November 1, 2026
A recent article published in Alzheimer’s & Dementia: The Journal of the Alzheimer’s Association by members of the IMPACT community including: Dae Hyun Kim, MD, ScD, Kevin W. McConeghy, PharmD, PhD, Jennifer L. Lund, PhD, Ryan Mace, PhD, David H. Lynch, MD, Elizabeth M. White, APRN, PhD, Hua Xu, PhD, Rafael Samper-Ternent, MD, PhD, Phillip Wang, MD, DrPH, and Ellen P. McCarthy, PhD, MPH addresses the ongoing challenge of using routinely collected healthcare data for measuring mild cognitive impairment (MCI) in embedded pragmatic clinical trials.
The authors explore three approaches for measuring MCI using routinely collected healthcare data: documentation-based approaches which identify clinician-recognized MCI, clinical criteria-based approaches which reconstruct the diagnostic construct using available clinical information, and prediction-based approaches, which estimate the likelihood of MCI using statistical or machine learning models. They describe the differences between these approaches including validity, scalability and susceptibility to bias.
Prioritizing clinical criteria-based approaches whenever feasible was recommended by authors. Documentation-based approaches can provide complementary measures of clinician-recognized MCI, while prediction-based approaches are best suited for screening and risk stratification rather than defining trial outcomes.
However, cognitive testing and functional status are often incompletely documented in electronic health records and administrative claims, while differences in healthcare access and documentation practices may also affect who is identified. The authors emphasize the need for validating measures across diverse populations, ensuring consistent measurement across trial groups, and promoting transparent reporting to support accurate and equitable MCI measurement.