Effectiveness of a community program for older adults with type 2 diabetes and multimorbidity: a pragmatic randomized controlled trial

Primary Objective: To compare the effect of a 6-month community-based intervention versus usual care on physical functioning in older adults with type 2 diabetes mellitus and 2 or more comorbidities.

Secondary Objective: To compare the effect of a 6-month community-based intervention versus usual care on mental health, depressive symptoms, anxiety, self-efficacy, self-management, and healthcare costs in older adults with type 2 diabetes mellitus and 2 or more comorbidities.

Participants were recruited from three Primary Care Networks (PCNs) in the Edmonton, Alberta, Canada census metropolitan area. PCNs are primary care clinics comprised of family physicians, nurses, nurse practitioners, dietitians, pharmacists, social workers and mental health professionals. The three PCNs were selected for this study because they were similar to one another in terms of primary care services for people with diabetes, provider skills, and older adults served. Study recruitment was conducted from September, 2015 to October, 2016. Clients who had been enrolled at the PCN within the past 24 months were screened (n = 725) by employees of the PCNs for potential inclusion and were eligible to participate if they were 65 years or older, living in the community, able to speak English, diagnosed with T2DM by a physician, self-reported at least 2 other chronic conditions (see Supplemental File - Table 1 for full list of conditions), not planning to leave the community during the 6-month study period, if they passed a cognitive screening assessment (achieved at least 5 correct responses on the Short Portable Mental Status Questionnaire [23], and if they were referred to or participated in a chronic disease management program at the participating PCNs within the previous 24 months. Clients were excluded if they were living in a long-term care facility or if someone in the same household was also enrolled in the study.

Of the 608 eligible clients, 132 (22%) provided their written consent and were enrolled in the study. The study flow is outlined in Fig. 1. Reasons for exclusion from study participation and for declining study participation are reported in Fig. 1. Within each PCN, participants were assigned to either the intervention or the usual care group using permuted block randomization (random block size sequences of 2, 4 and 6) administered by a centralized web-based service (RedCap) independent of the research term that allocated clients at each site to the 2 groups in accordance with the sequence and using a 1:1 ratio.

The Aging, Community and Health Research Unit-Community Partnership Program (ACHRU-CPP) was delivered by an interprofessional team consisting of a Registered Nurse (RN), Registered Dietitian (RD), and Program Coordinator (PC). RNs and RDs were employees of the PCNs. PCs were employees of community partner organizations where the monthly group sessions were hosted. Community partner organizations were selected based on pre-existing relationships with respective PCNs. Below we provide a brief summary of the intervention, and refer those seeking further information to the prior publications [20, 21].

The six-month intervention was grounded in Bandura’s Social Cognitive Theory [24], which recognizes the central role of self-efficacy in achieving (self-management) behavior change. The intervention was co-developed with older adults, family caregivers, and community service providers (e.g., family physicians, RNs) and designed using the guidelines for developing complex interventions [25], best practice guidelines for diabetes [26], and empirical evidence on managing diabetes in older adults, The intervention included several interacting components: 1) up to three in-home visits by the RN, RD or both; 2) six monthly group sessions at the community site; 3) monthly intervention team case conferences; and 4) care coordination led by the RN who worked collaboratively with the other interventionists and PCN providers (e.g., family physicians) and linked the client to other relevant health or social services. Family caregivers were invited to attend the in-home visits and monthly group sessions, to recognize their key role in supporting the older adult participants. The monthly group sessions included education offered by the RN or RD, exercise led by the PC, a light meal, and informal peer support (i.e., participants in the group session supported one another by sharing experiences, challenges and strategies for managing diabetes within the context of multimorbidity). Monthly case conferences enhanced team collaboration and communication by providing the provider team with regular opportunities to share observations about participants strengths, challenges and goals related to managing diabetes and other chronic conditions; identify needs related to other community-based services; and prepare for upcoming group sessions. The intervention was tailored to patient needs and preferences and the local context. For example, patients could decline any number of home visits or group sessions, and all participants continued to have access to the programs and services normally offered by the PCN. At one of the PCNs, in order to be consistent with usual care, the intervention team also included a Pharmacist and Kinesiologist. In this PCN, the kinesiologist took on a similar role to the PC in the other sites.

