The Economic Evaluation of Team Clinic—Group Approach to a Care Model of Early Adolescents With Type 1 Diabetes

Abstract

Introduction:

Glycemic control is challenging for adolescents with type 1 diabetes (T1D). Team Clinic, a shared medical appointment model, has improved psychosocial outcomes in middle school patients with T1D. We aimed to evaluate the costs of delivering Team Clinic.

Method:

Participants were randomized into Team Clinic (n = 44) or usual care (n = 42) groups.

Results:

We found no significant difference in 6-month total costs per subject ($3,204 [intervention] vs. $3,476 [control]. No significant differences were found in health care use, test strip use, or continuous glucose monitoring and/or pump. The intervention had more clinic visits (2.41 vs. 1.52 times) and a longer length of visit (2.34 vs. 0.74 hr, but no difference in provider time per patient per visit (median, 0.67 vs. 0.68 hr).

Discussion:

The Team Clinic care model may help young adolescents with T1D improve psychosocial outcomes and increase completion of clinical visits without increasing costs.

INTRODUCTION

Type 1 diabetes (T1D) is one of the most common worldwide childhood chronic diseases (Floyd et al., 2017; Noordman & van Dulmen, 2013; Soltesz, Patterson, Dahlquist, & EURODIAB Study Group, 2007). Its incidence is rising. In 2020, the Centers for Disease Control and Prevention reported a 30% increase in T1D diagnosis in the United States, with youth cases growing most sharply among diverse populations (Juvenile Diabetes Research Foundation International, 2020). Despite advances in monitoring, pharmacology, and technology, glycemic control is still a significant challenge, especially for youth with T1D. The proportions of them who achieved a target American Diabetes Association (ADA) hemoglobin A1c (HbA1c) goal of < 7.5% are only 23% for those aged 6–12 years and 17% for those aged 13–17 years (Miller et al., 2015). Furthermore, glycemic control continues to deteriorate during adolescent development from 10- to 19-years-old, the poorest control among age groups with T1D (Miller et al., 2015).

Multilevel challenges and barriers contribute to elevated HbA1c for young adolescents (Anderson & McKay, 2011; Anderson et al., 2002; Luyckx et al., 2013), including (1) developmental issues such as establishing independence which hinders adherence to the T1D treatment regimen, (2) psychological issues such as depression and eating disorders, and (3) parent-adolescent conflict resulting in challenges with the T1D treatment regimen. To overcome these challenges, the ADA recommends that adolescents be treated by a multidisciplinary team of specialists to meet the educational, nutritional, behavioral, and emotional needs of the growing child and family (American Diabetes Association [ADA], 2019; ADA, 2020; ADA, 2021). The multidisciplinary team shall be trained in pediatric diabetes management and sensitive to the challenges youth with T1D encounter. The team should provide routine diabetes care for this unique population (ADA, 2019; ADA, 2020; ADA, 2021).

Shared medical appointments (SMAs) are one way to meet this ADA recommendation (Floyd et al., 2017). SMAs integrate peer, health care provider, and family support into longer appointment times (Floyd et al., 2017). It provides education and clinical care in an interactive environment with patient groups and multidisciplinary teams and permits sharing experiences through social learning and/or care modeling (Floyd et al., 2017). SMAs have successfully increased patient and provider satisfaction and improved outcomes in other challenging patient populations (Edelman, Gierisch, McDuffie, Oddone, & Williams Jr., 2015; Kirsh et al., 2017; Menon et al., 2017; Ridge, 2012; Wadsworth et al., 2019) and recently in the adolescent T1D population (Floyd et al., 2017; Mejino, Noordman, & van Dulmen, 2012; Pascual, Pyle, Nieto, Klingensmith, & Gonzalez, 2019). In a recently completed randomized controlled trial, the Team Clinic model, a SMAs approach for middle school-aged participants with T1D (Raymond et al., 2015), found improved psychosocial outcomes such as participants’ self-reported depressive symptoms and family conflict (Majidi et al., 2021).

Although the clinical and psychological effects of SMAs are becoming clearer, there have been, to date, few economic evaluation studies of SMAs for the young adolescent population with T1D. Understanding the cost implications of SMAs are important for health systems considering the adoption of this form of care. Using data from the Team Clinic randomized clinical trial, we compared the cost of SMAs versus usual care on middle school-aged participants with T1D.

