Scalable Technologies for the Assessment and Rescue of Brain Health: Brain-START - PROJECT SUMMARY Alzheimer's Disease and Related Dementias (ADRD) disproportionately affect low- and middle- income countries (LMICs), where approximately 60% of cases reside. Current diagnostic approaches, primarily developed in high-income countries, have not been validated in LMIC settings, where unique sociocultural and linguistic characteristics affect test performance and interpretation. Additionally, the performance characteristics of blood-based biomarkers for Alzheimer's disease (AD-BBMs) remain unknown in genetically heterogeneous populations like Kenya and Pakistan, creating a critical barrier to early and accurate detection. The study has three specific aims: (1) Validate an integrated ADRD detection toolkit against gold-standard measures (CSF and FDG-PET) in urban hospital settings (N=300); (2) Implement and evaluate the toolkit in rural community clinics using a task-shifting approach (N=1,000 in Kenya; N=200 pilot in Pakistan); and (3) Develop population-specific risk assessment models incorporating traditional and LMIC-specific risk factors. The research design employs a mixed-methods approach combining clinical validation, implementation science, and advanced analytics. For Aim 1, we will recruit participants through hospital-based referral networks and community outreach, collecting comprehensive cognitive, biomarker, and clinical data. The toolkit features first-of-its-kind digital speech analysis that captures acoustic features robust to the unique language use patterns of multilingual participants. This is integrated with multi-domain digital cognitive assessment targeting memory, executive function, and visuospatial abilities to support differential diagnosis of ADRD subtypes. The toolkit also includes olfactory testing and AD- BBMs (plasma Aβ42/40 and p-tau217), with establishment of the first reference standards for these biomarkers in East African and South Asian populations. Aim 2 implements a two-stage assessment strategy in rural settings, with initial screening by community health workers followed by comprehensive evaluation of screen-positive cases. Implementation outcomes will be evaluated using the RE-AIM framework. Aim 3 applies machine learning techniques to identify novel risk factors and interaction effects specific to LMIC populations. This work will generate new knowledge about the adaptation requirements for valid ADRD assessment in multilingual populations, establish region-specific biomarker reference standards, and develop implementation frameworks for resource-limited settings. Results will inform strategies for serving similar populations globally, including resource-limited communities in high-income countries, through validated early detection tools, population-specific risk assessment models, and healthcare delivery frameworks utilizing task-shifting approaches.
