Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Predictors of mHealth engagement in a mindful eating intervention for Type 2 diabetes
(USC Thesis Other)
Predictors of mHealth engagement in a mindful eating intervention for Type 2 diabetes
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
PREDICTORS OF MHEALTH ENGAGEMENT IN A MINDFUL EATING INTERVENTION FOR TYPE 2 DIABETES by Kinnari (Nina) Jhaveri, M.A. A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (PSYCHOLOGY) December 2022 Copyright 2022 Kinnari (Nina) Jhaveri PREDICTORS OF MHEALTH ENGAGEMENT ii Acknowledgments A daughter once asked her mother, “What’s a ‘legacy’?” Her mother explained it can be many things. Legacy, she said, is something that’s passed on from one person to another, from one generation to another. Legacy is larger than any one individual. It transcends place, time, even death. Her daughter furrowed her brow. “I don’t get it.” Her mother smiled. “You will when you help someone in need. Or when you teach a class of students. Or if you have kids of your own someday. It’s the mark you will leave on the world.” Today, I write about the people whose legacy this dissertation represents. To my advisor, Dr. Stan Huey: I thank you for your unwavering dedication to the minds and lives of your students. You make me an immeasurably better thinker, writer, and scientist. I’m living my dream because you took a chance on me. I will continue to cherish our discussions, debates, and chuckles as we embark on new projects together with the omnipresent <awk> and <wc> that grace our manuscripts. To Dr. Ashley Mason: Thank you for your fierce advocacy and boundless support through the years. I cannot and do not want to imagine this journey without you. Here’s to many more years of collaboration and friendship. To my amazing committee, Stan, Ashley, and Drs. Gayla Margolin, Mark Lai and Britni Belcher: My deepest gratitude for everything you do. From attending the defense while on vacation to coaching me through a nail-biting switch in statistical software halfway through analysis, you offered me the support and scaffolding I needed while challenging me to go beyond my comfort zone. To my wonderful lab mates: Thank you for your brilliance and your compassion as you supported me through the roller coaster of a Ph.D. program. To Annemarie, Crystal, Mariel and Shubir: You are the best cohort I could have asked for. I have no doubt we will continue to share in each other’s laughter, tears, and transformative moments as we move forward to make an impact on this world. PREDICTORS OF MHEALTH ENGAGEMENT iii Legacy transcends place, time, even death. To the women in Salumbar and Kelwara, India: Thank you for welcoming into your homes a young foreign woman with foreign ideas about women’s health. Words fail as I attempt to honor the ways you have shaped me. You are why I strive to be better, do more, and dedicate myself to what matters. धन्यवाद (thank you) । To my Aajeevika, Indicorps and Berkeley families: How lucky I am to have you in my life. The many jokes about my having been a lifelong student never get old. To my parents: Mamma, all my life, I have stood on your shoulders. You made it possible for me to accomplish the ambitions you had to abandon. Our bond, unique in its strength, brilliance, humor, and light-heartedness, marks your legacy. Thank you for always cheering me on. Pappa, your quiet wisdom has taken me farther than you will ever know. Above all, thank you for expressing your love through the gift of song. Music elevates my joy, brings meaning to my sorrow, and helps me through the challenging moments of life. To my parents-in- law, Mom and Dad: You gave me unconditional love and support when I decided to go back to school. You always ask us for what we need before you think of your own needs. I am blessed to have you as my family. And, to my partner, Ankit. We have been through and accomplished more together than I could have imagined. As my honorary co-author, you remind me of my resilience when I struggle. You took on the million other aspects of our lives as I wrote this dissertation, all the while navigating the demands of your own career. You fill our lives with playfulness and a sense of curiosity, you teach me to be empathetic, patient and kind, you don’t hesitate to plunge into the new and the unexpected, and you went from never wanting a pet to adoring Curie the kitty more than any man has ever loved a cat. Thank you, for everything. PREDICTORS OF MHEALTH ENGAGEMENT iv Finally, this legacy begins with my grandparents, Dada and Ba. I carry forth the values you instilled in me for education and learning. I hope to support the education of others as you did for so many. Dada, you spoke to me days before your passing that I should do whatever I wish most to do in my life. You did not ask me to conform. I dedicate this dissertation to you, and to the many individuals like you who were not availed the privilege of a higher education and who persevered to forge their own paths. PREDICTORS OF MHEALTH ENGAGEMENT v Table of Contents Acknowledgments........................................................................................................................ ii List of Tables ............................................................................................................................... vi List of Figures .............................................................................................................................. vii Abstract ........................................................................................................................................ viii General Introduction .................................................................................................................... 1 Study 1: Does Education Level impact Engagement through Impulsivity in a Mobile Health Intervention for Type 2 Diabetes? .................................................................... 8 Methods.................................................................................................................... 12 Results ...................................................................................................................... 22 Discussion ................................................................................................................ 28 Study 2: Do Food Craving and Mindfulness Interact to Impact Engagement with a Mindful Eating mHealth Intervention? ......................................................................... 33 Methods.................................................................................................................... 39 Results ...................................................................................................................... 48 Discussion ................................................................................................................ 53 General Discussion ...................................................................................................................... 57 References .................................................................................................................................... 63 Tables ........................................................................................................................................... 75 Figures.......................................................................................................................................... 91 Appendices ................................................................................................................................... 94 Appendix A: Patient Health Questionnaire.............................................................. 94 Appendix B: Food Cravings Questionnaire-Trait-Reduced .................................... 95 Appendix C: Five Factor Mindfulness Questionnaire ............................................. 96 PREDICTORS OF MHEALTH ENGAGEMENT vi List of Tables Table 1 ......................................................................................................................................... 75 Table 2 ......................................................................................................................................... 76 Table 3 ......................................................................................................................................... 77 Table 4 ......................................................................................................................................... 78 Table 5 ......................................................................................................................................... 79 Table 6 ......................................................................................................................................... 80 Table 7 ......................................................................................................................................... 81 Table 8 ......................................................................................................................................... 83 Table 9 ......................................................................................................................................... 84 Table 10 ....................................................................................................................................... 85 Table 11 ....................................................................................................................................... 86 Table 12 ....................................................................................................................................... 87 Table 13 ....................................................................................................................................... 88 Table 14 ....................................................................................................................................... 89 PREDICTORS OF MHEALTH ENGAGEMENT vii List of Figures Figure 1 ........................................................................................................................................ 91 Figure 2 ........................................................................................................................................ 92 Figure 3 ........................................................................................................................................ 93 PREDICTORS OF MHEALTH ENGAGEMENT viii Abstract Mobile health interventions delivered through smartphones (“mHealth”) provide access to behavioral support tools with more ease and frequency than in-person interventions. Engagement, broadly defined as a participant’s use of an mHealth intervention, is a critical factor in achieving improved health outcomes. However, mHealth studies report variability in engagement, highlighting the need to study engagement and identify factors that influence engagement. The two present studies examined adherence (the extent to which a person’s behavior corresponds with prescribed recommendations) and usage (an individual’s total use of intervention tools) as indices of engagement in an mHealth mindful eating intervention targeting craving-related eating in type 2 diabetes. Study 1 tested education level as a predictor of mHealth engagement, and whether delay discounting (an index of cognitive impulsivity) explained the relationship between education and engagement. Education was negatively associated with weekly and overall adherence and impulsivity did not mediate this relationship. No direct or indirect effects were found for mHealth usage. Study 2 examined trait food craving as a predictor of engagement and tested trait mindfulness as a moderator of this relationship. Trait mindfulness interacted with trait food craving to impact overall adherence but not weekly adherence or daily usage. These studies advance the literature by exploring objective mHealth engagement data and testing predictors of engagement that may inform the development of tailored interventions to optimize engagement and improve health outcomes. PREDICTORS OF MHEALTH ENGAGEMENT 1 Predictors of mHealth Engagement in a Mindful Eating Intervention for Type 2 Diabetes General Introduction Mobile Health and Engagement Smartphone technology represents a rapidly emerging delivery channel for interventions that aim to change health by promoting particular behaviors (Michie & Abraham, 2004). Interventions delivered via smartphone applications (“apps”) are termed mobile health (“mHealth”) interventions (WHO Global Observatory for eHealth, 2011). mHealth interventions may ease barriers specific to in-person treatment engagement such as scheduling, travel, and time commitments (Haliwa et al., 2021). mHealth also enables individuals ongoing access to the intervention by allowing them to carry intervention tools with them wherever they go, potentially enabling individuals to use behavioral support tools with greater ease and frequency compared to in-person interventions (Free et al., 2013). Therefore, mHealth is purported to enhance accessibility by integrating interventions into the daily lives of users. This may be particularly relevant for self-care behaviors that involve repeated, ongoing maintenance to achieve desired health benefits, such as improving eating behavior. Although mHealth is touted as a promising intervention delivery platform, recent reviews suggest mixed evidence of mHealth effectiveness (Byambasuren et al., 2018). Some mHealth apps demonstrate efficacy in improving target health outcomes (e.g., Proudfoot et al., 2013, 2013; Roepke et al., 2015), while others show limited to no effectiveness (see, e.g., Direito, Jiang, Whittaker, & Maddison, 2015; Laing et al., 2014; Turner-McGrievy & Tate, 2011). Researchers have proposed variability in user engagement as a possible explanation for these mixed effects (Pham et al., 2019). While user engagement has been defined in many ways, one definition refers broadly to a participant’s use of the intervention. Within this context, studies PREDICTORS OF MHEALTH ENGAGEMENT 2 employ varying terms, such as adherence or compliance (the extent to which a person’s behavior corresponds with the requirements of the intervention), attrition (referring to individuals who drop out of the intervention), and usage or dosage (the level of exposure to, or use of, intervention tools; see, e.g., Hebert et al., 2010; Leung et al., 2017). Regardless of definition, low levels of engagement are commonly reported in mHealth trials (Szinay et al., 2020). Eysenbach (2005) coined the term “law of attrition,” which emphasizes early and rapid decline in engagement as an inherent problem in technology-delivered interventions. While research has generally focused on mHealth intervention outcomes, the role of engagement, specifically in the case of interventions promoting ongoing self-care behavior (e.g., eating behavior, meditation, or physical activity), is proposed as a critical factor in achieving improved health outcomes (Young et al., 2019). Preliminary investigations suggest that mHealth engagement is associated with target intervention outcomes in several areas of health, including weight loss, physical activity and stress reduction (Carter et al., 2013; Mattila et al., 2013, 2016). However, engagement is a largely understudied mechanism in achieving desired health outcomes and is an emerging priority in mHealth research (Yardley et al., 2016). Application of mHealth: Mindful Eating in Type 2 Diabetes mHealth engagement may be particularly relevant in the management of Type 2 diabetes mellitus, a chronic disorder of the endocrine system characterized by an inability to properly use or produce the hormone insulin (Olokoba et al., 2012). Individuals with diabetes are at a higher risk of heart disease, stroke, kidney disease, and premature mortality (Centers for Disease Control and Prevention, 2019). Successful management of diabetes includes eating a healthy diet, which is shown to improve glycemic control (i.e., control of blood glucose levels) and reduce the risk of diabetes-related complications (Wing et al., 2001). PREDICTORS OF MHEALTH ENGAGEMENT 3 Maintaining a healthy diet requires frequent and sustained engagement with healthy eating behavior. One type of intervention that targets eating behavior is based on the principles of mindful eating. Mindful eating is the practice of bringing awareness in a nonjudgmental manner to the experience of eating, which includes physical hunger, satiety cues, and environmental or emotional triggers to eat (Miller et al., 2014). There is substantial evidence that mindful eating interventions reduce episodes of overeating and improve eating regulation (Daubenmier et al., 2011; Kristeller et al., 2014; Kristeller & Hallett, 1999; Mason et al., 2018). Mindful eating may be associated with improved eating patterns such as increased fruit and vegetable consumption and reduced intake of energy-dense foods (Beshara et al., 2013; Jordan et al., 2014). Through practice over time, eating mindfully is posited to interrupt habitual eating behaviors and help regulate food choice, and there has been a call for applying mindful eating to diabetes management (Miller, 2017). In a study conducted among adults with type 2 diabetes, a mindful eating intervention incorporated training in mindful meditation, mindful eating, and bodily awareness of hunger and satiety cues (Miller et al., 2012). Participants evidenced a significant reduction in weight and significant improvement in glycosylated hemoglobin, a measure of glycemic control. These findings indicate that training in mindful eating may help people with diabetes develop more healthy eating patterns. Engagement is particularly relevant in mindfulness-based interventions as regular practice is theorized as essential for developing mindfulness skills, defined as the awareness that arises when paying attention in the present moment, on purpose and non-judgmentally (Carmody & Baer, 2008; Kabat‐Zinn, 2003). In a sample of women with binge eating disorder who completed a 6-week mindful eating intervention, the time spent in the practice of eating-related mindfulness exercises was significantly associated with degree of improvement in binge eating PREDICTORS OF MHEALTH ENGAGEMENT 4 (Kristeller & Hallett, 1999). A study testing a mindful eating intervention in a nonclinical sample offered support for a dose response effect, defined as the relationship between the applied dose/concentration (e.g., the level of exposure to an intervention) and the effect that is observed (e.g., changes in target treatment outcomes). Specifically, this study found that participants who engaged for greater amounts of time in the mindful eating intervention had greater improvements in a measure of behavioral flexibility, the primary intervention outcome (Janssen et al., 2018). These findings are corroborated in studies of digital mindfulness-based interventions in which greater levels of usage are related to significantly greater gains in target outcomes (Carmody & Baer, 2008; Trompetter et al., 2015). The present research investigated participant engagement with a mindful eating mobile health intervention designed to target eating in response to food cravings (Mason et al., 2019). Food cravings are defined as intense urges or desires to eat specific types of foods and can pose challenges to adherence to diet recommendations, including for people with diabetes (Weingarten & Elston, 1991; Yu et al., 2013). Training in mindfulness may equip individuals with skills to help them gain awareness of and observe their physical and psychological experience of food cravings without eating in response to them (Kabat‐Zinn, 2003; Tapper, 2018). For example, a pilot study of an mHealth mindful eating intervention found that individuals who completed an average of two or more app modules per week over 3 months evidenced a stronger association between craving-related eating and weight compared to participants who watched fewer modules per week (Mason et al., 2018). Such findings, along with the reported prevalence of food cravings among individuals with diabetes (Brod et al., 2016; Yu et al., 2013), underscore the importance of investigating engagement in the mHealth intervention under study. PREDICTORS OF MHEALTH ENGAGEMENT 5 Intervention-Specific Measures of Engagement As noted earlier, there are multiple definitions of engagement in the literature. It is often argued that engagement should be defined and operationalized in relation to the purpose of the specific mHealth intervention (Yardley et al., 2016). Therefore, we define engagement in the context of the mindful eating program under study, below. While there is ongoing debate about how to best measure engagement, a few studies have focused on two key dimensions, namely, adherence and usage. Adherence is the extent to which a person’s behavior corresponds with the requirements of the intervention (van Dulmen et al., 2007). For example, researchers of an mHealth intervention may prescribe completing a specific set of activities every week. Adherence measures the extent to which a participant follows this recommendation. Usage is defined as an individual’s overall utilization of intervention tools or, put another way, their level of activity within the intervention (Donkin et al., 2013; Mattila et al., 2016). For example, the intervention may include optional activities in addition to the aforementioned “prescribed” activities, all of which fall under the individual’s usage of the intervention. Usage metrics may be calculated in various ways, such as total number of app- based activities completed over a given time period, total time (e.g., in minutes) spent using the app, or the average minutes spent in the app each time a user logged into the app. The present research studied engagement in “Eat Right Now” (ERN), a mindfulness- based app targeting craving-related eating (Mason et al., 2018). ERN has a total of 28 psychoeducational modules and a number of additional mindfulness-based tools and exercises. Study instructors asked participants to watch 2-3 psychoeducational modules per week over a 12-week intervention period, which corresponds to completing all 28 modules within 12 weeks. We operationalized adherence by measuring: 1) the number of modules completed per week, PREDICTORS OF MHEALTH ENGAGEMENT 6 herein referred to as “weekly adherence” and 2) the total number of modules completed within the stipulated 12-week period, herein referred to as “overall adherence”. To conceptualize usage, we turned to mindfulness research. Studies indicate that regular practice is essential for developing mindfulness skills and they recommend integrating mindfulness into daily life in order to experience therapeutic benefits (Carmody & Baer, 2008). Furthermore, the ERN app targets eating behavior – specifically, food cravings and eating in response to food cravings. Research