This analysis was done as part of a study designed to develop a tool to identify caregivers of people living with dementia through the EHR for pragmatic trials related to medication management. Caregivers of people living with dementia and clinical staff from 2 health systems in Maryland and Colorado were recruited from March to May 2022. Research staff screened the EHR to identify patients aged 65 years or older with dementia (based on International Classification of Diseases, Tenth Revision [ICD-10] codes or dementia medications; Multimedia Appendix 1) and polypharmacy (defined as taking 5+ medications), who received care at primary care or memory clinics. The dementia diagnosis was confirmed by reviewing clinic notes. Our approach was designed to be pragmatic to characterize the real-world caregiving arrangements of people with dementia. For this reason, we sought to include caregivers regardless of how much time they spent caring for the people living with dementia. To maximize generalizability, we used broad eligibility criteria for caregivers and allowed them to self-identify. Recruitment letters addressed to “family members or friends” were mailed to the homes of eligible people living with dementia. We then called the people living with dementia to identify a caregiver interested in participating. Staff who provided ambulatory care to people living with dementia were identified through lists provided by clinics and recommendations of clinic administrators. They were recruited by email. We stopped recruitment once no new themes were emerging in interviews.

Chart review of people living with dementia whose caregivers participated focused on characterizing whether caregiver identities, contact information, and roles were documented in the EHR (and if so, where) and whether caregivers had registered for proxy access to the patient portal. We developed a chart abstraction form in Microsoft Excel. Two research coordinators (RQR and AED) reviewed charts in the EHR and recorded their findings. These data were discussed with the principal investigator (ARG) during team meetings to identify and adjudicate discrepancies. We conducted semistructured interviews with 22 caregivers and 16 clinical staff (eg, clinicians, nurses, medical assistants, and practice managers). An interview guide was developed based on the research questions. We asked caregivers about their roles and caregiving arrangements and compared this with the information that was documented in the EHR. We also asked whether the patient’s health care team ever routinely asked them about caregiving roles and responsibilities and, if so, when. We asked staff how they currently identify caregivers and record their information. Both groups were asked about challenges related to this process and how it could be improved. All participants were also asked for demographic information. Four research team members (RQR, AED, KSG, and TETM) experienced in qualitative research conducted the interviews via phone or videoconference; all interviews were conducted in private locations. Two coders (KSG and TETM) developed an initial codebook based on the Consolidated Framework for Implementation Research and the study aims. Three study team members then refined the codebook by reading transcripts from 5 caregivers and 5 staff. Open coding allowed inductive identification of new themes in addition to deductive coding guided by the Consolidated Framework for Implementation Research. The entire team met to discuss and finalize the codebook. Using this version, a single team member coded approximately 25% (n=8) of the remaining transcripts, which were reviewed by another coder. Discrepancies were discussed at study team meetings until 100% agreement was reached. The remaining transcripts were then coded by a single coder, and the study team met weekly to discuss emerging themes.

The Johns Hopkins University School of Medicine Institutional Review Board (IRB) served as the single IRB and approved this research (IRB protocol 00297952); the Kaiser Permanente Colorado IRB ceded. All participants provided verbal consent. Caregivers at both health systems and clinical staff in Maryland received a US $25 gift card for participating; clinical staff in Colorado volunteered without compensation, as company policy did not allow it. All interviews were audio recorded and transcribed verbatim.

Participant characteristics are described in Table 1. People with dementia (n=22) had a mean age of 84.1 (SD 6.5) years. Most were male (n=15, 68%) and White (n=15, 68%) and had a mean of 8 (SD 3.5) medications.

Caregivers had a mean age of 63 (SD 12.6) years. The majority were female (n=21, 96%) and White (n=16, 73%). Most were the spouse, partner, or adult child of the people with dementia (n=18,82%) and lived with the people with dementia (n=14, 64%). Only 1 caregiver had proxy access to the patient portal. As shown in Table 2, caregiver names and their relation to the people with dementia were usually recorded in the “contact information” or “patient contacts” tab of the chart (n=20, 91%). Caregivers were frequently mentioned in clinical notes (n=21, 96%), where their relations and roles were described with a varying degree of detail (eg, patient lives independently and son helps). One of the health systems had a “lay caregiver” field that was used inconsistently (n=3, 27%). Most people with dementia (n=13, 59%) had an advance directive in the chart that named a medical decision maker. When an advance directive was present, it identified the same caregiver as interviewed only slightly more than half of the time (n=12, 55%). Of the 22 caregivers interviewed, the majority (n=17, 77%) reported that the people with dementia had additional caregivers. Most charts (n=16, 73%) contained documentation implying the presence of multiple caregivers; this information was recorded unsystematically in various locations throughout the chart, and caregivers’ individual contributions were not explicitly described. Two major themes emerged from interviews (described with illustrative quotations in Table 3).

