A within-subjects, pre-post design was used to pilot MAP in 22 adults with T2D who were portal naïve. We developed the intervention protocol after seeking feedback from stakeholders on barriers and facilitators to patient portal use and logistics to consider optimizing the implementation of MAP in CHCs [25].

Participants were recruited from 2 CHCs which have been previously described [25]. The clinics are located in Connecticut which is a small and densely populated state with prominent health disparities. In Connecticut, Black residents are nearly 4 times more likely than white residents to have a diabetes-related lower extremity amputation, and among Latino individuals, the rate is nearly 3 times higher than for non-Latino White individuals [26]. Inclusion criteria for this study were as follows: (1) established patient at 1 of the 2 CHCs, (2) age 21-65 years, (3) diagnosed with T2D >6 months, (4) most recent HbA_1c measure >7.5%, (5) no use of the patient portal in the past year, (6) no intention of moving or changing clinic within 6 months, and (7) self-reported ability to read in English or Spanish. Exclusion criteria included cognitive impairment (≥3 incorrect answers on the Six Item Screener) [27] or current gestational diabetes.

Participants were recruited from select primary care provider panels at each clinic. A designated clinic staff member reviewed the weekly schedule for potentially eligible recruits and introduced the study to recruits in order to determine preliminary interest. If interested, a trained research assistant explained the study and determined eligibility with a screening questionnaire. If eligible, an appointment was scheduled in person or via Zoom (Zoom Video Communications) for informed consent, baseline data collection, and study enrollment. Written informed consent and data collection were completed in the language preference of the participant (English or Spanish).

Upon completion of baseline data collection, participants were scheduled for the first intervention session. All participants received standard T2D care at the CHC, which followed the guidelines for T2D management as recommended by the American Diabetes Association [28] (eg, quarterly appointments with primary care providers, medical management, and referrals to specialists as indicated). At the study clinics, trained nurses provided diabetes education individually during clinic appointments, as needed. All participants received the MAP intervention from community health workers (CHWs) and nurses employed at the clinics. Clinics were compensated for the salary of the nurses and CHWs for training, delivering the intervention, and completing study tasks.

Before delivering MAP to study participants, CHWs and nurses were provided a 1-day, in-person, interactive training on the study protocol. They were also provided supportive supervision throughout the study (weekly to biweekly). Training covered orientation to research and goals of the study; human subjects protection; protocol and documentation; and team roles, responsibilities, and supervision. Training also covered details of the intervention content through a study manual and specific strategies for working with low-literacy or low-numeracy individuals with diabetes [29]. Such strategies include using the teach-back method; asking open-ended questions; avoiding unclear statements (“your test was positive”); keeping sessions brief; presenting small chunks of information; encouraging practice between sessions; using nonnumerical measures (eg, 1 serving butter=size of the tip of the thumb); using pictures when possible; using plain language (no jargon or acronyms); reducing the reading level of written materials; and, using a friendly tone. In addition, they were trained in principles of autonomy support. Autonomy support refers to a patient’s perception that their health care provider recognizes the person’s personal agency, encourages self-efficacy, and supports their self-care choices [30].

The MAP intervention intentionally and directly intervened on the 4 intermediary determinants of disparities as outlined by the WHO Health Equity framework (material circumstances, psychosocial factors, behavioral and biological factors, and the health care system (Multimedia Appendix 1). All study participants received a tablet (which they were allowed to keep at the end of the study) along with internet access for the 6 months of the study.

The intervention was sequenced to first have CHWs assess SDoH needs using a questionnaire specific to each clinic (each clinic had a slightly different form already in use) and connect the participant to relevant community resources (eg, SNAP benefits). CHWs then provided training on how to gain access to the patient portal and on to mastery of the tablet and portal functionality. In addition to lack of training and lack of encouragement to use the portal, other barriers to patient use of portals addressed by CHWs included doubt about portal usefulness, lost passwords, anxiety about viewing medical information, and privacy concerns [31]. Once the technology training and social determinant referrals were completed, participants were referred to the clinic nurse (Multimedia Appendix 2).

Next, the clinic nurse contacted the participant via the portal to provide DSMS. According to the American Diabetes Association, DSMS is “the support that is required for implementing and sustaining coping skills and behaviors needed to self-manage on an ongoing basis” [32,33]. The nurse initially met individually with study participants to establish rapport, assess diabetes self-management (DSM) behaviors, and develop a DSMS plan collaboratively with the participant. Nurses were asked to communicate with patients via the portal at least twice weekly during the first 3 months and weekly for the latter 3 months, individualizing interactions based on participant needs. In these interactions, nurses assessed challenges and successes with the DSM plan of care and provided education, support, and encouragement. Nurses were also provided with a variety of electronic health education resources that could be sent to participants via the portal. Resources were assembled by the researchers from a thorough review of written and video materials available in English and Spanish for adults with low health literacy.

