Background: Digital and mobile health interventions using personalization via reinforcement learning algorithms have the potential to reach large number of people to support physical activity and help manage diabetes and depression in daily life. Objective: The Diabetes and Mental Health Adaptive Notification and Tracking Evaluation (DIAMANTE) study tested whether a digital physical activity intervention using personalized text messaging via reinforcement learning algorithms could increase step counts in a diverse, multilingual sample of people with diabetes and depression symptoms. Methods: From January 2020 to June 2022, participants were recruited from 4 San Francisco, California–based public primary care clinics and through web-based platforms to participate in the 24-week randomized controlled trial. Eligibility criteria included English or Spanish language preference and a documented diagnosis of diabetes and elevated depression symptoms. The trial had 3 arms: a Control group receiving a weekly mood monitoring message, a Random messaging group receiving randomly selected feedback and motivational text messages daily, and an Adaptive messaging group receiving text messages selected by a reinforcement learning algorithm daily. Randomization was performed with a 1:1:1 allocation. The primary outcome, changes in daily step counts, was passively collected via a mobile app. The primary analysis assessed changes in daily step count using a linear mixed-effects model. An a priori subanalysis compared the primary step count outcome within recruitment samples. Results: In total, 168 participants were analyzed, including those with 24% (40/168) Spanish language preference and 37.5% (63/168) from clinic-based recruitment. The results of the linear mixed-effects model indicated that participants in the Adaptive arm cumulatively gained an average of 3.6 steps each day (95% CI 2.45-4.78; P<.001) over the 24-week intervention (average of 608 total steps), whereas both the Control and Random arm participants had significantly decreased rates of change. Postintervention estimates suggest that participants in the Adaptive messaging arm showed a significant step count increase of 19% (606/3197; P<.001), in contrast to 1.6% (59/3698) and 3.9% (136/3480) step count increase in the Random and Control arms, respectively. Intervention effectiveness differences were observed between participants recruited from the San Francisco clinics and those recruited via web-based platforms, with the significant step count trend persisting across both samples for participants in the Adaptive group. Conclusions: Our study supports the use of reinforcement learning algorithms for personalizing text messaging interventions to increase physical activity in a diverse sample of people with diabetes and depression. It is the first to test this approach in a large, diverse, and multilingual sample.

@en

Cognitive behavioral therapy (CBT) is efficacious to treat depression, however more research is needed to understand its functions among Latinxs. This study analyzed qualitative responses that were paired with a mood rating (1–9 scale) from daily ecological momentary assessments via text-messaging of 52 low-income, Spanish-speaking patients to assess the relationship between word use and changes in mood during group CBT. Based on previous research, we chose 11 linguistic dimensions from the Linguistic Inquiry and Word Count text analysis software that conceptually related to core CBT treatment elements and sociocultural factors of depression in Latinxs. Results showed that the use of words from the categories of Friends, Religion, Positive Emotions, and Leisure (proxy for behavioral activation) were significantly associated with a significant increase in mood. The use of Negative Emotions and Health words were significantly associated with a significant decrease in mood. Post-hoc analysis revealed that Certainty (proxy for cognitive inflexibility) words were related to a significant decrease in mood when Negative Emotional words were present. Findings contribute to our understanding of the role of sociocultural factors and core CBT elements in changes in mood among Latinxs. Lastly, this paper demonstrates the potential for analyzing language content during a digital health intervention to better understand user experiences.@en

Background: Text messaging interventions can be an effective and efficient way to improve health behavioral changes. However, most texting interventions are neither tested nor designed with diverse end users, which could reduce their impact, and there is limited evidence regarding the optimal design methodology of health text messages tailored to low-income, low-health literacy populations and non-English speakers. Objective: This study aims to combine participant feedback, crowdsourced data, and researcher expertise to develop motivational text messages in English and Spanish that will be used in a smartphone app-based texting intervention that seeks to encourage physical activity in low-income minority patients with diabetes diagnoses and depression symptoms. Methods: The design process consisted of 5 phases and was iterative in nature, given that the findings from each step informed the subsequent steps. First, we designed messages to increase physical activity based on the behavior change theory and knowledge from the available evidence. Second, using user-centered design methods, we refined these messages after a card sorting task and semistructured interviews (N=10) and evaluated their likeability during a usability testing phase of the app prototype (N=8). Third, the messages were tested by English- and Spanish-speaking participants on the Amazon Mechanical Turk (MTurk) crowdsourcing platform (N=134). Participants on MTurk were asked to categorize the messages into overarching theoretical categories based on the capability, opportunity, motivation, and behavior framework. Finally, each coauthor rated the messages for their overall quality from 1 to 5. All messages were written at a sixth-grade or lower reading level and culturally adapted and translated into neutral Spanish by bilingual research staff. Results: A total of 200 messages were iteratively refined according to the feedback from target users gathered through user-centered design methods, crowdsourced results of a categorization test, and an expert review. User feedback was leveraged to discard unappealing messages and edit the thematic aspects of messages that did not resonate well with the target users. Overall, 54 messages were sorted into the correct theoretical categories at least 50% of the time in the MTurk categorization tasks and were rated 3.5 or higher by the research team members. These were included in the final text message bank, resulting in 18 messages per motivational category. Conclusions: By using an iterative process of expert opinion, feedback from participants that were reflective of our target study population, crowdsourcing, and feedback from the research team, we were able to acquire valuable inputs for the design of motivational text messages developed in English and Spanish with a low literacy level to increase physical activity. We describe the design considerations and lessons learned for the text messaging development process and provide a novel, integrative framework for future developers of health text messaging interventions.@en

