Developing and testing an integrated patient mHealth and provider dashboard application system for type 2 diabetes management among Medicaid-enrolled pregnant individuals based on a user-centered approach: Mixed-methods study

The Double Diamond Framework

Our study activities followed the double diamond framework, which is a user-centered approach widely used in information technology development. Iterative development cycles are utilized to produce user-specified needs and preferences. Figure 1 shows the roadmap of our design framework, starting with a diagnosis phase and followed by four sequential double diamond phases: discover, define, develop, and deliver [38]. Our study utilizes the double diamond process to: a) define problems faced by Medicaid-enrolled individuals and their providers for effective diabetes management during pregnancy, b) develop prototypes of our mHealth and dashboard applications, and c) elicit reactions from users with regard to usability and acceptance.

• Download figure
• Open in new tab

Figure 1.

ACHIEVE pilot study activities through the lens of a double diamond framework and participatory mindset

The shape of the double diamond figure signifies the process of collecting divergent insights during discovery and subsequently translating those insights into concepts by convergence and definition [39]. This conceptual framework has been successfully used for complex health care problems [40,41]. For example, Asbjørnsen and colleagues designed an eHealth intervention that aimed to support behavior change for long-term weight loss maintenance in the general population [39]. We complemented our double diamond framework thought process with a participatory design mindset and included system requirements generated by end-users to ensure expectations for the final product were met [42].

The section that follows describes each of the double diamond phases as they pertain to our study. Our team used brain mapping exercises aided with a Miro board to outline ideas, collaborate with other team members, identify themes, and design prototypes for the study.

Diagnose

As a first step, three members of our research team (EB, AL, GO) conducted a scoping review of the literature for key problems and existing gaps related to diabetes management in pregnant individuals. Our research focused on two domains: 1) health disparities due to clinical and non-clinical factors and 2) technology use. Both areas focused on individuals who are at risk, members of minority populations, and on public health insurance. We examined various sources that included peer-reviewed articles and registered clinical trials for scientific evidence. Significant findings were reported on our study Miro board for reference and discussed as a team, which included gaining input from our clinician investigator (KV) for validation from a health care provider perspective.

Phase I

Study Sample

Define (Data Analysis)

This phase consisted of thematically organizing patient and provider interviews by means of open coding. Interviews were analyzed inductively to categorize our findings and reveal emergent themes (Fereday & Muir-Cochrane, 2006). Three team members (EB, YW, NF) reviewed at least 20% of each transcript, and two members reviewed (PS, NF) the entire transcripts independently to identified emergent themes. These two members met frequently to ensure consistency of coding and agreement about definitions of new codes as they emerged during the coding process, consistent with the grounded theory approach [43]. A third researcher team member (RS) independently reviewed the thematic codes and reconciled any differences. Information from our literature review and thematic analyses, in concert with the fundamental constructs of social cognitive theory (SCT) [44,45], was used to develop elements that formed high-level specifications for the mHealth and dashboard prototypes. Microsoft Word was used to analyze quotes and STATA 17SE was used to analyze all quantitative data for the study.

Phase II

Deliver (Usability Assessment)

In the final phase of our pilot study, we used the high-fidelity clickable prototypes of the mHealth and dashboard applications that were developed based on the feedback provided by study participants and our user experience design team. Key features present in the initial low-fidelity prototypes, although not intended to be comprehensive, represented a core set of functions that would promote behavioral modifications as guided by the principles of SCT. Following the approach undertaken by Yardley and colleagues, our research team met multiple times over the course of several weeks to plan, prioritize, and highlight aspects of the prototype features that were crucial [46].

We conducted the Phase II interviews with the corresponding prototypes tested to obtain feedback from the users who participated in our Phase I interviews. We conducted usability testing and think-aloud sessions with patients and providers. The aim of these interviews was to systematically collect feedback about the prototypes using a deductive approach based on our prior work on decision support tools [47,48].We categorized identified themes into one of three categories: (1) function, issues that are related to meaningful and intuitive navigation of an application; (2) content, problems with information provided in an application that complicate or lead to misinterpretation of information; and (3) aesthetics, concerns that impede an application from having a minimalist design that effectively communicates information.

