Changes among Patients with Co-Occurring Diabetes and Serious Mental Illness Receiving Registry-Managed Care Coordination and Personalized Diabetes Education

Abstract

Objective:

Longitudinal changes in health outcomes of patients with serious mental illness and co-occurring diabetes were examined following introduction of an intervention involving electronic disease management, care coordination, and personalized patient education.

Methods:

This observational cohort study included 179 patients with serious mental illness and diabetes mellitus type 2 at a behavioral health home in Chicago. The intervention employed a care coordinator who used a diabetes registry to integrate services along with personalized patient education. Outcomes included glucose, lipid, and blood pressure as assessed by A1C, LDL, triglycerides, and systolic/diastolic values from electronic medical records, and completion of specialty visits confirmed with ophthalmologists and podiatrists. Interrupted time series segmented random regression models tested for level changes in the 8 study quarters following intervention implementation compared to 8 pre-implementation study quarters, controlling for clinic site and pre-intervention secular trends.

Results:

There were significant level declines in glucose, lipid, and blood pressure indicators post-implementation. In addition, completed optometry referrals increased by 44% and completed podiatry referrals increased by 60%.

Conclusions:

Significant improvement in medical outcomes was found among behavioral health home patients with co-morbid diabetes and mental illness after introduction of a multicomponent care coordination intervention, regardless of which clinic they attended.

The prevalence of diabetes among adults with serious mental illness is two- to three-fold higher than the general population (1,2), yet only one-third are diagnosed and treated (3), and care is often sub-par (4,5). While the benefits of integrated primary and behavioral healthcare for this population are acknowledged (6,7), less is known about how to integrate specialized diabetes care (8). Integration strategies tested in the general population include care coordination (9), use of electronic registries (10), and personalized diabetes self-management education (11). This study examined changes in patient-level outcomes among adults with diabetes and serious mental illness following introduction of an intervention incorporating these strategies.

Care coordination involves connecting patients with healthcare providers, monitoring their treatment plans, educating them about their conditions, and sharing information to enhance effective care (12,13). Studies show that introducing dedicated staff who coordinate healthcare for people with serious mental illness significantly improves the quality and outcomes of primary care (14). Recipients in one study of care coordination utilized more preventive services and more evidence-based cardio-metabolic care (15), while those in another study had significantly higher mean physical health scores after one year (16). Despite this promising evidence, care coordination for diabetes management in this population is rare.

Disease registries are electronic databases containing medical information on patients with specific types of chronic illnesses that is used to facilitate delivery of evidence-based care (17). Registries notify providers of abnormal test results and missed appointments, track the progress of high-risk patients, and enable health outcomes management at individual and clinic levels (18,19). One study of a diabetes registry introduced to primary care clinics (20) found significant reductions in glycosylated hemoglobin test (A1c) levels, while another study of a diabetes registry adopted by primary care practices (21) found significant improvement in A1c, low-density lipoprotein (LDL), and blood pressure outcomes. These studies suggest the promise of registries in the delivery of integrated care for patients with co-occurring diabetes and serious mental illness.

Diabetes patient education is designed to increase understanding of the disease and enhance skills and motivation for successful self-management (22). Effective delivery involves personalizing the information to patients’ specific medical situations, and incorporating their needs, goals, culture, and life experiences (23). Research shows that people with co-occurring diabetes and serious mental illness seldom receive diabetes education (5,24,25), yet evidence indicates its effectiveness for this population (26,27). However, there are few diabetes education programs geared toward the specific needs of people with mental illness.

Medical homes deliver primary care that is patient-centered, comprehensive, team-based, coordinated, and focused on quality and safety (28). Behavioral health homes integrate this kind of medical care with services for patients with mental illness (29,30). One type of behavioral health home involves co-locating primary care providers in community behavioral healthcare settings (31,32), with advanced practice nurses often delivering care (33,34). Patients report high satisfaction with co-location (35,36), while studies find significantly improved A1C, blood pressure, and LDL levels in patients using this model (37,38).

This study examined the medical outcomes of patients with type 2 diabetes mellitus at a behavioral health home with co-located primary care providers following introduction of a practice enhancement program including care coordination, registry-managed care, and personalized diabetes education. We hypothesized: 1) that there would be significant improvement over time in patients’ A1c, LDL, triglyceride, and blood pressure values, and 2) that significant increases would occur in proportions receiving dilated eye and comprehensive foot examinations.

Methods

Research setting.