Details regarding outcome measures are described in the study protocol [19]. The primary study outcome was the Physical Component Summary (PCS) score from the 12-item Medical Outcomes Study Short Form-12v1 Health Survey (SF-12) [29]. Secondary outcomes included the Mental Component Summary (MCS) score of the SF-12 [29], depressive symptoms assessed by the Centre for Epidemiological Studies Depression Scale (CESD-10) [30], anxiety assessed by the Generalized Anxiety Disorder (GAD-7) scale [31], self-efficacy assessed by the Self-Efficacy for Managing Chronic Disease scale [32], and self-management activities assessed by the Summary of Diabetes Self-Care Activities (SDSCA) scale [33]. Healthcare use was measured using the Health and Services Utilization Inventory (HSSUI) [34‐36]. The cost analysis applied unit costs to the service volumes reported in the HSSUI [37] and assumed a societal perspective in order to inform the broad allocation of resources in the public interest [38]. Outcome measures were assessed at baseline and at the 6-months follow-up.

Guidelines are available for judging clinical significance for only some study measures. SF-12 developers suggest minimally important difference (MID) of 3 for interpreting group mean summary score differences (PCS, MCS) and warn against comparing subdomain scores over time. The CESD-10, GAD-7, Self-Efficacy for Managing Chronic Disease, and SDSCA do not have established MIDs.

In order to reduce bias, the statistician and research assistants that collected the assessment data were blinded. It was not feasible to blind participants or providers.

Sample size was calculated based on the primary outcome measure, the PCS from the SF-12. The target enrollment was 160 clients to account for an expected 20% attrition and to ensure 80% power with medium effect size (0.50), and (two-sided) alpha equal to 0.05.

The reporting of this trial follows the CONSORT guideline (www.​consort-statement.​org). We applied intention-to-treat principle to analyze the data. Data are presented as means and standard deviations for continuous variables and as numbers and percentages for categorical variables. Analysis of covariance (ANCOVA) was used to test the differences in study outcomes between the intervention and usual care groups at 6 months. Separate ANCOVA models were run for each outcome where the 6-month outcome was the dependent variable, the group (intervention, usual care) was the independent variable, and the baseline outcome value was the covariate. Confidence intervals (95%) are reported for mean differences. Multiple imputation (n = 132) was employed using the procedure described by He [39]. A complete case analysis (n = 121), performed using only clients with complete outcome data, was conducted as a sensitivity analysis and the results are provided in Supplemental File (Table 2).

Subgroup analyses were proposed for the case where overall trial results achieved statistical significance, with the purpose being to examine the consistency of the treatment effect across subsets of participants. Subgroup analyses would be supplemental to main trial findings and restricted to consideration of the following five baseline factors selected a priori: age, number of chronic conditions, sex, self-efficacy, and diabetes duration. We hypothesized that the intervention may be less effective in participants that are older, female, or have more chronic conditions, lower self-efficacy, or have been living longer with diabetes. No subgroup analyses were proposed for the case where overall trial results did not achieve statistical significance, because these are considered inappropriate by the scientific community [40‐42].

The cost of health service use was compared between the intervention and usual care groups. Clients reported health service use at baseline for the six-month period preceding study enrolment and immediately following completion of the six-month intervention. The program cost was estimated and added to total health service cost for the intervention group and compared to the usual care group. Program costs included the payment to providers to deliver the intervention (e.g., time spent in in-home visits, group sessions, team conferences, travel). Mann-Whitney U-test was employed to evaluate differences in median costs between the two groups due to the positive skew of the data. All statistical analyses were performed using SAS Version 9.4 (SAS Institute, Inc., Cary NC).

