Designing for engagement with self-monitoring: A user-centered approach with low-income, Latino adults with Type 2 Diabetes

Abstract

Background and significance

Data-driven interventions for health can help to personalize self-management of Type 2 Diabetes (T2D), but lack of sustained engagement with self-monitoring among disadvantaged populations may widen existing health disparities. Prior work developing approaches to increase motivation and engagement with self-monitoring holds promise, but little is known about applicability of these approaches to underserved populations.

Objective

To explore how low-income, Latino adults with T2D respond to different design concepts for data-driven solutions in health that require self-monitoring, and what features resonate with them the most.

Material and methods

We developed a set of mockups that incorporated different design features for promoting engagement with self-monitoring in T2D. We conducted focus groups to examine individuals’ perceptions and attitudes towards mockups. Multiple interdisciplinary researchers analyzed data using directed content analysis.

Results

We conducted 14 focus groups with 25 English- and Spanish-speaking adults with T2D. All participants reacted positively to external incentives. Social connectedness and healthcare expert feedback were also well liked because they enhanced current social practices and presented opportunities for learning. However, attitudes were more mixed towards goal setting, and very few participants responded positively to self-discovery and personalized decision support features. Instead, participants wished for personalized recommendations for meals and other health behaviors based on their personal health data.

Conclusion

This study suggests connections between individuals’ degree of internal motivation and motivation for self-monitoring in health and their attitude towards designs of self-monitoring apps. We relate our findings to the self-determination continuum in self-determination theory (SDT) and propose it as a blueprint for aligning incentives for self-monitoring to different levels of motivation.

1. Background and significance

Informatics interventions driven by patient-generated health data (PGHD), collected by patients outside clinical settings, offer powerful opportunities to personalize self-management to each individual’s unique physiology and lifestyle [1–3]. Type 2 Diabetes (T2D) is an important use case because self-management of T2D is critical [6,7] but challenging in the setting of vast individual differences in blood glucose (BG) regulation [8]. Early evidence suggests that data-driven self-management interventions may improve clinical markers, including hemoglobin A1c [9–12]. In addition, a disproportionate number of the 422 million individuals living with T2D worldwide are from racial/ethnic minority groups and low income households [4,5]. The ubiquity of smartphones even among disadvantaged populations creates possibilities for data-driven self-management of T2D that may improve health equity [3,13–15].

Self-monitoring is necessary to generate PGHD of adequate quality and quantity to tailor self-management interventions. However, sustained engagement with self-monitoring is challenging to achieve, and abandonment of self-monitoring applications (apps) is high [3,16]. This problem is especially pronounced among underserved populations, who are less likely to engage in self-monitoring [17,18]. A “data divide” is emerging between those who can take advantage of interventions that use PGHD, and those who cannot [19,20], creating the potential for intervention-generated inequalities [21]. Additionally, lack of engagement may undermine the impact of data-driven interventions on important clinical outcomes [22]. If sustained engagement can be achieved, the potential for data-driven interventions to positively impact both health behaviors and clinical outcomes will be increased.

The informatics and human-computer interaction communities have recognized the need to improve sustained engagement and have accelerated efforts to motivate users to engage with these interventions. Thus far, attempts have largely consisted of incentives and gamification added post hoc to existing interventions, and have not substantially impacted sustained engagement [23,24]. Other research has shown that a variety of theory-driven and creative features hold promise in promoting motivation, including external incentives [25–27], health expert feedback [28,29], social connectedness [30–33], goal setting [34–36], self discovery [11,37,38], and decision support [39,40]. However, which, if any, of these features would motivate individuals from underserved populations to self-monitor is not well understood [41,42].