The behavioral health home consisted of two primary care clinics operated by the University of Illinois at Chicago (UIC) College of Nursing that were co-located in an outpatient mental health program on the north and south sides of Chicago called Thresholds (39). Serving over 900 patients annually (40), each clinic was staffed by 2–3 advanced practice registered nurses who are authorized to provide primary care services without oversight of a medical doctor in the state of Illinois. A medical assistant performed clerical duties such as reception and scheduling, and clinical functions such as taking vital signs and blood draws (41). Clinics used a team approach to care rather than individual patient assignments.

Patients with diabetes were asked to keep daily logs of their blood sugar levels but this occurred inconsistently. They often missed appointments or failed to follow through on prescribed treatments and referrals. Despite co-location, the electronic health record (EHR) used by clinic staff was separate from that used by the mental health program, and the former required specialized programming to generate automated reports for patient tracking and clinic management. Thus, a registry was needed to bridge this gap. The latest audit prior to study inauguration found that 24% of diabetes patients with co-occurring hypertension had uncontrolled blood pressure, and 33% did not meet the clinics’ standard for glycemic control. This presented an ideal situation for introduction of care coordination using an electronic registry, dedicated care manager, and personalized educational materials.

Study population.

The study included 179 clinic patients diagnosed with diabetes mellitus type 2. Time since diagnosis ranged from 6 months to 20 years, with high rates of co-occurring hypertension, coronary artery disease, kidney disease, and obesity. Most were covered by Medicaid (58%) or Medicare (32%), 5% had private insurance, and 5% were uninsured. At study inauguration, 88.8% of clinic patients were prescribed metformin, 65.1% ACE inhibitors, 27.5% insulin, 31.8% Glitazones, and 19.0% Sulfonylurea. Study procedures were approved by the institutional review board of UIC.

Intervention.

The registry contained EHR information specific to diabetes management, including medical and mental health diagnoses, laboratory results, appointment data, referrals for specialty care, and lifestyle factors (e.g., smoking, diet). Registry data were used to generate specialized reports for patients, providers, and administrators. For providers, patient-specific reports generated before each visit summarized recent lab values and test results, and flagged out-of-range values for further attention. For patients, user-friendly “report cards” showed trends over time in A1c, LDL, and blood pressure, and included reminders for upcoming eye or foot exams. Reports for administrators summarized patient outcomes and appointment attendance by clinic and across clinics.

A single care coordinator was based 50% time at each clinic, tasked with linking behavioral and primary care providers via email and telephone, enhancing communication between the two groups, and sorting out glitches in appointment scheduling and transportation.. She also populated the registry with lab values and appointment information, and generated all reports. Direct patient contact included appointment reminders by telephone, accompanying patients to specialty care visits, discussing report card results, and providing diabetes education. The coordinator also facilitated specialty care by negotiating specific blocks of appointment time with university outpatient eye and podiatry clinics, arranging transportation with mental health case managers, and scheduling appointments with eye and foot specialists willing to conduct examinations onsite.

The coordinator worked with primary care providers, mental health program staff, people with lived experience of diabetes and mental illness, medical students, and research staff to develop patient self-management education materials. These were packaged as an online Diabetes Education Toolkit of didactic information linked to care standards and related podcasts for use with patients in the clinic. A central part of the Toolkit was a “library” of one-page information sheets. These were written at grade school level to accommodate low levels of health literacy and numeracy (the patient’s ability to interpret and act on quantitative and probabilistic health information in making effective health decisions) commonly found among people with serious mental illness (42,43,44). The library covered a variety of topics linked to treatment regimens, co-occurring conditions, and strategies for diabetes self-management. To personalize education for each patient, the coordinator selected topics based on out-of-range lab values, poor health indicators, or clinical goals for that patient. Attaching educational materials to patient-specific reports prompted medical providers to review them with patients and send them home for sharing with family and other supporters. The same materials were sent by the coordinator to case managers with reminders to review them at the patient’s next behavioral health visit. Behavioral healthcare provider participation included reviewing and reinforcing diabetes education and treatment regimens, providing transportation to specialty appointments, and sharing treatment plans with the care coordinator and clinic medical staff.

Study design.

Due to the highly vulnerable nature of the patient population, random assignment was not considered practical or ethical. Instead, we used a one-group pre-post interrupted time series design, one of the strongest quasi-experimental research designs (45). For this approach, data on all participants are collected at equally spaced time points (in this case quarterly) before and after an intervention is implemented (46). The main objective is to examine whether data patterns observed post-implementation are different from those observed pre-implementation using segmented regression analysis (45,47). The pre-intervention segment acts as a control for secular trends in outcomes that may occur unrelated to the intervention (45,48).