METHODS

The study protocol was reviewed by the University of Chicago Institutional Review Board and determined to be non-human subject research.

Study Design and Overview

In this prospective pragmatic trial, participants aged 11–13 years with T1D were randomized into the Team Clinic intervention or usual care (control) groups at the Barbara Davis Center for Diabetes. Each subject was assessed at baseline, 3, and 6 months. We collected adolescents’ self-reported health care use for the prior 3 months at each visit. We also collected their visit times. Process measures included clinic attendance, insulin pump use, and continuous glucose monitoring (CGM). The clinical outcomes of interest included HbA1c, glucose test strip use, numbers of severe hypoglycemic and hyperglycemic events, blood pressure, and the number of study visits.

We adopted a formal health care sector perspective for a 6-month within-trial cost analysis. Per the Consolidated Health Economic Reporting Standards recommendation, we have provided a reporting checklist in Supplementary Table 1.

Team Clinic Intervention

The intervention consisted of (1) an individual appointment with a health care provider (nurse practitioner, physician assistant, and/or physician) before and after a group appointment; and (2) a group appointment with 3–5 adolescents facilitated by a certified diabetes educator, nurse, dietitian, or social worker for approximately 45–60 min. Although adolescents participated in the group component of Team Clinic, their caregivers participated in a separate, facilitator-led group session. The main trial paper provides a detailed description of the intervention, study design, and patient population (Berget et al., 2017; Majidi et al., 2021; McClain et al., 2018; Raymond et al., 2015).

Costs

The 6-month within-trial total costs included all direct costs associated with trial staff time as part of the study, health care use outside the study, and device use (CGM and/or pump). All cost assumptions are provided in Supplementary Table 2. All costs are expressed in 2016 U.S. dollars, as the study was mainly conducted in 2016.

We calculated costs by multiplying the U.S. Bureau of Labor Statistics median hourly wages (or prices per service) by hours spent (or the number of services used) in 6 months. Device costs: pump cost was estimated to be $12.81/day and includes costs from its three components (i.e., insulin pump, pump infusion set, and pump cartridge/syringe [Supplementary Table 3]). CGM cost was estimated to be $15.38/day, including costs from its three components (i.e., sensor, receiver, and transmitter [Supplementary Table 4]). These prices are estimated average allowable prices in the U.S. marketplace. Nondevice medical care costs: health care service use costs included routine office visits that occurred outside of the trial, after-hours clinic visits (urgent care), 911 calls, emergency room visits, hospitalizations, and daily test strip use for the prior 3 months. Direct clinical personnel costs: we included all staff time devoted to intervention or routine individual visits and excluded research time. In the Team Clinic group, provider time, other provider time, and total staff time were considered, including providers and nonprovider staff time for adolescents and caregivers.

Analysis Methods

We applied the intent-to-treat principle to all within-trial analyses. Costs, health care use and other outcomes were summarized at baseline and 6 months. The Wilcoxon rank-sum test compared the two groups regarding continuous outcomes, including costs. The Fisher exact test was used for group comparison in each categorical outcome. We used linear mixed models to test the effects of treatment, time, and their interaction, respectively, to model the continuous clinical outcomes repeatedly measured over time. In each model, we adjusted its baseline outcome and potential covariates (age, sex, and duration of T1D). To assess the homogeneity of the treatment effect, we also conducted subgroup analyses per baseline HbA1c level of 9.0% as a cutoff to investigate important subpopulations, which may differ in other important clinical outcomes.

We also made a baseline cost adjustment to account for the baseline imbalance in costs, as indicated by (van Asselt et al., 2009). We used the bootstrap method to calculate the mean difference-in-difference (DID; i.e., group difference in change costs from baseline) and its 95% confidence interval (CI) for group comparison.

We used the nearest neighbor imputation algorithm to fill in missing data such as health care use data because of multiple assessments in a short period. All tests were two-sided, and p values < .05 were considered significant. Analyses were conducted with SAS version 9.4 (SAS Institute, Cary, NC).