To fill this gap, we investigated the design space for data-driven interventions that would motivate adults with T2D from a primarily Latino and low-income urban community to engage in collection of their personal health data (i.e., PGHD). The primary objective for this research was to determine how individuals from disadvantaged communities respond to different design approaches intended to engage them in self-monitoring, and which approaches resonate with this population the most. We expected that participants’ lived experiences with T2D and technologies for self-management impact their responses to different design features, and thus are important to understand. Therefore, we also examined these experiences and included them into our analysis.

2. Materials and methods

2.1. Study design and participants

We conducted qualitative, in-person, English and Spanish focus groups to examine individuals’ perceptions and attitudes towards a set of mockups that incorporated different design features for promoting engagement. Participants were a subset of adults with T2D who previously participated in the Washington Heights/Inwood Informatics Infrastructure for Comparative Effectiveness Research (WICER) study and consented to be contacted for future research. WICER participants were recruited from the predominantly low-income, Latino neighborhood of Washington Heights/Inwood in New York City [43]. Inclusion criteria were: (1) self-reported diagnosis of type 2 diabetes mellitus, (2) age from 18 to 65 years, and (3) proficiency in either English or Spanish.

2.2. Designing app mockups

To elicit participants’ responses to different design features, we created a set of design mockups that represent different approaches to promoting engagement. Using the literature on design and engagement as a basis [11,25–40,44,45], we brainstormed and workshopped potential features to promote continuous engagement in self-monitoring until we arrived at mockups of six distinct designs. These included: providing external incentives, social connectedness, healthcare expert feedback, goal-setting, self-discovery, and personalized decision support in the form of post-meal BG forecasts (Table 1). We included features that could feasibly be implemented in an app given the state of technology and the healthcare system. For example, we chose registered dietitians (RDs) as the experts to provide feedback because they are the clinicians that patients most commonly consult about diet and lifestyle management in clinical practice. Additionally, we included designs created in our previous work on glucose forecasting described elsewhere [46,47]. All six mockups included the same features for data collection: tracking meals, and recording BG levels before and after these meals. We chose to limit diet tracking to its simplest form. In the absence of adequate solutions for passive diet tracking, pictures with text-based descriptions of meals are sufficient for post-processing using either automated image processing methods [48,49] or crowdsourcing mechanisms [50,51] and place the lowest burden on the users. Moreover, active BG tracking (i.e., periodically testing BG and entering levels into an app) is still necessary given that passive tracking (i.e., continuous glucose monitoring) is typically unavailable to low-income patients with T2D due to reimbursement policies [52].

Table 1.

Six mockups representing types of design features for promoting engagement.

Design Feature	Example	Description
External incentives		Users accrue rewards, either money (25 cents) or points (10 points), for each meal that they log. They receive a congratulatory message for each logged meal, as well. Users can view the total amount that they have accrued.
Social connectedness		Users share their logged meals and can view other user’s logged meals on a feed. They can also view details about the meal, “like” it, and leave comments. Each user has a profile with a small biography, profile picture, posts from others, and list of meals they have “liked.”
Healthcare expert feedback		Users estimate the macronutrient content of the meal in grams. A registered dietitian reviews the meal and provides their own macronutrient estimates of the meal. The feedback is shown as bar graphs comparing the two estimates and providing a score of the user’s accuracy assessing their own meal.
Goal setting		Users select one or more dietary goals, such as to include lean protein in a meal. They then self-assess whether each logged meal has met their goals. The progress towards each goal (for example, two of eight meals met this goal) is displayed cumulatively in a bar graph.
Self-discovery		The app visualizes users past trends in histograms and uses brushing[53] to show connections between properties of meals (e.g. amount of carbohydrate) and post-meal BG levels. For example, a user may want to view meals with 30–45 grams of carbohydrates. The app then highlights the area of a histogram of their average post-meal BG levels for meals that fall into the selected bin (e.g., meals with 30–45 grams of carbohydrates).
Personalized decision support		Users receive a computationally generated prediction of their post-meal BG forecast range for each logged meal. Users are required to enter pre-meal and post-meal BG for several meals to provide adequate data for this computation. Based on the forecast, users can change or keep their planned meal.

