Use of Multisystemic Therapy to Improve Regimen Adherence Among Adolescents With Type 1 Diabetes in Chronic Poor Metabolic Control: A randomized controlled trial

Adolescents with CPMC and their families were recruited from an endocrinology clinic within a tertiary care children’s hospital located in a major Midwestern metropolitan area. Inclusion criteria were 1) diagnosed with type 1 diabetes for at least 1 year; 2) an average HbA_1c [A1C] ≥8% during the year before study entry, as well as a most recent A1C ≥8%; 3) aged 10.0–17.0 years, and 4) sufficient mastery of English to communicate with therapists and complete study measures. The only study exclusion criteria were moderate/severe mental retardation or psychosis.

Of the 182 families eligible for participation, 33 families (18%) refused to participate. Five families (3%) indicated an interest in participating but asked to be recontacted later, and 17 families (9%) consented to participate but had not yet completed baseline data collection when study enrollment closed and were therefore not included. The final sample consisted of 127 adolescents and their families (70% of eligible families). Sixty-four participants were assigned to MST and 63 participants to the control condition. Nine of the 127 families (7%) dropped out of the study before completing follow-up data collection and another 8 (6%) did not complete data collection within the specified window (87% completion rate). Seven of the 17 were in the MST condition and 10 were in the control condition; there was no suggestion of differential loss to follow-up between the groups.

The study was a randomized controlled trial with a repeated-measures design. Families randomly assigned to MST received ∼6 months of home-based psychotherapy plus standard medical care, whereas families randomly assigned to the control condition received standard care only. Randomization to the treatment condition was completed immediately after baseline data collection by the project statistician. To ensure equivalence across treatment condition, randomization was stratified by level of A1C at the baseline visit. Data were collected at 7, 12, 18, and 24 months after baseline data collection. The present study reports on data from 7 months posttest, as data collection at later time points is not yet complete.

Adolescents assigned to the intervention condition received MST plus standard medical care (described below). MST is an intensive, family-centered, community-based treatment originally designed for use with adolescents presenting with serious antisocial behavior (27). As MST is designed to target the multiple systems within which youth with serious problems and their families are embedded, it does not follow a session-by-session treatment protocol. Rather, MST is specified through nine treatment principles that operationalize the parameters for designing and implementing interventions and a treatment manual focusing on the application of these principles. To promote treatment fidelity, therapists and their supervisor received formal, week-long training in MST techniques. MST interventions were also monitored for treatment fidelity using state-of-the-art quality assurance protocols, including weekly on-site clinical supervision and weekly phone consultation with an MST expert consultant (36,37). Therapists began by conducting a multisystemic assessment of the strengths and weaknesses of the family, then tailored treatment goals and interventions to each family to best treat the adherence problem based upon this assessment. Details of procedures to establish MST treatment fidelity, as well as data showing associations between fidelity to the MST approach and improved clinical outcomes, have been extensively reported elsewhere (38–40).

In the current study, therapists were expected to meet with families a minimum of two to three times per week at the beginning of treatment. Treatment was terminated when treatment goals were met rather than when a set number of sessions were completed. However, based on previous MST trials and our own prior experience, treatment was planned to last for ∼6 months. The mean length of treatment in the study was in fact 5.7 months. Twenty-five percent (16 of 64) of treatment families did not complete a full course of therapy. The mean number of sessions was 48 (SD 19) for treatment completers and 9 (8) for dropouts.

MST interventions targeted adherence-related problems within the family system, peer network, and the broader community systems within which the family was embedded. Therapists drew upon a menu of evidence-based intervention techniques that included cognitive-behavioral therapy, parent training, and behavioral family systems therapy. For example, individual interventions carried out in the present study included cognitive-behavior therapy with depressed adolescents. Family interventions included introducing systematic monitoring, reward, and discipline systems to decrease parental disengagement from the diabetes regimen; developing family organizational routines such as regular meal times; and teaching caregivers to communicate effectively with each other about the adolescent’s medical regimen. School interventions included improving family-school communication about the adolescent’s diabetes care needs and adherence behaviors (e.g., having school personnel report blood glucose readings from the school meter to parents weekly) and working with school personnel to monitor and support the adolescent’s regimen completion (e.g., finding a private place to test blood glucose). Peer interventions included enlisting the active support of peers regarding regimen adherence.

