Endocrinology · Medical Literature
Promoting Problem-Solving Among Low-Income Adults With Type 2 Diabetes: Cluster-Randomized Controlled Trial of a Mobile Health Intervention With SMS Text Messaging (Mobile Diabetes Detective).
Abstract
Problem-solving is essential for the self-management of type 2 diabetes but remains challenging for underserved individuals. Although mobile health (mHealth) interventions can improve diabetes self-management, few focus on problem-solving. This study evaluates the efficacy of Mobile Diabetes Detective (MoDD), a fully automated web-based intervention with SMS text messaging that provides problem-solving support tailored to self-monitoring data, for improving glycemic control among medically underserved adults with type 2 diabetes. This open-label, 1:1 cluster-randomized controlled trial was conducted in 2013-2018. Participants were adults with type 2 diabetes (glycated hemoglobin [HbA1c] >7.5%) receiving care at 8 Federally Qualified Health Centers serving medically underserved communities in the New York metropolitan area. The centers served as clusters and were randomized using computer-generated allocation. Recruitment and study sessions were conducted either in person or in a hybrid format. The intervention arm used MoDD for 12 months, whereas the control arm received standard diabetes education and routine care. The primary outcome was the change in HbA1c from baseline to 12 months, recorded from medical chart data. We hypothesized greater improvement in the intervention arm than in the control arm. Secondary outcomes included psychosocial measures. Outcomes were compared between groups using intention-to-treat analyses. This report presents the final analysis of the outcomes. This trial randomized 248 participants (intervention arm: n=126; control arm: n=122); 219 were included in the final analysis (intervention arm: n=111; control arm: n=108). Participants were predominantly female (147/219, 67.1%) and ethnically and racially diverse (112/219, 51.1%, Hispanic and 92/219, 42%, African American), with a mean baseline HbA1c of 9.9%. Overall, of the 111 participants, 44 (39.6%) engaged with MoDD at least once weekly in the first 30 days, and 22 (19.8%) engaged at least once weekly in the first 90 days. HbA1c did not differ significantly between groups at baseline (intervention: 9.81%, 95% CI 9.42%-10.20%; control: 9.95%, 95% CI 9.55%-10.34%; difference=0.14%, P=.63) or at 12 months (intervention: 9.36%, 95% CI 8.95%-9.78%; control: 9.58%, 95% CI 9.15%-10.01%; difference=-0.22%, P=.47). Both groups demonstrated reductions in HbA1c from baseline to 3 months. Sustained within-group improvement at 12 months was observed in the intervention group but not in the control group. No intervention-related adverse events were reported. This study evaluated the impact of a mobile intervention for problem-solving in diabetes. MoDD is innovative because it operates autonomously and tailors support to individuals' self-monitoring data. Although there was no significant between-group difference in HbA1c, the intervention group showed sustained within-group improvement at 12 months. These findings highlight the potential long-term benefits of autonomous mHealth interventions for problem-solving. The study observed an increase in diabetes distress, possibly reflecting heightened awareness of uncontrolled blood glucose levels. If implemented in clinical practice, MoDD could complement diabetes education and help improve glycemic control. ClinicalTrials.gov NCT02021591; https://clinicaltrials.gov/ct2/show/NCT02021591.
Clinical relevance
Among medically underserved adults with poorly controlled type 2 diabetes receiving care at Federally Qualified Health Centers, does the Mobile Diabetes Detective (MoDD)—a fully automated, SMS-supported, web-based problem-solving intervention—improve glycemic control (HbA1c) at 12 months compared with standard diabetes education and routine care? __SNAPSHOT__ {"studyDesign":"Open-label, 1:1 cluster-randomized controlled trial","setting":"8 Federally Qualified Health Centers (FQHCs) serving medically underserved communities in the New York metropolitan area, USA; conducted 2013–2018","sampleSize":"n=248 randomized (intervention: n=126; control: n=122); n=219 included in final analysis (intervention: n=111; control: n=108)","population":"Adults with type 2 diabetes and HbA1c >7.5% receiving care at FQHCs; predominantly female (67.1%); ethnically and racially diverse (51.1% Hispanic, 42% African American); mean baseline HbA1c approximately 9.8–9.95%","intervention":"Mobile Diabetes Detective (MoDD): a fully automated, web-based problem-solving support tool with SMS text messaging tailored to individual self-monitoring blood glucose data, used for 12 months. Participants received one-time technology training with up to four additional on-site support sessions as needed.","comparator":"Standard diabetes education and routine care","primaryOutcome":"Change in HbA1c from baseline to 12 months, ascertained from medical chart data","followUp":"12 months; intermediate assessments at 4 weeks and 3 months","fundingSource":"NIH (National Institutes of Health; R01-DK090372); government-funded"}
Key findings
The primary outcome—between-group difference in HbA1c change at 12 months—was not statistically significant. Baseline HbA1c did not differ significantly between groups (intervention: 9.81% [95% CI 9.42%–10.20%] vs. control: 9.95% [95% CI 9.55%–10.34%]; difference = 0.14%, P=.63). At 12 months, HbA1c remained statistically indistinguishable between arms (intervention: 9.36% [95% CI 8.95%–9.78%] vs. control: 9.58% [95% CI 9.15%–10.01%]; between-group difference = –0.22%, P=.47). Within-group, both groups demonstrated HbA1c reductions from baseline to 3 months; however, sustained within-group improvement at 12 months was observed only in the intervention arm and not in the control arm. The absolute within-group reduction in the intervention arm from baseline (9.81%) to 12 months (9.36%) was approximately 0.45 percentage points (calculated), and in the control arm from baseline (9.95%) to 12 months (9.58%) was approximately 0.37 percentage points (calculated); neither difference was reported with a formal p-value for within-group comparison. NNT cannot be calculated given the absence of a statistically significant between-group difference on the primary endpoint. Secondary outcomes included psychosocial measures; specific results for secondary endpoints are not reported in the available abstract. No intervention-related adverse events were reported.