Patient Characteristics

We recruited seven patients and seven providers in our study. All patients and providers completed Phase I, and three patients (42%) and four providers (57%) completed Phase II. Characteristics of our study samples are provided in Tables 1 and 2. As presented in Table 1, most of the patients identified as non-Hispanic Black, with a mean age of 29. Most of our patients had two previous pregnancies (gravida). Among our provider sample, we recruited four physicians and three registered nurses.

Diagnosis phase

Phase I

Patient-reported Needs and Preferences in an mHealth Application

Patient expressed a core set of expectations for an mHealth application: features that electronically captured and managed information, access to support systems, use of diabetes technology, and help through problem solving and personalized recommendations.

Electronic Capture and Management of Information

Our results indicated the emerging need to include an electronic feature to capture blood glucose levels. Existing processes called for recording blood glucose levels in paper logs and sending them to providers as emails or by fax, which was perceived as being cumbersome, as indicated by Patient 4: “something that can keep track of my, my sugar records. That would be nice, instead of having to write everything down and then send it on through email. It would be awesome just to have it on the app. That would definitely be a plus.” Patient 1 similarly noted: “Hey it’s time to take my sugar, let me go ahead and put it in my app instead of writing it on the paper. So we have kids, so the other kids get the paper or you could forget it, stuff like that, but our phones are always in our hands. So that would just make it easier, you know, than that paper. When they showed me the paper, I was like, oh God, I am not going to keep up with it, sorry, no. So that just replacing that would be helpful in general.”

In addition to capturing FSBG values, patients expressed the need to include a feature that captured details about their dietary intake and out-of-range values with means for providing open-ended responses. For example, Patient 4 mentioned: “it’d be nice to have like a schedule of like your, when you’re supposed to check your sugars and then being able to write down what your sugar was that day and then even going to the extent of what did you eat and explain why your sugar is that low or that high. I think that would be awesome.”

Support System

Patients suggested the need to include a feature that allows them to communicate with other individuals for support and to learn from their experiences. Patient 3 noted: “Maybe getting them connected with the right people on how to manage their care. Ways of controlling since they would know.” Patient 4 said: “What they eat [support group] is a good one because I get bored with the same old food so it’d be nice to know what other people are eating that’s been able to maintain their sugar. What they do, as far as exercise to where they don’t overdo it, but I know walking is a good one, but that would actually be something nice to know what other people is doing to maintain their weight and everything during, especially having diabetes during pregnancy.” They also said: “Well, it would definitely be nice to have a support system. Like something that I can go to if I need support, especially somebody that already knows what I am going through and how I am feeling. It’s nice to have somebody that can relate.”

Diabetes Technology

The availability of innovative technology for diabetes management has been rapidly increasing. Patients expressed interest in the use of such technologies, especially continuous glucose monitoring (CGM) devices. Patient 4 expressed interest in using CGM: “I do think one idea would be awesome would be having a monitor, which I know they have them out there, but for pregnant women … that is actually you’re able to check it like it’s inserted in you. I forget what it’s called, but I know they have some on that out there, but it’d be awesome if they can actually get one approved for pregnant women, to make it so much easier on them.”

Patient 1, who actively used CGM, indicated: “The scanner that I use is on my phone. I don’t have to have like an additional piece of equipment … I downloaded the app and just scan my arm from there, I don’t have to, you don’t have an extra piece of equipment, so it just makes it easier for them to get the numbers from directly from the app.” This patient also explained the motivation behind switching to CGM: “My PCP told me about it ’cause they were having so much trouble getting me the finger stick. It just hurts. Like, I won’t lie to you. I’d rather take the insulin than do the finger stick, it, it hurts that bad … even though my insurance doesn’t cover it, I still pay out of pocket for it, because it’s so much easier, I can just stick myself the one time, then I just scan my phone on my arm, way easier. As much as they want me to check my sugar, like I, I am not going to be able to stick myself. It’s something about sticking myself that hurts because you can’t see the needle and … it’s like a jack-in-the-box almost, you’re turning it, you’re turning it, you know it’s going to pop out, when is it going to pop out?”