on food cravings suggests that individuals may experience cravings on a daily basis (Richard et al., 2017). This indicated measuring usage per day. The daily frequency of mindfulness-based activities is conceptually tied to integrating mindfulness into daily life. This app specifically provides mindful eating training; given that most individuals consume meals or snacks multiple times per day, measuring the number of daily “activities” (i.e., app tools, exercises and modules) was indicated. This measurement is herein referred to as “daily usage”. In sum, the three engagement measures used include: total modules completed (“overall adherence”), modules per week (“weekly adherence”), and number of activities per day (“daily usage”). Of note, some researchers conceptualize intervention engagement as including direct use of the intervention as well as “offline” use (Yardley et al., 2016). In our case, this would correspond with a participant eating mindfully without using the app. While there are limitations of relying solely on mHealth app use to study engagement, several studies demonstrate a dose- response effect between app use and targeted outcomes, lending support to our approach to define engagement as the extent to which people use an mHealth intervention as objectively measured by app use data (Pham et al., 2019). The Present Research: Predictors of Engagement PREDICTORS OF MHEALTH ENGAGEMENT 7 In light of reported variability in mHealth engagement, it is important to identify individual-level factors that may predict engagement. Unfortunately, such research has been hampered by the lack of a systematic approach in defining engagement. Nevertheless, some studies have attempted to identify demographic, behavioral, and psychosocial predictors of mHealth engagement. For example, in an mHealth medication adherence promotion intervention for low-income adults with type 2 diabetes, engagement was measured by assessing participants’ responses to daily text messages and participation in weekly telephone calls. Researchers found that lower probability of engagement was associated with being non-White, reporting depressive symptoms and having lower health literacy (Nelson et al., 2016). Limitations of such trials include, in addition to inconsistent definitions of engagement, small sample sizes and short duration of follow-up. Overall, individual-level factors that impact engagement are not well understood in the context of mHealth. Identifying predictors of engagement can potentially inform the development of tailored interventions to optimize engagement and, correspondingly, improve health outcomes. This dissertation explored conceptually supported predictors of mHealth engagement among individuals with type 2 diabetes enrolled in an mHealth mindful eating intervention. Study 1 examined education as a predictor of mHealth engagement and tested whether impulsivity mediated this relationship. Study 2 examined trait food craving as a predictor of engagement and tested trait mindfulness as a moderator of this association. PREDICTORS OF MHEALTH ENGAGEMENT 8 Study 1: Does Education Level impact Engagement through Impulsivity in a Mobile Health Intervention for Type 2 Diabetes? Type 2 diabetes mellitus is a chronic disorder of the endocrine system characterized by an inability to properly use or produce the hormone insulin (Olokoba et al., 2012). Individuals with diabetes are at a higher risk of heart disease, stroke, kidney disease, and premature mortality (Centers for Disease Control and Prevention, 2019). The world’s prevalence of diabetes is estimated at 463 million people and is projected to rise to 700 million by 2045 (Saeedi et al., 2019). As prevalence rises, there is a need to better understand the determinants of diabetes and ways to improve diabetes management. Socioeconomic status (SES) is an example one such determinant. The inverse relationship between SES and diabetes prevalence is well documented across a number of socioeconomic status indicators (e.g., income, education, occupation;(Agardh et al., 2011; Houle et al., 2015; Robbins et al., 2005). Although the causal pathways are not fully understood, numerous factors have been cited to explain the relationship between SES and chronic illnesses such as diabetes, including inadequate access to health care, nutritional inadequacies and other inequalities in material circumstances, and unhealthy behaviors, to name a few. Several studies focus on access and quality of health care services as the main determinants; however, the impact of SES on health appears to persist even after accounting for access to health care services (Marmot et al., 2008). Researchers have also proposed that disparate measures of SES cannot be used interchangeably, that each measure acts through unique causal pathways, and that the most appropriate measures of SES may vary by the health outcome in question (Festin et al., 2017). Educational status is shown to be an important socioeconomic determinant of health. Although education covaries with other SES measures, research suggests that schooling has a PREDICTORS OF MHEALTH ENGAGEMENT 9 substantial independent effect on health (Feinstein et al. 2006). For example, a widely cited meta-analysis found that education is negatively correlated with disease-related mortality even after statistically controlling for other indicators of SES (Baker et al., 2011). For diabetes in particular, a large population study demonstrated that education is the strongest predictor of increased diabetes-related morbidity over income or occupation (Geyer et al., 2006). In another large cohort study of all-cause and cause-specific death rates, diabetes was among the diseases for which the inverse relation between education and mortality was the strongest (Steenland et al., 2002). These findings support the use of education as an appropriate SES indicator in the specific context of diabetes. Research also suggests differences across education levels in engagement with health promoting behaviors. Such behaviors are particularly relevant in the case of diabetes as successful management includes eating a healthy diet, which is shown to improve glycemic control (i.e., control of blood glucose levels) and reduce the risk of diabetes-related complications (Wing et al., 2001). Maintaining a healthy diet, in turn, requires frequent and sustained healthy eating behavior. Although engagement – broadly defined as a participant’s use of an intervention – is infrequently reported in behavioral interventions targeting diabetes, studies that do report engagement in educationally diverse samples show that fewer years of schooling are associated with poorer adherence. For example, in a study that investigated diabetes treatment adherence across a large national survey as well as a randomized clinical trial, the authors found that higher levels of education were associated with greater treatment adherence (Goldman & Smith, 2002). Another report of adherence rates from three different Internet-based trials of health-promoting interventions (diabetes self-management, smoking PREDICTORS OF MHEALTH ENGAGEMENT 10 cessation, and use of online personal health records) demonstrated educational status as a positive predictor of intervention engagement (Wangberg et al., 2008). Diminishing socioeconomic disparities in smartphone ownership (Pew Research Center, 2019) have given rise to a prevailing opinion that interventions delivered via mobile phones, termed mobile health or “mHealth”, will emerge as a solution to equalize access to and engagement with behavioral health interventions. mHealth purportedly addresses a number of barriers specific to in-person interventions such as scheduling, travel, and time commitment (Syed et al., 2013). However, the association between educational status and mHealth engagement is largely unknown, for several reasons. First, there is a dearth of educationally diverse samples in the mHealth literature, limiting the ability to study whether education levels are associated with engagement. Second, engagement is a relatively understudied construct in mHealth; therefore, studies that do include educationally diverse individuals generally do not measure engagement. It is possible that less educated individuals may face barriers to engagement that are unlikely to be fully addressed by an mHealth platform, leading to the question of whether having less education predicts poorer engagement in mHealth. A proposed mechanism explaining the relationship between education and health behavior is that schooling enables the development of higher-order cognitive skills that impact health promoting behaviors (Baker et al., 2011). Evidence suggests that formal schooling is associated with increases in domain-general cognitive processes (i.e., working memory, inhibitory control, and attention-shifting processes) that translate into improved decision-making (Ceci, 1991; Peters et al., 2010). One such decision-making process is delay discounting, the tendency to choose smaller, more quickly available rewards rather than larger rewards available later. Such choices are considered as indicators of impulsivity given that waiting would result in PREDICTORS OF MHEALTH ENGAGEMENT 11 a larger reward (Teuscher & Mitchell, 2011). There is clear evidence from multiple studies that higher levels of education are associated with less impulsivity as measured by delay discounting (Green et al., 1994, 1996; Reimers et al., 2009). These processes may be particularly relevant in the study of intervention engagement, since poor engagement may partly result from attributing a lower value to longer-term health as compared to other more immediate and proximal goals. Researchers hypothesize that an individual with poor engagement may be someone who prefers the immediate reward of not engaging with the intervention (e.g., extra time, rest, other activities) over the future implications of engagement, which is prevention of long-term disease complications (Lebeau et al., 2016; Reach, 2008). Since commitment to engagement requires a person to alter current behaviors with the goal of preventing a future outcome, discounting the future may be a risk factor for not engaging in preventive health behaviors. Indeed, a study of over 1000 participants from the Health and Retirement Survey showed that greater delay discounting (i.e., greater preference for smaller immediate rewards) was a predictor of poorer engagement with health promoting behavior (Bradford, 2010). Separate bodies of literature have linked education to cognitive responses such as impulsivity and impulsivity to engagement with health behavior. Considering these findings, does impulsivity mediate the association between education and mHealth engagement? The present study explored this question by studying engagement in “Eat Right Now” (ERN), a mobile mindful eating intervention (Mason et al., 2018). We hypothesized that having less education would be associated with poorer mHealth engagement. Moreover, we expected that impulsivity would mediate the association between education and engagement, such that less PREDICTORS OF MHEALTH ENGAGEMENT 12 educational attainment would predict greater impulsivity, which in turn would predict poorer mHealth engagement. While there are inconsistent approaches to defining engagement in the literature, researchers argue that engagement should be operationalized in context of the specific mHealth intervention (Yardley et al., 2016). In the present study, we measured engagement as adherence (the extent to which a person’s behavior corresponds with prescribed recommendations) and usage (an individual’s total use of intervention tools). ERN has a total of 28 psychoeducational modules and a number of additional mindfulness-based tools and exercises. Study facilitators instructed participants to complete all 28 modules over 12 weeks by watching 2-3 psychoeducational modules per week. We operationalized adherence by measuring the number of modules completed per week (“weekly adherence”) and the total number of modules completed within the stipulated 12-week period (“overall adherence”). To operationalize usage, we tracked all daily activities participants completed in the app (i.e., app tools, exercises and modules) and analyzed usage per day, or “daily usage”. Exploring daily usage was indicated as most individuals consume meals or snacks multiple times per day and may experience cravings on a daily basis (Richard et al., 2017). Methods Overview The present investigation derived from the Diabetes Education to Lower Insulin, Sugars, and Hunger (DELISH) Study, a randomized controlled trial targeting diabetes-related outcomes among individuals with type 2 diabetes (Mason et al., 2019). All procedures for Human Subjects research were approved by the institutional review board at the University of California San Francisco (UCSF). The DELISH Study compared the effect of two interventions: a nutrition PREDICTORS OF MHEALTH ENGAGEMENT 13 education alone arm (Ed) and a nutrition education with mindfulness-based intervention components (Demby) arm. Participants in both arms attended 12 weekly group-based in-person classes. Ed+MBI participants also received the Eat Right Now (ERN) program, a mindful eating intervention in the form of a mobile app (Mason et al., 2018). The ERN program was administered in combination with the weekly group classes. The Delish Study was conducted across two trials: the R61 trial randomized 60 participants in a 1:1 ratio to the Ed + MBI and Ed arms. The R33 trial randomized 120 participants in a 1:1 ratio to these arms. Across the R61 and R33 trials, investigators conducted a total of eight rounds of participant recruitment, enrollment, and intervention assessment. Each round consisted of running a single Ed + MBI group and a single Ed group. The present study investigated mHealth intervention engagement in the Ed + MBI arm of the R61 and R33 trials, totalling 91 Ed + MBI participants across all eight rounds. Participants Participants were English speaking adults (≥ 18 years) who screened positive for type 2 diabetes based on glycosylated hemoglobin (HbA1c) levels of 6.5%≤ HbA1c<12.0%. Additional eligibility criteria included: 1) Eating in response to food cravings at least twice over the course of three days, 2) Being able to engage in light physical activity, 3) Being willing and able to participate in the interventions, and 4) Having a smartphone and being willing to use it on a regular basis. Participants were recruited from several sources, including Facebook, Nextdoor, and Craigslist as well as within UCSF clinics. Procedures Mindful Eating Intervention PREDICTORS OF MHEALTH ENGAGEMENT 14 The 28-module ERN app is designed to deliver mindful eating training and has been tested in women with overweight who report overeating in response to food cravings (Mason et al., 2018), defined as intense desires or longings to eat a specific food (Weingarten & Elston, 1991). Participants in that study experienced reductions in craving-related eating as well as reductions in trait-level measures of overeating, including food cravings. ERN modules teach participants to attend to three aspects of eating: why they eat, including environmental and emotional triggers; what types of food are most likely to lead to cravings; and how to eat with awareness and mindful attention. Each module lasts approximately 5 to 10 minutes and users can access only one new module per day (after completing the previous module). Users have unlimited access to previous modules. DELISH study instructors asked participants to complete 2 or 3 modules per week during the 12-week intervention period. In addition to the modules, participants could access additional tools and exercises to aid in mindfully “riding out” food cravings. After the 28 modules were completed, all tools and previous modules remained available for the duration of the trial (see Mason et al., 2018 for a description of intervention content). As noted earlier, the DELISH Study included weekly in-person group classes that involved a dietary education component and mindful eating component. Participants in both study arms (Ed and Ed+MBI) attended the dietary education portion. Ed+MBI participants also attended the mindful eating portion. The mindful eating content was not didactic in nature but rather centered around discussing participants’ experiences with mindful eating practices, troubleshooting obstacles, and engaging in and reflecting on group exercises. Study Visits. Participants completed several measures at a baseline study visit. While the DELISH study collected data at multiple pre- and post-intervention visits, the present research PREDICTORS OF MHEALTH ENGAGEMENT 15 utilized questionnaire data collected at the baseline visit in addition to app engagement data collected over the 12-week intervention period. Measures We collected a range of self-report, computerized, and clinical measures as well as mHealth engagement measures. We describe the subset of measures used in the present study, below. Demographics. During the baseline visit, participants reported education level. Only three participants reported highest education attained as completing high school or the GED. Therefore, we combined this category with participants who had completed some college (re- coded categories were some college or less, associate's or 2-year degree, bachelor’s degree, and master's/doctoral degree). In addition, participants reported age, gender, race/ethnicity (re-coded to African American/Black, Asian/Pacific Islander, White/Caucasian, Latino/Hispanic, and Other), employment status (re-coded to employed full-time, employed part-time/student, unemployed, and retired), household income, marital status, total household size and number of children in the household. Patient Health Questionnaire (PHQ-8). The PHQ-8 is a standardized, well-validated 8- item measure of depressive symptoms (Kroenke et al., 2009; Appendix A). Participants responded on a 4-point scale: Not at all (0); Several days (1); More than half the days (2); Nearly every day (3). Higher scores indicated higher levels of depressive symptoms. The PHQ-8 demonstrated good reliability in the present sample (Cronbach’s alpha = 0.86). Validity of the PHQ-8 in several populations has been well established and the measure has demonstrated good construct validity and criterion validity (Kroenke et al., 2009). PREDICTORS OF MHEALTH ENGAGEMENT 16 Delay Discounting Task. The Delay Discounting Task is a cognitive measure of impulsivity. It assesses the extent to which individuals value delayed versus immediate rewards. Individuals who discount delayed rewards at a high rate are more likely to engage in behaviors such as overeating (Appelhans et al., 2011). A 5-trial adaptation of the original Delay Discounting task has been validated and is an effective way to assess discount rates while reducing participant burden (Bickel & Marsch, 2001; Koffarnus & Bickel, 2014). This task was administered to participants using a computerized program that asked participants to make choice decisions between smaller, immediate monetary rewards and larger, delayed monetary rewards. Delay discounting was calculated as the area under the empirical discounting function (AUC; Myerson, Green, & Warusawitharana, 2001). The AUC for delay discounting ranged from 0 (steepest possible discounting) to 1 (no discounting) and is herein referred to as “delay discounting AUC”. Therefore, higher delay discounting AUC scores indicated lower levels of impulsivity. mHealth Engagement. mHealth engagement was defined as participants’ use of the mindful eating app and was measured with objective user log data downloaded from the app. As noted earlier, we collected measures of overall adherence (total modules completed), weekly adherence (modules completed per week), and daily usage (activities per day) over the 12-week intervention period. Overall Adherence. Participants completed the app-related portion of the study intervention by watching all 28 psychoeducational modules within the 12-week intervention period. To measure overall adherence, we summed the total number of modules watched, ranging from 0 to 28. PREDICTORS OF MHEALTH ENGAGEMENT 17 Weekly Adherence. We calculated weekly adherence, or number of modules completed per week, by summing the total number of new modules a participant watched per week. The app allowed participants to watch a maximum of seven modules over a 7-day timespan. Daily Usage. We calculated daily usage, or number of activities completed per day, by summing the number of times a participant completed optional tools or exercises or watched a psychoeducational module within a 24-hour time span, defined from 12:00 am to 11:59 pm for each day. Data Management and Univariate Testing Across our analyses, we operationalized mHealth engagement as measures of overall adherence, weekly adherence, and usage, referred together as “engagement”. The overall adherence variable summed all modules a participant completed over the study period to provide an overall count. This variable was measured at the participant level and did not vary over time. The weekly adherence variable captured the number of modules a participant completed in a week over the 12-week study period and was measured at the week level. The daily usage variable captured the number of activities a participant completed within a day over the 78-day study period and was measured at the day level. Unlike overall adherence, both weekly adherence and daily usage varied across time. We conducted univariate assumptions tests on all study variables to assess normality and identify extreme outliers. We used visual assessments such as histograms and normal QQ-plots for normality testing. Using boxplots, we identified extreme outliers as any observation greater than three times the inter-quartile range (Field, 2013; Tabachnick & Fidell, 2007). These univariate tests indicated that there were non-normal distributions in all three of the outcome variables in addition to delay discounting AUC (the area under the curve for delay discounting). PREDICTORS OF MHEALTH ENGAGEMENT 18 The overall adherence variable had considerable left skew with most participants having completed all 28 modules (68.13%). We recoded this variable into a binary variable that assigned any participant who completed all 28 modules a one (1), and any modules less than 28 a zero (0). The weekly adherence variable demonstrated some right skew but approximated a normal distribution. Finally, the daily usage variable demonstrated considerable right skew with a substantial percentage of the distribution at zero, due to there being days when participants did not complete any activities. We did not find extreme outliers (i.e., outside three times the interquartile range) except for those variables that had extreme skew (overall adherence and daily usage). We addressed outlier issues by recoding overall adherence into a binary variable and modeling the skewed distribution of daily usage (negative-binomial). We grand-mean centered