At the community level, interventions included developing strategies to monitor and promote the youth’s diabetes care while participating in extracurricular activities or in other settings (e.g., visiting extended family members). Interventions within the health care system included helping the family resolve barriers to keeping appointments, and working with the family and the diabetes treatment team to promote a positive working relationship. For example, therapists were required to accompany families to their medical appointments to assist the family and members of the medical team in determining ways to improve regimen adherence.

Adolescents randomly assigned to the control condition received standard medical care. Standard care at the hospital where adolescents were cared for consisted of quarterly medical visits with a multidisciplinary medical team composed of an endocrinologist, nurse, dietitian, social worker, and psychologist. No restrictions were placed on receipt of mental health services such as outpatient psychotherapy or psychiatric management during study participation for those adolescents randomly assigned to receive standard care. However, only three adolescents randomly assigned to the control condition were reported by their parents to have received such services during the study.

Adolescents completed the Twenty-Four Hour Recall Interview (41,42), a semistructured interview used to assess adherence behaviors by obtaining a temporal sequence of events from the previous day, including all diabetes care tasks. Although in Johnson and colleagues’ original work several recall interviews were obtained during the course of a week to assess consistency of adherence behaviors over time, the nature of the population under study (i.e., erratic family schedules and lack of phones) made collection of more than one recall interview problematic. Therefore, only one recall interview was obtained from each adolescent. Recalls were coded for insulin adherence, dietary adherence, and blood glucose testing adherence (34). The insulin and dietary adherence variables were not coded for adolescents using insulin pumps, as the measure was not designed for use with this population and did not adequately capture adherence behaviors related to pump management.

Frequency of blood glucose testing, a specific adherence behavior, was also obtained directly from the adolescent’s blood glucose meter. For the present study, data were recorded regarding frequency of testing during the 14-day period immediately preceding data collection and an average daily testing frequency was subsequently calculated.

For the majority of participants, A1C (reference range = 4–6.4%) was calculated by the medical center laboratory from total GHb. This approach was taken by the laboratory because of the high prevalence rate of abnormal hemoglobin variant carriers in the population served, which is primarily African American. Total GHb was analyzed by boronate affinity chromatography using a GHb and protein analyzer model CLC385 (Primus, Kansas City, MO). However, 5.5% of A1C values were calculated during a medical appointment in the diabetes clinic with a DCA 2000 system (Bayer, Elkhart, IN), which uses an immunoglobulin-agglutination methodology. DCA results were used when adolescents were seen for a clinical appointment on the day of a research visit and refused a venipuncture or when families missed the study visit but completed their clinical visit.

Hospital utilization data for the present study were obtained from the hospital’s medical information system, as all participants enrolled in the study obtained their diabetes care from a single children’s hospital facility. Information regarding the number of hospitalizations and emergency department visits was requested for each participant from the medical information system for the 6-month window before study entry and the first 6 months of study participation. This approach was taken to allow a baseline rate of hospital utilization to be calculated and to assess a time window that approximated the length of the first treatment period. The variables used in the analyses therefore represent the number of emergency department visits and hospitalizations per adolescent during each 6-month window. Participants’ hospital records were first coded as diabetes related or not, based upon ICD-9 codes. Only visits that were related to diabetes were included in the present analyses. All records that were ambiguous were reviewed and coded by a pediatric endocrinologist who was blind to the participant’s randomization status. If a participant was admitted to an inpatient unit from the emergency department, the visit was coded only as an admission for the purpose of calculating rate of utilization. Admissions and emergency department visits clearly related to diabetes (e.g., hospitalizations for DKA) comprised the majority of the medical records. For example, 82% of hospital admissions were associated with an ICD-9 code of DKA.

Power calculations were based on outcomes from our preliminary studies regarding the effect of the MST intervention on metabolic control. The trial was powered to detect a 1.10% difference in A1C between the two groups using a two-tailed α of 0.01 and power = 0.80, with a total of 120 subjects (60 per group).