Measures.

Registry data were extracted for all patients with Type 2 diabetes served in the 8 quarters prior to intervention implementation (April 1, 2010 through March 31, 2012) who also had one or more lab values in the 8 quarters following implementation (April 1, 2012 through March 31, 2014), with extraction ending March 31, 2014. Values were grouped by study quarter, with the last measure in the quarter used when there were multiple measures. Values included A1c, LDL, triglycerides, and systolic and diastolic blood pressure. Occurrence of eye and foot examinations was verified with patients’ podiatry and ophthalmology providers and calculated as the proportion completed annually per American Diabetes Association guidelines (49).

Analysis.

Standard descriptive analysis was conducted to examine all variables’ distributions and measures of central tendency. The characteristics of each clinic’s patients were compared and tested for significant differences. Next, paired t-tests were computed to assess change following intervention implementation using the approach of Fesseha and colleagues (50) in which baseline A1C was defined as the final measure during the pre-implementation period and compared to the nadir A1C defined as the single lowest post-implementation measurement. Use of nadir LDL, triglyceride, and blood pressure values also followed published analyses (51,52,53). Multivariable interrupted time series segmented random effects regression models were used to examine changes in values over 16 quarters of data (8 pre- and 8-post implementation). The models accounted for autocorrelation of repeated measures using first-order autoregressive covariance structures, and controlled for study site to adjust for differences in clinic populations, and for number of outcome measurements. Potential seasonal effects on outcomes were accounted for by use of 4 years of data with equal seasonal exposures before and after the intervention (41). We posited an impact model that would show no significant pre-intervention trends in outcome measures and a statistically significant post intervention level change in outcomes. To measure change in specialty eye and foot care we calculated the percentage completing each examination during the 12-month period at the end of the first 8 quarters (April 1, 2011-March 31, 2012) and the end of the second 8 quarters (April 1, 2013-March 31, 2014). Analyses were conducted in IBM SPSS Statistics 25 and SAS 9.4.

Results

Table 1 presents the background characteristics of 179 patients in total and by clinic location. Around two-thirds were male, averaging 51.2 years of age. Around a third (30.9%) did not complete high school. Most had schizophrenia or schizoaffective disorder (62.0%), 19.0% had bipolar disorder, and 19.0% had major depressive disorder. Clinic patient populations were highly similar except for significantly higher proportions of African American patients at the south location, and Whites at the north, reflecting the racial makeup of their surrounding communities. Comparing this group of 179 with the 48 patients excluded because they lacked lab values in the post-implementation period, there were no significant differences (p<.05) in age, gender, race, education, or psychiatric diagnosis.

Table 1:

Baseline characteristics of study participants with co-occurring diabetes and serious mental illness (N=179)

Characteristics	Total Population (N=179)		North Clinic (N=88)		South Clinic (N=91)
	N	%	N	%	N	%
Gender
Male	119	66.4	60	68.2	59	64.8
Female	60	33.5	28	31.8	32	35.2
Race
Black	107*	61.8	40	46.5	67	77.0
White	49*	28.3	36	40.9	13	14.9
Hispanic	5	2.9	1	1.2	4	4.6
Asian	4	2.3	3	3.5	1	1.1
Other	8	4.6	6	7.0	2	2.3
Age (M±SD)	51.2± 9.8		52.5± 9.6		50.0± 9.8
Education
< High School	48	30.9	18	25.3	30	35.8
High School Graduate	65	41.9	36	50.7	29	34.5
Some College	30	19.4	11	12.5	19	22.7
College Graduate	12	7.8	6	8.4	6	7.2
Diagnosis
Schizoaffective Disorder	43	24.0	16	18.2	27	29.7
Schizophrenia	68	38.0	37	42.0	31	34.1
Bipolar Disorder	34	19.0	16	18.2	18	19.8
Depressive Disorder	34	19.0	19	21.6	15	16.5
Baseline Medical Outcomes (M, SD)
A1c	7.36	2.20	7.24	1.90	7.47	2.46
LDL	96.15	34.06	99.98	33.15	93.70	36.12
Triglycerides	134.74*	81.25	154.46	91.81	120.83	72.33
Blood Pressure
Systolic	123.96	14.84	125.94	16.30	120.64	13.36
Diastolic	81.35	8.53	80.59	9.12	82.64	8.47

^*p < .05 in chi-square or independent t-tests

Table 1 also presents patient medical outcomes, showing that the mean A1C level was 7.4±2.2 which exceeded the recommended level of <7 in the 2010 ADA standards in effect at the time of baseline (49,54). The average LDL level was 96.2±34.1 which was below the recommended level of <100 in the standards. The average triglycerides level was 134.7±81.2 which was below the recommended level of <150. The average systolic blood pressure was 124.0±14.8 and the average diastolic was 81.4±8.5 which exceeded the recommended level at baseline of <130/80 for the diastolic value but not the systolic. The only significant difference by clinic concerned triglycerides which were higher at the north clinic (154.5±91.8), exceeding the recommended level.