variables to improve interpretation. Analyses indicated missing data in the educational status, ethnicity/race, income, and the adherence or usage variables, constituting 1.90% of the overall proportion of missing data at the participant level, 1.17% of the overall proportion of missing data at the week level, and 2.15% of the overall proportion of missing data at the day level. We employed multiple imputation for missing value replacement to preserve power and to reduce potential bias due to missingness (Little & Rubin, 2020). Using MPlus, we conducted imputation using multiple chained equations and creating 10 imputed datasets to pool the results of the analysis models. Analytic Plan We conducted descriptive analysis of the frequencies and percentages of selected categorical study variables. Similarly, we calculated means, standard deviations, and minimum/maximum values for continuous study variables. Preliminary analyses examined PREDICTORS OF MHEALTH ENGAGEMENT 19 whether potential covariates (age, gender, race, household income, employment status, depression), independent, and mediator variables were associated with each other to explore relationships in the data and establish if multi-collinearity was an issue. We also conducted nonparametric equivalent tests for each preliminary analysis to ensure that effects were still statistically significant in comparison to parametric tests (e.g. Kruskal-Wallis for ANOVA and Spearman’s rho for correlation). As participants completed the intervention across eight distinct rounds, we investigated whether participant round was significantly associated with the outcome variables of interest; analyses indicated no significant association. The primary research aim was to determine whether education predicted mHealth engagement and to examine impulsivity (measured by delay discounting AUC) as a mechanism by which education was linked to engagement. We tested the relationship between the independent variable (education) and the outcome (engagement). To test delay discounting AUC as a mediator of this relationship, we employed the joint-significance test of mediation that tests individual components of the indirect effect (Yzerbyt et al., 2018). Studies show that the joint- significance test helps address inflated Type 1 errors that are relatively more inherent in tests of a single mediational index. We ran separate models testing for the relationship between education and delay discounting AUC (the “a-path”) and for the relationship between delay discounting AUC and engagement (the “b-path”). We tested for the indirect effects using a joint significance test of the a- and b-paths, using the distribution-of-product method to calculate the 95% confidence intervals to determine whether the indirect effect was statistically significant (Tofighi & MacKinnon, 2011). The joint test was conducted using the a-path and b-path model parameter coefficients and standard errors at the following website (https://amplab.shinyapps.io/MEDCI/), PREDICTORS OF MHEALTH ENGAGEMENT 20 which employs RMediation, an R software package for mediation analysis confidence intervals (R Core Team, 2020; Tofighi & MacKinnon, 2011). These analyses involved a combination of multi-level and ordinary regression models. The nested structure of time-varying engagement measures results in non-independence of observations due to clustering effects (Raudenbush & Bryk, 2002). Specifically, mHealth engagement as measured by daily usage and weekly adherence varied within individuals (Level 1) such that an individual may have a different value for each measure depending on the day (or week). The weekly adherence and daily usage variables were the only within-participant variables that were time-varying in addition to the time trend variable (week or day). All other measures varied between but not within individuals (Level 2). To account for the nested structure of these data, we utilized a series of multi-level model analyses (Singer & Willett, 2003). Specifically, when analyzing daily usage or weekly adherence, a 2-2-1 model was conducted to structure the analysis for the independent variable and mediator at Level 2 and the outcome variable at Level 1. One of our adherence measures, overall adherence, varied between but not within individuals. Given the lack of nested data in this case, multi-level models were not indicated, and we performed ordinary regression analyses when testing for the association between education and overall adherence. As noted earlier, the literature examining predictors of mHealth engagement is limited. Some studies of digital health engagement that do explore such predictors have shown age, gender, race, income, employment status, and depressive symptoms to impact engagement (Nelson et al., 2016; Reinwand et al., 2015; Rung et al., 2020). Therefore, we added these variables as covariates to the models. PREDICTORS OF MHEALTH ENGAGEMENT 21 The three engagement measures required different analytic approaches due to their varied distributions. For overall adherence, we applied binary logistic regression to assess the probability of a participant “completing all modules”. Due to an approximate normal distribution for weekly adherence, we used a linear mixed model with a random intercept, along with the maximum likelihood robust (MLR) estimator using robust standard errors in MPlus (Muthén & Muthén, 2017). For daily usage, which demonstrated right skew with a substantial percentage of the distribution at zero, we used a zero-inflated negative binomial multilevel model with a random intercept. The negative binomial model helps account for the overdispersion of the variance exceeding the mean. In addition, the zero-inflated model was necessary to account for the high proportion of zero counts in the outcome (Cameron & Trivedi, 1998). Due to the multilevel structure of the data, we grand-mean centered the predictor variables to provide accurate interpretation of the results with predictor variables that have no natural zero-point (Algina & Swaminathan, 2011). Accordingly, we reported effect sizes to interpret the strength and direction of the effects, including unstandardized beta coefficients, odds ratios (OR) for logistic regression, and incident rate ratios (IRR) typical of count data. We reported model fit statistics using relative fit statistics typical of multilevel modeling, including the log-likelihood value, Akaike information criterion (AIC), and Bayesian information criterion (BIC). We computed intra-class correlations (ICC) for the multi-level models to assess the level of variance in the outcome explained by the effect of cluster, in this case participant. The ICC varies between 0 and 1 and is calculated by dividing the cluster variance by the total variance (cluster and within variance). An ICC closer to 1 indicates a stronger effect of the cluster on the outcome and a greater justification for a multilevel model. PREDICTORS OF MHEALTH ENGAGEMENT 22 An ICC closer to 0 indicates a weaker effect of cluster on the outcome and less of a justification for the use of a multilevel model (Raudenbush and Bryk 2002). We conducted descriptive and preliminary analysis in Stata version 17 (StataCorp., 2021) analysis software. The primary analyses for this study used MPlus version 8.8 (Muthén & Muthén, 2017) analysis software. We employed MPlus for the primary analyses as it could better accommodate the complexity of multiple imputation and allows for multilevel mediation analysis with zero inflated count data. Results Descriptive Statistics Table 1 presents descriptive statistics for all relevant categorical demographic variables. The majority of the sample was female (62.64%). The largest ethnic group identified as White/Caucasian (46.15%) followed by Asian/Pacific Islander (16.48%). Over half (52.75%) of participants reported an annual household income of $75,000 or greater. Approximately 40% of participants were employed full-time and 29.67% of participants were retired. Participants reported level of education with approximately 26.37% completing some college or less. Most participants completed all 28 of the modules (68.13%). Table 2 presents means and standard deviations for all continuous variables. Participant age ranged from 21 to 80 years (M = 59.09, SD = 10.52). Delay discounting AUC scores ranged from .02 to .99 with a mean score of .71 (SD = .27); higher scores indicated lower levels of impulsivity. Depression scores ranged from 0 to 21 (M = 6.43, SD = 4.76). Participants averaged 25.93 total modules (SD = 5.09) out of a maximum of 28 modules, 2.16 modules per week (SD = 1.93), and 2.92 activities per day (SD = 4.26). Preliminary Analysis PREDICTORS OF MHEALTH ENGAGEMENT 23 Table 3 presents Pearson’s chi-square crosstabulations of relationships between categorical demographic variables and education level. The relationship between income levels and education was significant (χ 2 (21) = 41.67, p = .005, Cramer’s V = .41). Greater proportions of participants with higher levels of education tended to earn higher income levels and lower proportions of participants with less levels of education tended to earn lower income levels. The relationships of gender (χ 2 (3) = 1.54, p = .672, Cramer’s V = .13), race/ethnicity (χ 2 (12) = 9.96, p = .620, Cramer’s V = .19), and employment status (χ 2 (9) = 10.25, p = .331, Cramer’s V = .19) with education did not reach statistical significance. Table 4 summarizes mean scores for continuous measures by education level using analyses of variance (ANOVA) model statistics and comparative means. The relationship between age and education level was statistically significant (F (3,86) = 6.80, p < .001). The mean age for participants with some college or less (M = 51.71, SD = 11.85) was significantly lower than the mean age for participants completing a bachelor’s degree (M = 61.41, SD = 9.47) or master’s/doctoral degree (M = 63.15, SD = 8.05). There was a significant relationship between delay discounting AUC and education (F (4,86) = 3.25, p = .026). However, there were no statistically significant differences between any particular education level and delay discounting AUC. Analyses revealed no significant relationship between depression and educational status. A series of ANOVAs examining the relationships between categorical demographics and delay discounting AUC mean scores indicated no significant relationships (all p > .05), suggesting no significant relationship between delay discounting and the demographics evaluated. An additional series of ANOVAs revealed no statistically significant relationships between depression and demographics (all p >.05). Finally, Pearson’s correlations (r) indicated no significant relationships between mean scale scores on the continuous study variables. PREDICTORS OF MHEALTH ENGAGEMENT 24 Primary Analysis We tested for the association between education and engagement and examined delay discounting AUC as a mediator of the effect of education on engagement. These analyses controlled for relevant participant demographics as well as depression with the engagement outcome. Overall Adherence Table 5 presents model statistics for overall adherence for all model pathways. The c-path model had the following model fit statistics: loglikelihood = -40.27, AIC = 110.53, BIC = 148.19. Education was a significant predictor of overall adherence, with participants who had some college or less being significantly less likely to complete all modules (B = -1.825, OR = .161, p = .037) compared to participants who had a graduate degree. Participants who had some college or less were 6.203 times less likely to complete all modules compared to participants with a graduate degree. In addition to education, income level had a negative relationship with overall adherence (B = -.614, OR = .541, p = .003) suggesting that as income level increased, the likelihood of completing all modules decreased. Also, depression had a significant negative association with overall adherence (B = -.257, OR = .773, p = .001), suggesting that as levels of depression increased, the likelihood of completing all modules decreased. No other covariates were significant in the model. For the mediation model, the a-path captures the relationship of education with delay discounting AUC which was treated as an outcome variable for the purposes of this regression and measured on a continuous scale. This model had the following model fit statistics: loglikelihood = -3.084, AIC = 16.17, BIC = 28.72. The results showed a significant effect of education on delay discounting AUC, indicating that participants that had some college or less PREDICTORS OF MHEALTH ENGAGEMENT 25 had lower delay discounting AUC (i.e., higher levels of impulsivity) compared to participants who had graduate degrees (B = -.152, p = .028). In addition, participants with a bachelor’s degree had lower delay discounting AUC compared to participants with graduate degrees (B = -.169, p = .017). The b-path (delay discounting AUC with engagement) had the following model fit statistics: loglikelihood = -39.89, AIC = 111.78, BIC = 151.95. The b-path was not statistically significant, indicating no relationship between impulsivity and overall adherence. In terms of covariate effects on overall adherence, income level continued to show a negative association with overall adherence (B = -.664, OR = .515, p = .004) as well as depression (B = -.257, OR = .773, p = .001). An assessment of the indirect paths indicated that the indirect effects were not statistically significant. In the absence of significant effects of both the a- and b-paths, or an indirect effect, it can be assumed that no mediation effect exists. Weekly Adherence Table 6 presents model statistics for all model pathways. The c-path model had the following model fit statistics: loglikelihood = -2061.49, AIC = 4158.97, BIC = 4248.90, ICC = .458. The intra-class correlation suggests that 45.8% of the variation in weekly adherence was explained by the effect of participant or the between participant variation. Education was a significant predictor of weekly adherence with participants who had some college or less being significantly less likely to complete modules (B = -.425, p = .019) compared to participants who had a graduate degree. The beta coefficient suggests that participants who had some college or less completed .425 fewer modules per week compared to participants with a graduate degree. In addition to education, the time trend variable (by week) was statistically significant indicating that there was a steady decrease in the number of modules completed per week over the study PREDICTORS OF MHEALTH ENGAGEMENT 26 period (B = -.307, p < .001). A graphic depiction of this linear decrease in weekly adherence is shown in in Figure 1. The only other predictor that was significant in the model was income level, which had a negative relationship with weekly adherence (B = -.064, p = .009), suggesting that as income level increases, there is a .064 reduction in modules per week. No other covariates were significant in the model. For the mediation model, the a-path captures the relationship of education with impulsivity and had the following model fit statistics: loglikelihood = -2072.00, AIC = 4162.01, BIC = 4206.97, ICC = .455. The results showed a significant effect of education on delay discounting AUC, indicating that participants that had some college or less had lower delay discounting AUC (i.e., higher levels of impulsivity) compared to participants who had graduate degrees (B = -.161, p = .017). In addition, participants that had a bachelor’s degree had lower delay discounting AUC compared to participants with graduate degrees (B = -.175, p = .011). The b-path (delay discounting AUC with weekly adherence) had the following model fit statistics: loglikelihood = -2061.41, AIC = 4160.83, BIC = 4255.75, ICC = .459. The intra-class correlation suggests that 45.9% of the variation in weekly adherence was explained by the effect of participant. The b-path was not statistically significant, indicating no relationship between delay discounting AUC and weekly adherence. In terms of covariate effects on weekly adherence, the time trend continued to show a significant downward trend (B = -.304, p < .001) and income level continued to show a negative association with weekly adherence (B = -.067, p = .008). We assessed the indirect paths to establish whether the indirect effect was statistically significant. Results indicate that the indirect effects were not statistically significant. In the absence of an indirect effect or significant effects of both the a and b-paths, it can be assumed that no mediation effect exists. PREDICTORS OF MHEALTH ENGAGEMENT 27 Daily Usage Table 7 presents model statistics for all model pathways. The c-path model had the following model fit statistics: loglikelihood = -14260.12, AIC = 28558.23, BIC = 28688.71, ICC = .640. Education was a not a significant predictor of daily usage suggesting that there was no direct impact of education on per day activity completion rate. Chances of remaining in the zero state (i.e., the chances of no daily activities) were unrelated to education. The time trend variable (by day) was statistically significant, indicating that there was a steady decrease in activities per day over the study period (B = -.013, IRR = .987, p = .001). A graphic depiction of this linear decrease in daily usage is shown in Figure 2. Being male was negatively associated with daily usage (B = -.154, IRR = .857, p = .037). The statistical non-significance of education suggests that a direct effect of education on daily usage did not exist. The a-path tested the relationship of education with delay discounting AUC with the following model fit statistics: loglikelihood = -14271.04, AIC = 28562.08, BIC = 28630.76, ICC = .641. The results showed a significant effect of education on delay discounting AUC, indicating that participants with some college or less had lower delay discounting AUC (i.e., higher levels of impulsivity) compared to those with graduate degrees (B = -.160, p = .018). In addition, participants with a bachelor’s degree had lower delay discounting AUC compared to those with graduate degrees (B = -.179, p = .009). The b-path (delay discounting AUC with daily usage) had the following model fit statistics: loglikelihood = -14259.76, AIC = 28559.53, BIC = 28696.88, ICC = .639. The b-path was not statistically significant, indicating no relationship between delay discounting AUC and daily usage. Chances of remaining in the zero state were unrelated delay discounting AUC. In terms of covariates effects, the time trend showed a significant downward trend (B = -.013, IRR PREDICTORS OF MHEALTH ENGAGEMENT 28 = .987, p < .001) and being male was negatively associated with daily usage (B = -.162, IRR = .850, p = .026). We assessed the indirect paths to establish whether the indirect effect was statistically significant. Results indicated that the indirect effects were not statistically significant. In the absence of an indirect effect or significant effects of both the a and b-paths, it can be assumed that no mediation effect exists. Discussion Mobile health is designed to provide greater access to behavioral support tools, yet variability in user engagement highlights the need to study factors that may influence intervention adherence and usage. This investigation of individual-level predictors of engagement in an mHealth mindful eating intervention tested impulsivity as a mechanism linking educational status to intervention adherence and usage. Findings demonstrate that less education was associated with lower levels of weekly and overall adherence compared to individuals with higher educational status. This study provides initial empirical support for education as an individual-level predictor of mHealth adherence, thus highlighting a potential opportunity for intervention among a subset of individuals who may be at risk of lower adherence. Although shown to be linked to measures of adherence, educational status did not predict daily usage, measured as number of activities completed per day. This study also demonstrates that less education was linked with greater impulsivity as measured by valuing smaller immediate rewards over larger future rewards (i.e., delay discounting), lending empirical support for a well- studied relationship (Reimers et al., 2009). However, impulsivity did not impact adherence or usage, failing to lend support to impulsivity as a mechanism by which education impacts intervention engagement. PREDICTORS OF MHEALTH ENGAGEMENT 29 Although mHealth is touted as a solution to improve access to diverse populations, the impact of educational status on mHealth adherence remains largely unstudied. This investigation is among the first to show that similar to the trends reported for in-person interventions, individuals with less education may experience barriers to adherence despite the relative ease of access of a mobile health platform. The inverse relationship between educational status and adherence aligns with the limited number of digital intervention studies that document a similar association (Wangberg et al., 2008) and, more broadly, with the substantial research showing that schooling has a substantial effect on health and health behaviors (Goldman & Smith, 2002). However, this investigation should be considered preliminary given that most participants were highly educated in the present sample, thus impacting the generalizability of our findings. There is limited research or theory to adequately address why education would predict adherence but not usage. With this limitation in mind, we put forth a few possible explanations. One factor that may explain less adherence being associated with less educational status concerns access-related barriers. Less educational attainment is shown to be associated with being “smartphone dependent”, defined as having a smartphone but not a home broadband connection (Perrin, 2021; Pew Research Center, 2021). Despite rising numbers of smartphone ownership, smartphone dependent individuals are more likely to face canceled or suspended smartphone service because of financial constraints or reaching their smartphone plans’ maximum data caps (Anderson & Horrigan, 2016). It is possible that some participants in our study may have faced interruptions in their smartphone service or otherwise limited their use of the app to manage smartphone data limitations, which may in turn have impacted adherence. Why, then, was a similar pattern not observed for daily usage? Overall adherence captured the completion of psychoeducational modules that consisted of streaming several relatively data- PREDICTORS OF MHEALTH ENGAGEMENT 30 intensive videos. The daily usage variable included these video modules, but also included several tools and exercises that rely on far less smartphone data. It is possible, therefore, that the impact of education on adherence may reflect challenges related to data access that may not necessarily be reflected in daily usage in the same manner. Another