RESULTS

Eighty-six adolescents were enrolled and randomized into the study (44 participants in the intervention group and 42 in the control group). The self-reported surveys were completed at 6 months by 37 (84%) participants in the intervention group and 35 (83%) in the control group. The main baseline characteristics, including basic demographics, clinical characteristics, and health care use for the last 3 months, were summarized in Table 1. The two groups were generally balanced, except that the intervention group tended to have more pump use (89% vs. 74%; p = .07) and less CGM use (9% vs. 26%; p = .048) than the control group.

TABLE 1.

Baseline characteristics of the study populations

Characteristics	Intervention, n = 44	Control, n = 42	P Value
Demographics
Gender, female, n (%)	20 (45)	24 (57)	.29
Race, n (%)			.27
White	31 (70)	33 (78)
Black	2 (5)	4 (10)
Other	11 (25)	5 (12)
Ethnicity, n (%)			1.00
Hispanic/Latino	5 (11)	4 (10)
Others	39 (89)	38 (90)
Age, years			.18
Mean ± SD	11.9 ± 0.9	11.6 ± 0.8
Range	10–14	10–13
Insurance, n (%)			.95
Private	33 (75)	31 (74)
Public, Medicaid	10 (23)	9 (21)
Military, Tricare	1 (2)	1 (2)
None	0 (0)	1 (2)
Health self-care at baseline for the last 3 months
Daily strip test			.60
Mean ± SD	5.8 ± 2.2	5.7 ± 2.6
Range	2.1–11.4	2.4–15
Pump use, yes, n (%)	39 (89)	31 (74)	.07
Continuous glucose monitoring use, yes, n (%)	4 (9)	11 (26)	.048
Clinical characteristics at baseline
Hemoglobin A1c			.80
Mean ± SD	8.8 ± 1.2	8.8 ± 1.1
Range	6.6–11.6	6.2–11.4
Body mass index			.15
Mean ± SD	19.1 ± 2.2	20.2 ± 3.1
Range	15.4–26.4	15.7–28.3
Systolic blood pressure			.85
Mean ± SD	104.2–6.8	104.3 ± 9.5
Range	91–120	77–120
Diastolic blood pressure			.69
Mean ± SD	66.0 ± 5.3	66.3 ± 7.1
Range	55–77	46–81
No. of patients having severe hypoglycemia in the last 6 months
n (%)	0	0 (0)	1.00
No. of patients having severe hyperglycemia in the last 6 months
n (%)	2 (5)	3 (7)	.42
Type 1 diabetes duration, year			.53
Mean ± SD	4.5 ± 2.9	4.8 ± 2.7
Range	0.9–10.5	1.0–11.0
Smoke, yes, n (%)	0 (0)	0 (0)	1.00
Albuminuria, yes, n (%)	0 (0)	0 (0)	1.00

During the trial, mean 6-month total costs for the intervention group (Table 2) were $138 lower per subject than in the control group (intervention: $3,343 ± 1,137 vs. control: $3,481 ± 1,860), but this difference was not statistically significant (p = .84). The total costs were attributable to their similar health care use (all p > .57) and strip test use (p = .33). Combined device costs (pump use and CGM) were also similar. The intervention group had higher costs associated with trial staff effort (intervention: $227 vs. control: $76; p < .01). This was because the intervention group had more clinic visits (2.41 ± 0.62 vs. 1.52 ± 1.19; < .01), longer staff time (median: 0.36 vs. 0), and similar provider time (median: 0.67 vs. 0.68; p = .18), than the control group.

TABLE 2.

Within-trial results in costs, health care use, health self-care, staff time, and clinical outcomes