Problem Solving and Personalized Recommendations

Pregnancy is a unique experience for every patient and requires guidance that accounts for specific challenges faced by a pregnant individual. Patients reported the need for direction on how to actively plan activities and meals that aligned with effective blood glucose management. Patient 3 indicated that “Like about what’s good diet and exercising plans and how to portion control and ’cause I know a lot of people don’t know that when they become diabetic. They don’t, they don’t know that. Because they’re used to kind of eating whatever they want.”

Patient 1 mentioned: “Maybe like suggestions for different things like they want me to eat three meals and three snacks and there is no way I can keep all that down half the time, so maybe like suggestions or like what could be a snack and things like that and meal suggestions ’cause you don’t want me eating carbs and go easy on the sugar.” They also mentioned: “They tell me what they want them to be, like they want it to be 125 after a meal and I can’t get 125 to save my life. So, I’m just thinking like how am I going to get to this level. So they have me at like 10, 10 units of insulin; I think using the app would be easier with that part ’cause if like that 10 units of insulin is not enough, it needs to increase. I think being able to keep track of that in the app would be helpful too. Like hey if this meal has this much carbs, take this much insulin, ’cause I know it’s not like a sliding scale like it was before. So now they’re like you need to take this set amount, so I think just having some more reference material for the sugars in general. Like how the insulin makes you feel, what you should do if it’s low, things like that, you know, because sometimes what, and when it’s low I don’t know what I, what I should drink, what I should eat because before if it fell low I would get some juice or something, but now I’m scared the juice is going to raise it too high, you know what I mean? So that would be kind of helpful too.”

Provider-reported Needs and Preferences in a Dashboard Application

Providers indicated a general set of expectations for a dashboard application including features such as dynamic and actionable data, a method to capture unmet social needs, and enhanced patient-provider communication.

Define

Upon review of the transcripts, our team identified problems and possibilities faced by our patient population. Three study members used sticky notes to operationalize important concepts onto the Miro board following pre-defined color codes (i.e., orange for keywords, yellow for general summaries/insights, blue for ideas for prototyping, pink for barriers, azure for manual processes, and violet for analysis/decisions) (see appendix figures S1-S4). The study members worked collaboratively to manually sort notes into clusters of similar functional areas (e.g., communication, glucose management). This was conducted for both the patient mHealth and provider dashboard applications.

Study team members further synthesized information into high-level user specification maps (Figure 2) to translate the user requirements into an actionable development plan, highlighting the essential functions and their subthemes for each application. This process also facilitated forming the structure on how the mHealth and dashboard applications would achieve information exchange.

• Download figure
• Open in new tab

Figure 2.

Initial user specifications for mHealth and dashboard applications

The mHealth application centered on the reporting and management of blood glucose values and social needs. Customized patient goals and care pathway cards (clinical and social needs-oriented activities that would help patients achieve their specific goals) were included as primary features, along with other essential elements such as a calendar, video conferencing with the provider team, and medication information. The dashboard focused on showing the trend of a patient’s glycemic control, social needs information, and other critical vitals (e.g., gestational age, other medications) that would present a 360-degree view of the patient. Providers would also use the dashboard to monitor patient activity, actively communicate with them, and tailor their care pathways based on the feedback gathered about a patient from the system.

Phase II

Develop

We continued further analysis of the interview data using swim lane diagrams. A swim lane diagram represents how different roles perform their tasks, take on their responsibilities, and interact with others along a timeline, which generates a holistic view of the system to help identify gaps and discover intervention opportunities. We used the diagram to map the current clinical workflow based on the journey of a Medicaid-enrolled patient with T2D during pregnancy (Figure 3) and developed an improved clinical workflow with the engagement of the applications (Figure 4).

• Download figure
• Open in new tab

Figure 3.

Swim lane of “as-is” clinical workflow (figured scaled to illustrate complexity of the system)

• Download figure
• Open in new tab

Figure 4.