factor that may contribute to these findings relates to the regimented nature of health behavior that is captured by adherence measures. Living with type 2 diabetes involves maintaining complex health regimens that are prescribed to manage this chronic illness. Adherence to specific targets (in the case of the present intervention, to watch a prescribed number of modules within a given week) arguably requires planning and behavioral management as well as an understanding of the medical necessity for certain health behaviors, which are factors that may be affected by differences in educational status (Goldman & Smith, 2002). It is conceivable, therefore, that educational status may impact how individuals engage with the intervention as intended by the program instructors. It is critical that such considerations be interpreted not as shortcomings on part of the participant, but rather a call to deliver more suitable and effective intervention programs that meet the unique needs of all individuals. Contrary to adherence, daily usage captured the use of various components of the app that participants were free to engage with anytime and as often as they chose. This inherently represents more “free form” and less structured use of the intervention that may not involve a need for planning and behavioral management to the same degree as adherence does, and therefore may not be impacted in the same manner by differences in educational status. We tested several variables as covariates that may impact engagement levels. Findings indicated that income was associated with adherence. Specifically, higher income levels were negatively associated with weekly as well as overall adherence. It is possible that individuals PREDICTORS OF MHEALTH ENGAGEMENT 31 with higher income may have had the opportunity to utilize other resources to support diabetes management, thus contributing to a lesser perceived need for the content of the present intervention. However, this finding is in contrast with studies of interventions that demonstrate a positive association between income and engagement (Barnett et al., 2012; Rung et al., 2020). This finding also highlights the importance of income and education as distinct indicators of socioeconomic status may act through unique causal pathways and that cannot be used interchangeably (Festin, Thomas, Ekberg, & Kristenson, 2017). In addition, depression was significantly associated with overall adherence. Participants who reported higher depression were less likely to complete all 28 modules. This finding aligns with the broader literature that documents the strong impact of depression on adherence in type 2 diabetes and other chronic illnesses. In the context of mHealth interventions, a study of eighty patients with type 2 diabetes reported that individuals with more depressive symptoms were least likely to be engaged in a 3-month mHealth intervention designed to address medication adherence (Nelson et al., 2016). However, depression did not appear to be related with weekly adherence or daily usage. These findings may inform the design of future intervention procedures that identify individual endorsing baseline depression who may be more likely to disengage. This may allow for providers to tailor the intervention to their needs, potentially improving engagement and, correspondingly, intervention outcomes. This study provides initial evidence of education as a predictor of mHealth engagement in individuals with diabetes, and as such had several limitations. First, we observed limited diversity in the study sample, which was predominantly female and highly educated with several participants demonstrating low levels of impulsivity. Despite the limited representativeness of the group sample, however, we observed significant differences in adherence by educational PREDICTORS OF MHEALTH ENGAGEMENT 32 status and a significant association between educational status and impulsivity. Second, testing multiple measures of engagement inflated the risk of Type 1 error. Results should therefore be interpreted with caution. Finally, the finding that education impacted adherence and not usage highlights the importance of distinguishing these measures as distinct constructs and developing more refined hypotheses specific to each engagement measure. Despite these limitations, this study provides important preliminary evidence of the potential need to address mHealth adherence among individuals with different educational backgrounds living with type 2 diabetes. Future studies should further assess other relevant factors and more comprehensive models to predict differences in mHealth engagement. PREDICTORS OF MHEALTH ENGAGEMENT 33 Study 2: Do Food Craving and Mindfulness Interact to Impact Engagement with a Mindful Eating mHealth Intervention? Food cravings are defined as intense desires or longings to eat a specific food (Weingarten & Elston, 1991) and are often experienced on a daily basis (Richard et al., 2017). Although food cravings and hunger often co-occur, an energy deficit has not been shown to be a prerequisite for experiencing food cravings, which can occur even in the absence of hunger (Pelchat & Schaefer, 2000). Food cravings can pose challenges to adherence to diet recommendations, perhaps in part because diets are shown to generally increase the likelihood of food cravings (Hill, 2007). Food cravings, and eating in response to food cravings, may be particularly relevant in diabetes management. Successful management of diabetes includes eating a healthy diet which is shown to improve glycemic control (i.e., control of blood glucose levels), insulin resistance, and body weight, and reduce the risk of diabetes-related complications (Wing et al., 2001). However, in a study of adults with uncontrolled (glycosylated hemoglobin [HbA1c] > 8.0%) and well- controlled (HbA1c < 7.5%) diabetes, food cravings were considered a “very or extremely difficult” barrier to diabetes control by 67% and 64% of participants, respectively (Brod et al., 2016). Moreover, carbohydrate food cravings are positively associated with poorer glycemic control in diabetes (Yu et al., 2013). It is not surprising, perhaps, that food cravings are associated with dietary nonadherence in people with diabetes (Schlundt et al., 1994; Yu et al., 2013). Given that successful treatment of diabetes requires long-term dietary adherence, interventions that target food cravings and eating in response to cravings may be warranted in supporting diabetes management. PREDICTORS OF MHEALTH ENGAGEMENT 34 One type of intervention that may help address food cravings, and therein support dietary adherence, is based on the principles of mindful eating. Mindful eating is the practice of bringing awareness in a nonjudgmental manner to the experience of eating, which includes physical hunger, satiety cues, and environmental or emotional triggers to eat (Miller et al., 2014). This practice is based on the concept of mindfulness, which can be defined as paying attention in the present moment, on purpose and non-judgmentally (Kabat‐Zinn, 2003). In a study conducted among adults with type 2 diabetes, a mindful eating intervention incorporated training in mindful meditation, mindful eating, and bodily awareness of hunger and satiety cues (Miller et al., 2012). Participants evidenced a significant reduction in weight and significant improvement in glycosylated hemoglobin, a measure of glycemic control. Mindful eating is posited to provide individuals skills to recognize and observe their experiences of food cravings without eating in response to them (Mason et al., 2019). By equipping individuals with skills to prevent eating in response to cravings, mindful eating may help support adherence to dietary recommendations. Intervention engagement, broadly defined as an individual’s use of intervention tools and techniques, may be particularly relevant in mindfulness-based interventions as regular practice is theorized as essential for developing mindfulness skills (Carmody & Baer, 2008). A study testing a mindful eating intervention in a nonclinical sample offered support for a dose response effect, defined as the relationship between the applied dose/concentration (e.g., the level of use of an intervention) and the effect that is observed (e.g., changes in target treatment outcomes). Specifically, this study found that participants who engaged for greater amounts of time in the mindful eating intervention had greater improvements in a measure of behavioral flexibility, the primary intervention outcome (Janssen et al., 2018). However, levels of participant engagement with mindfulness interventions are shown to be inconsistent (Canby et al., 2021). PREDICTORS OF MHEALTH ENGAGEMENT 35 Several researchers have turned to mobile health (“mHealth”) to deliver mindfulness- based interventions through smartphones (Haliwa et al., 2021; Lyzwinski et al., 2019). mHealth is designed to provide access to behavioral tools more continuously and cost-effectively than in- person therapies (Covolo et al., 2017; Free et al., 2013). mHealth enables individuals ongoing access to the intervention by allowing them to carry intervention tools with them wherever they go, potentially allowing individuals to access behavioral support tools with greater ease and frequency compared to in-person interventions (Free et al., 2013). Therefore, mHealth is purported to enhance accessibility by integrating interventions into the daily lives of users. This may be particularly relevant for targeting dietary adherence and food craving experiences in diabetes, which involves repeated, ongoing maintenance to achieve desired health benefits. While mHealth appears to be a promising modality, studies report variability in user engagement (Pham et al., 2019). Low levels of engagement are commonly reported in mHealth trials (Szinay et al., 2020). As noted earlier, there is evidence of dose response relationships in mindful eating interventions, underscoring the importance of adequate intervention engagement. In light of these findings, it is important to identify individual-level factors that may predict engagement. Identifying predictors of engagement can potentially inform the development of tailored interventions to optimize engagement and, correspondingly, promote improved health outcomes. Little is known about intervention-specific predictors of engagement in mindful eating interventions. Researchers studying engagement in mindfulness interventions argue that a somewhat problematic relationship may exist, in that the psychological symptoms that mindfulness interventions intend to address may make it more difficult for participants to engage with the intervention (Crane & Williams, 2010; Kerr et al., 2019). For example, greater negative PREDICTORS OF MHEALTH ENGAGEMENT 36 affect, difficulties with emotion regulation, and lower baseline mindfulness are associated with poorer engagement in mindfulness interventions (Atkinson & Wade, 2012; Liu & Rice, 2018). In such cases, individuals who engage less in mindfulness interventions may be the ones who could benefit the most had they engaged. Therefore, identifying factors associated with engagement is crucial to enhance intervention uptake among those who may need it most. In the case of mindful eating targeting food cravings, one such predictor of interest may be baseline food craving. Food cravings can be experienced and assessed as a trait. According to a cognitive-emotional model of food craving known as the Elaborated Intrusion Theory (EIT), cues to eat, whether cognitive, emotional, or physiological, can trigger spontaneous thoughts of images of the properties of the craved item (May et al., 2012). Per the EIT, imagining and thinking about the craved item at first creates a sense that one is actually consuming the item, which can elicit emotions of pleasure and relief typically associated with previous consumption of the craved item. Prolonged mental elaboration is associated with enhanced awareness of unpleasant physical deficits or negative affect, which can also escalate cravings. Elaborated craving thoughts and imagery, and associated planning for food consumption, can be experienced as irritating or uncomfortable. Mindful eating is theorized to improve eating behavior by targeting food cravings. Instead of changing, suppressing, resisting, or avoiding cravings, mindfulness is designed to help individuals accept and paradoxically move closer to the thoughts, emotions, and body sensations that make up cravings (Brewer et al., 2018). It is possible that when engaging in mindfulness practice, people with more severe food cravings may experience greater discomfort when attempting to “sit with” and move closer to the craving experience, or they may experience an exacerbation of the craving itself (Alberts et al., 2013; Tapper, 2018). This may lead them to PREDICTORS OF MHEALTH ENGAGEMENT 37 engage less with mindful practice. Furthermore, for those who may have tried to manage their cravings by suppressing or avoiding them, the radically different approach of “moving closer” to the craving experience may pose challenges. Similar explanations have been put forth for other intervention-specific predictors of engagement. For example, in a randomized controlled trial testing mindfulness-based cognitive therapy for depression, rumination and worry predicted intervention dropout (Crane & Williams, 2010). Researchers theorized that individuals who persistently tried to suppress negative thoughts may have found mindfulness to be a daunting effort. Additionally, the practice of bringing awareness to these negative states may lead to discomfort and negative experiences (Banerjee et al., 2018). Indeed, qualitative studies report that some participants have difficulty with engaging in mindfulness practice due to feelings of discomfort (Lomas et al., 2015). The ability to identify people who are mostly likely to disengage may allow for providers to tailor the intervention to their needs, potentially improving engagement and, correspondingly, intervention outcomes. Trait-level food cravings have neither been examined in mindful eating interventions nor in interventions delivered on an mHealth platform. An important difference between in-person mindfulness interventions and mHealth is that mHealth-based interventions often deliver shorter mindfulness exercises designed to be used “in-the-moment.” Whether this difference impacts the hypothesized impact of cravings on engagement is unknown. However, mHealth interventions are often theoretically and conceptually aligned with in-person mindful eating interventions (Brewer et al., 2018), suggesting that individuals with greater trait-based food craving may experience challenges in a similar manner to what has been posited for in- person programs. PREDICTORS OF MHEALTH ENGAGEMENT 38 The degree to which trait food craving predicts engagement may vary based on trait mindfulness, which refers to a person’s baseline or average mindfulness (Siegling & Petrides, 2014). There are a number of possible mechanisms by which trait mindfulness may moderate the craving – engagement relationship. Trait mindfulness is associated with greater positive affect, and individuals high in positive affect may approach experiences with a nonreactive orientation (Polizzi et al., 2018). It is possible, therefore, that when experiencing uncomfortable food cravings, those higher in mindfulness may be less likely to reactively turn away from the experience. Trait mindfulness has also been shown to be associated with personality factors that include greater openness to experience (Polizzi et al., 2018). Individuals with high levels of openness are shown to have less parasympathetic and affective reactivity to stressful situations (Williams et al., 2009). Thus, it is possible that individuals with greater trait mindfulness might experience less challenge or discomfort with the process of attending to their food craving experience. Through this process, trait mindfulness may buffer the negative impact of trait food craving on engagement, such that food craving is negatively associated with engagement among those with lower levels of mindfulness, but not associated with engagement for those with higher levels of mindfulness. However, no studies to our knowledge have examined whether trait mindfulness interacts with trait food craving to influence engagement in a mindful eating intervention. The present study was designed to address this gap by examining engagement in an mHealth mindful eating intervention, called “Eat Right Now” (ERN), among people with type 2 diabetes. We examined trait food craving as a predictor of engagement and tested trait mindfulness as a moderator of this association. We hypothesized that greater levels of food craving would be associated with poorer mHealth engagement and that trait mindfulness would PREDICTORS OF MHEALTH ENGAGEMENT 39 moderate the craving – engagement relationship. Specifically, we posited that craving would be negatively associated with engagement for individuals with lower levels of trait mindfulness and would not be associated for individuals with higher levels of trait mindfulness. Methods Overview The present investigation derived from the Diabetes Education to Lower Insulin, Sugars, and Hunger (DELISH) Study, a randomized controlled trial targeting diabetes-related outcomes among individuals with type 2 diabetes (Mason et al., 2019). All procedures for Human Subjects research were approved by the institutional review board at the University of California San Francisco (UCSF). The DELISH Study compared the effect of two interventions: a nutrition education alone arm (Ed) and a nutrition education with mindfulness-based intervention components (Ed+MBI) arm. Participants in both arms attended 12 weekly group-based in-person classes. Ed+MBI participants also received the Eat Right Now (ERN) program, a mindful eating intervention in the form of a mobile app (Mason et al., 2018). The ERN program was administered in combination with the weekly group classes. The Delish Study was conducted across two trials: the R61 trial randomized 60 participants in a 1:1 ratio to the Ed + MBI and Ed arms. The R33 trial randomized 120 participants in a 1:1 ratio to these arms. Across the R61 and R33 trials, investigators conducted a total of eight rounds of participant recruitment, enrollment, and intervention assessment. Each round consisted of running a single Ed + MBI group and a single Ed group. The present study investigated mHealth intervention engagement in the Ed + MBI arm of the R61 and R33 trials, totalling 91 Ed + MBI participants across all eight rounds. Participants PREDICTORS OF MHEALTH ENGAGEMENT 40 Participants were English speaking adults (≥ 18 years) who screened positive for type 2 diabetes based on glycosylated hemoglobin (HbA1c) levels of 6.5%≤ HbA1c<12.0%. Additional eligibility criteria included: 1) Eating in response to food cravings at least twice over the course of three days, 2) Being able to engage in light physical activity, 3) Being willing and able to participate in the interventions, and 4) Having a smartphone and being willing to use it on a regular basis. Participants were recruited from several sources, including Facebook, Nextdoor, and Craigslist as well as within UCSF clinics. Procedures Mindful Eating Intervention The ERN app is designed to deliver mindful eating training over 28 psychoeducational modules and has been tested in women with overweight who report overeating in response to food cravings (Mason et al., 2018). Participants in that study experienced reductions in craving- related eating as well as reductions in trait-level measures of overeating, including food cravings. ERN modules teach participants to attend to three aspects of eating: why they eat, including environmental and emotional triggers; what types of food are most likely to lead to cravings; and how to eat with awareness and mindful attention. Each module lasts approximately 5 to 10 minutes, and users can access only one new module per day (after completing the previous module). Users have unlimited access to previous modules. DELISH study instructors asked participants to complete 2 or 3 modules per week during the 12-week intervention period. In addition to the modules, participants could access additional tools and exercises to aid in mindfully “riding out” food cravings. After the 28 modules were completed, all tools and previous modules remained available for the duration of the trial (see Mason et al., 2018 for a description of intervention content). PREDICTORS OF MHEALTH ENGAGEMENT 41 As noted earlier, the DELISH Study included weekly in-person group classes that involved a dietary education component and mindful eating component. Participants in both study arms (Ed and Ed+MBI) attended the dietary education portion. Ed+MBI participants also attended the mindful eating portion. The mindful eating content was not didactic in nature but rather centered around discussing participants’ experiences with mindful eating practices, troubleshooting obstacles, and engaging in and reflecting on group exercises. Study Visits. Participants completed several measures at a baseline study visit. While the DELISH study collected data at multiple pre- and post-intervention visits, the present research utilized questionnaire data collected at the baseline visit in addition to app engagement data collected over the 12-week intervention period. Measures We collected a range of self-report, computerized, and clinical measures as well as mHealth engagement measures. We describe the subset of measures used in the present study, below. Demographics. During the baseline visit, participants reported the following demographics information: age, gender, race/ethnicity (re-coded to African American/Black, Asian/Pacific Islander, White/Caucasian, Latino/Hispanic, and Other), educational attainment (re-coded to some college or less, associate's or 2-year degree, bachelor’s degree, and master's/doctoral degree), employment status (re-coded to employed full-time, employed part- time/student, unemployed, and retired), household income, marital status, total household size and number of children in the household. Patient Health Questionnaire (PHQ-8). The PHQ-8 is a standardized, well-validated 8- item measure of depressive symptoms (Kroenke et al., 2009; Appendix A). Participants PREDICTORS OF MHEALTH ENGAGEMENT 42 responded on a 4-point scale: Not at all (0); Several days (1); More than half the days (2); Nearly every day (3). Higher scores indicated higher levels of depressive symptoms. The PHQ-8 demonstrated good reliability in the present sample (Cronbach’s alpha = 0.86). Validity of the PHQ-8 in several populations has been well established and the measure has demonstrated good construct validity and criterion validity (Kroenke et al., 2009). Food Craving Questionnaire - Trait Reduced Version (FCQ-T-R). The 15-item FCQ- T-R assesses experiences of food craving as a trait (Meule, Hermann, & Kübler, 2014; Appendix B). FCQ-T-R scores are positively