2.3. Data collection

We obtained approval from Columbia University Medical Center (CUMC) Institutional Review Board. The WICER study coordinator recruited participants by phone. Participants provided informed consent before each session and received $20 and a light meal for their time. We retrieved demographic information from the WICER database. English- and Spanish-speaking moderators conducted focus groups following a language-specific semi-structured focus group guide that facilitated exploration of the six mockups. During the session, participants viewed printed-paper mockups that illustrated interactivity of each of mockup (4–8 separate screens) and discussed their impressions and suggestions for improvement. We conducted focus groups until reaching data saturation. All focus groups were audio recorded, translated into English if needed, and transcribed verbatim using professional transcription service. A Spanish-speaking research coordinator checked Spanish focus group transcripts for accuracy.

2.4. Data analysis and ensuring rigor

We analyzed transcripts in NVivo 11 software using directed content analysis, a qualitative method that uses a pre-determined framework to guide analysis [54,55]. Following this approach, we created a preliminary codebook based on the six mockups. Three coders (MRT, EH, ML) began with a two-hour session to collaboratively code transcripts using the codebook, and then completed all remaining analyses independently. All authors met weekly to compare emerging results and resolve coding discrepancies through discussion. Analysis continued until consensus by all authors regarding the final set of categories was reached.

3. Results

3.1. Description of focus groups and the sample

A total of 25 adults with T2D participated in 14 focus groups (9 English, 5 Spanish) with a mean duration of 90 min. Their demographic characteristics (Table 2) are reflective of the larger WICER cohort from which participants in this study were recruited [43].

Table 2.

Demographic Characteristics (n = 25).

Age	Mean: 51.2 (±8.70); Range: 20–61
Gender	Female: 22 (88%); Male: 3 (12%)
Race	Hispanic/Latino or Black/African American: 23 (92%); White: 2 (8%)
Highest Education Level	High school or less: 23 (92%); College: 2 (8%)
Birth country	Dominican Republic or Costa Rica: 22 (88%); United States: 3 (12%)

3.2. Participant backgrounds: experiences and attitudes towards self-monitoring

Most study participants had previous experience with self-monitoring diet and BG levels in T2D, which they regarded as the two critical components of diabetes self-management. However, participants reported many barriers to systematic diet monitoring. These included lifestyle factors (busy schedules, financial constraints) and the need to identify and capture nutrition in complex meals. As a result, they felt that routine self-monitoring of diet was overwhelming: “Figuring out the carbs… Somebody told me how to do with the 15 –and it was like, no way. Don’t even tell me, because it’s so complicated. I’m not going to do it.” –FG1 (English). Instead, the participants followed a simpler visual system that relied on broad principles, such as using the MyPlate visual depiction of proportions of different food groups on a plate (www.myplate.gov).

Similarly, most participants expressed reluctance to self-monitor BG regularly, citing both the high cost of test strips and the pain associated with pricking their fingers. However, they also acknowledged that gaining, or regaining, control over their BG most often comes with increased self-monitoring. Even so, most perceived self-monitoring of BG as a problem-solving activity for a pre-specified period of time, either at a clinicians’ request or in response to perceived symptoms of high or low blood sugar. One participant stated, “The only way that I will do it, if my sugar has been up and down…just for me to have a better sense of how can I manage” -FG7 (English).

Most study participants’ previous experience tracking personal health data was in the form of paper diaries, which they found to be tedious and inconvenient: “It gets annoying if you’re out and you forget your book and you don’t have it.” –FG1 (English). Therefore, all participants reported being amenable to using technology for self-monitoring because they felt apps are constantly accessible for recording or retrieving information and could enable more precise capture of nutrition in meals. At the same time, the participants had many practical concerns, mostly importantly the associated financial burden using the app, data plan, and BG testing strips required for more frequent BG monitoring. They expressed willingness to use an app if costs would be defrayed. Moreover, Spanish-speaking participants requested that apps would be available in Spanish and technical assistance would be provided: “If you explain to me, I’ll do it.” –FG2 (Spanish).