Swim lane of “to-be” clinical workflow with ACHIEVE tools

In the “as-is” clinical workflow (Figure 3), five roles (1-Pregnant T2D Patients, 2-OB/GYN or MFM provider, 3-Diabetes Nurse Educator, 4-Physician, 5-Community Health Worker) are presented in lanes with distinguished colors, followed by the actions taken from a patient’s perinatal period through their post-partum period. The arrows and lines indicate the flow of interaction. We used color-coded notes to capture findings in the workflow, which include issues/barriers, possible features, questions, analysis/decision points, and follow-up questions. During this activity, a variety of gaps were revealed. Major ones included: 1) the patient’s adherence to their goals, care plan, and medication regimen; 2) the communication and information exchange between the patient and the provider, as well as among members of the care team; and 3) the need to address barriers to social needs occurring throughout pregnancy.

Based on the study results from Phase I and our additional analyses, our design team designed high-fidelity prototypes for the mHealth application and dashboard. With features reflecting patient and provider generated requirements with the goal of achieving relevance through a useable, useful, and desirable product [63]. Figure 5 displays the mHealth application prototype with five different pages that cover critical domains identified for the application. Figure 6 displays the provider dashboard prototype with a patient page displayed. Both prototypes were used for our usability testing.

• Download figure
• Open in new tab

Figure 5.

mHealth application prototype. The application allows a patient to view a to-do list that offers an overview of their tasks for the day; report patient reported outcomes; track their glucose levels; manage and learn about their care pathways; and track their goal achievements over time (names are not real).

• Download figure
• Open in new tab

Figure 6.

Dashboard application prototype. The dashboard presents patient profiles, scheduling information, and options to communicate with the patient or provider; track A1c, proportion of time in the target range for glucose level, their goal achievement, and social needs; and see notification messages about the patient’s health status and care pathway selections for the patient given their specific patient reported outcome responses (names are not real).

Deliver

We conducted usability assessments of the mHealth and dashboard applications with patient and provider participants, respectively. We synthesized and categorized feedback from participants and their actions into the content, function, and aesthetics domains based on the application tested (Tables 3 and 4).

Principal Findings

We identified three themes that exemplified pregnancy experience among Medicaid enrolled individuals with T2D: managing exhaustion, adherence to a new regimen, and preparing for pregnancy. Exhaustion and fatigue are among the commonly reported feelings during pregnancy that may exacerbate a chronic comorbid condition, such as diabetes [65]. During pregnancy with diabetes, specific lifestyle adjustments are recommended over a short period of time. [7] Individuals with T2D who were previously on oral pharmacotherapy or diet-based therapy now generally require insulin to achieve glucose control using strict parameters, including upwards of four daily FSBG [66,67], and patients are challenged to learn to regulate both the quantity and quality of carbohydrate consumption while simultaneously adopting new medication use and exercise behaviors. The strict medication and dietary requirements and increased home-based activities such as glucose monitoring and reporting escalate the burden of disease management contributing to pregnancy fatigue. Moreover, multiparous women taking care of older children may face additional barriers that compound this preexisting exhaustion.

Diabetes in pregnancy requires economic and social stability. T2D management in pregnancy is expensive, with >$7,000 in excess pregnancy-related costs, and individuals often report financial planning as part of diabetes management strategy [68]. Likewise, social support is equally important for good quality of life during pregnancy and requires planning in the earlier stages. The themes we identified align with previous finding. The associated disease novelty, especially among nulliparous women living with T2D, and dynamic physiological and clinical changes require preparation for upcoming lifestyle demands [69,70].

Participants expressed needs and preferences in an mHealth application that were represented by five themes: electronically captured and managed information, access to support systems, use of diabetes technology, help through problem solving, and personalized recommendations. The most desired feature in an mHealth application was the ability to electronically capture blood glucose data. Pregnant individuals with diabetes often find it cumbersome to manually record, save, and communicate their glucose values with their care team providers. An electronic platform with the ability to capture and communicate blood glucose values in a meaningful manner emerged as a desired feature by both patients and providers. As opposed to manual data entry, an mHealth application provides patients with the opportunities to experience streamlined and routinized data collection that requires less effort on their part. For example, use of a CGM device that syncs with an mHealth application with a user-friendly interface may promote long-term use of the application and device [71].

