´óÏóÊÓƵ

Slide 1

Implementing Team Approaches for Improving Diabetes Care in Health Centers

Hector P. Rodriguez, PhD, MPH (hrod@berkeley.edu)

UC Berkeley School of Public Health

Images: Logos of the Community Health Partnership and CPCA: California Primary Care Association; photograph showing hands applying a glucosometer test strip to a drop of blood on a pricked fingertip.

Slide 2

iCARE (Innovative Care Approaches through Research & Education)
Acknowledgements

Other Research Team Members (University of California and RAND)

• Dylan H. Roby, PhD, MPP, Ana E. Martinez, MPH, Arturo Vargas-Bustamante, PhD, MPP, Mark Friedberg, MD, MPP, Philip van der Wees, PhD, Marc N. Elliott, PhD, Allen Fremont, MD, PhD, Xiao Chen, PhD, Nigel Lo, and Sean Wu.

QI/Interventions (Community Health Partnership, UCSF Center for Excellence in Primary Care, and CA Primary Care Association)

• Kent Imai, MD, Elena Alcala, MPH, Tom Bodenheimer, MD, MPH, Dolores Alvarado, MSW, MPH, Kat Contreras, Val Sheehan, MPH, Alpana Verma-Alag, MD, MBA.

Clinic Organizations (Intervention staff, IT/Data staff, primary care teams, leadership)

• North East Medical Services, Gardner Family Health Network, MayviewCommunity Health Center, Indian Health Center, Salud Para la Gente.

Funded by the ´óÏóÊÓƵ (AHRQ), under the American Recovery and Reinvestment Act (ARRA) (1R18HS020120-01).

Slide 3

The Effectiveness of QI Strategies: Findings from a Recent Review of Diabetes Care

Image: Chart illustrates the variety of QI strategies attempted and their effectiveness.

Shojania, K. G. et al. JAMA 2006;296:427-440.

Slide 4

Community Health Centers and Chronic Care Management

• Prior to the iCARE trial, the largest multi-site trial of diabetes care improvement in the safety net was conducted as part of the Diabetes Health Disparities Collaborative (BPHC HRSA).
• Chart review of 969 patients, 17 health centers.
• Processes of care improved (testing for HbA1c, foot exams, eye exams, and lipids).
• HbA1c control improved somewhat (borderline significant).

Note:  Chin MH et al, Diabetes Care, 2004

Slide 5

Primary Research Aims

1. To compare the effectiveness of 1) office-based medical assistant panel managers and 2) community-based health workers in improving diabetes care quality, patient self-management, and patients’ experiences of primary care in CHCs.
2. To clarify the organizational facilitators and barriers to the effective integration of the strategies into routine care in CHCs.

Slide 6

Comparing Two Team-Based Approaches to Diabetes Care Management

	Medical Assistant Panel Manager	Community Health Worker
Pre-visit
Discussing the patient case with the physician	X
Agenda setting with the patient	X
Ordering routine services	X	X
History tracking	X	X
During the Visit
Document physician findings	X
Send electronic prescriptions to pharmacy	X
Write prescriptions for the physician to sign	X
Post-visit
Discuss patients' concerns	X	X
Recapitulate the advice given by the physician	X	X
Set goals with the patient	X	X
Make sure that patients can navigate the system	X	X
Between Visits
Provide culturally appropriate and accessible health education and information	X	X
Assure that people with diabetes receive the services they need	X	X
Follow up over the phone	X	X
Offer informal counseling and social support	X
Provide information to families to support the lifestyle changes of patients with diabetes	X
Build individual and community capacity	X

 

Based on Bodenheimer’s teamlet model of primary care.

Slide 7

Cluster Randomized Design

Image: Tree chart shows the following information:

• 14 participating clinics (originally 17): Santa Clara, San Francisco, and San Benito counties (Bay Area) California:
  • Medical Assistant panel management, N=3.
  • Community Health Worker health coaching, N=3.
  • Control, N=11.