associated with low dieting success and higher body mass index. Components of food craving that are covered by this measure are its emotional and environmental triggers, mental imagery and cognitive elaboration, and behavioral consequences, that is, searching for and consuming food. Participants responded on a 6-point scale from 1 (never) to 6 (always). Food craving was measured as the overall score with higher scores indicating greater food craving. The FCQ-T-R demonstrated excellent reliability in the present sample (Cronbach’s alpha = 0.94). FCQ-T-R scores have been shown to predict cue-elicited food craving, are positively associated with body mass index and are negatively associated with self- perceived dieting success, in support of the measure’s validity (Meule et al., 2014). Five Factor Mindfulness Questionnaire (FFMQ). The FFMQ assesses trait-like tendencies to be mindful in experiences of daily life (Baer, Smith, Hopkins, Krietemeyer, & Toney, 2006; Appendix C). Participants responded to 24 items on a 5-point scale ranging from 1 (never or very rarely true) to 5 (very often or always true). Trait mindfulness was measured as the overall score with higher scores indicating greater mindfulness. The FFMQ has good reliability in the current sample (Cronbach’s alpha = 0.88). This measure has demonstrated construct validity (Baer et al., 2008) and convergent validity with measures of psychological PREDICTORS OF MHEALTH ENGAGEMENT 43 well-being (PWB) and psychological symptoms (SCL-90R), two constructs theoretically linked to mindfulness (Goldberg et al., 2016). mHealth Engagement. We defined mHealth engagement as participants’ use of the mindful eating app based on objective user log data downloaded from the app. We examined adherence (the extent to which a person’s behavior corresponds with prescribed recommendations) and usage (an individual’s total use of intervention tools) as indices of engagement. Specifically, ERN has a total of 28 psychoeducational modules as well as additional mindfulness-based tools and exercises. Study instructors asked participants to watch 2-3 modules weekly and complete all 28 modules within 12 weeks. Therefore, for adherence we measured the number of modules completed per week as well as the total number of modules completed. In terms of usage, we measured usage per day, i.e., the total number of activities participants completed in the app on a daily basis. This was warranted as not only do studies recommend integrating mindfulness into daily life in order to experience therapeutic benefits (Carmody & Baer, 2008) but individuals often experience cravings on a daily basis (Richard et al., 2017). In sum, the three engagement measures included the following: Overall Adherence. To measure overall adherence, we summed the total number of modules watched ranging from 0 to 28. Weekly Adherence. We calculated weekly adherence, or number of modules completed per week, by summing the total number of new modules a participant watched per week. The app allowed participants to watch a maximum of seven modules over a 7-day timespan. Daily Usage. We calculated daily usage, or number of activities completed per day, by summing the number of times a participant completed optional tools or exercises or watched a PREDICTORS OF MHEALTH ENGAGEMENT 44 psychoeducational module within a 24-hour time span, defined from 12:00 am to 11:59 pm for each day. Data Management and Univariate Testing Across our analyses, we operationalized mHealth engagement as measures of overall adherence, weekly adherence, and usage, referred together as “engagement”. The overall adherence variable summed all modules a participant completed over the study period to provide an overall count. This variable was measured at the participant level and did not vary over time. The weekly adherence variable captured the number of modules a participant completed in a week over the 12-week study period and was measured at the week level. The daily usage variable captured the number of activities a participant completed within a day over the 78-day study period and was measured at the day level. Unlike overall adherence, both weekly adherence and daily usage varied across time. We conducted univariate assumptions tests on all study variables to assess normality and identify extreme outliers. We used visual assessments such as histograms and normal QQ-plots for normality testing. Using boxplots, we identified extreme outliers as any observation greater than three times the inter-quartile range (Field, 2013; Tabachnick & Fidell, 2007). These univariate tests indicated that there were non-normal distributions in all three of the outcome variables. The other measures (food craving, mindfulness, and depression) approximated normal distributions. The overall adherence variable had considerable left skew with most participants having completed all 28 modules (68.13%). We recoded this variable into a binary variable that assigned any participant who completed all 28 modules a one (1), and any modules less than 28 a zero (0). The weekly adherence variable demonstrated some right skew but approximated a normal PREDICTORS OF MHEALTH ENGAGEMENT 45 distribution. Finally, the daily usage variable demonstrated considerable right skew with a substantial percentage of the distribution at zero, due to there being days when participants did not complete any activities. We did not find extreme outliers (i.e., outside three times the interquartile range) except for those variables that had extreme skew (overall adherence and daily usage). We addressed outlier issues by recoding overall adherence into a binary variable and modeling the skewed distribution of daily usage (negative-binomial). We grand-mean centered variables to improve interpretation. Analyses indicated missing data in the educational status, ethnicity/race, income, and the adherence or usage variables, constituting 1.90% of the overall proportion of missing data at the participant level, 1.17% of the overall proportion of missing data at the week level, and 2.15% of the overall proportion of missing data at the day level. We employed multiple imputation for missing value replacement to preserve power and to reduce potential bias due to missingness (Little & Rubin, 2020). Using MPlus, we conducted imputation using multiple chained equations and creating 10 imputed datasets to pool the results of the analysis models. Analytic Plan We conducted descriptive analysis of the frequencies and percentages of selected categorical study variables. Similarly, we calculated means, standard deviations, and minimum/maximum values for continuous study variables. Preliminary analyses examined whether potential covariates (age, gender, race, household income, employment status, depression), independent, and moderator variables were associated with each other to explore relationships in the data and to establish if multi-collinearity was an issue. We also conducted nonparametric equivalent tests for each preliminary analysis to ensure that effects were still PREDICTORS OF MHEALTH ENGAGEMENT 46 statistically significant in comparison to parametric tests (e.g., Kruskal-Wallis for ANOVA and Spearman’s rho for correlation). As participants completed the intervention across eight distinct rounds, we investigated whether participant round was significantly associated with the outcome variables of interest; analyses indicated no significant association. The primary research aim was to examine whether trait-based food craving predicted mHealth engagement and to test the role of trait mindfulness as a moderator of this relationship. We tested for the moderating effect of mindfulness on the association between food craving and engagement with the interaction term craving*mindfulness. We included the main effects of the independent and moderator variables in the model in addition to the interaction effect to assess for moderation. We plotted significant interaction effects to visualize the impact of the moderator on engagement. These analyses involved a combination of multi-level and ordinary regression models. The nested structure of time-varying engagement measures results in non-independence of observations due to clustering effects (Raudenbush & Bryk, 2002). Specifically, mHealth engagement as measured by daily usage and weekly adherence varied within individuals (Level 1) such that an individual may have a different value for each measure depending on the day (or week). The weekly adherence and daily usage variables were the only within-participant variables that were time-varying in addition to the time trend variable (week or day). All other measures varied between but not within individuals (Level 2). To account for the nested structure of these data, we utilized a series of multi-level model analyses (Singer & Willett, 2003). All predictors for the multi-level models were grand mean centered to improve interpretation of the effects of variables that may not have a natural zero point. One of our adherence outcomes, overall adherence, also varied between but not within individuals. Given the lack of nested data PREDICTORS OF MHEALTH ENGAGEMENT 47 in this case, multi-level models were not indicated and we performed ordinary regression analyses when testing for the association between baseline food craving and overall adherence. As noted earlier, the literature examining predictors of mHealth engagement is limited. Some studies of digital health engagement that do explore such predictors have shown age, gender, race, income, employment status, and depressive symptoms to impact engagement (Nelson et al., 2016; Reinwand et al., 2015; Rung et al., 2020). Therefore, these variables were added as covariates to the models. The three engagement measures required different analytic approaches due to their varied distributions. For overall adherence, we applied binary logistic regression to assess the probability of a participant “completing all modules”. Due to an approximate normal distribution for weekly adherence, we used a linear mixed model with a random intercept, along with the maximum likelihood robust (MLR) estimator using robust standard errors in MPlus (Muthén & Muthén, 2017). For daily usage, which demonstrated right skew with a substantial percentage of the distribution at zero, we used a zero-inflated negative binomial multilevel model with a random intercept. The negative binomial model helps accounts for the overdispersion of the variance exceeding the mean. In addition, the zero-inflated model was necessary to account for the high proportion of zero counts in the outcome (Cameron & Trivedi, 1998). Due to both the multilevel structure of the data and the moderation analyses, we grand-mean centered the predictor variables to provide accurate interpretation of the results with predictor variables that have no natural zero-point (Algina & Swaminathan, 2011). Accordingly, we reported effect sizes to interpret the strength and direction of the effects, including unstandardized beta coefficients, odds ratios (OR) for logistic regression, and incident rate ratios (IRR) typical of count data. We reported model fit statistics using relative fit statistics PREDICTORS OF MHEALTH ENGAGEMENT 48 typical of multilevel modeling, including the log-likelihood value, Akaike information criterion (AIC), and Bayesian information criterion (BIC). We computed intra-class correlations (ICC) for the multi-level models to assess the level of variance in the outcome explained by the effect of cluster, in this case participant. The ICC varies between 0 and 1 and is calculated by dividing the cluster variance by the total variance (cluster and within variance). An ICC closer to 1 indicates a stronger effect of the cluster on the outcome and a greater justification for a multilevel model. An ICC closer to 0 indicates a weaker effect of cluster on the outcome and less of a justification for the use of a multilevel model (Raudenbush and Bryk 2002). We conducted descriptive and preliminary analysis in Stata version 17 (StataCorp., 2021) analysis software. The primary analyses for this study used MPlus version 8.8 (Muthén & Muthén, 2017) analysis software. Results Descriptive Statistics Table 8 presents descriptive statistics for all relevant categorical demographic variables. The majority of the sample was female (62.64%). The largest ethnic group identified as White/Caucasian (46.15%) followed by Asian/Pacific Islander (16.48%). Over half (52.75%) of participants reported an annual household income of $75,000 or greater. Approximately 40% of participants were employed full-time and 29.67% of participants were retired. Participants reported level of education with approximately 26.37% completing some college or less. Most participants completed all 28 of the modules (68.13%). Table 9 presents means and standard deviations for all continuous variables. Participant age ranged from 21 to 80 years (M = 59.09, SD = 10.52). Depression scores ranged from 0 to 21 (M = 6.43, SD = 4.76), food craving scores ranged from 23 to 87 (M = 52.24, SD = 13.08), and PREDICTORS OF MHEALTH ENGAGEMENT 49 mindfulness ranged from 43 to 106 (M = 79.87, SD = 12.07). Participants averaged 25.93 total modules (SD = 5.09) out of a maximum of 28 modules, 2.16 modules per week (SD = 1.93), and 2.92 activities per day (SD = 4.26). Preliminary Analysis We conducted preliminary analyses to explore relationships between the study variables. First, we conducted a series of one-way analyses of variances (ANOVAs) to compare trait food craving mean scores by categorical demographic variables (Table 10). Results revealed no significant relationships between the pertinent demographics and food craving (all p > .05). Second, we analyzed trait mindfulness mean scores by demographics using a series of one-way ANOVAs. Results revealed a marginally significant relationship between race/ethnicity and mindfulness (F (4,85) = 2.41, p = .056). However, using Bonferroni adjustments for multiple tests, none of the individual categorical comparisons were significant. There was no relationship between gender, income level, or employment status with mindfulness (all p > .05). Third, we ran an additional series of ANOVAs to examine relationships between depression mean scores and the categorical demographics. Results revealed no statistically significant relationships between depression and demographics (all p >.05), suggesting no mean differences in depression by specified demographics. Finally, we conducted Pearson’s correlations (r) that to examine the relationship between mean scale scores on the continuous study variables (Table 11). Results revealed that higher levels of mindfulness were associated with lower levels of craving (r = -.448). Additionally, higher levels of depression were associated with higher levels of craving (r = .389) and lower levels of mindfulness (r = -.509). Primary Analysis PREDICTORS OF MHEALTH ENGAGEMENT 50 Analytic models were blocked with a basic model (with no moderation) and a moderation model to examine the interaction effect of trait food craving on mindfulness to predict engagement. Overall Adherence Table 12 presents model statistics. The basic model had the following model fit statistics loglikelihood = -39.86, AIC = 113.71, BIC = 156.39. While the effect of food craving was not a significant predictor of engagement, mindfulness was a significant predictor (B = -.020, OR = .980, p = .043) which suggests that a one unit increase in mindfulness resulted in a decrease in the likelihood of completing all modules by a factor of 1.02. In terms of covariate effects, depression (B = -.263, OR = .769, p = .009) and income (B = -.632, OR = .532, p = .006) were significantly negatively associated with overall adherence. In addition, individuals with some college or less were significantly less likely to complete all modules than those with a graduate degree (B = -1.891, OR = .151, p = .043). No other covariates were significant in the model. The moderation model had the following model fit statistics: loglikelihood = -37.89, AIC = 111.78, BIC = 156.98. The main effects of food craving and mindfulness were not statistically significant (all p < .05); however, the interaction effect was statistically significant (B = -.005, OR = .995, p = .017). Figure 3 presents the interaction effect on overall adherence. The figure suggests that participants with lower food craving levels who also reported higher mindfulness levels were more likely to complete all modules compared to those with lower levels of mindfulness. Lower mindfulness levels appeared to be associated with a lower probability of completing all modules. The higher probability of completing all modules at higher levels of mindfulness seemed to drop off as levels of food craving increased. At higher craving levels, PREDICTORS OF MHEALTH ENGAGEMENT 51 there eventually appeared to be little to no impact of mindfulness on the probability of completing all modules. Regarding covariate effects in the moderation model, depression (B = -.252, OR = .777, p = .017) and income (B = -.671, OR = .511, p = .005) were significantly associated with overall adherence. In addition, participants with some college or less were significantly less likely to complete all modules (B = -1.848, OR = .158, p = .050) compared to those with a graduate degree. Weekly Adherence Table 13 presents model statistics. The basic model had the following model fit statistics: loglikelihood = -2060.70, AIC = 4161.41, BIC = 4261.32, ICC = .459. The intra-class correlation suggests that 45.9% of the variation in weekly adherence was explained by the effect of participant. Neither food craving nor mindfulness were significant predictors of weekly adherence. The time trend variable was significant (B = -.307, p < .001) showing a decline in modules per week over the study period. Income level was negatively associated with weekly adherence (B = -.070, p = .006). Participants with some college or less completed fewer modules per week (B = -.438, p = .015) compared to those with a graduate degree. No other covariates were significant in the model. The moderation model had the following model fit statistics: loglikelihood = -2060.66, AIC = 4163.32, BIC = 4268.23, ICC = .458. The intra-class correlation suggests that 45.8% of the variation in weekly adherence was explained by the effect of participant. The main effects of food craving and mindfulness were not statistically significant (all p < .05). The interaction effect was not statistically significant. In terms of covariate effects, the time trend variable remained a significant negative predictor of weekly adherence (B = -.307, p < .001). Income level showed a PREDICTORS OF MHEALTH ENGAGEMENT 52 negative association with weekly adherence (B = -.070, p = .006). In addition, those with some college or less were likely to complete fewer modules (B = -.434, p = .017) compared to those with a graduate degree. Daily Usage Table 14 presents model statistics. The basic model had the following model fit statistics: loglikelihood = -14258.80, AIC = 28559.60, BIC = 28703.82, ICC = .639. Neither food craving nor mindfulness were statistically significant predictors of daily usage. Chances of remaining in the zero state (i.e., the chances of no daily activities) were unrelated to craving (B = .002, IRR = 1.002, p = .872) or mindfulness (B = .017, IRR = 1.017, p = .190). The time trend variable was significant (B = -.013, IRR = .987, p < .001) showing a decline in activities per day over the study period. Being male was negatively associated with daily usage (B = -.177, IRR = .838, p < .013). The moderation model had the following model fit statistics: loglikelihood = -14258.71, AIC = 28561.42, BIC = 28712.51, ICC = .637. Neither food craving nor mindfulness were statistically significant predictors of daily usage. The interaction effect was not statistically significant, indicating that there was not interaction between craving and mindfulness with daily usage. Chances of remaining in the zero state were unrelated to craving (B = .001, IRR = 1.001, p = .930) or mindfulness (B = .018, IRR = 1.018, p = .180). In terms of covariates effects, the time trend variable remained a significant negative predictor of daily usage (B = -.013, IRR = .987, p < .001) and being male was negatively associated with daily usage (B = -.177, IRR = .838, p < .012). PREDICTORS OF MHEALTH ENGAGEMENT 53 Discussion The present study examined the impact of trait food craving and trait mindfulness on engagement with a mindful eating mHealth intervention targeting craving-related eating. To our knowledge, this is the first investigation that tests the impact of food craving on mHealth engagement, thus contributing to our understanding of how baseline symptoms may play a role in the extent to which participants use interventions. Findings demonstrated no significant direct relationship between food craving and engagement measures. However, results suggest that trait mindfulness interacted with trait food craving to impact overall adherence. Among those with greater mindfulness, food craving appeared to be negatively related to overall adherence. In addition, there appeared to be less overall adherence among those with lower mindfulness compared to participants with higher mindfulness levels. Overall, the subset of participants who seemed to demonstrate greater overall adherence endorsed lower cravings and higher mindfulness. It is notable that this subset, who arguably were in least need of behavioral support tools, seemed to demonstrate more adherence, lending a degree of support to similar trends that have been reported in some interventions. While the main effect of food craving on adherence was not significant, there appeared to be a negative relationship between craving and adherence among those with greater mindfulness. A number of factors may explain this finding. Prior research shows that trait mindfulness facilitates an openness to experience (Polizzi et al., 2018). Openness can facilitate curiosity about one’s subjective experience (Giluk, 2009) and also may allow an individual to better maintain goal-consistent behaviors (Tronieri et al., 2020) which may drive continued adherence with the mindful eating intervention. Openness may also help individuals sustain through some of the challenge or discomfort that may accompany mindfully attending to food cravings, also allowing PREDICTORS OF MHEALTH ENGAGEMENT 54 for relatively more adherence (Forman et al., 2013). This may the case, however, only if food craving levels are below a certain threshold. Contrary to our hypothesis, adherence appeared to decrease with increased food craving despite greater mindfulness levels. Individuals with greater trait food craving may be more likely to experience intense desires or longing to eat particular foods. Studies have demonstrated that self-reported trait food craving is associated with increases in state food craving during cognitive tasks that involve food stimuli (Meule et al., 2014). Such elaborated cravings may be experienced as