Outcomes	Intervention, n = 44		Control, n = 42		p Valuea
	Mean ± SD or n (%)	Median (IQR or Range)	Mean ± SD or n (%)	Median (IQR or Range)
Per patient costs, $
Nonstudy health care use	31 ± 104	0 (0–0)	18 ± 77	0 (0–0)	.52
Trial staff for intervention/control	227 ± 71	234 (183–253)	76 ± 64	96(0–108)	< .01
Strip test use	943 ± 414	881 (624–1,148)	1,024 ± 464	978(643–1,212)	.58
Continuous glucose monitoring use	176 ± 597	0 (0–0)	800 ± 1,140	0 (0–2,584)	< .01
Pump use	1,859 ± 747	2,152(2,152–2,152)	1,563 ± 957	2,152(0–2,152)	.10
Total costs	3,343 ± 1,137	3,407(3,046–3,781)	3,481 ± 1,860	3,279(2,148–4,727)	.84
Health care use because ofT1D for 6 months
911 calls	0±0	0 (0–0)	0±0	0 (0–0)	1.00
Ambulance use	0±0	0 (0–0)	0±0	0 (0–0)	1.00
Emergency department visits	0.05 ± 0.21	0 (0–2)	0.02 ± 0.15	0(0–1)	.57
Urgent care visits	0±0	0 (0–0)	0±0	0 (0–0)	1.00
Hospitalizations	0±0	0 (0–0)	0±0	0 (0–0)	1.00
Nonstudy outpatient visits	0.09 ± 0.29	0(0–1)	0.07 ± 0.26	0 (0–2)	.72
Health self-care for 6 months
No. of clinic visits	2.41 ± 0.62	2(1–3)	1.52 ± 1.19	2 (0–3)	< .01
Daily strip tests	4.92 ± 2.26	4.7(1.5–11)	5.59 ± 2.79	5.1 (1.7–16)	.33
Pump use, yes	38 (86)		30(71)		.11b
Continuous glucose monitoring use, yes	2(5)		14 (33)		< .01b
Staff time per patient per visit
Total staff time, hrc	2.42 ± 0.82	2.13(1.58–5.17)	0.74 ± 0.12	0.68(0.67–1.18)	< .01
Clinician time, hr	0.72 ± 0.11	0.67(0.67–1.22)	0.74 ± 0.12	0.68(0.67–1.18)	.18
Other clinician time, hrd	–	–	0.02 ± 0.05	0(0–0.11)
Clinical outcomes at 6 months
Hemoglobin A1c at 6 monthse	8.99 ± 1.25	9.0(6.5–11.7)	8.96 ± 1.69	8.8(6.8–15)	.70f
Systolic blood pressuree	105.7 ± 8.4	107(90–122)	106.6 ± 8.5	105.5(93–125)	.06f
No. of patients having severe hypoglycemic events, %	0		0 ± 0		1.00b
No. of patients having severe hypergycemic events, %	1 ± 2		0 ± 0		1.00b
Subgroup analyses
In the subgroup with high baseline Hemoglobin A1c(≥ 9.0%; n = 41)
Hemoglobin A1cat6 monthse	9.92 ± 0.89	9.85(8.4–11.7)	10.07 ± 1.64	9.6(8.4–15)	.49f
Systolic blood pressure	106.3 ± 9.2	107.5(90–122)	108.2 ± 8.8	107(93–125)	.08f
Daily strip tests	4.20 ± 1.80	4.1 (1.5–7.0)	5.26 ± 3.15	4.6(2.5–16)	.28f
No. of clinical visits	2.38 ± 0.67	2(1–3)	1.7 ± 1.1	2 (0–3)	.06
Pump use—yes, %	18 (86)		13 (65)		.16b
Continuous glucose monitoring use—yes, %	1 (5)		4 (20)		.18b
In the subgroupwith low baseline HbA1c (< 9.0%; n = 45)
Hemoglobin A1cat6 monthse	8.00 ± 0.68	8.0 (6.5–9.1)	7.86 ± 0.77	7.9 (6.8–9.6)	.03f
Systolic blood pressure	105.2 ± 7.7	106(93–120)	104.9 ± 8.1	103(94–123)	.57f
Daily strip tests	5.75 ± 2.50	6.0(2.7–11.2)	5.90 ±2.45	5.4(1.7–10.3)	.09f
No. of clinical visits	2.43 ± 0.59	2(1–3)	1.36 ± 1.26	2 (0–3)	<.01
Pump use, yes	20 (87)		17(77)		.46b
Continuous glucose monitoring use, yes	1 (4)		10(45)		<.01b

Note. T1D, type 1 diabetes.

IQR used to summarize costs data and Range used to summarize the other outcomes.

^aThe default statistical method was Wilcoxon’s test.