User Needs and Preferences for T2D Management during Pregnancy

The interrelated themes we identified from patients and providers revealed specific end-user specifications that aligned with the theoretical underpinnings of SCT. SCT posited that successful performance of a behavior depends on an individual’s behavioral capability as well as cognitive and environmental influences on behavior via three domains: 1) skills; 2) knowledge and beliefs; and 3) self-efficacy [44]. The themes identified from our qualitative work highlight the importance of these three domains specific to individuals with T2D during pregnancy. Individuals require educational information that clearly explains how behaviors and specific clinical and social needs pathways will help them achieve the primary outcome of glycemic control (knowledge and beliefs). Patients need skills and coaching to engage with the diabetes care team, which can be accomplished by collecting and synthesizing detailed information about the pregnant individual from the mHealth application to the provider dashboard to better communicate with the diabetes care team (skills). Team-based coaching based on CGM output, personalized information, and care pathways may help the pregnant individual learn more about T2D in pregnancy and better adhere to their T2D care regimen. Closing the loop (e.g., having a community health worker document the individual’s successful securing of healthy food through a community food pantry) could ensure that an individual’s new skills yield meaningful outcomes and enhance their confidence (self-efficacy).

Design Requirements and Refinements Needed to Meet End-User Preferences and Needs

Our research team recognized the importance of leveraging technology that is highly relevant to our study population and their providers. We utilized applications that did not require high costs to maintain in the long term and were evidenced-based. In our exploration of the evidence base for mHealth and provider dashboard applications tailored to similar populations, we quickly recognized the challenges faced in prior implementations of each application [21]. Specifically, these previous approaches to improving care for Medicaid-enrolled patients with diabetes have been siloed and fragmented [11]. We therefore identified an intervention that would be systems-oriented and addressed the pitfalls associated with the use of each application by itself. We focused more on the gestalt effect of these applications and their ability to collectively transform health care by providing a comprehensive, 360-degree view of the patient using information that was dynamic and closed the loop between the provider and the patient (e.g., the provider will be aware that a patient completed an assigned care pathway).

Our analysis of the “as-is” and the “to-be” workflows identified major gaps: 1) patient adherence to their goals, care plan, and medication regiment; 2) the communication and information exchange between the patient and the provider, as well as among members of the care team; and 3) the need to address barriers to social needs occurring throughout the pregnancy. These were critical to an ideal workflow. Our applications considered not just the clinical and psycho-social requirements for the patient, but also these realities in our clinical setting. We pursued a deeper understanding of the requirements of applications by testing them with patients and providers, and although initial impressions were highly positive (e.g., TLX scores for both applications were in the 20th percentile of national TLX scores across all devices), we obtained notable feedback that will be used to improve the content, function, and aesthetics of our applications before a full deployment. Our favorable scores on the TLX surveys for both our prototypes should also be highlighted because participants did not receive any pre-training to use the applications. This suggests the intuitiveness of our prototypes, especially the interfaces. The implications for such design performance may include reductions in cost and effort associated with training and greater engagement from individuals with lower technology literacy.

Implications and Recommendations for Future Design and Practice

A key consideration when it comes to the use of digital health tools is the reversal of newly established behaviors. [72,73] Whether our approach – which is supported by prior studies, our own mixed-methods findings, and theory – results in long-term behavioral modifications is not demonstrable in this paper and requires further inquiry. A critical component of such an inquiry would involve gauging the cost-benefit ratio that individuals mentally construct in determining their desire to balance the use of our applications to manage blood glucose levels during their pregnancy. Relatedly, the timing of our intervention is another area of consideration, especially given our focus on pregnancy. Extending the use of our application before pregnancy (in preparation for pregnancy) and postpartum use (especially during the interconception period) could have implications for the successful management of blood glucose beyond the perinatal period.

Strengths and Limitations

An important limitation to our study is that it is not tied to the evaluation of health outcomes. It reflects realistic perspectives of patients and providers whose behaviors and interactions could have an impact on health outcomes. Our study is also focused on a specific population during pregnancy and does not capture their views before or after pregnancy. Our study involves perspectives of patients and providers who are highly committed and engaged in the use of our applications, which may influence study findings [74]. Nevertheless, our team was able to gather both positive and negative impressions on the final prototypes. More work is required to obtain generalizable perspectives beyond our sample. Our approach to developing applications using a double diamond framework with a participatory design mindset is generalizable and is already utilized in multiple domains beyond health care including government, finance, insurance, and other general consumer markets.