Practice implications
This trial established that a 12-month fully automated mHealth problem-solving intervention (MoDD) did not produce a statistically significant improvement in HbA1c versus standard care in medically underserved adults with poorly controlled type 2 diabetes. The study did not establish superiority of MoDD on the primary glycemic endpoint. The observed between-group difference (–0.22%) was small and not statistically significant (P=.47), indicating the intervention did not meet its prespecified primary hypothesis. A pattern of sustained within-group HbA1c reduction in the intervention arm—versus attenuation in controls—is of potential interest but is an exploratory observation and cannot support causal inference or practice change on its own. The population studied (predominantly Hispanic and African American adults attending FQHCs, mean HbA1c ~9.9%) is clinically important and underrepresented in prior mHealth trials; however, because the primary endpoint was not met, findings are not directly actionable for individual patient management. Patients with higher baseline HbA1c, different social determinants of health, or those outside the NYC FQHC setting may respond differently. The trial was open-label, which precludes blinding of participants or providers, and HbA1c ascertainment from chart data introduces risk of missing or inconsistent measurement. Only the abstract was available for this review; full secondary outcome data, subgroup analyses, engagement metrics, and attrition details were not available for appraisal. __LIMITATIONS__ Null primary outcome: No statistically significant between-group HbA1c difference was demonstrated (P=.47); the trial did not achieve its primary hypothesis.,Small number of clusters (n=8 FQHCs): Cluster-randomized designs with few clusters are vulnerable to imbalance and have limited power to adjust for cluster-level confounders; the intraclass correlation and its impact on power are not reported in the abstract.,Open-label design: Neither participants nor providers were blinded, introducing risk of performance bias and differential attention effects (Hawthorne effect in both arms).,HbA1c ascertainment from medical records: Chart-extracted HbA1c values introduce risk of missing data and non-standardized measurement intervals, which may bias within-group and between-group estimates.,Long trial duration (2013–2018): A five-year enrollment window introduces risk of secular changes in background diabetes care, technology access, and standard-of-care practices that could confound results.,Attrition: 29 of 248 randomized participants (11.7%) were excluded from the final analysis; reasons for exclusion and differential attrition by arm are not reported in the abstract.,Within-group versus between-group framing: Reporting of 'sustained within-group improvement' in the intervention arm without a significant between-group difference risks overinterpretation; regression to the mean cannot be excluded.,Limited secondary outcome reporting: Psychosocial and problem-solving outcome data are referenced but not detailed in the available abstract, precluding assessment of mechanism or secondary benefit.,Generalizability: Restricted to FQHCs in one metropolitan region; findings may not generalize to rural, suburban, or non-US underserved populations, or to patients with less severe baseline hyperglycemia.,Fully automated intervention design: Absence of human coaching may limit engagement and problem-solving skill development, especially in populations with variable health literacy and technology experience.,Only abstract available for this review: Full details on randomization balance, protocol deviations, sensitivity analyses, and subgroup effects could not be independently verified. __BOTTOM_LINE_CATEGORY__ not_ready_for_clinical_application __BOTTOM_LINE__ MoDD, a fully automated SMS-supported mHealth problem-solving intervention, did not significantly reduce HbA1c compared with standard care at 12 months in medically underserved adults with poorly controlled type 2 diabetes (between-group difference –0.22%, P=.47). A within-group signal of sustained glycemic improvement in the intervention arm warrants further investigation but is insufficient to support practice change. Current evidence does not support adopting MoDD as a superior alternative to standard diabetes education for glycemic control in this population.