Characteristics of each CHC required that MAP be integrated into clinic operations in a tailored fashion. First, one CHC did not employ CHWs and their care coordinators (who usually addressed patient social needs) were not available to deliver the MAP intervention. Therefore, at that clinic, the study nurse was trained to complete both CHW and nursing components of MAP. Second, the 2 clinics used different portal platforms. Therefore, training procedures and written materials were designed to be equivalent between, but tailored to, each clinic and its platform.

Data were collected and managed using REDCap (Research Electronic Data Capture; Vanderbilt University) an National Institutes of Health supported, Food and Drug Administration compliant electronic data capture application for data collection and storage. Data were collected from participants at baseline, 3 months, and 6 months including point-of-care HbA_1c values. Research assistants entered data at the time of data collection into the database via tablets or a computer. All participants received a gift card after each wave of data collection—US $40 at baseline, US $40 at 3 months, and US $60 at 6 months. Nurses at the clinics extracted clinical data and information technology specialists at the clinics extracted patient portal data from the electronic health record.

This study (NCT05180721) was approved by the institutional review board at Yale University (IRB# 2000031753; approved on December 21, 2022). All patients provided informed consent before participating in this study.

All data were downloaded from REDCap onto a secure server. Descriptive analyses were performed to assess demographic and clinical characteristics of the sample. Distributions of outcome variables were examined for central tendency and dispersion. We estimated the Cohen d effect size of MAP on A1C and tested the statistical significance of the change from baseline using longitudinal models, including generalized linear mixed model (GLMM), a logistic model with random intercept (which incorporates the correlation within repeated measures), and a negative binomial model with random intercept. The coefficients of the categorial time variable (ie, baseline, 3 months, and 6 months) represent the average change of HbA_1c at 3 and 6 months from baseline. The GLMM included all participants with data at baseline and at least one postintervention value and missing data was handled using the maximum likelihood approach. The repeatedly measured secondary outcomes were analyzed with the same approach using GLMM, with each outcome analyzed individually. We used the GLMM identify link function for continuous variables, the logit link function for binary variables, and the log link function for count variables. Residuals were assessed for normality assumption and variables were transformed with an appropriate form when the normality assumption did not hold. Log-in data from one participant was excluded due to errors encountered in downloading it from the portal that our IT expert could not resolve.

Recruitment and data collection took place from May 2023 to July 2024. A total of 47 participants were approached about the study, 26 were eligible and provided informed consent, 23 received the intervention and one was lost to follow up (refer to Figure 1 for CONSORT [Consolidated Standards of Reporting Trials] diagram). The sample of completers (n=22) was recruited from 2 clinics, 68% (n=15) from 1 clinic, 32% (n=7) from the other clinic. Refer to Table 1 for demographic characteristics. The sample was predominately Latino or Hispanic (7/22, 77%) had a mean age of 56.32 (SD 10.93) years, 73% (16/22) were female, and 55% (12/22) were married or partnered. The majority reported low-income (86% <US $40,000/year, 19/22), low educational attainment (59%, 13/22 less than high school graduates), and no health insurance (55%, 12/22). All participants had access to a smartphone, but 91% (20/22) had never accessed a health app. Yet, they reported some confidence in using apps, setting up video chats, solving basic technical issues, and using a tablet (mean score 3.1, SD 1.0 with a scale ranging from 1 to 5). They also reported positive perceptions of portal ease of use, usefulness, and confidence in using the portal. The mean duration of diabetes was 11.75 (SD 9.09) years, 36% (8/22) were on insulin at baseline, 14% (3/22) were current smokers, 32% (7/22) reported a history of depression, 59% (13/22) reported severe diabetes distress, 68% (15/22) reported a history of hypertension, and baseline HbA_1c was 8.31% (1.05%; measured using the A1CNow system). In total, 73% (16/22) of participants had a baseline HbA_1c level greater than the recommended 7.0% for adults with T2D and 73% (16/22) rated their health as fair or poor while 27% (6/22) rated it as good or very good. Mean BMI was 32.21 (SD 5.98), with 23% (5/22) of the sample overweight and 68% (15/22) were categorized as obese. Other baseline clinical data are reported in Table 2 (Data extracted from medical records except for HbA_1c, which was collected by researchers for study assessment).

