Gaps Seen in Dementia AI Device Transparency
Researchers from several academic medical centers say that only half of artificial intelligence (AI)- and machine learning (ML)-enabled devices authorized by FDA for Alzheimer’s disease and related dementias (ADRD) have demographic details reported for their training or validation datasets. Writing in a JAMA Research Letter, the authors say “AI- and ML-enabled devices hold promise for earlier diagnosis and continuous symptom monitoring for patients [with ADRD], potentially improving disease management and outcomes. However, device performance may vary across patient groups. Data transparency, particularly regarding demographic representativeness of training and validation datasets, is essential to understanding performance variability and ensuring appropriate application in intended populations.”
The study identified 24 FDA-authorized devices that have been available since 1/2015. Among them, six devices had multiple authorizations.
Of the 24 studied devices, 12 had no publicly available information about their training or validation datasets in either FDA summaries or peer-reviewed publications, the authors say. Training data were described for 10 devices in FDA summaries and five devices in peer-reviewed articles, but reporting of key demographic details was said to be limited. Validation data were also sparse, the study says, being reported for two devices in agency summaries and 10 devices in peer-reviewed publications.
“Existing disparities in ADRD diagnoses and care among racially and ethnically minoritized groups underscore the risk of unintended harms from underrepresentation in training and validation data,” the authors say. “Enforcing guidelines for data transparency and device labeling would help regulators identify potential biases and promote safe and effective AI and ML deployment in dementia. This approach would additionally support patients and clinicians in informed decision-making, ensuring devices are appropriately selected for diverse disease states and patient populations.”