Table 2 presents paired t-tests of the change in clinical outcomes between pre- and post-implementation periods. Compared to their final pre-intervention value, individuals’ nadir measurements after intervention implementation were significantly lower for A1C (average declineSD = −0.68±1.16, t=7.55, df=168, p<0.001), for LDL (average decline±SD = −9.31±24.73, t=4.89, df=168, p<0.001), for triglycerides (average decline±SD) = −16.79±56.35, t=3.73, df=156, p<0.001), and for systolic and diastolic blood pressure (average systolic decline±SD= −12.65±10.54, t=15.59, df=168, p<0.001; average diastolic decline±SD= −9.20±7.14, t=16.75, df=168, p<0.001).

Table 2.

Changes in diabetes-related medical outcome measures from pre-intervention to post-intervention

Medical Outcomes	Average change between pre- and post-implementation
	M	SD	Test Statistica	Df	P
A1C	−0.68	1.16	t=7.55	168	<.001
LDL	−9.31	24.73	t=4.89	168	<.001
Triglycerides	−16.79	56.35	t=3.73	156	<.001
Blood pressure: Systolic	−12.65	10.54	t=15.59	168	<.001
Blood pressure: Diastolic	−9.20	7.14	t=16.75	168	<.001

^aPaired t-tests compare the final value from the pre-implementation period with the nadir value from the post-intervention period

Table 3 presents the results of multivariable interrupted time series random regression analyses. Significant post-implementation level decline was found in A1C (−0.75±0.35, p=0.032), LDL (−19.75±8.07, p=0.015), triglycerides (−47.88±22.75, p=0.037), and blood pressure >=130/80 (−2.07±0.71, p=0.004). There were no significant secular trends occurring in outcomes prior to the intervention with the exception of blood pressure which showed a small pre-intervention decline (−0.12±0.04, p=0.001), and there were no significant differences in utcomes associated with study site except for blood pressure which was lower at the north than south clinic (−0.74±0.24, p=0.002).

Table 3.

Interrupted time series random effects regression models of changes across 16 quarters in medical outcomes post-implementation, controlling for pre-implementation trends, clinic site, and number of observations among patients

Medical Outcomes	Estimate	SE	p	95% CI
A1C
Intercept	7.01	0.16	<.001	6.70, 7.31
Sitea	−0.20	0.18	.267	−0.55, 0.15
Pre-Intervention Secular Trend	−0.03	0.02	.117	−0.07, 0.01
Post-Intervention Level Change	−0.75	0.35	.032	−1.43, −0.06
Number of observations	0.07	0.04	.092	−0.12, 0.16
LDL
Intercept	97.61	3.71	<.001	90.32, 104.91
Sitea	−1.09	4.09	.791	−9.17, 7.00
Pre-Intervention Secular Trend	−0.51	0.49	.306	−1.48, 0.47
Post-Intervention Level Change	−19.75	8.07	.015	−35.62, −3.89
Number of observations	1.03	1.32	.433	−1.55, 3.62
Triglycerides
Intercept	130.34	9.82	<.001	110.98, 149.69
Sitea	14.30	9.62	.140	−4.77, 33.37
Pre-Intervention Secular Trend	−0.97	1.55	.534	−4.05, 2.11
Post-Intervention Level Change	−47.88	22.75	.037	−92.83, −2.93
Number of observations	3.60	4.49	.422	−5.23, 12.44
Blood Pressureb
Intercept	−0.87	0.24	<.001	−1.34, −0.41
Sitea	−0.74	0.24	.002	−1.22, −0.27
Pre-Intervention Secular Trend	−0.12	0.04	.001	−0.19, −0.05
Post-Intervention Level Change	−2.07	0.71	.004	−3.46, −0.68
Number of observations	0.02	0.08	.822	−0.13, 0.17

^aClinic site 1= North clinic, 0 = South clinic;

^bBlood Pressure, BP ≥ 130/80 = 1, < 130/80 = 0

Finally, we examined changes in the proportions of patients who completed specialty care appointments for monofilament foot and dilated eye examinations. Results (not shown) revealed significant increases in the proportion completing eye exams from 23.1% (N=40 of 173) pre implementation to 33.5% post-implementation (N=58 of 173) (χ²=4.46, N=173, df=1, p=0.023), and in the proportion completing foot examinations from 17.2% (N=30 of 174) to 27.6% (N=48 of 174) (χ²=5.35, N=174, df=1, p=0.014).