Importantly, most envisioned using self-monitoring apps to answer specific questions or focus on problem areas, rather than for regular self-monitoring: “Maybe I will select the meals that are more heavy…only the ones that I have to.” –FG7 (English).

3.3. Attitudes towards features for engagement

3.3.1. External incentives

All participants responded positively to the concept of receiving external incentives for using an app. They viewed money as a strong motivator that would translate into increased tracking despite the small amount of money the app offered (25 cents): “It’ll help with transportation. Getting a metro card…25 cents adds up.” – FG1 (English). Points were viewed as especially motivating if connected to a tangible benefit, such as discounts or coupons for groceries. While most participants were satisfied receiving incentives for simply logging a meal, one participant envisioned points connecting directly to healthy behaviors: “I imagine that this food didn’t receive a lot of points, because it’s not healthy. It has fat.” -FG3 (Spanish).

3.3.2. Social connectedness

Nearly all participants reacted positively to the concept of social connectedness with peers, and it was the favorite design for some. Participants reported liking this design because it enhanced their current social practices, such as trading recipes, providing emotional support, and sharing new information about health. They described these interactions as being frequent within their communities: “I don’t know people [in] the park sometimes and I’ll start talking to them…I try to help as many people as I can.” -FG6 (English). Participants were especially excited at the prospect of accessing culturally-tailored meals and recipes that also fit within their health-related constraints (for example, low carbohydrate diet). One participant stated: “Everyone has their own kitchen…For example Mexicans have their typical food that they eat. We Dominicans do as well.” -FG3 (Spanish). Finally, they were excited about reciving motivational and emotional support from peers, which one participant conceptualized as “teams…somebody you can just vent to.”- FG1 (English).

However, some participants were less enthusiastic about this feature, and wanted to limit the peer network so that it only included a small group of personal contacts. These participants were more guarded with their personal data, such as daily meal choices, and preferred not to share them widely. They also expressed heightened concern about privacy in web-based platforms in general: “Well, I don’t like putting all my information on chat or Facebook… I don’t know who I’m talking to. People get your information and then they might steal my [money].” –FG1 (English).

3.3.3. Healthcare expert feedback

Many participants reported liking the idea of receiving dietitian feedback because they viewed it as an opportunity to improve their knowledge, particularly their macronutrient estimation: “With the feedback that I receive I can make adjustments for the next meal.” –FG7 (English). When responding to this design in particular, they often lamented how difficult healthy eating can be in the setting of busy lifestyles, which one participant described as “so disorganized” (FG10 Spanish). They felt a dietican could guide them in overcoming these barriers, similar to a personal coach, and some envisioned ways for the dietitian to help them with specific areas in which they were struggling. Participants particularly liked being able to visualize feedback through the colored bar graphs: “This is helpful. it’s more visual and you can really identify [whereas] reading this information, you’re going to lose track and say okay whatever.” –FG7 (English). However, similar to the social connectedness design, some participants were hesitant to share their daily dietary choices, in part because they feared judgment about these choices, and preferred greater privacy.

3.3.4. Goal setting

Attitudes towards goal setting features were split within the sample. Whereas all participants were familiar with goal setting as a common strategy for self-management, only about half of them wanted to use an app to set and track goals. Participants who liked this feature felt it would help to improve aspects of self-monitoring they needed help with, such as keeping BG levels in normal range. These individuals were driven by setting concrete goals and visualizing their progress towards them. Towards this end, they envisioned ways to customize visualizations to learn more about meals they enjoyed: “If she likes that meal she would like to see the [macronutrient] summary for that one meal.” - FG10 (Spanish). Conversely, half of participants preferred to follow the goals they had already set for themselves and mentally track progress towards achieving these goals. In some cases, participants reported they had internalized relevant knowledge, making an app less necessary to aid in working towards a goal: “I know it by memory…I know how to include fruits daily–the banana, watermelon, and apple–all that good stuff.” –FG6 (English).