unpleasant or uncomfortable (May et al., 2012). Participants with higher levels of trait food craving may have faced greater discomfort interacting with the app when attempting to experientially “sit with” the craving experience as required by the intervention and may therefore have demonstrated less adherence. It is important to note, however, that we cannot be certain of the mechanisms driving these preliminary findings. While there is a plethora of research exploring the impact of engaging with mindfulness practice on craving and mindfulness, no studies to our knowledge have tested the impact of craving and mindfulness on engaging with mindfulness exercises, thus resulting in limited theoretical and empirical insights to guide our interpretation. Notably, the interaction of trait-level food craving and mindfulness impacted adherence but not usage, measured in our study as number of app activities completed per day. The mHealth intervention was designed to space out certain components of the program over time. In terms of adherence, participants were instructed to complete all 28 modules over the span of 12 weeks and were able to complete no more than one new module per day. The theoretical premise of this intervention structure was to enable participants time to assimilate information learned from each module and practice related mindfulness and mindful eating exercises before embarking on new information and related exercises. However, in terms of usage, participants PREDICTORS OF MHEALTH ENGAGEMENT 55 were free to engage with other components of the app and rewatch modules that were already completed as often as they chose. When exploring the interaction of food craving and mindfulness on engagement measures, it is unclear why no association was found with usage. Some studies have proposed that usage may vary in mHealth due to participants’ greater levels of discretion in using the app, unlike adherence that is more strictly specified, defined and prescribed (Donkin et al., 2011). Several studies have highlighted the importance of employing measures of adherence, which are often conceptually supported and defined, over those of usage. However, understanding the implications of the observed differences between adherence and usage as described above requires investigating the impact of each measure on intervention outcomes, which was not part of the current investigation and warrants attention in future studies. This study is a preliminary investigation and as such had several limitations. First, the small sample size impacted power to detect significant associations; additionally, findings should be interpreted with caution given limited sample diversity with predominantly female and highly educated participants. Second, we measured trait food craving and trait mindfulness at a single point in time – before the intervention commenced. However, varying levels of state-based mindfulness and food craving may have impacted engagement at different moments in time during the intervention period, thus highlighting the possible benefits of studying these relationships in an ecological momentary context. However, analyzing baseline predictors of engagement can help with early identification of participants who may struggle with adequate engagement, offering opportunities to plan and intervene appropriately. Third, the somewhat universal approach to our study hypotheses, that all three engagement measures would demonstrate similar relationships with the predictors being tested, was not supported by our PREDICTORS OF MHEALTH ENGAGEMENT 56 findings given observed differences between adherence and usage. Limited research in this area currently precludes the development of more refined hypotheses for each engagement measure. Finally, studying the impact of engagement on treatment outcomes is a critical component of engagement research but was beyond the scope of the present study. Future investigations should attempt to identify adherence and usage levels that correlate with optimal medical and behavioral outcomes for individuals managing type 2 diabetes. Overall, these findings do highlight a potential and major concern that individuals in greatest need of mindful eating techniques, who may stand to benefit the most from the intervention, may have poorer adherence. This concern remains despite the relatively greater accessibility to behavioral tools afforded by an mHealth-based program. There is a need to investigate ways to improve adherence among at-risk individuals and implement appropriate strategies to maintain adherence levels that optimize desired outcomes. PREDICTORS OF MHEALTH ENGAGEMENT 57 General Discussion Despite the rise of mHealth interventions, research demonstrates mixed intervention effectiveness. Participant engagement is an understudied yet critical factor that may influence treatment outcomes in mHealth. Studies that do report on engagement reveal variability in the extent to which participants utilize mHealth, highlighting the need to study engagement and identify factors that influence engagement. While mHealth is touted as an emerging solution to equalize access to and engagement with behavioral health interventions, little is known about how individuals from diverse backgrounds or those with greater symptom severity engage with mHealth interventions. Each of the studies in this dissertation examined predictors of engagement with a mindful eating mHealth app targeting Type 2 diabetes. Study 1 evaluated educational status as a predictor of engagement and impulsivity as a mechanism of this relationship. Study 2 tested whether trait food craving and trait mindfulness interact to impact engagement. Our findings provided some support for the impact of these individual-level predictors on mHealth engagement, while suggesting that different engagement measures such as adherence and usage should not be used interchangeably. These studies highlight the need for additional research to systematically explore predictors of engagement and carefully define conceptually supported engagement measures. The first study in this dissertation provides important contributions to the literature. Findings demonstrate that education predicts overall and weekly adherence such that individuals with less education are less likely to adhere to prescribed engagement recommendations. Notably, we observed no differences by educational status for a daily measure of usage, suggesting that despite differences in adherence, individuals across a range of educational levels demonstrate similar usage rates. These contributions address a critical gap in the mHealth PREDICTORS OF MHEALTH ENGAGEMENT 58 literature given that little is known about the impact of educational status on the use of mHealth. Taken together, these findings suggest that greater attention to educational differences is warranted to understand how such participant characteristics may influence mHealth engagement. Study 2 findings suggest that trait mindfulness interacts with trait food craving to impact overall adherence. Counter to our predictions, food craving appeared to be negatively related to overall adherence among those with greater mindfulness. Overall, the subset of participants who seemed to demonstrate greater overall adherence endorsed lower craving and higher mindfulness. It is notable that this subset of participants who arguably were in least need of behavioral support tools demonstrated greater likelihood for overall adherence. These results indicate a concerning possibility that greater baseline symptom severity may negatively impact mHealth adherence. These findings highlight the importance of studying ways to improve adherence among individuals in greatest need of behavioral interventions. Overall, these studies provided initial data about how certain individual-level factors may impact mHealth engagement. Taken together, the studies suggest that it may be beneficial to explore how such conceptually supported factors may influence engagement as well as treatment outcomes in mHealth. Insights from such research could inform tailoring strategies to better support participants who may be at risk of low adherence. Both studies shared several strengths, including objective longitudinal measurement of mHealth intervention engagement, conceptually supported definition and measurement of adherence and usage, and high completion rates of self- report measures. Other strengths included comparable study designs and consistency in outcome measures used across both studies, particularly in light of broader challenges to draw definitive PREDICTORS OF MHEALTH ENGAGEMENT 59 conclusions in the mHealth literature given inherent differences in study designs and the multitude of ways that engagement is assessed (Gandapur et al., 2016). Notably, findings varied when testing intervention adherence versus usage across both studies. Educational status was associated with weekly and overall adherence and the interaction of trait-level food craving and mindfulness was found to impact overall adherence. However, we did not observe these relationships with daily usage, measured as number of app activities completed per day. While the mHealth literature is inconsistent in its measures of engagement, the World Health Organization (WHO) defines adherence as the extent to which a person’s behavior corresponds with prescribed recommendations (Sabaté, 2003). Usage, on the other hand, is often defined as an individual’s total use of intervention tools. A review of digital health intervention studies reported that only 8% of studies defined and measured adherence by describing the intended use of the technology and justifying adherence using theory, evidence or rationale (Sieverink et al., 2017). In light of limited empirical insights on how the impact of individual-level predictors may vary for mHealth adherence vs. usage, this investigation put forth exploratory hypotheses that adherence and usage would be impacted in similar ways by the proposed predictors. That these predictors impacted adherence and not usage, however, highlight the importance of distinguishing these measures as distinct constructs that may be influenced by participant characteristics in different ways. Limitations Some common limitations arose across both studies. First, testing multiple relationships across multiple engagement measures inflated the risk of Type 1 error. In addition, there were limitations in applying universal measures of adherence (e.g., completion of all 28 modules) and usage, which implied “more is better”, across all participants. For instance, pharmaceutical trials PREDICTORS OF MHEALTH ENGAGEMENT 60 often utilize a dose–response curve to identify the optimal level of medication required to achieve a favorable response. Optimal adherence is defined based on this information (Donkin et al., 2011). Adherence that is tailored to achieving desired outcomes may be just as important a consideration in mHealth. Some participants, for instance, may not have required completion of all 28 intervention modules to realize gains in their eating behavior. Second, the sample used across both studies evidenced limited diversity with predominantly female and highly educated participants. Most group-based psychosocial intervention studies report similar challenges and sample representativeness and generalizability of results are pervasive concerns in this literature. Indeed, it is ironic that while the closing digital divide is touted as helping improve access to health care and interventions, most studies do not include diverse representation. The current study underscored the continued and critical need to actively recruit, and address barriers faced by, individuals of underrepresented groups. Third, the engagement measures used in both studies did not capture “offline” user behavior. Therefore, we were unable to account for moments during which an individual engaged in mindfulness or mindful eating practice without using the app. It is possible that as participants gained greater competency with mindful eating practices over the course of the intervention, they engaged in these practices in daily life without the use of app. However, there is a wealth of data that supports dose response effects between objectively measured app use and targeted outcomes, lending support to the current approach . Future research should employ multi-method data collection techniques such as user-initiated tracking and self-report of engagement. Fourth, the weekly group classes, led by intervention instructors, were likely to impact participant’s overall motivation for and engagement with the mHealth intervention, thereby PREDICTORS OF MHEALTH ENGAGEMENT 61 limiting generalizability. Studies have shown that apps coupled with human support are more likely to evidence greater engagement compared to apps with no human support (Perski et al., 2017; Szinay et al., 2020; Zhao et al., 2016). It is possible that engagement patterns in self-paced mHealth apps with no in-person support differ from those in the current study. Finally, the relationships and variables under study were likely impacted by other pathways that are not being tested in the proposed research. Future Directions This investigation provides initial data on testing conceptually supported predictors of mHealth engagement as well as defining specific engagement measures in context of the intervention under study. Exploring individual-level predictors of mHealth engagement is critical to identifying participants who could benefit from additional support or otherwise modified intervention approaches. Future studies should aim to replicate the present investigation in larger samples, as well as extend such research by exploring dose response effects to understand optimal levels of mHealth engagement required to achieve desired treatment outcomes. Indeed, researchers and practitioners have not yet defined an empirically supported threshold for optimal mHealth engagement. In addition, objective app data allows for detailed analyses of specific tools and content that participants use in the mHealth intervention, which can allow for a nuanced exploration of how engagement with specific intervention components impact treatment outcomes. Put together, such efforts can inform the development of dynamic and personalized interventions that optimize engagement based on participant characteristics. The current studies were conducted in an urban setting in the United States and contributes to the literature of mHealth studies based in high-income countries. The largest growth of mobile phones, however, has been in low-income countries, yet potential applications PREDICTORS OF MHEALTH ENGAGEMENT 62 of mHealth in these areas remain relatively unexplored (Feroz et al., 2018; WHO Global Observatory for eHealth, 2011). Studying predictors of mHealth engagement are particularly relevant in developing countries given the unique challenges and barriers faced by communities living at the bottom of the economic pyramid. Simply receiving an mHealth intervention does not necessarily translate to adequate uptake of that intervention, underscoring the need for research that untaps the potential of low-cost scalable interventions to populations that need them most. PREDICTORS OF MHEALTH ENGAGEMENT 63 References Agardh, E., Allebeck, P., Hallqvist, J., Moradi, T., & Sidorchuk, A. (2011). Type 2 diabetes incidence and socio-economic position: A systematic review and meta-analysis. International Journal of Epidemiology, 40(3), 804–818. https://doi.org/10.1093/ije/dyr029 Alberts, H., Thewissen, R., & Middelweerd, M. (2013). Accepting or suppressing the desire to eat: Investigating the short-term effects of acceptance-based craving regulation. Eating Behaviors, 14(3), 405–409. https://doi.org/10.1016/j.eatbeh.2013.06.008 Algina, J., & Swaminathan, H. (2011). Centering in two-level nested designs. In Handbook for advanced multilevel analysis (pp. 285–312). Routledge/Taylor & Francis Group. Anderson, M., & Horrigan, J. B. (2016). Smartphones help those without broadband get online, but don’t necessarily bridge the digital divide. Pew Research Center. https://www.pewresearch.org/fact-tank/2016/10/03/smartphones-help-those-without- broadband-get-online-but-dont-necessarily-bridge-the-digital-divide/ Appelhans, B. M., Woolf, K., Pagoto, S. L., Schneider, K. L., Whited, M. C., & Liebman, R. (2011). Inhibiting food reward: Delay discounting, food reward sensitivity, and palatable food intake in overweight and obese women. Obesity, 19(11), 2175–2182. https://doi.org/10.1038/oby.2011.57 Atkinson, M. J., & Wade, T. D. (2012). Impact of metacognitive acceptance on body dissatisfaction and negative affect: Engagement and efficacy. Journal of Consulting and Clinical Psychology, 80(3), 416–425. https://doi.org/10.1037/a0028263 Baer, R. A., Smith, G. T., Hopkins, J., Krietemeyer, J., & Toney, L. (2006). Using self-report assessment methods to explore facets of mindfulness. Assessment, 13(1), 27–45. https://doi.org/10.1177/1073191105283504 Baer, R. A., Smith, G. T., Lykins, E., Button, D., Krietemeyer, J., Sauer, S., Walsh, E., Duggan, D., & Williams, J. M. G. (2008). Construct validity of the five facet mindfulness questionnaire in meditating and nonmeditating samples. Assessment, 15(3), 329–342. https://doi.org/10.1177/1073191107313003 Baker, D. P., Leon, J., Greenaway, E. G. S., Collins, J., & Movit, M. (2011). The education effect on population health: A reassessment. Population and Development Review, 37(2), 307–332. https://doi.org/10.1111/j.1728-4457.2011.00412.x Banerjee, M., Cavanagh, K., & Strauss, C. (2018). Barriers to mindfulness: A path analytic model exploring the role of rumination and worry in predicting psychological and physical engagement in an online mindfulness-based intervention. Mindfulness, 9(3), 980–992. https://doi.org/10.1007/s12671-017-0837-4 PREDICTORS OF MHEALTH ENGAGEMENT 64 Barnett, J., Aguilar, S., Brittner, M., & Bonuck, K. (2012). Recruiting and retaining low-income, multi-ethnic women into randomized controlled trials: Successful strategies and staffing. Contemporary Clinical Trials, 33(5), 925–932. https://doi.org/10.1016/j.cct.2012.06.005 Beshara, M., Hutchinson, A. D., & Wilson, C. (2013). Does mindfulness matter? Everyday mindfulness, mindful eating and self-reported serving size of energy dense foods among a sample of South Australian adults. Appetite, 67, 25–29. https://doi.org/10.1016/j.appet.2013.03.012 Bickel, W. K., & Marsch, L. A. (2001). Toward a behavioral economic understanding of drug dependence: Delay discounting processes. Addiction, 96(1), 73–86. https://doi.org/10.1046/j.1360-0443.2001.961736.x Bradford, W. D. (2010). The association between individual time preferences and health maintenance habits. Medical Decision Making: An International Journal of the Society for Medical Decision Making, 30(1), 99–112. https://doi.org/10.1177/0272989X09342276 Brewer, J. A., Ruf, A., Beccia, A. L., Essien, G. I., Finn, L. M., van Lutterveld, R., & Mason, A. E. (2018). Can mindfulness address maladaptive eating behaviors? Why traditional diet plans fail and how new mechanistic insights may lead to novel interventions. Frontiers in Psychology, 9. https://doi.org/10.3389/fpsyg.2018.01418 Brod, M., Pfeiffer, K. M., Barnett, A. H., Berntorp, K., Vilsbøll, T., & Weissenberger, B. (2016). Perceptions of diabetes control among people with type 2 diabetes treated with basal insulin in Sweden, Switzerland, and the United Kingdom. Current Medical Research and Opinion, 32(10), 1653–1661. https://doi.org/10.1080/03007995.2016.1198311 Byambasuren, O., Sanders, S., Beller, E., & Glasziou, P. (2018). Prescribable mHealth apps identified from an overview of systematic reviews. Npj Digital Medicine, 1(1), 1–12. https://doi.org/10.1038/s41746-018-0021-9 Cameron, A. C., & Trivedi, P. K. (1998). Regression analysis of count data. Cambridge University Press. https://doi.org/10.1017/CBO9780511814365 Canby, N. K., Eichel, K., Peters, S. I., Rahrig, H., & Britton, W. B. (2021). Predictors of out-of- class mindfulness practice adherence during and after a mindfulness-based intervention. Psychosomatic Medicine, 83(6), 655–664. https://doi.org/10.1097/PSY.0000000000000873 Carmody, J., & Baer, R. A. (2008). Relationships between mindfulness practice and levels of mindfulness, medical and psychological symptoms and well-being in a mindfulness- based stress reduction program. Journal of Behavioral Medicine, 31(1), 23–33. https://doi.org/10.1007/s10865-007-9130-7 PREDICTORS OF MHEALTH ENGAGEMENT 65 Carter, M. C., Burley, V. J., Nykjaer, C., & Cade, J. E. (2013). Adherence to a smartphone application for weight loss compared to website and paper diary: Pilot randomized controlled trial. Journal of Medical Internet Research, 15(4). https://doi.org/10.2196/jmir.2283 Ceci, S. J. (1991). How much does schooling influence general intelligence and its cognitive components? A reassessment of the evidence. Developmental Psychology, 27(5), 703– 722. https://doi.org/10.1037/0012-1649.27.5.703 Centers for Disease Control and Prevention. (2019). National Diabetes Statistics Report. https://www.cdc.gov/diabetes/data/statistics/statistics-report.html Covolo, L., Ceretti, E., Moneda, M., Castaldi, S., & Gelatti, U. (2017). Does evidence support the use of mobile phone apps as a driver for promoting healthy lifestyles from a public health perspective? A systematic review of randomized control trials. Patient Education and Counseling, 100(12), 2231–2243. https://doi.org/10.1016/j.pec.2017.07.032 Crane, C., & Williams, J. M. G. (2010). Factors associated with attrition from mindfulness-based cognitive therapy in patients with a history of suicidal depression. Mindfulness, 1(1), 10– 20. https://doi.org/10.1007/s12671-010-0003-8 Daubenmier, J., Kristeller, J., Hecht, F. M., Maninger, N., Kuwata, M., Jhaveri, K., Lustig, R. H., Kemeny, M., Karan, L., & Epel, E. (2011). Mindfulness intervention for stress eating to reduce cortisol and abdominal fat among overweight and obese women: An exploratory randomized controlled study. Journal of Obesity, 2011, e651936. https://doi.org/10.1155/2011/651936 Direito, A., Jiang, Y., Whittaker, R., & Maddison, R. (2015). Smartphone apps to improve fitness and increase physical activity among young people: Protocol of the Apps for IMproving FITness (AIMFIT) randomized controlled trial. BMC Public Health, 15(1), 635. https://doi.org/10.1186/s12889-015-1968-y Donkin, L., Christensen, H., Naismith, S. L., Neal, B., Hickie, I. B., & Glozier, N. (2011). A systematic review of the impact of adherence on the effectiveness of e-therapies. Journal of Medical Internet Research, 13(3). https://doi.org/10.2196/jmir.1772 Donkin, L., Hickie, I. B., Christensen, H., Naismith, S. L., Neal, B., Cockayne, N. L., & Glozier, N. (2013). Rethinking the dose-response relationship between usage and outcome in an online intervention for depression: Randomized controlled trial. Journal of Medical Internet Research, 15(10). https://doi.org/10.2196/jmir.2771 Eysenbach, G. (2005). The law of attrition. Journal of