This study is an exploratory secondary data analysis of a single-blind pilot RCT (National Institutes of Health Stage Model, Stage 1b) of HMP-C, a digital MBI tailored for caregivers of persons living with dementia.

This trial is registered with ClinicalTrials.gov (NCT05732038), and all study procedures were approved by the affiliated institutional review board (2022P001601) [25,28]. Informed consent was collected electronically. A member of the study staff was present to answer any questions. Participants received a total of US $50 for completing the postassessment and follow-up assessments. The privacy and confidentiality of all participant data were maintained.

The primary single-blind feasibility pilot RCT recruited informal adult caregivers of a known person living with Alzheimer disease or a related dementia reporting elevated levels of perceived stress. Recruitment was carried out through flyers and online advertisements. Interested caregivers referred themselves via an online link included in the recruitment materials. Here, they completed a brief eligibility questionnaire (verified by a phone call with study staff) based on the following inclusion criteria: (1) >18 years old, (2) English literacy, (3) self-identify as an informal caregiver, (4) Perceived Stress Scale-10 score of ≥6 [29,30], (5) willing to be randomized, and (6) care recipient rated >1 on the Functional Assessment Staging Tool [31]. Exclusion criteria included (1) any planned change in psychotropic pharmacological treatment for the duration of the study, (2) use of any consumer-based mindfulness meditation app for more than 60 minutes per month in the past 6 months, (3) current participation in a meditation program, (4) major illness anticipated to worsen dramatically or require surgery in the next 20 weeks, (5) active treatment for cancer (eg, chemotherapy, radiation), (6) placement of the care recipient in a nursing home, (7) involvement in another clinical trial for caregivers, and (8) 4+ errors on the Short Portable Mental Status Questionnaire (ie, for caregivers older than 65 years of age) [32]. The primary feasibility RCT analyzed 90 participants across the HMP-C intervention (n=46) and a control (n=44) condition [25]. The present secondary analysis utilized participants in the intervention condition only, as the control condition did not teach mindfulness skills.

HMP-C is a digital MBI tailored for caregivers of persons living with dementia, based on content from HMP, a publicly available mobile app [25,33]. HMP consists of written and guided listening content that aims to help users develop the following mindfulness skills: (1) awareness (eg, mindfulness, attention, self-awareness), (2) connection (eg, appreciation, kindness, compassion), (3) insight (eg, self-inquiry, self-knowledge, transcending the self), and (4) purpose (eg, purpose clarity, embodying values, finding meaning) [8,33]. Based on the Dahl et al [8] framework, these skills are taught to listeners through educational content based on the science of well-being and procedural guided mindfulness practices [25].

Participants were directed to use the app for at least 10 minutes per day, or 70 minutes per week, for 12 weeks. Participants were reminded each week by text message to use their respective phone app for the prescribed amount of time. App utilization data were reviewed weekly by study staff, and participants who did not use at least 75% of their recommended amount in the past week were sent an additional standardized reminder message. Participants who did not complete their recommended utilization for 2 consecutive weeks were called by a study staff member to problem-solve potential barriers to adherence. Throughout the 12-week trial period, participants were emailed a REDCap (Research Electronic Data Capture) link to weekly single-item stress and mindfulness assessments. Participants who completed all assessments were offered a monetary incentive of up to US $50.

Table 1 outlines participant characteristics at baseline for the 42 eligible caregivers included in this study. The average age was 55 (SD 14) years. Most participants were female by sex (n=39, 93%) and gender identity (n=38, 90%). Most participants were White (n=34, 81%). Most participants were married and living with their partner (n=28, 67%). Participants in this study used HMP-C for a median of 70 (IQR 41.7-102.2) minutes per week.

Median mindfulness and stress ratings at baseline were 4 (IQR 3-6, range 0‐8) and 7 (IQR 6.5-8, range 4‐10), respectively (Figure 1). By week 6 (ie, midway through the study), median mindfulness ratings increased to 7 (range 2‐10) and stress ratings decreased to 5 (range 0‐8). This represents a net increase in median mindfulness of 3 points and a net decrease in median stress of 2 points between baseline and week 6. Mindfulness and stress ratings remained similar throughout the remainder of the trial, with a net increase in median mindfulness and stress of 1 point between week 6 and 12 of the trial. Overall, there was an increase in median mindfulness of 3 points and a decrease in median stress of 1.5 points across the full trial (ie, baseline to week 12).

According to Figure 1, stress ratings appear to have the greatest decrease between baseline and week 3. By week 3, the median stress rating was 5 (IQR 4, range 0‐10), which corresponds to a 2-point decrease from baseline. Mindfulness appears to experience the greatest increase between baseline and week 4. By week 4, the median mindfulness rating was 6.5 (IQR 3, range 2‐10), which corresponds to a 2.5-point increase from baseline.

