Aging
Estimating dementia risk: leveraging electronic health records among study participants lost to follow-up Joseph Oliver Fong* Joseph Oliver Fong Fong Fong Fong Fong Fong Fong Fong Fong Fielding School of Public Health, UCLA
In cohort studies of aging, poor health, including incipient dementia, may influence continued study participation. The Kaiser Healthy Aging and Different Life Experiences (KHANDLE) study is a multi-ethnic cohort study of long-term Kaiser Permanente Northern California members that aims to examine lifecourse determinants of late-life brain health and cognitive decline. Participants were members of Kaiser at enrollment and completed 4 cognitive assessments (~18 months between each assessment) over 6 years of follow-up as part of the study. We evaluated dementia risk (i.e. cumulative incidence) using diagnosis information from study assessments alone (“study diagnosis alone”) versus diagnosis information supplemented with diagnoses from electronic health records (“EHR-supplemented diagnosis”) among those who dropped out. Dropout was defined as active/passive withdrawal from the study or missing >=2 visits. For “study diagnosis alone” analyses, participants were followed until study dementia diagnosis, death, dropout, or end of follow-up (4th assessment). Case ascertainment and censoring were similar in “EHR-supplemented” analyses; however, in these analyses those who dropped out were followed for incident EHR-supplemented dementia diagnosis from time to study dropout to time to last EHR contact time or 6 years, whichever came first. We used the Aalen-Johansen estimator to calculate dementia risk, with dementia-free death as a competing event (study time as timescale). In our analytic sample of 1609 participants, n=329 (20%) were lost to follow-up over 6 years. Dementia risk over 6 years was lower in study diagnosis alone analyses (n=53, 3%) versus EHR-supplemented analyses (n=101, 6%) (Figure). We captured additional incident dementia cases by supplementing with EHR data among participants who dropped out of the study; in future work, we will evaluate concordance of study-based and EHR-based diagnoses during follow-up.