Discussion

Following introduction of a multicomponent intervention designed to improve patient outcomes and adherence to diabetes care standards, significant improvement was noted in glucose, lipid, and blood pressure indicators. In addition, patient completion of optometry referrals increased by 45% and completion of podiatry referrals increased by 60%. Ours is the first study to show that this combination of evidence-based intervention components, previously found effective for other populations, was associated with improvement in a population of people with co-occurring diabetes and serious mental illnesses.

More limited success was achieved with specialty care outcomes despite extensive planning with the directors of university eye and podiatry clinics. Accommodations included setting aside special dates and times for appointments, provision of transportation, telephone reminders, and support from the project’s care coordinator who “hung out” with patients in clinic waiting areas providing magazines and healthy snacks. This speaks to persistent barriers to receiving care outside the health home when patients are required to travel to unfamiliar treatment locations. The patient appointment no-show rate for the on-site clinics was a noteworthy 24%, so keeping off-site appointments at university outpatient clinics remained particularly challenging.

The care coordinator devoted considerable effort to engaging behavioral health staff in an understanding of each patient’s diabetes management goals. In particular, sharing patients’ report cards and individualized education materials with mental health staff, along with reminders to discuss them at the next meeting, helped provide a consistent message to patients about self-managing their diabetes. The report card motivated patients by visually illustrating how their medical test results were improving, remaining stable, or worsening over time. Dedicating the time of the care manager to maintaining the registry and generating specialized reports helped fill a gap in information that had been previously identified as problematic by the clinics’ nursing staff and management.

The Diabetes Education Toolkit was used in several ways. While primary care providers were well-educated about diabetes, they expressed appreciation for being able to access a variety of handouts written at a level their patients could understand. Since behavioral health staff were not well-educated about diabetes or diabetes self-management strategies, they used the Toolkit to increase their own and their clients’ awareness of diabetes basics, available treatments, common co-morbidities, and self-management strategies. As such, the Toolkit helped both sets of staff to deliver easily understandable and consistent messages to patients about how to better manage their diabetes. This Toolkit is updated annually and can be accessed along with a simple diabetes tracking spreadsheet in Microsoft Excel at http://www.cmhsrp.uic.edu/health/diabetes-library-home.asp.

Approximately 55% of clinic patients reported being current smokers, which may have interfered with their achievement of targeted health outcomes. While nurses reported that they encouraged smoking cessation at every visit, coordinated efforts to engage patients in accessible, evidence-based smoking cessation classes might have helped improve their diabetes and overall health outcomes (55,56). Similarly, 49% of patients were obese and another 14% were overweight. While nurses encouraged weight loss, patients did not have access to evidence-based weight management classes that were welcoming to people in mental health recovery (57).

A number of caveats apply to our findings. First, our study population came from a single behavioral health home and not a nationally representative sample which may limit the generalizability of our findings. Second, our data came from EMRs and the reports of podiatrists and ophthalmologists regarding specialty visit completion. Data gathered directly from patients would have shed light on their reactions to the educational materials and their satisfaction with primary care services. Third, in the absence of a randomized controlled trial, we cannot attribute the positive changes we observed to the intervention itself. Finally, use of a single care coordinator may also limit the generalizability of findings regarding this role’s impact.

Conclusions

Co-morbid mental illness and diabetes are associated with poor quality of life, low treatment adherence, inferior glycemic control, frequent use of emergency department and inpatient care, and high medical costs (6, 58). Regarding the latter, costs for patients with co-occurring psychiatric and endocrinal disorders are twofold or even higher (depending on treatment setting) than those incurred by patents with endocrinal disorders alone (1). To address these formidable obstacles, our findings and those of others can be used to create interventions that incorporate additional evidence-based practices. These include proven weight reduction strategies (59), smoking cessation models (60), and substance use treatment (61) that addresses the needs of lower-income populations with limited health literacy and additional medical and behavioral health comorbidities. Finally, research using rigorous designs will be required to further develop these types of interventions and evaluate their feasibility and effectiveness.