We approached 557 potential subjects, 70 of whom refused, 90 of whom were not eligible, and 58 of whom were potentially eligible but did not attend the study enrollment visit (Figure A3, available in an online appendix). Thus, 339 (61%) subjects enrolled and were randomly assigned to ATSM (n = 112), GMV (n = 113), or usual care (n = 114). Participants had a mean age of 56 years, more than half had some high school education or less, half lacked health insurance, and the majority reported incomes <$30,000 (). A roughly equal percentage spoke English or Spanish, followed by Cantonese. Among English and Spanish speakers, more than half had limited health literacy. A1C at enrollment was 9.5%, blood pressure was 140/77 mmHg, and BMI was 31.5 kg/m2. There were no statistically significant differences in baseline characteristics across arms.
Of the subjects, 305 (90%) completed follow-up interviews at 1 year (Figure A3). Three participants died during the study period in each of the three arms. Participants lost to follow-up were younger (51.7 vs. 56.5 years, P = 0.02) but otherwise were no different at P < 0.05. Paired values for A1C were available for 88.2% of the sample, blood pressure for 94.1%, and BMI for 92.3%.
Engagement with interventions
Of the 112 randomly assigned subjects, 105 (94%) completed ≥1 ATSM call. The mean number of ATSM calls completed among ever users was 21.9 of 39 automated calls delivered. Among ATSM users, 100 (95%) received ≥1 care manager call-back, and the mean number of call-backs was 9.2. Of these, 88 (88%) created ≥1 action plan, with a mean of 5.2. Partial or complete success was reported to care managers on a mean of 2.5 action plans.
Of the 113 randomized participants, 78 (69%) attended ≥1 GMV. The mean number of GMVs attended among ever users was 4.8 of 9 GMVs offered. Among GMV attendees, 69 (89%) created ≥1 action plan. This subgroup generated a mean of 3.2 action plans and reported partial or complete success to GMV facilitators on a mean of 1.6. Across both SMS interventions, exercise and/or diet constituted the majority of action plans.
Effects on structure and processes of care
Both ATSM and GMV participants demonstrated robust improvements relative to usual care participants in PACIC (P < 0.0001), with standardized effect sizes of 0.51 for ATSM versus usual care and 0.53 for GMV versus usual care (). No significant differences were observed between ATSM and GMV participants in overall PACIC change. For PACIC subscales, both ATSM and GMV participants demonstrated significant improvements relative to usual care participants in delivery system/practice design, goal setting, problem-solving, and follow-up/coordination. Only ATSM participants demonstrated significant improvements relative to usual care participants in patient activation (Table A1, available in an online appendix).
Comparison of baseline and 12-month outcomes
ATSM and GMV participants showed similar improvements in diabetes self-efficacy relative to usual care participants (effect size 0.41, P < 0.01, and 0.38, P < 0.01), with no significant differences between ATSM and GMV participants. In contrast, participants in ATSM reported improvements in interpersonal communication relative to both the usual care (0.50, P < 0.001) and GMV participants (0.31, P = 0.03). For the IPC subscales, ATSM yielded significant improvements relative to usual care in explanations of processes of care, explanations of self-care, and empowerment and significant improvements relative to both usual care and GMV in elicitation of patient problems and decision-making (Table A1).
Effects on behavior
Compared with usual care participants, ATSM and GMV participants showed significant increases in self-management behavior (). The increase was more robust for ATSM than for GMV participants (effect size 0.34, P = 0.02). For individual self-management domains (Table A1), both ATSM and GMV participants improved with respect to self-monitoring of blood glucose, but only ATSM participants improved foot care. Although ATSM and GMV participants increased relative to usual care participants for diet and exercise, only ATSM participants reported a significant increase in physical activity, with 2 more h/week relative to usual care participants (effect size 0.31, P = 0.03).
In a comparison of baseline and follow-up reports, a greater percentage of ATSM participants achieved weekly minimum recommendations regarding physical activity of ≥30 min three times per week (59.8 vs. 68.3%; odds ratio [OR] 1.5 [95% CI 0.9–2.4]). There was little change observed for GMV participants (60.2 vs. 59.6%, 1.0 [0.6–1.5]) and a reduction in usual care participants (58.8 vs. 53.3%; 0.8 [0.5–1.2]). The interaction between the ratio of those achieving standards of physical activity at baseline versus follow-up was significant in ATSM participants versus usual care participants (P = 0.05) but not in a comparison of ATSM versus GMV participants (P = 0.50).
Effects on functional outcomes
ATSM participants reported significant decreases in days restricted to bed compared with usual care participants (−1.7 days/month, rate ratio 0.5 [95% CI 0.3–1.0]) and with GMV participants (−2.3 days/month, rate ratio 0.4 [0.2–0.7]) (). ATSM participants were less likely to report that diabetes prevented them from carrying out daily activities: 15% reported activity restriction at baseline, compared with 6% at 1 year (OR 0.37 [95% CI 0.1–0.9]). Comparable values for GMV participants were 16 and 17% (1.0 [0.5–2.0]) and for usual care participants were 17 and 21% (1.3 [0.7–2.3]). The interaction between the proportion reporting restricted activity at baseline versus follow-up was not significant for ATSM versus GMV participants (P = 0.6) but was for ATSM versus usual care participants (P = 0.04). Although SF-12 physical health increased across all three groups, there were no statistically significant differences between ATSM, GMV, and usual care participants. SF-12 mental health differentially improved for ATSM relative to GMV (effect size 0.31, P = 0.03) and usual care (effect size 0.18, P = 0.2) participants.
Effects on metabolic outcomes
Glycemic control improved across all three arms, but there were no statistically significant differences in A1C change between ATSM, GMV, and usual care participants. Although SBP and DBP fell in both the ATSM and GMV arms relative to the usual care arm, these values did not reach statistical significance. Changes in BMI were not different across the three arms.
Post hoc analyses
To better understand our findings regarding the relative superiority of ATSM over GMV, we carried out two post hoc analyses. To test whether the different engagement rates might explain observed differences in behavioral and functional outcomes, we created a standardized, three-level variable that categorized participant engagement into low, medium, and high. Whereas greater engagement was associated with improvements in self-management behavior and functional status in both SMS arms, including this variable as a covariate in our models did not alter the size of the effect of ATSM relative to that of GMV for either outcome. To explore whether disproportionate improvements in IPC observed in ATSM versus GMV mediated differences in behavioral and functional outcomes, we reran self-management and bed-day models, including IPC scores as an additional covariate. Inclusion of IPC scores reduced the effect of ATSM versus GMV for self-management behavior from 0.33 days/week (P = 0.02) to 0.25 days/week (P = 0.07) and reduced the rate ratio in bed days from 0.35 (P < 0.01) to 0.45 (P = 0.03).