SDOH needs were assessed before beginning the technology training with 36% (8/22) of participants requiring a referral. Referrals needed were for assistance with utilities (n=4), transportation (n=3), food (n=3), and insurance (n=2). Training on the use of the portal averaged 84.55 (SD 49.32) minutes, with a range between 30-180 minutes. The portal training protocol implementation was high at 85% (524/616 tasks) fidelity across all protocol items and participants. Participants reported moderate confidence in using the portal after the training session with a mean confidence of 2.3 (SD 0.93) on a 3 point-scale (low, moderate, and high confidence). The majority of participants completed the portal training in one session; however, 4 participants (4/22, 18%) required technology support or additional training due to technical challenges using the portal. Nurse protocol implementation was high at 84% (222/264; 12 weeks × 22 participants) at 3 months; but decreased to 52% (137/264) at 6 months with fewer messages sent by nurses over time.

Barriers to integrating MAP into routine clinical care that was noted during regular supervision were varied and included the following: (1) turnover of CHWs, (2) MAP nurses not embedded in the participant’s clinical team, (3) portal or clinic constraints on messaging from patient to PCP, (4) difficulty tracking those portal messages from nurses to patients that were not opened and acknowledged, (5) cumbersome and error-prone portal features for transmitting glucose data from patient to clinician, and (6) few high-quality diabetes education videos suitable for low-literacy, low-numeracy, and Spanish-speaking individuals that could be delivered via the portal.

Program satisfaction was high at 3 and 6 months (mean 4.03, SD 0.81 and mean 4.14, SD 0.49 on a 5-point scale respectively). All participants indicated that they would recommend the MAP intervention to a friend. Participants also reported high autonomy support from nurses at 3 months which increased at 6 months. At 6 months, 100% (22/22) of participants reported that they would continue to use the portal for diabetes care and 100% (22/22) felt like the portal was helpful for their diabetes care.

Refer to Table 3 for outcomes over time. Portal activation was high with 100% (22/22) of participants creating a portal account and logging into the portal in the first month (n=21). Mean logins per week was 3.16 (SD 1.55) over the first 3 months and 1.45 (SD 0.93) between 3 and 6 months (Figure 2). Mean portal logins per month were 12.65 (SD 6.21) for the first 3 months and 5.79 (SD 3.74) for the last 3 months. Participant engagement (as defined by the literature) was high, with 96% (20/21) of participants logging into the portal at least twice per month in the first 3 months and 76% (16/21) of participants meeting this benchmark between 3 and 6 months. At baseline, participants perceived that the portal would be easy to use (mean score 4.0, SD 0.55 on a 5-point scale) and useful (mean score 4.25, SD 0.44) with no significant change over time. There was a significant increase in technology confidence over 6 months (P<.05), with a trend for increased digital health literacy at 6 months (P=.08).

In the mixed effect model, diabetes distress (PAID) significantly decreased from baseline to 3 months; it continued to decrease significantly from 3 months to 6 months (P<.01; Table 3). The proportion of clinically elevated diabetes distress (ie, PAID>40) was 50% (11/22) at baseline, and it decreased to 36.4% (8/22) at 3 months (P=.20) and 22.7% (5/22) at 6 months (P=.03). Diabetes self-efficacy increased from baseline to 3 months and continued to increase until 6 months when it became statistically significant. A similar pattern of significant change from baseline to 6 months was found for confidence in using technology and digital health literacy. HbA_1c was slightly decreased at 3 months with a small effect size (Cohen d=–0.17) but returned to baseline level at 6 months (Cohen d=–0.07). There was no significant change in perceived autonomy support or diabetes self-care behaviors.

This pilot study evaluated an intervention to increase patient portal use in adults with T2D who access health care at CHCs. We reached a sample of adults of diverse races or ethnicities, low income, and low educational attainment with low use of the patient portal. The MAP intervention was specifically designed to address multiple levels of challenges identified by CHC patients and providers to patient portal usage. Thus, to enhance successful implementation, participants received a tablet and data plan, SDOH screening, training in patient portal navigation with ongoing support, and diabetes education and support by nurses. Overall, feasibility, acceptability, and improvement in portal and health outcomes of clinical significance were demonstrated yet integration into clinical care was challenging.