The hypothesis that the MST group would have improved outcomes relative to the standard care group was tested using two approaches: intent-to-treat analyses and per protocol analyses in which those families who received no or almost no treatment were excluded from the analyses. In the intent-to-treat analyses, all randomly assigned participants were included. Noncompleters of the posttest data collection (7 MST patients and 10 control subjects) were treated conservatively as if no change occurred, and the participant’s baseline score for each outcome variable was forwarded to the follow-up point. Mixed design 2 × 2 (treatment × time) ANOVAs were conducted with A1C, 24-h recall insulin compliance, dietary compliance, glucose testing compliance, and inpatient admissions and emergency department visits serving as the dependent variable in separate analyses. Treatment (MST or standard care) served as the between-subjects factor, and time (baseline to follow-up) served as the within-subjects factor. The effect of interest was the treatment × time interaction term. Basing the test on change from baseline scores improves the sensitivity of the analysis and controls for any potential baseline differences between groups.

The per protocol analyses used general linear mixed effects modeling for repeated-measures data (43) to evaluate the same set of study outcomes. Four families who had been randomly assigned to MST who received zero to two sessions of treatment (i.e., ≤1 week of treatment given at the two to three sessions per week frequency) were excluded. General linear mixed effects modeling is a generalization of the usual mixed design ANOVA that allows unbalanced and missing data. It uses maximum likelihood procedures to account for missing data for the purpose of parameter estimation.

Sample characteristics are presented in Table 1. Consistent with the demographics of the clinic where participants were recruited and the known overrepresentation of minorities among the population of adolescents with CPMC, 63% of the sample was African American. The mean (±SD) A1C for adolescents at study entry was 11.3 ± 2.3%, confirming that adolescents in the sample were in poor metabolic control. The majority of adolescents participating in the study were managed with injected insulin, whereas 8% used an insulin pump. Comparability between the two treatment conditions was tested using Student’s t test for continuous variables and Fisher’s exact test for categorical variables. There were no significant differences between the MST and control groups on study entry A1C or on the majority of demographic variables such as duration of diabetes, regimen type, child or parent age, family composition, ethnicity, or income (Table 1). However, significantly more males were randomly assigned to MST.

Table 2 shows results of the intent-to-treat analyses and the effect sizes associated with the treatment × time interaction for each outcome variable. For adherence measures, a significant treatment × time interaction was found for frequency of blood glucose testing as assessed by blood glucose meter and 24-h recall. Adolescents receiving MST significantly increased their frequency of blood glucose testing from baseline to 7-month follow-up compared with the control group, for whom no change was evident (F[1,125] = 16.75, P = 0.001; F[1,125] = 6.70, P = 0.011). No significant treatment × time interaction was found for insulin adherence (F[1,125] = 1.90, P = 0.171) or eating adherence F[1,125] = 0.04, P = 0.847). For A1C, the treatment × time interaction was significant at the trend level F[1,125] = 3.04, P = 0.085). Adolescents who received MST had improvements in metabolic control from baseline to posttest of 0.61% compared with control subjects with the improvement being 0.09%. For hospital utilization measures, a significant treatment × time interaction was found for inpatient admissions. Adolescents receiving MST had a significantly reduced number of admissions, whereas the number of admissions increased for control subjects (F[1,125] = 6.25, P = 0.014). There was no significant interaction for number of emergency department visits (F[1,125] = 2.81, P = 0.101). Effect sizes ranged from moderate to large (44).

Table 2 shows the results of the per protocol analyses, which excluded four families who received no or almost dose of no intervention (6% of the sample randomly assigned to MST). Similar to intent-to-treat analyses, a significant treatment × time interaction was found for frequency of blood glucose testing as assessed by blood glucose meter and 24-h recall (F[1,105] = 14.93, P = 0.000; F[1,105] = 9.11, P = 0.003, respectively), with adolescents receiving MST showing improvements and control subjects showing no change. No significant treatment × time interactions were found for insulin compliance or eating compliance. Compared with the intent-to-treat analyses, per protocol analyses showed a significant treatment × time interaction for A1C (F[1,114] = 4.03, P = 0.047). Adolescents who received MST had improvements in metabolic control from baseline to posttest of 0.77%, compared with control subjects who had a 0.01% decline in A1C. Per protocol analyses produced findings similar to those for the intent-to-treat analyses for hospital utilization measures. A significant treatment × time interaction was found for inpatient admissions (F[1,121] = 6.47, P = 0.012), with reductions in admissions for adolescents receiving MST and increases in admissions for control subjects. There was no significant interaction for rates of emergency department visits. Effect sizes again ranged from moderate to large (Table 2).