3.3.5. Self-discovery

Participants were markedly less enthusiastic about features for self-discovery, primarily for two reasons. First, these features, which employed histograms displaying trends over time, caused confusion for most participants and often required several minutes of explanation before attitudes towards the feature could be explored. Second, once the general concept was understood, participants struggled to interpret the data and determine how it could be useful to their efforts. They reported lacking skills and knowledge to identify an appropriate next step, and overwhelmingly preferred that a healthcare provider interpret the data for them: “I may be wrong and I would like to have somebody that has knowledge on that field to be able to correct me.” –FG7 (English). Nonetheless, a small minority of participants embraced this design because they felt that viewing relationships and trends had the potential to help them troubleshoot problematic meals: “All that you know is that it [BG] is high… so I would be interested in what carbs are going to do to my blood sugar.” –FG6 (English).

3.3.6. Personalized decision support

Similar to the self-discovery feature, the personalized decision support mockup was met with low enthusiasm from most participants. This mockup also seemed to cause confusion and required several minutes of clarification, particularly around the idea that it was showing a prediction rather than a prior trend. Once clarified, participants again reiterated the need for more guidance about next steps they should take after seeing their forecasted BG. In fact, several participants asked specifically for personalized recommendations: “Based on my weight and my height, I would ask for a weekly meal [plan] …with the right amount of carbs, protein, fat, fiber, and calories.” –FG1 (English). A small minority of participants also embraced this feature, however, and envisioned ways to experiment with their meal choices using it: “Can this provide—okay, I’m grabbing for something sweet, and that’s what I’m going to be eating. So can you put like whether to eat those two cookies versus eating just one because sometimes it’ll change the decision.” –FG7 (English).

4. Discussion

In this study of underserved adults with T2D, we found that participants reported similar experiences and attitudes towards self-monitoring and the use of self-monitoring technologies. Nearly all participants perceived benefits in self-monitoring. Moreover, nearly all saw self-monitoring apps as preferable to the more traditional paper diary approaches, and were amenable to using apps if they aided their short-term, problem-solving efforts (e.g., troubleshooting during periods of high BG levels). Most were less enthusiastic when considering self-monitoring as a long-term routine. However, participants’ attitudes towards different design features for engagement varied drastically, and no single set emerged as an obvious favorite, outside of external incentives. Many preferred the social connectedness and healthcare expert feedback features, which resonated with their sense of community and desire for interpersonal support. Some preferred the goal setting feature that reinforced their internal perceptions of self-management goals and self-regulation. A few preferred self-discovery and personalized decision support features, appealing to their interest in self-experimentation and higher self-awareness.

Nonetheless, there emerged broad trends in participants’ attitudes towards different design mockups overall. Notably, the more popular mockups generally relied on external factors for engaging in self-monitoring, such as financial incentives or feedback from peers and healthcare providers. In contrast, the less popular mockups implied a higher degree of internal motivation to engage in self-monitoring for its own sake or for its intrinsic benefits for one’s health. This finding is consistent with the analysis of different factors that motivate human behaviors in self-determination theory (SDT). SDT is a well-established theory that has been widely used in health sciences research to understand and motivate various health-related behaviors since it was first introduced in the 1980s [56–58]. SDT posits that motivation to perform a behavior exists on a continuum from external motivation (driven by external rewards and incentives) to internal motivation (driven by intrinsic desires to perform the behavior) [56]. Notably, this continuum broadly aligns with the popularity of mockups seen in our study; most participants in our study did not exhibit intrinsic motivation to self-monitor in health, and, concordantly, most gravitated towards mockups with externally driven features (Fig. 1).

Fig. 1.

