Improving Medication Adherence and Outcomes in Adult Kidney Transplant Patients Using a Personal Systems Approach

SystemCHANGE Results of the MAGIC Randomized Clinical Trial

Cynthia L. Russell; Donna Hathaway; Laura M. Remy; Dana Aholt; Debra Clark; Courtney Miller; Catherine Ashbaugh; Mark Wakefield; Sangbeak Ye; Vincent S. Staggs; Rebecca J. Ellis; Kathy Goggin


American Journal of Transplantation. 2020;20(1):125-136. 

In This Article


Patients were enrolled between January 12, 2015 and April 14, 2017. Of 1096 persons contacted for participation, 580 (52.9%) agreed to participate and 438 (75.5%) completed screening. Figure 1 delineates the flow of participants in the study and reasons for follow-up losses. Of those who completed screening, 281 (64.2%) were adherent and excluded from the study. In total, 156 (35.6%) patients were nonadherent and assessed for eligibility for the intervention phase of the study. A total of 130 participants were randomized, 65 into SystemCHANGE™ and 65 into attention control. After randomization, 25 participants in SystemCHANGE™ and 16 in the attention control groups declined to participate in the study. Consequently, 89 participants were enrolled in the study. During the 6-month intervention phase, 5 withdrew from the study (3 from SystemCHANGE™ and 2 from the attention control intervention), yielding a sample of 84 (94% retention) for the 6-month intervention phase analysis. The 6-month maintenance phase was completed by 73 participants (82% retention).

Figure 1.

Participant flow diagram

Table 3 delineates baseline sample demographics of patients randomized into the interventions (n = 89). The average age was 51.8 years (standard deviation SD 10.5) with 58.4% (n = 52) male and nearly two-thirds African American (61%, n = 54). Marital status differed somewhat between the two groups and was included as a covariate in the regression analysis. Baseline MA scores were comparable between the two groups.

Table 4 compares SystemCHANGE™ and attention control intervention MA scores at the 6-month intervention and 6-month maintenance phases. Using an intent-to-treat analysis, at the completion of the 6-month intervention phase, mean MA for SystemCHANGE™ (median 0.91, interquartile range [IQR] 0.76–0.96) and attention control (median 0.67, IQR 0.52–0.72) patients differed markedly (difference in medians 0.24, 95% confidence interval [CI] 0.13–0.30, P < .001). Similar results were found at the conclusion of the 6-month maintenance phase, with MA for SystemCHANGE™ (median 0.77, IQR 0.56–0.94) and attention control (median 0.60, IQR 0.44–0.73) patients remaining meaningfully different (difference in medians 0.17, 95% CI 0.06–0.33, P = .004). In a "completers" analysis, where only those with MA data at the end of the 6-month intervention phase were compared, MA for SystemCHANGE™ (median 0.93, IQR 0.78–0.96) and attention control (median 0.68, IQR 0.55–0.79) patients again differed widely (difference in medians 0.25, 95% CI 0.14–0.30, P < .001). Likewise, in a "completers" analysis at the end of the 12-month maintenance phase, MA for SystemCHANGE™ (median 0.77, IQR 0.56–0.94) and attention control (median 0.60, IQR 0.44–0.73) patients remained substantially different (difference in medians 0.17, 95% CI 0.06–0.33, P < .001).

Table 5 delineates the regression model results for adherence at 6 and 12 months. Controlling for marital status, race, and perceived health, membership in the intervention group was associated with an estimated 20% higher MA at 6 months and 16% higher at 12 months (P < .001).

Figure 2 depicts the pattern of MA over the course of the study including both intervention and maintenance phases using a random intercept model. Initially, a model with both a random intercept and random time slope were both fit, but the random time slope was deemed unnecessary due to its small variance. A group-by-time interaction was considered, but due to a strong similarity in the decaying trends in the SystemCHANGE™ and attention control groups (SC slope −0.0056; PE slope −0.0056), the interaction term was dropped.

Figure 2.

Random intercept model of SystemCHANGE™ and patient education interventions. AC, attention control group; SC, SystemCHANGE™ group

Figure 3 shows the means and frequencies for creatinine, BUN, and infections. Only 3 acute and chronic rejections, no kidney failures, and one death occurred during the study so these data are not presented. No adverse events were reported during this study.

Figure 3.

Patient outcome data. AC, attention control group; SC, SystemCHANGE™ group. Study Phase 1 was baseline; phase 2 was 6 mo and phase 3 was 12 mo. There are two depths of shaded areas in the upper figures. The 95% confidence interval bands are displayed as shaded area around each line. Lightly shaded areas pertain to the line that is closest. In both trend plots, the dotted line and the solid lines are close enough to each other such that each line's shaded area overlaps to form the darker shaded areas