Based on established relationships with health system and community partners, we will purposefully select interested parties for two groups: (1) health care administrators and frontline clinicians, and (2) people living with dementia and their care partners that meet eligibility criteria (see Table 1). Each group will form a design team and have 5-7 participants, which is within the standard range used in participatory design research [25]. The small number of participants in groups reduces the potential for groupthink (ie, by having products from one group evaluated by the other) and ensures that all participants have ample time to provide their perspectives [26].

Teams will complete 5 co-design videoconference sessions occurring in parallel across 4 months, with 2-3 weeks between each session. In session one, we will present the original CHAT. We will lead discussion, interpretation, and respectful debate among design team members to begin to generate the needs of dementia care partners and health care system process changes to optimize their involvement in care. We will ensure progress toward solution convergence from session to session. The design teams will conduct sessions independently, but in between sessions, they will swap solutions and provide feedback, which we will summarize and present in subsequent sessions. Each web-based session will last approximately 90 minutes and will be audio recorded for subsequent analysis. The design process can be conceptualized as a funnel in which we start with divergent ideas and then work toward convergence through the sessions. Participants will receive US $20 after completing each session.

Within the 2-3 weeks between each session, we will review audio recordings as well as any team and participant observation notes, case scenarios, and sketches. Two members of our team will use the Rapid Identification of Themes From Audio Recordings (RITA) method [27] to identify design specifications. The same 2 team members will then combine the themes from the RITA with the synthesized team and participant observation notes, use of case scenarios, and sketches to provide as inputs to participants at the following design session. Following the design session’s completion, we will adapt the CHAT for dementia care, applying input from the final design session, and performing member checking by reviewing changes with members from the groups. This process will produce a converged-upon D-CHAT that is ready for stage 1B feasibility testing [21].

This single-blind pilot randomized controlled trial [28] will occur on medical and surgical units at a large academic health care system in the Midwest. This system serves more than 700,000 patients each year with approximately 1849 physicians and 21,000 staff at 7 hospitals. Care partners will be asked to participate if they meet the following inclusion criteria: (1) provide unpaid care to a hospitalized adult relative or partner to help them take care of themselves because of a dementia diagnosis, (2) be at least 18 years or older, and (3) speak and understand English. Care partners of hospitalized adult relatives or partners that are observation or same-day stay patients will be excluded. There will be no exclusion based on sex, gender, race, or ethnicity.

Before recruitment begins, we will provide training on the D-CHAT to clinicians in medical and surgical units. To reduce contamination [29], we will instruct trained clinicians to not use the D-CHAT when treating those in the control arm and to not discuss details of the D-CHAT with nontrained clinicians. Once this training is completed, a member of the study team will work with unit coordinators to schedule morning, afternoon, and evening shifts for data screening, consent, and data collection. Institutional review board approval will be obtained to access electronic health records of hospitalized people living with dementia. We will use electronic health records to identify patients who have a documented care partner (or legally authorized representative). Someone who is part of the patient’s care team will then approach the patient’s present care partners to determine their interest in the study. If a care partner expresses interest in participating, a study team will then share study details, obtain written consent, and complete the baseline visit (see Figure 2). A member of the study team will monitor protocol deviations like unintended crossovers by completing weekly, in-person check-ins with the trained clinicians.

The baseline visit will be completed in a designated room in the hospital to limit distractions. After completing consent, participants will complete descriptive measures and the primary clinical outcome measure (see Table 2). All measures will be administered by a trained, blind assessor. Immediately after these procedures, participants will be randomized to either the intervention or control group, and the clinical team will be notified of the assignment. These efforts will minimize bias and increase internal validity [28].

Care partner participants randomized to the intervention group will receive the D-CHAT and usual care delivered through the patient’s care team. Aftercare partners complete the D-CHAT, and their responses will be given to the unit coordinator to review and make suggested referrals to specified clinicians. These clinicians then use the responses to develop and complete information and skill preparation with the care partners. For example, the D-CHAT could reveal that a care partner does not have the necessary diabetes knowledge or skill set to help give her husband living with Alzheimer disease and diabetes his insulin injections. Information gleaned from the D-CHAT would trigger a diabetes care and education specialist to address the care partners’ caregiving needs.

The comparison condition will be the usual care that hospitalized people living with dementia and their care partners receive as part of the large academic health care system. No standardized clinical decision-support tool that facilitates the identification and preparation of care partners is currently used for routine hospital-based dementia care. These care partner participants will complete the same measures as participants in the intervention condition on the same timeline using identical assessment procedures.

Participants will be assessed on the measures listed in Table 2. Participants will be compensated US $25 after each time point is completed. Baseline measures will be collected in person at the hospital, and follow-up assessments will be administered via phone by a study team member. The primary clinical outcome we will measure is a change in care partner preparedness. The Preparedness for Caregiving Scale [30] is a self-rated measure that consists of 8 items that ask care partners how well prepared they believe they are for multiple domains of caregiving. Preparedness is defined as perceived readiness for multiple domains of the caregiving role, such as providing physical care, providing emotional support, setting up in-home support services, and dealing with the stress of caregiving. Responses are rated on a 5-point scale, with scores ranging from 0 (not at all prepared) to 4 (very well prepared). The scale is scored by calculating the mean of all items answered with a score range of 0 to 4. The higher the score, the more prepared the care partner feels for caregiving. The items have an internal consistency rating of 0.88 to 0.93. We will also collect secondary outcome measures, including satisfaction of care, caregiving burden, and depression, using standardized measures.

An estimated sample size of 128 will be recruited for the main trial based on the 95% CI width around the effect of the D-CHAT on care partner preparedness as measured by the Preparedness for Caregiving Scale [31]. Since this is a novel tool that will be adapted as part of phase 1, the variability of the response is unknown, and therefore justification is based on general terms of SD. With 64 subjects in both groups, the 95% CI width of the estimated effect will be 0.35 × SD. We expect to see a medium effect size (Cohen d=0.5). With this 95% CI width and expected effect size, we will have a sufficiently narrow CI of the true effect size to apply toward the planning of the future efficacy trial. If there was a ≈12% loss to follow-up such that 56 in each group have follow-up data, the 95% CI width would only increase to 0.375 × SD.

Analyses will be conducted using SAS (SAS Institute Inc) [32]. We will screen univariate distributions for normality and influential outliers and confirm that the randomization process balanced demographic characteristics between groups. We will also document the number of participants who are lost in each group, the reasons for the loss, and unintended crossovers. This information will influence the design of a future trial, including the frequency and mode of communication with clinicians and participants. The effect of the D-CHAT on care partner preparedness will be estimated as the difference in mean change from baseline to study completion between groups. The 95% CI will be calculated using a t distribution. Cohen d effect size will be calculated as the effect of the D-CHAT divided by the pooled standard deviation. This will give us a sense of the magnitude of the effect in terms of SDs. If normality assumptions in the raw data are not met and a simple transformation of the data (logarithmic or square root) does not sufficiently normalize the distribution, then we will use a nonparametric equivalent methodology [33]. If greater than 10% of the participants are noncompliant with their group assignment, then a per-protocol sensitivity analysis will be conducted to assess how influential the noncompliance was on the results of the study. Noncompliance is based on whether the group assignment was followed.