objective quantification methods. In this study, we aim to develop a novel factor analytic approach that quantifies the richness of BPA data in the Framingham Heart Study (FHS). Method Data included BPA error scores (n = 172) derived from 10 baseline NP tests from 2238 Offspring participants. The outcome variable was dementia diagnosis. Factor analyses were conducted using Kemeny covariance matrix with maximum likelihood decomposition. Dwyer’s (1937) method was used to estimate factor loadings for demographic variables (sex, education, and age) and dementia status. Results Participants’ average age was 67.3 ± 9.3 years, 54.5% were female, 41.3% had education ≥ college degree, and 92.5% were without dementia. A bifactor model demonstrated adequate fit (TLI = 0.95, RMSEA = 0.03), similar to the best multi-factor model (TLI = 0.95, RMSEA = 0.03), and improved over a single factor model (TLI = 0.87, RMSEA = 0.05). Omega estimates revealed saturation in general factor versus total factor loadings (ratio: 0.90). Dementia status loaded highly on the general factor (0.82, h2 = 0.81), as did sex (0.44, h2 = 0.27). Conclusions BPA data fit a bifactor model, with most variation accounted for by the general factor. This study quantifies the richness of BPA measures into a single factor score that can be used as a predictor for several outcomes in clinical research.