^bFisher exact test was used to compare the groups.

^cTotal staff time includes (1) patients’ group appointments with a facilitator and (2) parents/caregivers’ group discussion with a facilitator. Time was calculated by dividing the total amount of time over the total number of patients at a group visit.

^dIn addition to receiving usual care (seeing a nurse and primary care provider), some in the usual care group might see additional providers such as a psychologist, dietitian, and social worker.

^eBased on observed data at 6 months, basic descriptive statistics were provided per group, but the p value was from a linear mixed model (Footnote b).

^fA linear mixed model was used to compare the groups, adjusting its baseline outcome and other covariates such as age, gender, and duration of diabetes.

The p value is the smaller p value for either the main group effect or the interaction effect between group and visit.

To adjust for the imbalanced nonnormal baseline costs (shown in Supplementary Table 5), we calculated the mean DID costs and its 95% CI using the bootstrap method (Table 3). The intervention group had lower mean DID total costs (−$490) than the control group, but the difference was not statistically significant (95% CI [–$2,805 to $833]). The results by the bootstrap method were consistent with the actual within-trial costs results in Table 2.

TABLE 3.

Comparison of study groups in health care use and costs by the bootstrap method

Variables	Mean DID (intervention – control)		95% CI of DID
	By original dataset	By bootstrap	Lower	Upper
Health care use for 6 months before and after intervention
911 calls	0.00	0.00	0.00	0.00
Ambulance use	0.00	0.00	0.00	0.00
Emergency department visits	0.10	0.10	−0.05	0.21
Urgent care visits	0.00	0.00	0.00	0.00
Hospitalizations	−0.02	−0.02	− 0.16	0.08
Nonstudy outpatient visits	0.29	0.26	0.01	0.48
Per patient costs, $
Total costs	−503	−490	−2,805	833
Trial staff for intervention/control	150	153	124	178
Nonstudy health care use	−281	−279	−1,923	1994
Strip test use	−101	−101	−291	82
Pump use	−23	−24	−221	319
Continuous glucose monitoring use	−182	−185	−424	−32

Note. CI, confidence interval; DID, difference-in-difference.

In addition to the key economic evaluation results, we also evaluated several important clinical outcomes (Table 2). No significant differences were observed in HbA1c, systolic blood pressure, and the number of severe acute events (hypoglycemia and hyperglycemia).

To investigate consistency and/or potential heterogeneity of the treatment effect across subgroups, we conducted subgroup analyses on the basis of the baseline HbA1c with a cutoff of 9.0% (Table 2). Among participants with high baseline HbA1c (n = 41), compared with the control group, the intervention group had no significant improvement in HbA1c (p = .49; Supplementary Figure), systolic blood pressure (p = .08), and other self-care measures. Among participants with a low baseline HbA1c subgroup (n = 45), compared with the control group, the intervention group had slightly worse HbA1c (p = .03; Supplementary Figure). However, this HbA1c difference at 6 months (intervention: 8.0% vs. control: 7.86%; Table 2) were not > 0.5%, which is routinely considered clinically significant. In addition, the intervention group had more study visits (p < .01) and less CGM use (p < .01) than the control group. All the other results of the subgroup analyses are consistent with the findings of the main analyses.

DISCUSSION

Adolescence is a difficult period for young people with T1D, both in clinical and psychological terms (Łuczyński et al., 2019), and the Team Clinic model was designed to address the complex needs of this population. Although SMAs are slowly increasing in popularity, there are very few randomized controlled trials in a young adolescent population with T1D. Given the rising costs of diabetes care (ADA, 2018), economic evaluation is critical to their long-term adoption by health care clinicians, insurers, and payers. To our knowledge, this study is the first economic evaluation of the SMAs care model in English-preferring adolescents with T1D based on a formal health care sector perspective.

We found no statistical difference in health care use and total costs between the study groups during the randomized trial because of huge data variability. Despite no statistical difference in total costs, from the economic perspective, Team Clinic saved $490 net costs per subject for 6 months after adjusting for imbalanced baseline costs. The total net per subject savings has already accounted for intervention implementation spending ($150). Furthermore, for 6 months, Team Clinic maintained the clinical attendance every 3 months as recommended by the ADA. The intervention group had about one visit more than the usual care group. However, there were no significant improvements in glucose control, pump use, or CGM use. In addition, despite the Team Clinic increasing staff time, clinician time was very similar between groups.