Interviews revealed that most people living with dementia had multiple people who helped them, including family, friends, and community members. Some people who provided instrumental support did not view themselves as caregivers, complicating health system efforts to identify them. Caregivers said that their needs (eg, skills or resources needed for caregiving) were rarely assessed or recorded by the health system. Staff said caregiver information was not collected systematically and that caregiver roles, such as which caregiver helps with medical decision-making and daily activities, including medication administration, were usually described in clinical notes rather than readily accessible fields. This often made it time-consuming and difficult to identify the appropriate caregiver to contact for information exchange. Even when a caregiver was identified by name or relation to the patient, it was often not easy to determine if that caregiver was the relevant person for a particular clinical decision, such as a medication change.

Caregivers expressed a desire to be identified and have their needs routinely assessed and addressed. Clinical staff recommended developing workflows to identify caregivers and assess their needs at multiple touchpoints, including standardized questions in the waiting room, the patient portal, and during scheduling. Staff said that it would be useful to collect information on what tasks caregivers perform and suggested adding a caregiver tab to the EHR to standardize what information is collected and where it is recorded. Caregivers and staff suggested that proxy access to patient portals could facilitate improved identification, engagement, and support of caregivers. However, the complicated proxy enrollment process, which requires downloading and completing forms and providing proof of legal representative status if the patient lacks decision-making capacity, was seen as a barrier.

This exploratory chart review and qualitative study found that people living with dementia frequently have multiple caregivers, whose roles and needs are inconsistently captured in the EHR. Our findings were similar in both health systems. Caregivers emphasized the importance of having their needs routinely evaluated and addressed. Clinical staff suggested creating processes to identify caregivers and assess their needs at various points, such as using standardized questions in the clinic waiting room, through the patient portal, and during appointment scheduling. Participants recommended simplifying the process of obtaining proxy access to patient portals, particularly for caregivers of people living with dementia.

Similar to our findings, a previous study of 211 people living with dementia that did not involve interviews or determine whether people living with dementia had multiple caregivers found that 89% of charts identified caregivers; however, “an extensive search through individual notes” was required to locate caregiver information; few charts documented caregiver needs [8]. These findings have important implications. Caregiver engagement and support are essential components of high-quality dementia care. Patients with poorly supported caregivers face a heightened risk for inappropriate medication use [9], hospitalization, and burdensome treatments [10]. Pragmatic trials may exclude people with dementia when there is no documentation of a caregiver, limiting generalizability [11].

Earlier research has shown that patient portals can be a valuable resource for supporting patients with serious illness and their caregivers [12]. This is illustrated by the fact that patient portal use, such as secure messaging, tends to increase following a dementia diagnosis, reflecting the substantial need for information and communication of people living with dementia and their caregivers [13]. In a pilot study, caregivers of people living with dementia perceived that they received more information about resources from the doctor after completing a caregiver wellness questionnaire in the patient’s EHR [14]. Another study tested an EHR-driven process to identify veterans with unpaid caregivers based on receipt of home- and community-based referrals; however, the process was inefficient [4]. Shared access to the patient portal would facilitate the identification and enable the engagement of multiple caregivers, reflecting the reality of caregiving arrangements for people living with dementia [15]. However, health systems should simplify the complicated proxy enrollment process, which is a barrier [16,17]. To address the concerns identified in our study, health systems should develop and test workflows to identify caregivers, assess their needs at multiple touchpoints, and record their roles and contact information in standardized, clearly defined, and readily extractable fields. These workflows could include fielding caregiver questions as part of the Medicare Annual Wellness Visit, creating a communication plan at the time of dementia diagnosis, and testing different modes of outreach to people living with dementia and caregivers [15,16].

This study has some limitations. The small sample from 2 health care systems may not be generalizable to other practice settings. In addition, caregivers who could be identified during recruitment and who agreed to participate may be more likely to be engaged and identified in the EHR. Our broad inclusion criteria may have resulted in people who only spent a few hours caregiving being classified as caregivers. Although both health systems used the Epic EHR, the tabs and fields were different in each system, and the locations where we found caregiver information may not be generalizable. As a qualitative study, this work is designed to be exploratory. However, it is valuable because it provides an in-depth understanding of caregiving arrangements for people living with dementia and of the lived experiences of dementia caregivers in their interactions with the EHR. In so doing, the study generates important hypotheses about how health systems can better support caregivers. Finally, we did not seek to interview paid caregivers, who may be responsible for the day-to-day care of the people living with dementia and accompany them to medical appointments [18].

This qualitative study and exploratory chart review found that caregivers of people living with dementia are identified inconsistently in the EHR. If their roles and needs are documented, this information is not easy to locate. Future studies could develop and test workflows to identify caregivers, assess their needs, and record their information in extractable EHR fields.