Medical Internet Research, 7(1), e11. https://doi.org/10.2196/jmir.7.1.e11 Feroz, A., Kadir, M. M., & Saleem, S. (2018). Health systems readiness for adopting mHealth interventions for addressing non-communicable diseases in low- and middle-income countries: A current debate. Global Health Action, 11(1), 1496887. https://doi.org/10.1080/16549716.2018.1496887 PREDICTORS OF MHEALTH ENGAGEMENT 66 Festin, K., Thomas, K., Ekberg, J., & Kristenson, M. (2017). Choice of measure matters: A study of the relationship between socioeconomic status and psychosocial resources in a middle- aged normal population. PLoS ONE, 12(8), e0178929. https://doi.org/10.1371/journal.pone.0178929 Field, A. (2013). Discovering statistics using IBM SPSS Statistics (4th ed.). SAGE Publications Ltd. Forman, E. M., Butryn, M. L., Juarascio, A. S., Bradley, L. E., Lowe, M. R., Herbert, J. D., & Shaw, J. A. (2013). The mind your health project: A randomized controlled trial of an innovative behavioral treatment for obesity. Obesity, 21(6), 1119–1126. https://doi.org/10.1002/oby.20169 Free, C., Phillips, G., Galli, L., Watson, L., Felix, L., Edwards, P., Patel, V., & Haines, A. (2013). The effectiveness of mobile-health technology-based health behaviour change or disease management interventions for health care consumers: A systematic review. PLoS Medicine, 10(1), e1001362. https://doi.org/10.1371/journal.pmed.1001362 Gandapur, Y., Kianoush, S., Kelli, H. M., Misra, S., Urrea, B., Blaha, M. J., Graham, G., Marvel, F. A., & Martin, S. S. (2016). The role of mHealth for improving medication adherence in patients with cardiovascular disease: A systematic review. European Heart Journal. Quality of Care & Clinical Outcomes, 2(4), 237–244. https://doi.org/10.1093/ehjqcco/qcw018 Geyer, S., Hemström, Ö., Peter, R., & Vågerö, D. (2006). Education, income, and occupational class cannot be used interchangeably in social epidemiology. Empirical evidence against a common practice. Journal of Epidemiology and Community Health, 60(9), 804–810. https://doi.org/10.1136/jech.2005.041319 Giluk, T. L. (2009). Mindfulness, Big Five personality, and affect: A meta-analysis. Personality and Individual Differences, 47(8), 805–811. https://doi.org/10.1016/j.paid.2009.06.026 Goldberg, S. B., Wielgosz, J., Dahl, C., Schuyler, B., MacCoon, D. S., Rosenkranz, M., Lutz, A., Sebranek, C. A., & Davidson, R. J. (2016). Does the Five Facet Mindfulness Questionnaire measure what we think it does? Construct validity evidence from an active controlled randomized clinical trial. Psychological Assessment, 28(8), 1009–1014. https://doi.org/10.1037/pas0000233 Goldman, D. P., & Smith, J. P. (2002). Can patient self-management help explain the SES health gradient? Proceedings of the National Academy of Sciences of the United States of America, 99(16), 10929–10934. https://doi.org/10.1073/pnas.162086599 Green, L., Fry, A., & Myerson, J. (1994). Discounting of delayed rewards: A life-span comparison. Psychological Science, 5(1), 33–36. https://doi.org/10.1111/j.1467- 9280.1994.tb00610.x PREDICTORS OF MHEALTH ENGAGEMENT 67 Green, L., Myerson, J., Lichtman, D., Rosen, S., & Fry, A. (1996). Temporal discounting in choice between delayed rewards: The role of age and income. Psychology and Aging, 11(1), 79–84. https://doi.org/10.1037/0882-7974.11.1.79 Haliwa, I., Ford, C. G., Wilson, J. M., & Shook, N. J. (2021). A mixed-method assessment of a 10-day mobile mindfulness intervention. Frontiers in Psychology, 12, 722995. Hebert, E. A., Vincent, N., Lewycky, S., & Walsh, K. (2010). Attrition and adherence in the online treatment of chronic insomnia. Behavioral Sleep Medicine, 8(3), 141–150. https://doi.org/10.1080/15402002.2010.487457 Hill, A. J. (2007). The psychology of food craving: Symposium on ‘Molecular mechanisms and psychology of food intake.’ Proceedings of the Nutrition Society, 66(2), 277–285. https://doi.org/10.1017/S0029665107005502 Houle, J., Beaulieu, M.-D., Chiasson, J.-L., Lespérance, F., Côté, J., Strychar, I., Bherer, L., Meunier, S., & Lambert, J. (2015). Glycaemic control and self-management behaviours in Type 2 diabetes: Results from a 1-year longitudinal cohort study. Diabetic Medicine, 32(9), 1247–1254. https://doi.org/10.1111/dme.12686 Janssen, L. K., Duif, I., van Loon, I., de Vries, J. H. M., Speckens, A. E. M., Cools, R., & Aarts, E. (2018). Greater mindful eating practice is associated with better reversal learning. Scientific Reports, 8(1), 5702. https://doi.org/10.1038/s41598-018-24001-1 Jordan, C. H., Wang, W., Donatoni, L., & Meier, B. P. (2014). Mindful eating: Trait and state mindfulness predict healthier eating behavior. Personality and Individual Differences, 68, 107–111. https://doi.org/10.1016/j.paid.2014.04.013 Kabat‐Zinn, J. (2003). Mindfulness-based interventions in context: Past, present, and future. Clinical Psychology: Science and Practice, 10(2), 144–156. https://doi.org/10.1093/clipsy.bpg016 Kerr, D. C., Ornelas, I. J., Lilly, M. M., Calhoun, R., & Meischke, H. (2019). Participant engagement in and perspectives on a web-based mindfulness intervention for 9-1-1 telecommunicators: Multimethod study. Journal of Medical Internet Research, 21(6). https://doi.org/10.2196/13449 Koffarnus, M. N., & Bickel, W. K. (2014). A 5-trial adjusting delay discounting task: Accurate discount rates in less than one minute. Experimental and Clinical Psychopharmacology, 22(3), 222–228. https://doi.org/10.1037/a0035973 Kristeller, J., & Hallett, C. B. (1999). An exploratory study of a meditation-based intervention for binge eating disorder. Journal of Health Psychology, 4(3), 357–363. https://doi.org/10.1177/135910539900400305 Kristeller, J., Wolever, R. Q., & Sheets, V. (2014). Mindfulness-Based Eating Awareness Training (MB-EAT) for binge eating: A randomized clinical trial. Mindfulness, 5(3), 282–297. https://doi.org/10.1007/s12671-012-0179-1 PREDICTORS OF MHEALTH ENGAGEMENT 68 Kroenke, K., Strine, T. W., Spitzer, R. L., Williams, J. B. W., Berry, J. T., & Mokdad, A. H. (2009). The PHQ-8 as a measure of current depression in the general population. Journal of Affective Disorders, 114(1–3), 163–173. https://doi.org/10.1016/j.jad.2008.06.026 Laing, B. Y., Mangione, C. M., Tseng, C.-H., Leng, M., Vaisberg, E., Mahida, M., Bholat, M., Glazier, E., Morisky, D. E., & Bell, D. S. (2014). Effectiveness of a smartphone application for weight loss compared with usual care in overweight primary care patients: A randomized, controlled trial. Annals of Internal Medicine, 161(10_Supplement), S5. https://doi.org/10.7326/M13-3005 Lebeau, G., Consoli, S. M., Le Bouc, R., Sola-Gazagnes, A., Hartemann, A., Simon, D., Reach, G., Altman, J.-J., Pessiglione, M., Limosin, F., & Lemogne, C. (2016). Delay discounting of gains and losses, glycemic control and therapeutic adherence in type 2 diabetes. Behavioural Processes, 132, 42–48. https://doi.org/10.1016/j.beproc.2016.09.006 Leung, A. W. Y., Chan, R. S. M., Sea, M. M. M., & Woo, J. (2017). An overview of factors associated with adherence to lifestyle modification programs for weight management in adults. International Journal of Environmental Research and Public Health, 14(8). https://doi.org/10.3390/ijerph14080922 Little, R. J. A., & Rubin, D. B. (2020). Statistical analysis with missing data. Wiley. Liu, B., & Rice, V. J. (2018). Characteristics of United States military personnel and veterans who complete mindfulness training. Advances in Human Factors and Ergonomics in Healthcare and Medical Devices, 590, 22–32. https://doi.org/10.1007/978-3-319-60483- 1_3 Lomas, T., Cartwright, T., Edginton, T., & Ridge, D. (2015). A qualitative analysis of experiential challenges associated with meditation practice. Mindfulness, 6(4), 848–860. https://doi.org/10.1007/s12671-014-0329-8 Lyzwinski, L. N., Edirippulige, S., Caffery, L., & Bambling, M. (2019). Mindful eating mobile health apps: Review and appraisal. JMIR Mental Health, 6(8). https://doi.org/10.2196/12820 Marmot, M., Friel, S., Bell, R., Houweling, T. A., & Taylor, S. (2008). Closing the gap in a generation: Health equity through action on the social determinants of health. The Lancet, 372(9650), 1661–1669. https://doi.org/10.1016/S0140-6736(08)61690-6 Mason, A. E., Jhaveri, K., Cohn, M., & Brewer, J. A. (2018). Testing a mobile mindful eating intervention targeting craving-related eating: Feasibility and proof of concept. Journal of Behavioral Medicine, 41(2), 160–173. https://doi.org/10.1007/s10865-017-9884-5 Mason, A. E., Saslow, L., Moran, P. J., Kim, S., Wali, P. K., Abousleiman, H., Hartman, A., Richler, R., Schleicher, S., Hartogensis, W., Epel, E. S., & Hecht, F. (2019). Examining the effects of mindful eating training on adherence to a carbohydrate-restricted diet in patients with Type 2 diabetes (the DELISH Study): Protocol for a randomized controlled trial. JMIR Research Protocols, 8(2), e11002. https://doi.org/10.2196/11002 PREDICTORS OF MHEALTH ENGAGEMENT 69 Mattila, E., Lappalainen, R., Välkkynen, P., Sairanen, E., Lappalainen, P., Karhunen, L., Peuhkuri, K., Korpela, R., Kolehmainen, M., & Ermes, M. (2016). Usage and dose response of a mobile Acceptance and Commitment Therapy app: Secondary analysis of the intervention arm of a randomized controlled trial. JMIR MHealth and UHealth, 4(3), e90. https://doi.org/10.2196/mhealth.5241 Mattila, E., Orsama, A.-L., Ahtinen, A., Hopsu, L., Leino, T., & Korhonen, I. (2013). Personal health technologies in employee health promotion: Usage activity, usefulness, and health- related outcomes in a 1-year randomized controlled trial. JMIR MHealth and UHealth, 1(2). https://doi.org/10.2196/mhealth.2557 May, J., Andrade, J., Kavanagh, D. J., & Hetherington, M. (2012). Elaborated Intrusion Theory: A cognitive-emotional theory of food craving. Current Obesity Reports, 1(2), 114–121. https://doi.org/10.1007/s13679-012-0010-2 Meule, A., Hermann, T., & Kübler, A. (2014). A short version of the Food Cravings Questionnaire-Trait: The FCQ-T-reduced. Frontiers in Psychology, 5, 190. https://doi.org/10.3389/fpsyg.2014.00190 Michie, S., & Abraham, C. (2004). Interventions to change health behaviours: Evidence-based or evidence-inspired? Psychology & Health, 19(1), 29–49. https://doi.org/10.1080/0887044031000141199 Miller, C. K. (2017). Mindful eating with diabetes. Diabetes Spectrum, 30(2), 89–94. https://doi.org/10.2337/ds16-0039 Miller, C. K., Kristeller, J. L., Headings, A., & Nagaraja, H. (2014). Comparison of a mindful eating intervention to a diabetes self-management intervention among adults with type 2 diabetes: A randomized controlled trial. Health Education & Behavior, 41(2), 145–154. https://doi.org/10.1177/1090198113493092 Miller, C. K., Kristeller, J. L., Headings, A., Nagaraja, H., & Miser, W. F. (2012). Comparative effectiveness of a mindful eating intervention to a diabetes self-management intervention among adults with type 2 diabetes: A pilot study. Journal of the Academy of Nutrition and Dietetics, 112(11), 1835–1842. https://doi.org/10.1016/j.jand.2012.07.036 Muthén, L. K., & Muthén, B. O. (2017). Mplus User’s Guide. Eighth Edition. Muthén & Muthén. Myerson, J., Green, L., & Warusawitharana, M. (2001). Area under the curve as a measure of discounting. Journal of the Experimental Analysis of Behavior, 76(2), 235–243. https://doi.org/10.1901/jeab.2001.76-235 Nelson, L. A., Mulvaney, S. A., Gebretsadik, T., Ho, Y.-X., Johnson, K. B., & Osborn, C. Y. (2016). Disparities in the use of a mHealth medication adherence promotion intervention for low-income adults with type 2 diabetes. Journal of the American Medical Informatics Association: JAMIA, 23(1), 12–18. https://doi.org/10.1093/jamia/ocv082 PREDICTORS OF MHEALTH ENGAGEMENT 70 Olokoba, A. B., Obateru, O. A., & Olokoba, L. B. (2012). Type 2 diabetes mellitus: A review of current trends. Oman Medical Journal, 27(4), 269–273. https://doi.org/10.5001/omj.2012.68 Pelchat, M. L., & Schaefer, S. (2000). Dietary monotony and food cravings in young and elderly adults. Physiology & Behavior, 68(3), 353–359. https://doi.org/10.1016/s0031- 9384(99)00190-0 Perrin, A. (2021, June 3). Mobile technology and home broadband 2021. Pew Research Center: Internet, Science & Tech. https://www.pewresearch.org/internet/2021/06/03/mobile- technology-and-home-broadband-2021/ Perski, O., Blandford, A., West, R., & Michie, S. (2017). Conceptualising engagement with digital behaviour change interventions: A systematic review using principles from critical interpretive synthesis. Translational Behavioral Medicine, 7(2), 254–267. https://doi.org/10.1007/s13142-016-0453-1 Peters, E., Baker, D. P., Dieckmann, N. F., Leon, J., & Collins, J. (2010). Explaining the effect of education on health: A field study in Ghana. Psychological Science, 21(10), 1369–1376. https://doi.org/10.1177/0956797610381506 Pew Research Center. (2019). Demographics of mobile device ownership and adoption in the United States. Pew Research Center: Internet, Science & Tech. https://www.pewresearch.org/internet/fact-sheet/mobile/ Pew Research Center. (2021, April 7). Mobile Fact Sheet. Pew Research Center Internet, Science & Tech. https://www.pewresearch.org/internet/fact-sheet/mobile/ Pham, Q., Graham, G., Carrion, C., Morita, P. P., Seto, E., Stinson, J. N., & Cafazzo, J. A. (2019). A library of analytic indicators to evaluate effective engagement with consumer mHealth apps for chronic conditions: Scoping review. JMIR MHealth and UHealth, 7(1), e11941. https://doi.org/10.2196/11941 Polizzi, C., Gautam, A., & Lynn, S. J. (2018). Trait mindfulness: A multifaceted evaluation. Imagination, Cognition and Personality, 38(2), 106–138. https://doi.org/10.1177/0276236618774321 Proudfoot, J., Clarke, J., Birch, M.-R., Whitton, A. E., Parker, G., Manicavasagar, V., Harrison, V., Christensen, H., & Hadzi-Pavlovic, D. (2013). Impact of a mobile phone and web program on symptom and functional outcomes for people with mild-to-moderate depression, anxiety and stress: A randomised controlled trial. BMC Psychiatry, 13(1), 312. https://doi.org/10.1186/1471-244X-13-312 R Core Team. (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.eea.europa.eu/data-and-maps/indicators/oxygen- consuming-substances-in-rivers/r-development-core-team-2006 PREDICTORS OF MHEALTH ENGAGEMENT 71 Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods. Sage Publications. Reach, G. (2008). A novel conceptual framework for understanding the mechanism of adherence to long term therapies. Patient Preference and Adherence, 2, 7–19. Reimers, S., Maylor, E. A., Stewart, N., & Chater, N. (2009). Associations between a one-shot delay discounting measure and age, income, education and real-world impulsive behavior. Personality and Individual Differences, 47(8), 973–978. https://doi.org/10.1016/j.paid.2009.07.026 Reinwand, D. A., Schulz, D. N., Crutzen, R., Kremers, S. P., & de Vries, H. (2015). Who follows eHealth interventions as recommended? A study of participants’ personal characteristics from the experimental arm of a randomized controlled trial. Journal of Medical Internet Research, 17(5), e115. https://doi.org/10.2196/jmir.3932 Richard, A., Meule, A., Reichenberger, J., & Blechert, J. (2017). Food cravings in everyday life: An EMA study on snack-related thoughts, cravings, and consumption. Appetite, 113, 215–223. https://doi.org/10.1016/j.appet.2017.02.037 Robbins, J. M., Vaccarino, V., Zhang, H., & Kasl, S. V. (2005). Socioeconomic status and diagnosed diabetes incidence. Diabetes Research and Clinical Practice, 68(3), 230–236. https://doi.org/10.1016/j.diabres.2004.09.007 Roepke, A. M., Jaffee, S. R., Riffle, O. M., McGonigal, J., Broome, R., & Maxwell, B. (2015). Randomized controlled trial of SuperBetter, a smartphone-based/internet-based self-help tool to reduce depressive symptoms. Games for Health Journal, 4(3), 235–246. https://doi.org/10.1089/g4h.2014.0046 Rung, A. L., Oral, E., Berghammer, L., & Peters, E. S. (2020). Feasibility and acceptability of a mobile mindfulness meditation intervention among women: Intervention study. JMIR MHealth and UHealth, 8(6), e15943. https://doi.org/10.2196/15943 Sabaté, E. (2003). Adherence to Long-term Therapies: Evidence for Action. World Health Organization. Saeedi, P., Petersohn, I., Salpea, P., Malanda, B., Karuranga, S., Unwin, N., Colagiuri, S., Guariguata, L., Motala, A. A., Ogurtsova, K., Shaw, J. E., Bright, D., & Williams, R. (2019). Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: Results from the International Diabetes Federation Diabetes Atlas, 9th edition. Diabetes Research and Clinical Practice, 157, 107843. https://doi.org/10.1016/j.diabres.2019.107843 Schlundt, D. G., Rea, M. R., Kline, S. S., & Pichert, J. W. (1994). Situational obstacles to dietary adherence for adults with diabetes. Journal of the American Dietetic Association, 94(8), 874–876, 879. https://doi.org/10.1016/0002-8223(94)92367-1 PREDICTORS OF MHEALTH ENGAGEMENT 72 Siegling, A. B., & Petrides, K. V. (2014). Measures of trait mindfulness: Convergent validity, shared dimensionality, and linkages to the five-factor model. Frontiers in Psychology, 5. https://doi.org/10.3389/fpsyg.2014.01164 Sieverink, F., Kelders, S. M., & Gemert-Pijnen, J. E. van. (2017). Clarifying the concept of adherence to eHealth technology: Systematic review on when usage becomes adherence. Journal of Medical Internet Research, 19(12), e402. https://doi.org/10.2196/jmir.8578 Singer, J. D., & Willett, J. B. (2003). Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. Oxford University Press. StataCorp. (2021). Stata Statistical Software: Release 17. https://www.stata.com/support/faqs/resources/citing-software-documentation-faqs/ Steenland, K., Henley, J., & Thun, M. (2002). All-cause and cause-specific death rates by educational status for two million people in two American Cancer Society cohorts, 1959– 1996. American Journal of Epidemiology, 156(1), 11–21. https://doi.org/10.1093/aje/kwf001 Syed, S. T., Gerber, B. S., & Sharp, L. K. (2013). Traveling towards disease: Transportation barriers to health care access. Journal of Community Health, 38(5), 976–993. https://doi.org/10.1007/s10900-013-9681-1 Szinay, D., Jones, A., Chadborn, T., Brown, J., & Naughton, F. (2020). Influences on the uptake of and engagement with health and well-being smartphone apps: Systematic review. Journal of Medical Internet Research, 22(5), e17572. https://doi.org/10.2196/17572 Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics. Pearson/Allyn & Bacon. Tapper, K. (2018). Mindfulness and craving: Effects and mechanisms. Clinical Psychology Review, 59, 101–117. https://doi.org/10.1016/j.cpr.2017.11.003 Teuscher, U., & Mitchell, S. H. (2011). Relation between time perspective and delay discounting: A literature review. The Psychological Record, 61(4), 613–632. https://doi.org/10.1007/BF03395780 Tofighi, D., & MacKinnon, D. P. (2011). RMediation: An R package for mediation analysis confidence intervals. Behavior Research Methods, 43(3), 692–700. https://doi.org/10.3758/s13428-011-0076-x Trompetter, H. R., Bohlmeijer, E. T., Veehof, M. M., & Schreurs, K. M. G. (2015). Internet- based guided self-help intervention for chronic pain based on Acceptance and Commitment Therapy: A randomized controlled trial. Journal of Behavioral Medicine, 38(1), 66–80. https://doi.org/10.1007/s10865-014-9579-0 PREDICTORS OF MHEALTH ENGAGEMENT 73 Tronieri, J. S., Wadden, T. A., Pearl, R. L., Berkowitz, R. I., Alamuddin, N., & Chao, A. M. (2020). Mindful eating, general mindful awareness, and acceptance as predictors of weight loss. Mindfulness, 11(12), 2818–2827. https://doi.org/10.1007/s12671-020-01493- 5 Turner-McGrievy, G., & Tate, D. (2011). Tweets, apps, and pods: Results of the 6-month mobile pounds off digitally (mobile POD) randomized weight-loss intervention among adults. Journal of Medical Internet Research, 13(4), e120. https://doi.org/10.2196/jmir.1841 van Dulmen, S., Sluijs, E., van Dijk, L., de Ridder, D., Heerdink, R., & Bensing, J. (2007). Patient adherence to medical treatment: A review of reviews. BMC Health Services Research, 7, 55. https://doi.org/10.1186/1472-6963-7-55 Wangberg, S. C., Bergmo, T. S., & Johnsen, J.-A. K. (2008). Adherence in Internet-based interventions. Patient Preference and Adherence, 2, 57–65. Weingarten, H. P., & Elston, D. (1991). Food cravings in a college population. Appetite, 17(3), 167–175. https://doi.org/10.1016/0195-6663(91)90019-o WHO Global Observatory for eHealth. (2011). mHealth: New horizons for health through mobile technologies. World Health Organization. Williams, P. G., Rau, H. K., Cribbet, M. R., & Gunn, H. E. (2009). Openness to experience and stress regulation. Journal of Research in Personality, 43(5), 777–784. https://doi.org/10.1016/j.jrp.2009.06.003 Wing, R. R., Goldstein, M. G., Acton, K. J., Birch, L. L., Jakicic, J. M., Sallis, J. F., Smith-West, D., Jeffery, R. W., & Surwit, R. S. (2001). Behavioral science research in diabetes: Lifestyle changes related to obesity, eating behavior, and physical activity. Diabetes Care, 24(1), 117–123. https://doi.org/10.2337/diacare.24.1.117 Yardley, L., Spring, B. J., Riper, H., Morrison, L. G., Crane, D. H., Curtis, K., Merchant, G. C., Naughton, F., & Blandford, A. (2016). Understanding and promoting effective engagement with digital behavior change interventions. American Journal of Preventive Medicine, 51(5), 833–842. https://doi.org/10.1016/j.amepre.2016.06.015 Young, C., Campolonghi, S., Ponsonby, S., Dawson, S. L., O’Neil, A., Kay-Lambkin, F., McNaughton, S. A., Berk, M., & Jacka, F. N. (2019). Supporting engagement, adherence, and behavior change in online dietary interventions. Journal of Nutrition Education and Behavior, 51(6), 719–739. https://doi.org/10.1016/j.jneb.2019.03.006 Yu, J. H., Shin, M. S., Kim, D. J., Lee, J. R., Yoon, S.-Y., Kim, S. G., Koh, E. H., Lee, W. J., Park, J.-Y., & Kim, M.