First, we demonstrated high feasibility by tailoring implementation strategies to meet the unique needs of each clinic. We also demonstrated low attrition, high patient portal activation (22/22, 100%), and high patient portal engagement over 6 months. High participant activation is not surprising as we had dedicated staff available to train participants in portal use and assist with any challenges. Although the measurement of patient portal engagement is not standardized in the literature, a common metric is the frequency of logins per month. Consistent portal use has been defined as two or more patient logins per month according to a recent systematic review of the measurement of patient portal use [36]. In another study, super-users were defined as those logging in twice or more per month [53]. In our study, consistent portal use was demonstrated in 96% of participants over 3 months, 76% of participants between 3 and 6 months, and 87% over a 6-month period among 21 patients. Thus, the results of our portal activation and engagement are encouraging. Factors that likely contributed to these positive outcomes were technology training or support for the duration of the study, participants’ development of a relationship with a clinic nurse, and consistent outreach by the nurse over 6 months via the portal to provide DSMS. While the provision of a tablet and a data plan may have also contributed to portal engagement over time in this study, we judge that increasing access to smartphones combined with free internet sites in community settings (eg, libraries and coffee shops) may be sufficient for future implementation efforts. Decline over time in consistent use may be partially explained by less consistent protocol implementation by nurses from 3 to 6 months and lack of full integration of MAP into clinic workflow.

Engagement over time was supported by improvements in other portal-related outcomes. Whereas there was no change in how participants viewed the portal per se (perceived usefulness and perceived ease of use), participant perceptions of their own interactions with the technology did improve. Increases were observed in both technology confidence (“I am confident in my ability to use MyChart”) and digital health literacy (“I can solve, or figure out how to solve, basic technical issues on my cell phone, computer, or other device”). Those improvements reached statistical significance at 6 months suggesting that individuals may need more than 3 months of supported use of the tablet and the portal to create change in confidence and digital health literacy. Also encouraging was the high level of acceptability including participant satisfaction with the MAP intervention and high intention to continue use of the portal after the MAP study ended.

Second, some clinically important health-related outcomes significantly improved in our small sample. The decrease in diabetes distress at 3 months, with a further decrease at 6 months, is important, given the body of research showing that diabetes distress is robustly associated with lower self-management and may be associated with higher HbA_1c [54,55]. The MAP intervention’s reduction in diabetes distress is consistent with findings from other interventions. A meta-analysis of 41 studies testing eHealth interventions for diabetes self-management found that such interventions are effective in reducing diabetes distress [56]. Our rate of elevated diabetes distress (59%) is higher than a meta-analysis of 55 studies, which showed an overall prevalence of 36% of diabetes distress in people with type 2 diabetes, with higher rates in predominantly female samples and those with more comorbidities [57]. Our small sample size, comprised of participants with elevated HbA_1c and who enrolled in a treatment study may account for this high rate of distress. Alternatively, diabetes self-management may be exceptionally distressing in the context of high SDoH needs, as is common among patients at CHCs. In this way, MAP may be particularly beneficial for adults with T2D who access health care at CHCs. One study in T2D found that patients who had high diabetes distress at baseline had a greater increase in self-management and a decrease in HbA_1c from mHealth DSMS compared to their low-distress counterparts [58].

Diabetes self-efficacy increased at 3 months but only became statistically significant at 6 months, which may reflect our small sample size since larger studies have shown increases by 3 months [59]. Alternatively, this sample with elevated HbA_1c, high diabetes distress, and high social needs may require 6 months of intervention to increase self-efficacy. Diabetes self-efficacy has been shown in a systematic review to be an important predictor of self-care behaviors [60] including for example medication adherence [61]. The improvement in diabetes self-efficacy observed in our study may prove important for downstream clinical outcomes since interventions that improve self-efficacy have been shown to directly and indirectly improve glycemic control [62].

We observed that HbA_1c and DSM improved at 3 and 6 months relative to baseline values but these improvements were not statistically significant in our small sample. Research on the effectiveness of patient portals on glycemic outcomes has produced inconsistent results [13,14,63-65]; yet, observational studies with adequate sample sizes do in fact provide consistent evidence that patient portal use has a beneficial impact on HbA_1c [66-68]. The consistency of portal use is also associated with benefits for HbA_1c. In a study that examined the consistency of patient portal use among adults with T2D (N=95,043), an increasing number of calendar months of patient portal use was associated with a significant decrease in HbA_1c levels [69]. In addition, several studies have highlighted the role of patient portals in improving diabetes self-management behaviors. A retrospective cohort study involving over 100,000 adults with diabetes found that when patients are given access to the patient portal via a mobile device, their adherence to oral antihyperglycemic medications improves significantly [66].

