Improved Diabetes Care Management Through a Text-Message Intervention for Low-Income Patients: Mixed-Methods Pilot Study

BACKGROUND: Diabetes is a major contributor to global death and disability. Text-messaging interventions hold promise for improving diabetes outcomes through better knowledge and self-management. OBJECTIVE: The aim of this study was to examine the implementation and impact of a diabetes text-messaging program targeted primarily for low-income Latino patients receiving care at 2 federally qualified health centers (FQHCs). METHODS: A mixed-methods, quasi-experimental research design was employed for this pilot study. A total of 50 Spanish or English-speaking adult patients with diabetes attending 2 FQHC sites in Los Angeles from September 2015 to February 2016 were enrolled in a 12-week, bidirectional text-messaging program. A comparison group (n=160) was constructed from unexposed, eligible patients. Demographic data and pre/post clinical indicators were compared for both the groups. Propensity score weighting was used to reduce selection bias, and over-time differences in clinical outcomes between groups were estimated using individual fixed-effects regression models. Population-averaged linear models were estimated to assess differential effects of patient engagement on each clinical indicator among the intervention participants. A sample of intervention patients (n=11) and all implementing staff (n=8) were interviewed about their experiences with the program. Qualitative data were transcribed, translated, and analyzed to identify common themes. RESULTS: The intervention group had a mean glycated hemoglobin (HbA(1c)) reduction of 0.4 points at follow-up, relative to the comparison group (P=.06). Patients who were more highly engaged with the program (response rate ≥median of 64.5%) experienced a 2.2 point reduction in HbA(1c), relative to patients who were less engaged, controlling for demographic characteristics (P<.001). Qualitative analyses revealed that many participants felt supported, as though “someone was worrying about [their] health.” Participants also cited learning new information, setting new goals, and receiving helpful reminders. Staff and patients highlighted strategies to improve the program, including incorporating patient responses into in-person clinical care and tailoring the messages to patient knowledge. CONCLUSIONS: A diabetes text-messaging program provided instrumental and emotional support for participants and may have contributed to clinically meaningful improvements in HbA(1c). Patients who were more engaged demonstrated greater improvement. Program improvements, such as linkages to clinical care, hold potential for improving patient engagement and ultimately, improving clinical outcomes.