LAUSR

To cite: Festa N, Moura LMVR, Blacker D, et al. BMJ EvidenceBased Medicine Epub ahead of print: [please include Day Month Year]. doi:10.1136/ bmjebm-2022-112134 © Author(s) (or their employer(s)) 2023. No commercial reuse. See rights and permissions. Published by BMJ. Festa and colleagues highlight underrecognised factors that may bias research, policy and population health strategies predicated on claims-based ascertainment of Alzheimer’s disease and related dementias within the USA Absent effective diseasemodifying therapies for Alzheimer’s disease and related dementias (ADRD), measures to prevent incident disease, delay symptom onset and prolong functional independence are paramount. Identifying effective preventive measures and care delivery models, in turn, requires improved information regarding modifiable risk factors and corresponding populationlevel interventions that may alter the trajectory of ADRD. In many countries, including the USA, administrative data offer a singular means of studying these relationships in large and representative populations. Researchers in the USA have increasingly drawn on administrative diagnostic information from the Centers for Medicare & Medicaid Services (CMS)—the major public payer for healthcare of the elderly—to ascertain ADRD status. For example, the often used CMS Chronic Conditions Warehouse (CCW) definition of ADRD has underpinned recent public health research identifying air pollution as a potentially modifiable contributor to neurodegenerative disease. The CMSCCW definition has also been used in health services research to examine the effects of local healthcare access on ADRDrelated outcomes. Moreover, clinical research has applied this definition to ascertain ADRD among persons for whom this diagnosis has important prognostic and treatmentrelated implications for comorbid conditions, such as cancer. Emerging applications of claimsbased ADRD ascertainment now include drugrepurposing and pharmacogenomics. Some limitations of ADRD ascertainment using claims have been enumerated, including selection bias due to underdiagnosis within administrative data. However, we argue that significant issues remain unaddressed, which may bias research, policy and population health strategies predicated on administrative data within the US Medicare population. Below, we discuss potential sources of bias that researchers should consider when using diagnostic claims to ascertain ADRD status. Our illustrative examples are focused within the US Medicare programme due to the proliferation of research measuring ADRD within this population. Many of the considerations that we outline, however, are universally applicable to research predicated on claimsbased ADRD ascertainment.