This study was conducted in accordance with the Tri-Council Policy Statement, Ethical Conduct for Research Involving Humans [43]. Institutional ethics approval was obtained from the Health Panel of the Health Research Ethics Board at the University of Alberta (Pro00054028) and renewed yearly as required. Operational approval to conduct the research study was obtained from each PCN. Written informed consent was obtained from all participants.

Table 1 provides information related to the characteristics of the study sites. Sites 1 and 2 served suburban/rural geographies. Half of the intervention teams at these two sites, were Certified Diabetes Educators, and both sites had PCs from a community partner organization to run group sessions. Site 3 served an urban/metropolitan geography, the intervention team included a Pharmacist and Kinesiologist in addition to RN and RD; the interventionists at this site did not have the Certified Diabetes Educator credentials, and the kinesiologist was replaced 3 months into the program.

In total, 22% (132/608) of eligible older adults consented and entered the study (Fig. 1). Of the 132 enrolled participants, 121 completed the 6-month study follow-up (92%, 65 in the intervention and 56 in the control group). Of the 11 participants that did not complete study follow-up at 6 months (< 9% attrition), six participants could not be contacted.

Baseline characteristics of participants are reported in Table 2. For both groups, more than 42% of participants were 75 years and older and more than half were married or living with a partner. Income differences were seen across groups, with one-third of participants in the intervention group reporting an annual income less than $40,000 (CAD) compared to half of the participants in the usual care group. Over 75% of both groups reported having six or more chronic conditions (in addition to T2DM). At least half of participants in both groups were diagnosed with T2DM between 5 and 20 years ago, and approximately 30% were diagnosed with T2DM more than 20 years ago. There were a higher number of females allocated to the intervention group (63%) than the usual care group (45%). Group differences at baseline were also seen for some outcomes, including higher scores for the intervention group for MCS (55.6 versus 52.5) and self-management (37.3 versus 34.2), and fewer depressive symptoms in the intervention group (5.0 versus 6.7). It is notable that the mean group difference for MCS exceeded the MID (3.0), whereas the mean group difference in PCS scores (0.70) was well below the MID. MIDs do not exist for the other outcome measures; however, mean baseline scores were similar across groups.

Of the 70 participants in the intervention group, 68 (97%) received at least one home visit and 61 (87%) attended at least one group session. Participants had an average of 1.8 (median = 2) out of a maximum of 3 home visits and attended an average of 3.7 (median = 4) out of a maximum 6 group sessions (data not shown).

Pooled results from five imputations are presented in Fig. 2. There was no significant difference between the intervention and usual care groups for the primary outcome, PCS (mean difference: -0.74; 95% CI: − 3.22 to 1.74). There were no significant differences between the groups for MCS score (mean difference: 1.24; 95% CI: − 1.12 to 3.60), depressive symptoms (mean difference: -0.25: 95% CI: − 2.03 to 1.53), anxiety (mean difference: -0.93; 95% CI: − 2.63 to 0.77), or self-management (mean difference: 0.79; 95% CI: − 2.37 to 3.96). Complete case analyses were performed as a sensitivity analysis, and the results are presented in Supplemental Table 1. The complete case findings are in agreement with the multiple imputation analysis for the primary and secondary outcomes (see Supplemental Table 1). Subgroup analyses were not done since statistical significance was not achieved (see Statistical Analysis above).

Given the relatively high MCS scores at baseline observed in the intervention group, we conducted an exploratory stratified analysis comparing the mean differences in the groups for different MCS baseline values. This analysis was done to see if improvements in MCS were higher among those with lower baseline MCS scores (defined as ≤50). This was indeed the case – i.e., the mean difference and 95% CI for those with a baseline MCS score of 50 or less compared to those with more than 50 was 2.85 (− 3.42 to 9.12) versus 0.42 (− 2.05 to 2.91). This analysis is supplemental and exploratory, and therefore should be interpreted with caution; however, it suggests that larger variation occurs in participants with lower baseline scores, with the potential for significant differences favouring the intervention (e.g., 9.12 upper CI limit which is more than 3 times the MID).