Our estimates of costs for the trial participants were very similar to the usual care costs reported in other studies. The ADA recently reported that the average per-person annual direct cost for children aged < 18 years with T1D or T2D in 2017 was $7,510, which is equivalent to $3,755 for 6 months (ADA, 2018). Our study’s 6-month direct costs of $3,343 (Team Clinic care) and $3,481 (usual care) align with the ADA’s cost estimate. The ADA’s total direct cost estimate includes costs associated with health care services, ambulance services, home health, diabetes supplies, and other equipment and supplies. Our study considered all these key components for estimating our total direct costs. A recent cost-evaluation study found that SMAs for Latino children with T1D yielded 1-year total per subject savings of > $2,000 (Gold et al., 2021). However, their study was based on a comparison with matched control subjects, and their study population had a wide age range (i.e., from 1 to 20 years). In addition, their total cost savings were based on net costs, not on statistical testing.

We found the heterogeneity of the intervention effect, which was varied by baseline HbA1c. Therefore, we conducted subgroup analyses. Among those whose baseline HbA1c > 9%, Team Clinic maintained their glucose control level, that is, model-adjusted HbA1c remaining at ~9.8% over the study, whereas the usual care group had worsened HbA1c from ~9.8% at baseline to 10% at 6 months. However, the difference between the two groups was not significantly different and may not be clinically meaningful.

Our study has limitations. First, the within-trial health care use and total costs may be underestimated because we relied on participants’ self-reported health care use, and insulin dose data were not collected during the trial. We are reassured that our total costs appear similar to those from other studies. Second, as recommended, conducting the cost-effectiveness of delivering SMAs is needed to fill the gaps in the literature (Ridge, 2012). The study lacked the quality of life data, and thus we were unable to conduct a cost-effectiveness analysis. The study also lacked data on work performance and self-management time for total indirect cost analysis. Both will become part of our future work. Third, although the Team Clinic had been continuing at the Barbara Davis Center, the participants’ self-reported data for conducting research were only collected for up to 6 months. Because behavioral changes, and eventual impact, often take time, a 6-month follow-up may not be long enough to capture the potential effect of SMAs. Fourth, although SMAs are becoming popular, additional research is needed to determine the optimal group size, format, and curriculum for group visits. Last, the Team Clinic model is an in-person SMA model. During this worldwide transition to telehealth, in-person SMAs may not always be available. A study of virtual SMAs for the adolescent population with T1D is currently underway, which would be a safe option during our global pandemic and has the potential to have a significantly different impact. Finally, our study may be underpowered because of missing data (≤ 17%), although the study team attempted to collect patient survey data by using electronic and alternative forms of communication (i.e., outside of the study visits). We used the nearest neighbor algorithm for data imputation.

The Team Clinic care model has been demonstrated to be a viable and sustainable option to strengthen the foundation of primary health care and provide better management for young adolescents with T1D. With open general discussions in a group setting, young adolescents potentially receive more education compared with during standard visits by hearing from their peers (Raymond et al., 2015), which may result in higher acceptance of suggestions from clinicians and staff (Clancy, Huang, Okonofua, Yeager, & Magruder, 2007). In addition, showing other group members’ commitment to their management may contribute to diminished anxiety (Clancy et al., 2007). The Team Clinic also created a platform for group discussions among parents and/or staff to understand the development of their adolescents. By contrast, the Team Clinic satisfied clinicians and staff more efficiently in delivering ongoing diabetes support and education (Berget et al., 2017). It can be manageable and completed without funding, donations, or additional staff (Raymond et al., 2015). The Team Clinic shall also be generalizable to other age groups with other chronic conditions.

Conclusions

Based on the single-center randomized trial in the young adolescent population with T1D, the Team Clinic model, an in-person, shared medical appointment, may maintain diabetes care use, increase diabetes visit frequency, and improve psychosocial outcomes with no cost increase in a short-time period. Additional trials with larger participant numbers, longer-term follow-ups, and virtual shared medical visit models are needed.