Although sometimes viewed as related to adolescent developmental issues and therefore as transitory, chronic poor adjustment to diabetes and CPMC during the adolescent and young adult years are predictors of poor metabolic control and diabetes complications in adult life (45,46). Despite these known negative trajectories for adolescents with CPMC and the associated high costs of care (4,5), adolescents with CPMC have historically been regarded as a difficult-to-treat population, and few clinical management programs have been described (47). They have also been neglected in the clinical trials literature. However, this randomized trial shows that MST can successfully improve health outcomes among adolescents with a history of chronically poor metabolic control. First, our results suggest that MST was successful in improving metabolic control. Although only trends to significance were found in the intent-to-treat analyses, when the 94% of families who received any significant dose of treatment were considered, A1C was found to decline an average of 0.8%. In addition to being statistically significant, improvement in A1C in this range is clinically meaningful and has been linked to reductions in complication rates (48). Many studies of behavioral interventions among adolescents with type 1 diabetes show either nonsignificant or small effects on metabolic control (49) and almost none have been reported to be effective for adolescents with CPMC. However, despite the intensity of the intervention, average A1C did not decrease to the range recommended for children with type 1 diabetes (≤8.5%). Continued follow-up of the sample will clarify whether behavioral changes were sustained and whether this resulted in additional declines in A1C over time.

Adolescents who received MST also had significant increases in frequency of blood glucose testing as assessed both by self-report and by blood glucose meter readings compared with control sujects. Frequent testing of blood glucose has been linked to better metabolic control (50,51) and may therefore account for the improvements in metabolic control experienced by the MST group. However, no significant effects were found for insulin adherence or eating adherence. Despite the fact that the sample patients were in very poor metabolic control, 87% of adolescents reported 100% compliance with taking their insulin injections at baseline. This may have limited our ability to detect changes in insulin adherence. However, as DKA admissions (which are primarily caused by insulin omission) decreased substantially in the MST group, insulin adherence may have in fact improved for the subset of adolescents for whom injections were frequently missed. Families receiving MST whose children were in the early stages of DKA may also have more effectively taken actions that prevented hospitalization (e.g., taking extra insulin or consuming extra fluids to stay hydrated).

Adherence to diet is one of the more difficult aspects of the diabetes regimen for adolescents, and therefore it is perhaps not surprising that changes were not documented in this aspect of regimen adherence, particularly given the challenging nature of the sample. Obtaining 24-h recall interviews on a single day rather than measuring adherence over 3 days may also have affected the psychometrics of the measure and reduced the likelihood that dietary changes could be detected.

In light of the fact that populations such as the one targeted in this study typically do not access outpatient behavioral health interventions (28–30), it is important to note that the current study showed high rates of recruitment (70%) into a home-based intervention as well as high retention rates (75%) in a treatment that lasted >5 months on average. This suggests that when behavioral interventions are provided in a way that increases access, they have a high likelihood of being accepted by such adolescents and their families.

As MST is an intensive intervention and therapist caseloads are low, it is relatively costly (52) compared with other mental health services such as outpatient psychotherapy. For MST to be a viable treatment in the real world, documentation of reductions in cost of diabetes care is important. Although the current article does not directly assess costs, adolescents receiving MST had a significantly reduced frequency of hospital admissions during the study. Prior work by our group has suggested that this effect occurred primarily through elimination of DKA admissions in the subset of adolescents with recurrent DKA (30). MST can substantially reduce costs of diabetes care by reducing such unnecessary hospitalizations (30). However, more comprehensive cost analyses are needed to clarify whether MST is in fact cost effective.

The current study reports on findings from data collected immediately after the termination of treatment. Although significant improvements were found in a number of domains, the only other randomized clinical trial conducted with this population reported that initial improvements in A1C were lost in the 6 months after the intervention was terminated (9). Therefore, additional follow-up of the sample is needed to determine whether treatment effects attenuate over time.

Results of the study suggest that MST holds promise in improving the diabetes management and metabolic control of adolescents with chronically poorly controlled type 1 diabetes. Home-based interventions provide an ecologically valid, family-centered means of engaging such adolescents and their parents in behavioral treatment (15) and may also allow the interventionist to better understand the “real-world” barriers to regimen adherence that such families face (29). Additional studies are needed to continue to identify effective and cost-efficient treatments to improve diabetes management and health outcomes among such high-risk adolescents.

This project was supported by Grant R01 DK59067 from the National Institute of Diabetes and Digestive and Kidney Diseases.