A comparison of health service use costs for the intervention and control groups is presented in Table 3. For all but diabetes care services, the cost analysis showed no significant differences in baseline to 6-month cost change between the two groups. As expected, the significant difference in diabetes care cost change between the groups was due to the inclusion of intervention program costs in the intervention group. However, this significant difference in diabetes care cost change did not result in a group difference in total health service costs (including or excluding acute care costs).

This study employed a rigorous pragmatic design. Personnel employed by the site and external to the research team delivered the program. Intervention fidelity was enhanced by multiple approaches, including standardized provider training., and regular meetings were held with the research team. The pragmatic design captured clients from a population typically seen in a clinic setting for chronic disease management. Utility of a RD in the ACHRU-CPP intervention allowed for a holistic approach to dietary counselling that extends beyond calorie restriction to goal setting, general healthy eating, and food preparation. This is particularly important in strategies for older adults as considerations for body mass distribution (lean tissue vs. visceral adiposity) may be more important than targeted weight loss alone [54]. Additionally, the inclusion of a RD in the intervention team is a strength as intensive dietary support is recommended for older adults, men, and individuals whom have been living with T2DM for a long time [55].

Study limitations include the use of self-reported data and the unavailability of clinical outcome data (e.g., HbA1c). However, concerns have been expressed regarding the relevance of clinical outcomes [56], recent work on patient-reported outcomes (PROMs) is motivated by these concerns [see http://​www.​healthmeasures.​net/​explore-measurement-systems/​promis/​intro-to-promis], and PROMs such as the PCS used in this study are increasingly used as primary outcomes in interventions for older adults with multimorbidity [50, 51]. We were unable to obtain the target sample (enrolled 132, needed 160), which could have led to the study’s non-significant findings. With the exception of outcomes from the SF-12 survey (PCS, MCS), MIDs have not been established for the other outcomes included in this study. Thus, distinguishing “negative” effects from “inconclusive” ones by comparing the 95% CI to MIDs could not be done for outcomes other than PCS & MCS, thereby limiting our understanding of the impact of the intervention. This advanced interpretation of CIs in clinical trials is particularly important because it ensures that potentially beneficial interventions are not prematurely abandoned [49]. Interpretation of subscales/subdomains can also be limited. For example, although there was improvement in the Specific Diet subscale score (see Supplemental Table 1, complete case analysis), Toobert and Johnson caution against over-interpretation because this subscale suffers from poor consistency and low internal reliability [33]. The duration of the intervention and the length of follow-up may have been insufficient to see improvements to physical health, and/or it may be that improvements to mental functioning/health are needed before improvements in physical health can be seen. Strong cross-effects between mental and physical health have been reported; however little is known about the potential pathways through which mental health affects physical health and vice versa [57] Subgroup analyses were not performed due to the overall non-significant findings, but as a result we cannot assess the potential impact of differences such as a gender imbalance among the groups or differences across sites. We did not have access to information on important covariates such as physical activity and nutritional intake, which may have influenced the study outcomes.

This study also suggests potential areas to pursue in future research studies. Some studies that have focused on physical exercise have been effective, thus exploring the impact of physical activity components that vary in type or intensity within our intervention may be beneficial. Similar comments apply to other components of our intervention (e.g., diet, social participation), since our intervention is complex making it challenging and potentially inappropriate to isolate the effectiveness of individual elements [25]. Inclusion of clinical outcome measures such as HbA1C could also help in understanding effects, providing careful attention is given to the levels/changes appropriate and safe for individual older adults enrolled in the study.