Beyond the Uniform Data Set (UDS): Benefits of incorporating additional items for the measurement of memory, executive functioning, and language from the University of Pittsburgh Alzheimer’s Disease Research Center

Abstract

AbstractBackgroundAlzheimer’s Disease Research Centers (ADRCs) administer prescribed cognitive batteries – the Uniform Data Set (UDS) – and report UDS data to the National Alzheimer’s Coordinating Center (NACC). ADRCs may administer other items beyond the UDS. We used data from the University of Pittsburgh ADRC (Pitt) to determine psychometric implications of UDS items alone versus UDS items plus additional Pitt‐specific items.MethodWe used confirmatory factor analyses to co‐calibrate Uniform Data Set 1 and 2 (UDS1/2, broadly overlapping, 2005–2015) and UDS3 (2015‐2020) data for memory, executive functioning, and language. We added Pitt‐specific items for each domain. We compared measurement properties of UDS scores vs. UDS plus Pitt scores, including measurement precision and projected sample sizes needed to show a 25% reduction in rate of decline over 12 months for people with Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD).ResultThere were 2,726 participants with 8,602 visits across UDS1, 2, and 3 (Table 1). Figure 1 summarizes findings for memory. Figure 1a shows boxplots for memory scores for normal cognition (NC), MCI, and AD for UDS1/2 (green) and UDS3 (purple). Figure 1b shows corresponding boxplots and standard error of measurement (SEM) curves for UDS1/2 (left) and UDS3 (right). Incremental improvements in measurement precision can be seen with the lower SEM curve for the UDS plus Pitt curves. Figures 2 and 3 summarize similar findings for executive functioning and for language. In each case UDS Plus Pitt has less measurement error than UDS alone. Proportions of scores with standard errors >0.30 (a commonly used threshold for individual decision‐making) during UDS1/2 and UDS3 are in Table 2. Pitt‐specific items reduced the proportion with imprecise scores for every comparison with the exception of executive functioning in UDS1/2. Sample sizes needed to show a 25% reduction in the rate of decline are shown in Table 3. Improved precision from additional items was associated with greater power to show change over time.ConclusionIntegrating additional data beyond UDS1/2 and UDS3 results in better measurement precision and increased statistical power.