Background: Social distancing is a crucial intervention to slow down person-to-person transmission of COVID-19. However, social distancing has negative consequences, including increases in depression and anxiety. Digital interventions, such as text messaging, can provide accessible support on a population-wide scale. We developed text messages in English and Spanish to help individuals manage their depressive mood and anxiety during the COVID-19 pandemic. Objective: In a two-arm randomized controlled trial, we aim to examine the effect of our 60-day text messaging intervention. Additionally, we aim to assess whether the use of machine learning to adapt the messaging frequency and content improves the effectiveness of the intervention. Finally, we will examine the differences in daily mood ratings between the message categories and time windows. Methods: The messages were designed within two different categories: behavioral activation and coping skills. Participants will be randomized into (1) a random messaging arm, where message category and timing will be chosen with equal probabilities, and (2) a reinforcement learning arm, with a learned decision mechanism for choosing the messages. Participants in both arms will receive one message per day within three different time windows and will be asked to provide their mood rating 3 hours later. We will compare self-reported daily mood ratings; self-reported depression, using the 8-item Patient Health Questionnaire; and self-reported anxiety, using the 7-item Generalized Anxiety Disorder scale at baseline and at intervention completion. Results: The Committee for the Protection of Human Subjects at the University of California Berkeley approved this study in April 2020 (No. 2020-04-13162). Data collection began in April 2020 and will run to April 2021. As of August 24, 2020, we have enrolled 229 participants. We plan to submit manuscripts describing the main results of the trial and results from the microrandomized trial for publication in peer-reviewed journals and for presentations at national and international scientific meetings. Conclusions: Results will contribute to our knowledge of effective psychological tools to alleviate the negative effects of social distancing and the benefit of using machine learning to personalize digital mental health interventions. Trial Registration: ClinicalTrials.gov NCT04473599; https://clinicaltrials.gov/ct2/show/NCT04473599 International Registered Report Identifier (IRRID): DERR1-10.2196/23592@en

Background: Social distancing and stay-at-home orders are critical interventions to slow down person-to-person transmission of COVID-19. While these societal changes help contain the pandemic, they also have unintended negative consequences, including anxiety and depression. We developed StayWell, a daily skills-based SMS text messaging program, to mitigate COVID-19–related depression and anxiety symptoms among people who speak English and Spanish in the United States. Objective: This paper describes the changes in StayWell participants’ anxiety and depression levels after 60 days of exposure to skills-based SMS text messages. Methods: We used self-administered, empirically supported web-based questionnaires to assess the demographic and clinical characteristics of StayWell participants. Anxiety and depression were measured using the 2-item Generalized Anxiety Disorder (GAD-2) scale and the 8-item Patient Health Questionnaire-8 (PHQ-8) scale at baseline and 60-day timepoints. We used 2-tailed paired t tests to detect changes in PHQ-8 and GAD-2 scores from baseline to follow-up measured 60 days later. Results: The analytic sample includes 193 participants who completed both the baseline and 60-day exit questionnaires. At the 60-day time point, there were significant reductions in both PHQ-8 and GAD-2 scores from baseline. We found an average reduction of –1.72 (95% CI –2.35 to –1.09) in PHQ-8 scores and –0.48 (95% CI –0.71 to –0.25) in GAD-2 scores. These improvements translated to an 18.5% and 17.2% reduction in mean PHQ-8 and GAD-2 scores, respectively. Conclusions: StayWell is an accessible, low-intensity population-level mental health intervention. Participation in StayWell focused on COVID-19 mental health coping skills and was related to improved depression and anxiety symptoms. In addition to improvements in outcomes, we found high levels of engagement during the 60-day intervention period. Text messaging interventions could serve as an important public health tool for disseminating strategies to manage mental health.@en