-S. (2013). Enhanced carbohydrate craving in patients with poorly controlled Type 2 diabetes mellitus. Diabetic Medicine: A Journal of the British Diabetic Association, 30(9), 1080–1086. https://doi.org/10.1111/dme.12209 PREDICTORS OF MHEALTH ENGAGEMENT 74 Yzerbyt, V., Muller, D., Batailler, C., & Judd, C. M. (2018). New recommendations for testing indirect effects in mediational models: The need to report and test component paths. Journal of Personality and Social Psychology, 115(6), 929–943. https://doi.org/10.1037/pspa0000132 Zhao, J., Freeman, B., & Li, M. (2016). Can mobile phone apps influence people’s health behavior change? An evidence review. Journal of Medical Internet Research, 18(11), e287. https://doi.org/10.2196/jmir.5692 PREDICTORS OF MHEALTH ENGAGEMENT 75 Tables Table 1 Study 1: Frequencies and Percentages for Categorical Variables Variables n % Gender Male 34 37.36 Female 57 62.64 Race/ethnicity* African American/Black 11 12.09 Asian/Pacific Islander 15 16.48 White/Caucasian 42 46.15 Latino/Hispanic 8 8.79 Other 14 15.38 Household income* Less than $20k 9 9.89 $20 to $35k 9 9.89 $35 to $50k 7 7.69 $50 to $75k 9 9.89 $75 to $100k 12 13.19 $100 to $150k 14 15.38 $150 to $200k 12 13.19 Greater than $200k 10 10.99 Employment status Employed full-time 36 39.56 Employed part-time/student 19 20.88 Unemployed 9 9.89 Retired 27 29.67 Education level* Some college or less 24 26.37 Associate's or 2-yr degree 12 13.19 Bachelor’s degree 27 29.67 Master's/Doctoral degree 27 29.67 Overall adherence Did not complete all 28 modules 29 31.87 Completed all 28 modules 62 68.13 Note. *Frequencies not summing to 91 reflect missing data. PREDICTORS OF MHEALTH ENGAGEMENT 76 Table 2 Study 1: Means and Standard Deviations for Continuous Variables Variables N M SD Min Max Depression 91 6.43 4.76 0 21 Delay discounting AUC 91 .71 .27 .02 .99 Age 91 59.09 10.52 21 80 Total modules per participant 91 25.93 5.09 0 28 Modules per week 1092 2.16 1.93 0 7 Activities per day* 6318 2.92 4.26 0 50 Note. *Ns not equal to 7098 reflect missing data for daily usage data. PREDICTORS OF MHEALTH ENGAGEMENT 77 Table 3 Study 1: Frequencies and Percentages for Categorical Demographic Variables by Education Variable Some College or Less Associate's/ 2-yr Degree Bachelor’s Degree Graduate Degree n % n % n % n % Gender Male 9 37.5 3 25.0 12 44.4 9 33.3 Female 15 62.5 9 75.0 15 55.6 18 66.7 Race/ethnicity African American/Black 5 20.8 1 9.1 2 7.4 3 11.1 Asian/Pacific Islander 4 16.7 2 18.2 5 18.5 4 14.8 White/Caucasian 7 29.2 5 45.5 12 44.4 17 63.0 Latino/Hispanic 3 12.5 1 9.1 4 14.8 0 .0 Other 5 20.8 2 18.2 4 14.8 3 11.1 Household income Less than $20k 4 16.7 3 30.0 1 4.2 0 .0 $20 to $35k 3 12.5 1 10.0 4 16.7 1 4.4 $35 to $50k 2 8.3 0 .0 2 8.3 3 13.0 $50 to $75k 6 25.0 0 .0 1 4.2 2 8.7 $75 to $100k 4 16.7 3 30.0 5 20.8 0 .0 $100 to $150k 2 8.3 1 10.0 7 29.2 4 17.4 $150 to $200k 2 8.3 2 20.0 3 12.5 5 21.7 Greater than $200k 1 4.2 0 .0 1 4.2 8 34.8 Employment Employed full-time 14 58.3 6 50.0 8 29.6 8 29.6 Employed part-time/student 4 16.7 2 16.7 7 25.9 6 22.2 Unemployed 3 12.5 2 16.7 2 7.4 2 7.4 Retired 3 12.5 2 16.7 10 37.0 11 40.7 PREDICTORS OF MHEALTH ENGAGEMENT 78 Table 4 Study 1: Means and Standard Deviations for Age, Depression, and Delay Discounting AUC by Education Variables by Education N M SD F p Depression by education 1.48 .226 Some college or less 24 7.46 a 4.79 Associate's or 2-yr degree 12 4.08 a 3.60 Bachelor’s degree 27 6.04 a 4.28 Master's/Doctoral degree 27 6.81 a 5.48 Delay discounting AUC by education 3.25 .026 Some college or less 24 .65 a .27 Associate's or 2-yr degree 12 .82 a .17 Bachelor’s degree 27 .63 a .30 Master's/Doctoral degree 27 .80 a .22 Age by education 6.80 < .001 Some college or less 24 51.71 a 11.85 Associate's or 2-yr degree 12 58.83 a,b 8.26 Bachelor’s degree 27 61.41 b 9.47 Master's/Doctoral degree 27 63.15 b 8.05 Note. Means with different superscripts differ significantly, p < .05, calculated using Bonferroni correction. PREDICTORS OF MHEALTH ENGAGEMENT 79 Table 5 Study 1: Mediation Analysis of Education through Delay Discounting AUC on Overall Adherence Variable c-path a-path b-path Indirect Effect Overall Adherence Delay Discounting AUC Overall Adherence 95% CI B p B p B p B LL UL Meditation Some college or less a -1.825 .037 -.152 .028 -1.858 .034 -.164 -.739 .284 Associate's/2-yr degree a -.307 .784 .022 .731 -.439 .708 .024 -.205 .309 Bachelor degree a .315 .708 -.169 .017 .491 .598 -.182 -.803 .312 Delay discounting AUC 1.076 .454 Covariates Male b .746 .274 .804 .279 African American/Black c .973 .414 1.030 .365 Asian/Pacific Islander c .342 .697 .356 .693 Latino/Hispanic c -1.456 .155 -1.388 .178 Other c -.444 .600 -.570 .521 Income level -.614 .003 -.664 .004 Employed part-time d -1.123 .209 -1.407 .194 Unemployed d -.470 .710 -.561 .664 Retired d -.665 .503 -.721 .468 Depression -.257 .001 -.257 .001 Age -.004 .920 -.009 .819 Note. Reported coefficients are unstandardized beta coefficients. a Compared to master’s/doctoral degree, b Compared to female, c Compared to White/Caucasian, and d Compared to Employed full- time. PREDICTORS OF MHEALTH ENGAGEMENT 80 Table 6 Study 1: Mediation Analysis of Education through Delay Discounting AUC on Weekly Adherence Variable c-path a-path b-path Indirect Effect Overall Adherence Delay Discounting AUC Overall Adherence 95% CI B p B p B p B LL UL Meditation Some college or less a -.425 .019 -.161 .017 -.423 .019 -.011 -.067 .038 Associate's/2-yr degree a -.106 .308 .016 .796 -.114 .271 .001 -.020 .025 Bachelor degree a -.001 .990 -.175 .011 .008 .918 -.012 -.072 .041 Delay discounting AUC .068 .638 Covariates Time (weeks) -.307 .000 -.307 .000 Male b .066 .431 .063 .459 African American/Black c -.043 .770 -.037 .799 Asian/Pacific Islander c -.022 .801 -.025 .779 Latino/Hispanic c -.169 .285 -.162 .301 Other c -.118 .423 -.124 .378 Income level -.064 .009 -.067 .008 Employed part-time d -.115 .417 -.127 .357 Unemployed d -.163 .353 -.165 .346 Retired d -.049 .701 -.053 .681 Depression -.013 .172 -.013 .183 Age -.002 .730 -.002 .708 Note. Reported coefficients are unstandardized beta coefficients. a Compared to master’s/doctoral degree, b Compared to female, c Compared to White/Caucasian, and d Compared to Employed full- time. PREDICTORS OF MHEALTH ENGAGEMENT 81 Table 7 Study 1: Mediation Analysis of Education through Delay Discounting AUC on Daily Usage Variable c-path a-path b-path Indirect Effect Daily Usage Delay Discounting AUC Daily Usage 95% CI B p B p B p B LL UL Mediation (Daily usage) Some college or less a -.043 .771 -.160 .018 -.039 .790 -.031 -.103 .009 Associate's/2-yr degree a .085 .576 .009 .886 .062 .691 .002 -.032 .039 Bachelor degree a .006 .954 -.179 .009 .034 .727 -.039 -.111 .01 Delay discounting AUC .220 .121 Time (days) -.013 .001 -.013 .000 Male b -.154 .037 -.162 .026 African American/Black c .046 .654 .063 .522 Asian/Pacific Islander c .047 .660 .042 .701 Latino/Hispanic c -.070 .370 -.047 .535 Other c .137 .259 .122 .307 Income level -.013 .536 -.021 .333 Employed part-time d -.174 .141 -.212 .082 Unemployed d .047 .716 .041 .745 Retired d .095 .391 .088 .424 Depression .013 .109 .013 .093 Age .003 .583 .002 .673 Zero state (Daily usage) Some college or less a .375 .479 .362 .494 Associate's/2-yr degree a .227 .550 .190 .613 Bachelor degree a -.519 .179 -.495 .206 Delay discounting AUC .284 .618 Time (days) .025 .003 .025 .000 Male b -.230 .436 -.219 .457 African American/Black c .120 .772 .128 .759 Asian/Pacific Islander c -.322 .413 -.304 .437 Latino/Hispanic c .808 .044 .841 .040 Other c .075 .890 .061 .908 Income level .049 .554 .037 .653 Employed part-time d .616 .223 .541 .291 Unemployed d .637 .174 .604 .200 Retired d .036 .937 .014 .975 PREDICTORS OF MHEALTH ENGAGEMENT 82 Variable c-path a-path b-path Indirect Effect Daily Usage Delay Discounting AUC Daily Usage 95% CI B p B p B p B LL UL Depression .029 .407 .031 .308 Age .005 .764 .003 .864 Note. Reported coefficients are unstandardized beta coefficients. a Compared to master’s/doctoral degree, b Compared to female, c Compared to White/Caucasian, and d Compared to Employed full-time. PREDICTORS OF MHEALTH ENGAGEMENT 83 Table 8 Study 2: Frequencies and Percentages for Categorical Variables Variables n % Gender Male 34 37.36 Female 57 62.64 Race/ethnicity* African American/Black 11 12.09 Asian/Pacific Islander 15 16.48 White/Caucasian 42 46.15 Latino/Hispanic 8 8.79 Other 14 15.38 Household income* Less than $20k 9 9.89 $20 to $35k 9 9.89 $35 to $50k 7 7.69 $50 to $75k 9 9.89 $75 to $100k 12 13.19 $100 to $150k 14 15.38 $150 to $200k 12 13.19 Greater than $200k 10 10.99 Employment status Employed full-time 36 39.56 Employed part-time/student 19 20.88 Unemployed 9 9.89 Retired 27 29.67 Education level* Some college or less 24 26.37 Associate's or 2-yr degree 12 13.19 Bachelor’s degree 27 29.67 Master's/Doctoral degree 27 29.67 Overall adherence Did not complete all 28 modules 29 31.87 Completed all 28 modules 62 68.13 Note. *Frequencies not summing to 91 reflect missing data. PREDICTORS OF MHEALTH ENGAGEMENT 84 Table 9 Study 2: Means and Standard Deviations for Continuous Variables Variables N M SD Min Max Depression (PHQ-8) 91 6.43 4.76 0 21 Food craving (FCQ-T-R) 91 52.24 13.08 23 87 Mindfulness (FFMQ) 91 79.87 12.07 43 106 Age 91 59.09 10.52 21 80 Total modules per participant 91 25.93 5.09 0 28 Modules per week 1092 2.16 1.93 0 7 Activities per day* 6318 2.92 4.26 0 50 Note. *Ns not equal to 7098 reflect missing data for daily usage data. PREDICTORS OF MHEALTH ENGAGEMENT 85 Table 10 Study 2: Means and Standard Deviations for Trait Food Craving by Categorical Demographic Variables Demographics N M SD F p Gender 1.28 .261 Male 34 50.24 a 13.93 Female 57 53.44 a 12.51 Race/ethnicity 1.32 .270 African American/Black 11 46.91 a 13.41 Asian/Pacific Islander 15 53.80 a 17.30 White/Caucasian 42 54.48 a 12.34 Latino/Hispanic 8 49.63 a 10.58 Other 14 47.93 a 8.70 Household income 1.85 .090 Less than $20k 9 50.22 a 10.16 $20 to $35k 9 44.00 a 8.51 $35 to $50k 7 57.14 a 13.04 $50 to $75k 9 56.78 a 10.32 $75 to $100k 12 52.00 a 14.17 $100 to $150k 14 51.86 a 15.31 $150 to $200k 12 60.50 a 11.45 Greater than $200k 10 47.30 a 14.08 Employment status 0.37 .772 Employed full-time 36 51.44 a 13.11 Employed part-time/student 19 50.95 a 9.37 Unemployed 9 56.00 a 19.20 Retired 27 52.96 a 13.37 Note. Means with different superscripts differ significantly, p < .05, calculated using Bonferroni correction. PREDICTORS OF MHEALTH ENGAGEMENT 86 Table 11 Study 2: Bivariate Correlations between Continuous Variables of Interest Variable Food Craving Mindfulness Depression r p r p r p Food craving (FCQ-T-R) - Mindfulness (FFMQ) -.448 .001 - Depression (PHQ-8) .389 .001 -.509 .001 - Age -.065 .538 .136 .198 -.168 .111 PREDICTORS OF MHEALTH ENGAGEMENT 87 Table 12 Study 2: Moderation Analysis of Food Craving with Mindfulness on Overall Adherence Variable Basic Model Moderation Model Overall Adherence Overall Adherence B p B p Moderation Food craving -.024 .564 -.030 .383 Mindfulness -.020 .043 .004 .920 Food craving*Mindfulness -.005 .017 Covariates Some college or less a -1.891 .043 -1.848 .050 Associate's or 2-yr degree a -.085 .944 .194 .891 Bachelor degree a .383 .663 .420 .664 Male b .710 .301 .749 .256 African American/Black c .914 .443 .304 .806 Asian/Pacific Islander c .411 .633 .272 .756 Latino/Hispanic c -1.370 .174 -1.421 .168 Other c -.385 .658 -.310 .736 Income level -.632 .006 -.671 .005 Employed part-time/student d -1.200 .201 -1.482 .165 Unemployed d -.434 .749 -.862 .496 Retired d -.700 .495 -.793 .444 Depression -.263 .009 -.272 .017 Age -.003 .926 .006 .877 Note. Reported coefficients are unstandardized beta coefficients. a Compared to master’s/doctoral degree, b Compared to female, c Compared to White/Caucasian, and d Compared to Employed full- time. PREDICTORS OF MHEALTH ENGAGEMENT 88 Table 13 Study 2: Moderation Analysis of Food Craving with Mindfulness on Weekly Adherence Variable Basic Model Moderation Model Weekly Adherence Weekly Adherence B p B p Moderation Food craving -.001 .578 -.001 .542 Mindfulness -.007 .058 -.007 .067 Food craving*Mindfulness .000 .660 Covariates Time (weeks) -.307 .000 -.307 .000 Some college or less a -.438 .015 -.434 .017 Associate's or 2-yr degree a -.078 .478 -.077 .485 Bachelor degree a .016 .833 .016 .826 Male b .052 .531 .053 .524 African American/Black c -.057 .690 -.069 .642 Asian/Pacific Islander c -.050 .574 -.052 .565 Latino/Hispanic c -.129 .404 -.131 .394 Other c -.107 .468 -.110 .465 Income level -.070 .006 -.070 .006 Employed part-time/student d -.106 .449 -.108 .444 Unemployed d -.170 .318 -.174 .306 Retired d -.076 .557 -.080 .544 Depression -.020 .084 -.021 .088 Age -.001 .899 .000 .936 Note. Reported coefficients are unstandardized beta coefficients. a Compared to female a Compared to master’s/doctoral degree, b Compared to female, c Compared to White/Caucasian, and d Compared to Employed full-time. PREDICTORS OF MHEALTH ENGAGEMENT 89 Table 14 Study 2: Moderation Analysis of Food Craving with Mindfulness on Daily Usage Variable Basic Model Moderation Model Daily Usage Daily Usage B p B p Moderation (Daily usage) Food craving -.005 .105 -.005 .101 Mindfulness -.004 .173 -.004 .176 Food craving*Mindfulness .000 .853 Time (days) -.013 .000 -.013 .000 Some college or less a -.045 .766 -.043 .777 Associate's or 2-yr degree a .133 .409 .133 .407 Bachelor degree a .018 .860 .019 .853 Male b -.177 .013 -.177 .012 African American/Black c .013 .893 .009 .934 Asian/Pacific Islander c .036 .739 .035 .743 Latino/Hispanic c -.048 .560 -.049 .545 Other c .136 .253 .134 .247 Income level -.013 .545 -.013 .546 Employed part-time/student d -.164 .158 -.164 .165 Unemployed d .060 .645 .059 .650 Retired d .096 .399 .094 .410 Depression .013 .122 .013 .141 Age .003 .563 .003 .554 Zero state (Daily usage) Food craving .002 .872 .001 .930 Mindfulness .017 .190 .018 .180 Food craving*Mindfulness -.001 .528 Time (days) .025 .000 .025 .000 Some college or less a .408 .426 .447 .378 Associate's or 2-yr degree a .163 .688 .173 .666 Bachelor degree a -.564 .149 -.554 .158 Male b -.196 .512 -.196 .513 African American/Black c .143 .722 .077 .861 Asian/Pacific Islander c -.238 .548 -.252 .526 Latino/Hispanic c .713 .088 .698 .096 Other c .055 .918 .038 .943 Income level .066 .455 .066 .459 Employed part-time/student d .589 .238 .578 .262 PREDICTORS OF MHEALTH ENGAGEMENT 90 Variable Basic Model Moderation Model Daily Usage Daily Usage B p B p Unemployed d .650 .166 .632 .190 Retired d .117 .798 .087 .855 Depression .048 .266 .045 .308 Age .002 .936 .003 .863 Note. Reported coefficients are unstandardized beta coefficients. a Compared to female a Compared to master’s/doctoral degree, b Compared to female, c Compared to White/Caucasian, and d Compared to Employed full-time. PREDICTORS OF MHEALTH ENGAGEMENT 91 Figures Figure 1 Study 1: Weekly Adherence over Time Note. Blue area around the center line are 95% confidence intervals. PREDICTORS OF MHEALTH ENGAGEMENT 92 Figure 2 Study 1: Daily Usage over Time Note. Blue area around the center line are 95% confidence intervals. PREDICTORS OF MHEALTH ENGAGEMENT 93 Figure 3 Study 2: Interaction Effect of Levels of Food Craving with Levels of Mindfulness on Overall Adherence PREDICTORS OF MHEALTH ENGAGEMENT 94 Appendices Appendix A: Patient Health Questionnaire Over the last 2 weeks, how often have you been bothered by any of the following problems? 1. Little interest or pleasure in doing things 2. Feeling down, depressed, or hopeless 3. Trouble falling or staying asleep, or sleeping too much 4. Feeling tired or having little energy 5. Poor appetite or overeating 6. Feeling bad about yourself, or that you are a failure, or have let yourself or your family down 7. Trouble concentrating on things, such as reading the newspaper or watching television 8. Moving or speaking so slowly that other people could have noticed. Or the opposite — being so fidgety or restless that you have been moving around a lot more than usual PREDICTORS OF MHEALTH ENGAGEMENT 95 Appendix B: Food Cravings Questionnaire-Trait-Reduced 1. When I crave something, I know I won't be able to stop eating once I start. 2. If I eat what I am craving, I often lose control and eat too much. 3. Food cravings invariably make me think of ways to get what I want to eat. 4. I feel like I have food on my mind all the time. 5. I find myself preoccupied with food. 6. Whenever I have cravings, I find myself making plans to eat. 7. I crave foods when I feel bored, angry, or sad. 8. I have no will power to resist my food crave. 9. Once I start eating, I have trouble stopping. 10. I can't stop thinking about eating no matter how hard I try. 11. If I give in to a food craving, all control is lost. 12. Whenever I have a food craving, I keep on thinking about eating until I actually eat the food. 13. If I am craving something, thoughts of eating it consume me. 14. My emotions often make me want to eat. 15. It is hard for me to resist the temptation to eat appetizing foods that are in my reach. PREDICTORS OF MHEALTH ENGAGEMENT 96 Appendix C: Five Factor Mindfulness Questionnaire Below is a collection of statements about your everyday experience. Using the scale below, please indicate how frequently or infrequently you have had each experience in the last month. Please answer according to what really reflects your experience rather than what you think your experience should be. 1. I’m good at finding the words to describe my feelings 2. I can easily put my beliefs, opinions, and expectations into words 3. I watch my feelings without getting carried away by them 4. I tell myself that I shouldn’t be feeling the way I’m feeling 5. It’s hard for me to find the words to describe what I’m thinking 6. I pay attention to physical experiences, such as the wind in my hair or sun on my face 7. I make judgments about whether my thoughts are good or bad 8. I find it difficult to stay focused on what’s happening in the present moment 9. When I have distressing thoughts or images, I don’t let myself be carried away by them 10. Generally, I pay attention to sounds, such as clocks ticking, birds chirping, or cars passing 11. When I feel something in my body, it’s hard for me to find the right words to describe it 12. It seems I am “running on automatic” without much awareness of what I’m doing 13. When I have distressing thoughts or images, I feel calm soon after 14. I tell myself I shouldn’t be thinking the way I’m thinking 15. I notice the smells and aromas of things 16. Even when I’m feeling terribly upset, I can find a way to put it into words 17. I rush through activities without being really attentive to them 18. Usually when I have distressing thoughts or images I can just notice them without reacting 19. I think some of my emotions are bad or inappropriate and I shouldn’t feel them 20. I notice visual elements in art or nature, such as colors, shapes, textures, or patterns of light and shadow 21. When I have distressing thoughts or images, I just notice them and let them go 22. I do jobs or tasks automatically without being aware of what I’m doing 23. I find myself doing things without paying attention 24. I disapprove of myself when I have illogical ideas
Abstract (if available)
Abstract
Mobile health interventions delivered through smartphones (“mHealth”) provide access to behavioral support tools with more ease and frequency than in-person interventions. Engagement, broadly defined as a participant’s use of an mHealth intervention, is a critical factor in achieving improved health outcomes. However, mHealth studies report variability in engagement, highlighting the need to study engagement and identify factors that influence engagement. The two present studies examined adherence (the extent to which a person’s behavior corresponds with prescribed recommendations) and usage (an individual’s total use of intervention tools) as indices of engagement in an mHealth mindful eating intervention targeting craving-related eating in type 2 diabetes. Study 1 tested education level as a predictor of mHealth engagement, and whether delay discounting (an index of cognitive impulsivity) explained the relationship between education and engagement. Education was negatively associated with weekly and overall adherence and impulsivity did not mediate this relationship. No direct or indirect effects were found for mHealth usage. Study 2 examined trait food craving as a predictor of engagement and tested trait mindfulness as a moderator of this relationship. Trait mindfulness interacted with trait food craving to impact overall adherence but not weekly adherence or daily usage. These studies advance the literature by exploring objective mHealth engagement data and testing predictors of engagement that may inform the development of tailored interventions to optimize engagement and improve health outcomes.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
The role of primary care in the use of specialty mental health services: an investigation of utilization patterns among African Americans and non-Hispanic Whites
PDF
Savoring the service: a mixed-methods study of Latine community health workers’ health promotion efforts
PDF
The effects of acceptance and commitment therapy-based exercises on eating behaviors in a laboratory setting
PDF
Applications of mindfulness toward promoting psychosocial health in diverse communities
PDF
Links between discrimination and sleep difficulties: romantic relationships as risk and resilience factors
PDF
Family environment as a moderator of the association between theory of mind and social functioning in people with schizophrenia
PDF
Theory of mind processing in expectant fathers: associations with prenatal oxytocin
PDF
System dynamics of in-home family eating behavior: insights from intensive longitudinal data using Ecological Momentary Assessment and wearable sensors
PDF
Reducing gang involvement through employment: a pilot intervention
PDF
Prison of Peace in a jail setting: a randomized trial of a conflict resolution program for incarcerated women
PDF
Zooming in on mindfulness and gaming: the application of a brief, standardized, zoom-based, mindfulness intervention on videogame play as compared to a treatment as usual approach
PDF
The effects of mindfulness on adolescent cigarette smoking: Measurement, mechanisms, and theory
PDF
Perspective taking behavior and social outcomes in late adolescence
PDF
Opportunity beliefs and behavioral outcomes in Latinx youth: an exploration of Ogbu’s cultural-ecological model
PDF
Effectiveness of a heart plus brain health-focused nutrition intervention among Latinas: a randomized controlled trial of Buenos Hábitos Alimenticios Para Una Buena Salud (Good Eating Habits for ...
PDF
Social norms intervention to promote help-seeking with depressed Asian and European Americans: A pilot study
PDF
The successful African American Parenting (SAAP) Program: an evaluation of a culturally sensitive parenting intervention
PDF
Sleep health and variability in youth: a real-time data capture study to examine influences on daily dietary intake patterns and longitudinal weight trajectories
PDF
An investigation of the factors associated with electronic aggression among college students
PDF
Pregnancy in the time of COVID-19: effects on perinatal mental health, birth, and infant development
Asset Metadata
Creator
Jhaveri, Kinnari
(author)
Core Title
Predictors of mHealth engagement in a mindful eating intervention for Type 2 diabetes
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Psychology
Degree Conferral Date
2022-12
Publication Date
09/08/2022
Defense Date
08/18/2022
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
Diabetes,health behavior,health equity and access,mindful eating,mobile health,OAI-PMH Harvest
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Huey, Stanley Jr. (
committee chair
), Belcher, Britni (
committee member
), Lai, Mark (
committee member
), Margolin, Gayla (
committee member
), Mason, Ashley (
committee member
)
Creator Email
kinnarii@gmail.com,kinnarij@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC111939544
Unique identifier
UC111939544
Legacy Identifier
etd-JhaveriKin-11190
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Jhaveri, Kinnari
Type
texts
Source
20220909-usctheses-batch-979
(batch),
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
health behavior
health equity and access
mindful eating
mobile health