In addition to the small sample size, our study may not have produced significant improvements in HbA_1c and DSM because of the low degree of integration of MAP into clinical care. We ensured that MAP was delivered in a clinical setting by embedded clinicians, yet we encountered a number of barriers to a deeper level of clinical integration. For example, in one clinic, participants were not able to send portal messages directly to their PCPs because that portal functionality had not yet been activated clinic-wide. As another example, at the other clinic, the MAP nurse was not part of the patient’s own clinical team. Whereas MAP nurses did send monthly progress notes to the patients’ respective PCPs, it is unknown how the PCPs made use of the reports. In our study, one nurse was bilingual in English or Spanish; however, not all were and one participant had Polish as a first language. In these situations, translation of messages was required. Thus, MAP’s good metrics of implementation were only made possible with creative approaches and “work arounds” to clinic and portal idiosyncrasies. It remains untested how effectively MAP could improve glycemic control and DSM when fully integrated into clinical care.

Health care providers and adults with diabetes in CHCs continue to face barriers that limit their use of portals. Well-documented challenges include low health- and technical literacy, lack of regular access to internet-connected mobile phones, limited language concordance between providers and patients, and other SDoH-related issues [25,70]. CHCs also face the challenge of lacking standardized implementation strategies for rolling out portals and supporting patients to use portals. The findings from the current pilot study provide preliminary evidence of what may work in CHC settings to mitigate these barriers and improve portal use among adults with T2D.

Our findings have several clinical implications. First, having clinicians recommend the use of the patient portal and having dedicated staff to provide training and technical support to patients in CHCs can be useful in improving portal use. Racial and ethnic differences in portal use attenuate when clinicians offer portals to all patients irrespective of their racial and ethnic background [71]. In addition, a recent systematic review to summarize research priorities and best practices in patient portals found that studies that report high rates of portal use often have dedicated staff to enroll and assist patients [72]. Evidence-based techniques to educate socially vulnerable patients, such as screening for health literacy, using the teach-back method, and using bilingual messaging, were used in our intervention and can also be adopted by clinicians in CHCs. Many CHCs incorporate nursing case management for patients with chronic conditions. Nurse case management communication could also be expanded to use the patient portal for ongoing check-ins, support, and patient education, which may enhance portal engagement and health outcomes. In a systematic review of the effectiveness of patient education through patient portals, significant improvements in knowledge, self-management, health behaviors, mental health, and select health outcomes were demonstrated [73].

Lastly, clinic administrators at CHCs could work with their electronic health record support team to eliminate the need for email accounts to activate the patient portal. Patients who access care at CHCs may not have email accounts, which creates an additional barrier to portal use among CHC patients [74]. Alternative approaches such as the use of phone numbers to authenticate users should be encouraged, as the majority of patients in CHCs have access to mobile phones.

Findings should be interpreted with caution given study limitations, the primary of which is a design without a comparison group or randomization to treatment. Future research should test MAP using a more rigorous study design. Implementation research should identify strategies to more fully integrate interventions like MAP into complex clinical settings. Common metrics of portal use should be used as outcomes, in addition to study-specific indicators, in order to facilitate comparison across studies. Our sample size at each clinic was insufficient to test for differences between clinics in sample characteristics or differences in outcomes. Study samples should also be large enough to be powered to test for clinically meaningful improvements in HbA_1c and to include adequate representation of medically underserved individuals so that intervention effects on disparities in portal use—both ameliorating and exacerbating—can be examined. A longer duration of follow-up would allow for investigation of the durability of any treatment effect.

There is also a need for research to develop portal platforms and functionality that better meet the needs of patients and clinicians. This could include, for example, graphical and pictorial data displays, audiovisual capabilities, and libraries of patient education materials. Beyond the patient portal, our preparation for the intervention implementation revealed a paucity of high-quality, patient-friendly diabetes educational materials (pictorial, video, interactive, and gaming) that are designed for low-literacy, low-numeracy, and Spanish-speaking individuals. Whereas a wide variety of materials exist, few if any meet all the above criteria, and those that try tend to be extremely cursory.

In this pilot study among people with diabetes receiving care at CHCs, the MAP intervention produced high activation and engagement in portal use as well as meaningful improvement in psychosocial outcomes and promising changes in clinical outcomes. Numerous challenges were also identified that can be addressed in future research. Increasing portal use specifically in health disparities populations may be one key to ameliorating disparities in diabetes outcomes.