BACKGROUND: People with cognitive impairment commonly use central nervous system-active potentially inappropriate medications (CNS-PIM), increasing risk of adverse outcomes. Patient portals may be a promising tool for facilitating medication-related conversations. Little is known about portal use by this population related to medications. OBJECTIVE: To target portal interventions, we sought to identify individuals with cognitive impairment and CNS-PIM exposure who discussed medications through the portal and to determine how frequently their messages described possible adverse effects. METHODS: We used electronic health record (EHR) data from an academic health system in Maryland (Site 1) from 2017 to 2022 and pharmacy and EHR data from an integrated health system in Colorado (Site 2) in 2022 to identify people with cognitive impairment and CNS-PIM exposure who communicated about medications through the portal. At Site 1, message threads were manually categorized based on content. At Site 2, messages were categorized using natural language processing (NLP). RESULTS: The Site 1 cohort included 5543 patients aged ≥65 with cognitive impairment and ≥2 outpatient visits from 2017 to 2022. Over half (n = 3072
55%) had CNS-PIM prescriptions. Most with CNS-PIM prescriptions had portal use (n = 1987
65%)
1320 (66%) of those patients sent messages during possible CNS-PIM exposure. Coding of a 5% random sample of message threads revealed that 3% mentioned CNS-PIM and possible adverse effects, while 8% mentioned possible adverse effects without referencing CNS-PIM. At Site 2, 4270 people had cognitive impairment and CNS-PIM exposure in 2022
of these, 1984 (46%) had portal use and 1768 (41%) sent medication-related messages during CNS-PIM exposure. NLP identified 663 (8%) messages that mentioned CNS-PIM and possible adverse effects, while 726 (41%) mentioned possible adverse effects without referencing CNS-PIM. CONCLUSIONS: People with cognitive impairment and care partners frequently send portal messages about medications and possible adverse effects. Identifying such messages can help target deprescribing interventions.