Introduction: Cognitive behavioral therapy (CBT) is an established treatment for depression, but its success is often impeded by low attendance. Supportive text messages assessing participants' mood in between sessions might increase attendance to in-clinic CBT, although it is not fully understood who benefits most from these interventions and how. This study examined (1) user groups showing different profiles of study engagement and (2) associations between increased response rates to mood texts and psychotherapy attendance. Methods: We included 73 participants who attended Group CBT (GCBT) in a primary care clinic and participated in a supportive automated text-messaging intervention. Using unsupervised machine learning, we identified and characterized subgroups with similar combinations of total texting responsiveness and total GCBT attendance. We used mixed-effects models to explore the association between increased previous week response rate and subsequent week in-clinic GCBT attendance and, conversely, response rate following attendance. Results: Participants could be divided into four clusters of overall study engagement, showing distinct profiles in age and prior texting knowledge. The response rate to texts in the week before GCBT was not associated with GCBT attendance, although the relationship was moderated by age; there was a positive relationship for younger, but not older, participants. Attending GCBT was, however, associated with higher response rate the week after an attended session. Conclusion: User groups of study engagement differ in texting knowledge and age. Younger participants might benefit more from supportive texting interventions when their purpose is to increase psychotherapy attendance. Our results have implications for tailoring digital interventions to user groups and for understanding therapeutic effects of these interventions. @en

Background: The COVID-19 pandemic has propelled patient-facing research to shift to digital and telehealth strategies. If these strategies are not adapted for minority patients of lower socioeconomic status, health inequality will further increase. Patient-centered models of care can successfully improve access and experience for minority patients. Objective: This study aims to present the development process and preliminary acceptability of altering in-person onboarding procedures into internet-based, remote procedures for a mobile health (mHealth) intervention in a population with limited digital literacy. Methods: We actively recruited safety-net patients (English- and Spanish-speaking adults with diabetes and depression who were receiving care at a public health care delivery system in San Francisco, United States) into a randomized controlled trial of text messaging support for physical activity. Because of the COVID-19 pandemic, we modified the in-person recruitment and onboarding procedures to internet-based, remote processes with human support. We conducted a preliminary evaluation of how the composition of the recruited cohort might have changed from the pre-COVID-19 period to the COVID-19 enrollment period. First, we analyzed the digital profiles of patients (n=32) who had participated in previous in-person onboarding sessions prior to the COVID-19 pandemic. Next, we documented all changes made to our onboarding processes to account for remote recruitment, especially those needed to support patients who were not very familiar with downloading apps onto their mobile phones on their own. Finally, we used the new study procedures to recruit patients (n=11) during the COVID-19 social distancing period. These patients were also asked about their experience enrolling into a fully digitized mHealth intervention. Results: Recruitment across both pre-COVID-19 and COVID-19 periods (N=43) demonstrated relatively high rates of smartphone ownership but lower self-reported digital literacy, with 32.6% (14/43) of all patients reporting they needed help with using their smartphone and installing apps. Significant changes were made to the onboarding procedures, including facilitating app download via Zoom video call and/or a standard phone call and implementing brief, one-on-one staff-patient interactions to provide technical assistance personalized to each patient's digital literacy skills. Comparing recruitment during pre-COVID-19 and COVID-19 periods, the proportion of patients with digital literacy barriers reduced from 34.4% (11/32) in the pre-COVID-19 cohort to 27.3% (3/11) in the COVID-19 cohort. Differences in digital literacy scores between both cohorts were not significant (P=.49). Conclusions: Patients of lower socioeconomic status have high interest in using digital platforms to manage their health, but they may require additional upfront human support to gain access. One-on-one staff-patient partnerships allowed us to provide unique technical assistance personalized to each patient's digital literacy skills, with simple strategies to troubleshoot patient barriers upfront. These additional remote onboarding strategies can mitigate but not eliminate digital barriers for patients without extensive technology experience.@en

Introduction Depression and diabetes are highly disabling diseases with a high prevalence and high rate of comorbidity, particularly in low-income ethnic minority patients. Though comorbidity increases the risk of adverse outcomes and mortality, most clinical interventions target these diseases separately. Increasing physical activity might be effective to simultaneously lower depressive symptoms and improve glycaemic control. Self-management apps are a cost-effective, scalable and easy access treatment to increase physical activity. However, cutting-edge technological applications often do not reach vulnerable populations and are not tailored to an individual's behaviour and characteristics. Tailoring of interventions using machine learning methods likely increases the effectiveness of the intervention. Methods and analysis In a three-arm randomised controlled trial, we will examine the effect of a text-messaging smartphone application to encourage physical activity in low-income ethnic minority patients with comorbid diabetes and depression. The adaptive intervention group receives messages chosen from different messaging banks by a reinforcement learning algorithm. The uniform random intervention group receives the same messages, but chosen from the messaging banks with equal probabilities. The control group receives a weekly mood message. We aim to recruit 276 adults from primary care clinics aged 18-75 years who have been diagnosed with current diabetes and show elevated depressive symptoms (Patient Health Questionnaire depression scale-8 (PHQ-8) >5). We will compare passively collected daily step counts, self-report PHQ-8 and most recent haemoglobin A1c from medical records at baseline and at intervention completion at 6-month follow-up. Ethics and dissemination The Institutional Review Board at the University of California San Francisco approved this study (IRB: 17-22608). We plan to submit manuscripts describing our user-designed methods and testing of the adaptive learning algorithm and will submit the results of the trial for publication in peer-reviewed journals and presentations at (inter)-national scientific meetings. Trial registration number NCT03490253; pre-results.

@en