UCSF

This dissertation applies methodological innovations in real-world data to examine three areas: (1) the extent of detection bias in electronic health record (EHR)-based research on clinical predictors of dementia, (2) the potential impact of COVID-19-related brain MRI changes on long-term dementia incidence, and (3) the association between the age at onset of herpes viral infection and dementia incidence. All three chapters utilize data from the UK Biobank, with Chapter 1 also incorporating EHR data from the All of Us Research Program. Each chapter uses a unique analytical approach. Chapter 1 presents the first empirical evidence of detection bias in EHR-based dementia research, demonstrating that healthcare encounters for common conditions increase the likelihood of dementia detection, leading to spurious associations between clinically diagnosed conditions and subsequent dementia risk. Chapter 2 finds that naturally occurring variations in imaging-derived phenotypes linked to COVID-19 are associated with higher dementia incidence, suggesting a potential neurobiological link between COVID-19 and dementia. Chapter 3 finds that individuals with a herpes simplex virus (HSV) infection before age 65 have an increased incidence of dementia compared to those without an HSV infection. This dissertation contributes to dementia epidemiology by addressing key methodological challenges and leveraging large-scale real-world data to enhance our understanding of infectious disease-related risk factors for dementia.