Objectives: To investigate risk factors for progression to Alzheimer's disease and related dementias (ADRD) in African Americans and non-Hispanic Caucasians in a large US cohort.
Design: A matched case-control design using electronic health records (EHRs) from 2000 - 2017.
Setting: Cerner EHRs database covering more than 600 Cerner client hospitals.
Participants: 79,120 patients aged 65 and older (#ADRD=39,560, #non ADRD older adults=39,560) from an initial cohort of 49,826,000 patients.
Measurements: We converted ICD9 or ICD10 diagnosis codes into PheWas codes to increase clinical relevance. Then we detected ADRD as having both ADRD diagnosis codes and medications. We considered PheWas codes for Alzheimer's disease, dementia with cerebral degenerations, senile dementia, and vascular dementia. We considered ADRD medications including acetylcholine and memantine.
Results: Using two-step propensity score matching, we built an African American cohort of 4,429 and a 4,570-person matched Caucasian cohort that was similar in terms of onset age, observation length, sex, and known ADRD risks (diabetes, vascular disease, heart disease, head injury, and obesity). Older African Americans had a statistically significant progression from cerebrovascular risk (transient ischemic attack) to ADRD incidence (treatment effect coefficient = 0.0978, p-value <0.000) whereas the matched Caucasians did not (treatment effect coefficient = 0.403, p-value = 0.196).
Conclusion: Our extensive causal analysis using a nationwide EHR discovered disease progression pathways to ADRD. The carefully matched cohorts from different racial groups showed different progression, which partly explains the racial disparities in ADRD incidence.