Figure 2 demonstrates the variability in the correlation between weekly mindfulness and stress ratings. Table 2 reports the results of the GLMM characterizing the relationship between mindfulness and stress. There was a significant fixed effect of mindfulness on stress, such that every additional point of mindfulness above the mean at baseline was correlated with a decrease in participant stress by an average of 0.5 points (P<.001). Similarly, participant stress was reduced by 0.06 (P=.007), on average, for each week throughout the duration of the study. However, the significant interaction between mindfulness and study week (β=.05, P=.001) showed that this association attenuated over time (ie, the stress-reducing effect of mindfulness was most potent during the early weeks of the trial and less potent during the later weeks in the trial). Random effects showed variability in baseline stress (τ₀₀=2.04) and the association between mindfulness and stress (τ₁₁=0.08) among participants. Notably, there was a high correlation (ρ₀₁=0.86) between intercepts and slopes, meaning that participants with higher baseline stress had a higher association between mindfulness and stress reduction over the course of the trial. Simulation results for the interaction term in this model show approximately 88.80% (95% CI 86.70-90.70) power to detect effects.

Figure 3 details the variability in the correlation between HMP-C utilization and stress and mindfulness ratings each week through the 12-week trial. Correlations largely reflect an inverse relationship, where utilization generally correlates with higher mindfulness and lower stress. This is reflected in Table 3, where the fixed effects of utilization on mindfulness and stress were both significant. For every additional 10 minutes of HMP-C utilization between baseline and week 1, participants rated their mindfulness as 0.14 points higher on average (P<.001). Similarly, for every additional 10 minutes of HMP-C utilization, participants rated their stress as 0.14 points lower on average (P<.001). Both models demonstrated an effect of time, where for every week of utilization, participants’ mindfulness increased by 0.10 (P<.001) and stress decreased by 0.07 (P<.001) on average. Both models had statistically significant interaction terms with study week. However, the effect sizes of these interaction terms were very small for both mindfulness (β=.02, P<.001) and stress (β=.01, P=.01). This suggests that while there is a statistically significant attenuation of the impact of HMP-C utilization on stress and mindfulness ratings, the size of that impact was very small, meaning that this correlation was mostly stable throughout the trial. Simulation results for the interaction term show approximately 99.50% (95% CI 98.84-99.84) power and 68.90% (95% CI 65.93-71.76) power in the mindfulness and stress models, respectively.

This secondary data analysis aimed to characterize weekly mindfulness and stress ratings among caregivers of persons living with dementia and explore how digital MBI utilization was associated with those ratings throughout 12 weeks of a feasibility trial. We found that, on average, participants’ mindfulness and stress ratings improved throughout the initial weeks of the study and then sustained that improvement throughout the remainder of the study duration. Further, we found that there was a significant association between mindfulness and stress ratings throughout the study duration among our sample of caregivers of persons living with dementia, such that in weeks when mindfulness was high, stress was low. However, this association became weaker as the study progressed, suggesting that perceived stress and mindfulness are less intertwined over time. Finally, HMP-C utilization was associated with changes in mindfulness and stress ratings, suggesting that continued utilization may be helpful for maintaining these improvements over time.

The sample of participants from this study in some ways reflects the broader population of caregivers of persons living with dementia and in other ways potentially deviates from the average caregiver of persons living with dementia. On the one hand, participants overwhelmingly identified as women. Women have historically undertaken the role of caregiving for persons living with dementia, and this finding is consistent across geographic regions. Further, women often spend more time caregiving than their male caregiver counterparts and often experience a higher level of caregiver burden [36]. Conversely, the present sample of participants is largely White, has attained a graduate degree, and is earning more than US $100,000 per year. Past work has shown that approximately two-thirds of caregivers of persons living with dementia are White (compared to roughly 4 in 5 caregivers in the present sample). Further, past work has shown that the average caregiver of persons living with dementia earns US $50,000 or less per year. Additionally, the present sample represents a cohort of individuals positioned to have greater access to and literacy with technology due to their demographic characteristics (eg, race, income, education, and age) [37]. The complex intersectionality of underresourced identities and technological access and literacy can impact digital intervention engagement. Indeed, past reviews have shown that technological literacy is a significant barrier to implementing digital interventions for caregivers of persons living with dementia, despite their high health literacy [38-44].

Both mindfulness and stress ratings improved throughout the first 6 weeks of the study, with most of the improvement occurring between the first 3 to 4 weeks. From there, mindfulness and stress ratings leveled off throughout the remainder of the study. These findings suggest that participants may have developed some level of mastery in mindfulness skills within the first 4 weeks of HMP-C utilization, with more minor improvements occurring after week 4. Since Dahl et al [8] conceptualize mindfulness practice through a series of developable skills, it would be expected that once those skills are mastered, continued practice would lead to maintenance (ie, not a linear increase). This study reflects this trajectory, where caregivers improved mindfulness and stress early (weeks 1‐4) and then stabilized by week 6. This may reflect acquisition of new skills followed by subsequent consolidation of those skills. It remains plausible that continued practice would yield additional benefit at longer follow-up. Future work should evaluate this by extending assessments over a longer period of time.