Spectrum of participant attitudes towards features for engagement.

Our study builds on prior work in which SDT has been used to explain motivation to engage with self-monitoring technologies [57,59]. The connection between the popularity of the design mockups observed in this study and the self-determination continuum in SDT suggests that this continuum may also be a useful blueprint for the design of self-monitoring technologies that target individuals with different levels of motivation. For example, it may be possible to tailor different interventions to different levels of motivation, using external incentives and social sharing for those who lack internal motivation for self-tracking, and focusing on features for data exploration and decision support for those who already have well-developed internal motivation. Further, SDT suggests that individuals can move along the self-determination continuum and transition toward internal motivation by increasing their autonomy (independent volition over one’s behaviors), relatedness (sense that behaviors are part of a larger social movement), and competence (skills and knowledge necessary to perform the behaviors) [56]. Therefore, it may be possible to use such techniques as Sequential Multiple Assignment Randomized Trial (SMART) design to adapt intervention design to an individual’s level of internal motivation and any changes that occur in this motivation over time.

Overall, our findings suggest the need for new approaches for supporting individuals with low motivation to engage in self-monitoring. For example, discovery with personal health data is a core component of most data-driven interventions. Yet many participants were reluctant to engage this process and instead wished for more directed guidance for action, such as personalized meal recommendations. Building on work from outside of the healthcare space [60,61], recommender systems that use personal health data to guide health behavior hold promise, but further research is required to examine applicability of such approaches to health-related behaviors. Moreover, participants’ conceptualization of self-monitoring as a short-term problem-solving activity and general reluctance to engage in sustained self-monitoring presents constraints for data-driven interventions targeting this population. Specifically, there is a need for computational data analysis methods capable of performing well with sparse irregular data. Recent research has begun to address such challenges by infusing human expert knowledge into computational data analysis [46,62]. However, more research examinining data-driven interventions in the context of potential data sparsity is needed.

This study had several limitations. First, by intentionally recruiting from WICER, our sample mirrored the demographics of the larger WICER cohort (predominantly Latino, female, and middle-aged), limiting the generalizability of our findings. In addition, participants viewed paper mockups rather than electronic versions. This may have resulted in more negative attitudes towards more complex designs (e.g., histogram of trends over time), which may be better understood through interaction. On a related note, it is possible that dissatisfaction with some mockups was due to limitations of the actual designs, rather than underlying features they were meant to illustrate. For example, other ways to communicate trends in self-monitoring data, for example as natural language sentences, may be more appealing for individuals with low numeracy or graph literacy. These approaches, however, require additional research and exploration. Moreover, the healthcare provider arm only included Registered Dietitians and limited feedback to nutritional composition of meals; other healthcare providers, such as physicians or Nurse Practitioners, and other types of feedback could have led to different attitudes. Finally, this qualitative study did not examine extensive psychosocial and clinical characteristics of the participants, which could help to stratify their attitudes towards different mockups.

5. Conclusion

Underserved adults with T2D included in this study expressed clear preferences for how they engage with self-monitoring as well as informatics interventions that rely on personal health data. Behavior change theories that explain human motivation such as self-determination theory may be useful in guiding design based on motivation. Future work exploring this possibility holds promise to improve engagement among underserved individuals, towards the goal of equity in the adoption and use of data-driven informatics interventions for health.

Summary points.

What was already known on the topic

• Sustained engagement with self-monitoring interventions is frequently low, especially among underserved individuals.

• Various features have been developed to attempt to promote engagement in the general population, though not specifically for underserved individuals.

What this study added to our knowledge

• Even when coming from similar cultural and socioeconomic contexts, underserved individuals may prefer very different design features for engagement.

• Underserved individuals may engage with their personal health data in unexpected ways, and therefore may require features for engagement with self-monitoring that are tailored accordingly.

• Behavior changes theories, including self-determination theory, may be useful in interpreting preferences and aligning features for engagement.