Improvements in weekly mindfulness ratings correlated with stress reductions throughout the study duration. However, there is context to this finding. First, there was variability in baseline stress and the association between mindfulness and stress across participants. Notably, there was a positive correlation between baseline stress and the effect of mindfulness on stress reductions, meaning those who were the most stressed at baseline benefited the most from increases in mindfulness. This suggests that mindfulness skills may have offered these participants a strategy to immediately manage stressful situations in their lives. However, the association between mindfulness and stress decreased over time, regardless of baseline stress levels. This trajectory in mindfulness and stress ratings reflects several previous findings, where an initial effect of mindfulness skills on the primary outcome is observed, followed by a period of attenuation [45-48]. These findings underscore the importance of considering both individual differences at baseline and the temporal dynamics of mindfulness practice. The early pronounced association for those with higher stress highlights mindfulness as a potentially powerful mechanism for alleviating acute distress (ie, state mindfulness for coping). At the same time, the subsequent attenuation of effects suggests that the role of mindfulness may shift from producing large, immediate reductions in stress to supporting the maintenance of stress reduction over time (ie, trait mindfulness). This pattern is consistent with theoretical models, such as Dahl et al [8], that view mindfulness as a set of skills that, once acquired, foster a sense of resiliency and stability, rather than indefinite linear improvement. Taken together, these results suggest that digital MBIs should emphasize intensive early engagement while also providing education and strategies to sustain and deepen benefits during later phases of practice. Future research should aim to assess the mediating role of mindfulness in stress reduction for caregivers of persons living with dementia as compared to an attention-placebo control condition in a fully powered efficacy trial.

HMP-C utilization was broadly associated with increases in mindfulness and reductions in stress ratings throughout the duration of the study. Despite a statistically significant interaction of HMP-C utilization and time in study weeks in both utilization models, the effect sizes of these interactions were very small. While there may be some minor attenuation of the association between HMP-C utilization on mindfulness and stress over time, the association was largely maintained throughout the study duration. This suggests that continued utilization of HMP-C may serve as an important tool for mindfulness skill maintenance, such that utilization may be protective against returning to baseline levels of mindfulness and stress. Therefore, continued use of the digital MBI would be recommended for achieving and sustaining improvements in mindfulness and stress among caregivers of persons living with dementia. These findings suggest several directions for future work. First, to examine if and to what extent a dose-response relationship exists between HMP-C utilization and mindfulness or stress, a future study could randomize participants to incrementally different “dosages” of HMP-C utilization (eg, 45, 60, 75, and 90 min/wk). Further, future research is needed to address whether mindfulness skills and their effect on stress can be maintained without continued practice or the same amount of continued practice. For example, future studies may choose to randomize participants to continued, reduced, or no HMP-C utilization after an initial period of 4 weeks. These would provide the rigorous groundwork needed to tune digital MBI utilization recommendations.

There are several limitations to consider when interpreting the findings from this study. First, the participants in this study were a small sample from a stage 1b feasibility RCT. In many ways, the sample was not representative of caregivers of persons living with dementia broadly with respect to race, income, education, and potentially digital literacy. Due to this, the generalizability of these findings to the broader population of caregivers may be limited. Further, analytically, we were not powered to control for possible sources of confounding, such as digital and health-related literacy, which may have led participants to be more adept at improving from mindfulness skill practice. Second, this study does not include a control or comparator arm in the analysis. This precludes making causal statements about the efficacy of HMP-C or digital MBIs with respect to changes in mindfulness and stress. Possible sources of confounding due to lack of randomization include seasonal changes, regression toward the mean, or simply random fluctuations. Third, single-item weekly ratings are less explanatory than more comprehensive measures. For example, from this study, we cannot ascertain exactly what stressors participants were experiencing or were being mitigated. However, single-item measures have been shown to be comparable to full-scale measures as broad measures of stress or mindfulness, suggesting that the overall trends observed are likely valid [49-51]. Fourth, week-over-week changes in mindfulness and stress may be reflective of the Hawthorne effect or assessment reactivity, as opposed to true changes in the measured constructs.

This study is an exploratory secondary analysis of a feasibility pilot RCT of HMP-C, a digital MBI tailored for caregivers of persons living with dementia. We found that mindfulness and stress ratings improved throughout the first 3 to 4 weeks of the trial before leveling off throughout the remainder of the trial. Particularly, in these first 4 to 6 weeks of skill acquisition, mindfulness and stress were significantly negatively related to one another. This association was higher for those with high baseline stress. Together, these findings highlight the potential benefit of digital MBIs for stress reduction among caregivers of persons living with dementia and suggest that mindfulness skills may be acquired within the first 4 to 6 weeks of the intervention and then maintained over a longer period of time. Future research is needed to move closer to precise dosing recommendations for mobile apps with this population.