Slide 8

Image: Model shows the process of CHP Confidential Data File Production.

Slide 9

Project Data Sources

• 17 CHC sites in Northern CA with over 10K diabetic patients.
• Practice Climate Survey (n=249; RR=81%) in 2011.
• Clinical Quality, Demographic, and Diagnostic information for all adult patients with diabetes (n=6,111) in 2011 and 2012.
• Patient Experience Survey (random samples of patients with 2+ visits) (2012 RR= 45%, n=907; 2013 RR=63%, n=714).
• Key Informant Interviews of practice stakeholders in early (2012) and late (2013) intervention period (n=24).
• Practice structural capabilities survey (RR=100% in 2011 and 2013).

Slide 10

Focus on Intervention Effects

• Change Over Time Results:
  • Intermediate Outcomes of Diabetes Care:
    • Hemoglobin A1c.
    • LDL-Cholesterol.
    • Blood pressure.
  • Patients’ Experiences of Care:
    • CG-CAHPS Communication (k=6).
    • Patient Assessment of Chronic Illness Care (n=11).
• Key Implementation Insights for Health Centers Implementing Team-Based Diabetes Care Approaches with MAs and/or CHWs.

Slide 11

Primary Care Clinicians and Staff Occupations (n=249)

Images: Two pie charts show clinician and staff occupations:

• Nurse Practitioner: 32.2%.
• Physician: 52.1%.
• Physician's Assistant: 15.7%.

• Care Coordinator: 5.7%.
• Clinic Manager: 8.5%.
• Licensed Vocational Nurse: 8.9%.
• Medical Assistant: 32.3%.
• Non-Clinical Office Staff: 27.6%.
• Other Clinical Staff: 5.7%.
• Registered Nurse: 5.1%.
• Unknown: 6.1%.

Slide 12

Key informant role	Intervention (5 clinics) N	Control (7 clinics) N	Total (12 clinics) N
Practice leader (coordinator, medical director)	3	2	5
Clinician (physician, nurse practitioner)	4	4	8
Medical assistant	3	3	6
Community health worker	3	0	3
Other (nutritionist and registered nurse)	0	2	2
Total	13	11	24

 

Slide 13

Image: Three large ovals captioned Clinic Workload, Clinic Functionality, and Teamwork Perceptions are interconnected by double-headed arrows. Below four smaller ovals captioned Staff Relationships, Quality Improvement Orientation, Manager Readiness for Change, and Staff Readiness for Change; arrows point down from Clinic Functionality to these four ovals.

Slide 14

Care Team Functioning and Organizational Readiness for Change

Composite measure	Intervention Mean (Range)	Control Mean (Range)	Overall Mean (Range)
Staff relationships	64.8 (46.9 to 78.2)	65.7 (56 to 81.7)	65.2 (46.9 to 81.7)
Quality improvement	66.6 (49.4 to 76.5)	62.7 (49.1 to 76)	65.2 (49.1 to 76.5)
Manager readiness for change	63.8 (42.5 to 77.5)	63.9 (38.1 to 80.8)	63.9 (38.1 to 80.8)
Staff readiness for change	70.0 (60.4 to 81.2)	67.6 (52.8 to 74.6)	69.2 (52.8 to 81.3)
Teamwork attitude	55.0 (50.0 to 60.0)	55.0 (52.5 to 57.5)	55.0 (50.0 to 60.0)
Clinic workload	44.9 (33.6 to 60.1)	41.7 (30.4 to 76)	43 (30.4 to 60.1)

 

Composite scores (range 0-100) based on clinician/staff survey prior to the intervention to measure cultural aspects of the control and intervention sites.

Source: Van der Wees PJ, Friedberg MW, Alcala E, Ayanian JZ, Rodriguez HP. Comparing the implementation of team approaches for improving diabetes care in community health centers. BMC Health Services Research.  In press.

Slide 15

Structural capability	Intervention (n=5)	Control (n=8)
Checklist or flow-sheet for:
HbA1c testing	4/5	6/8
Cholesterol testing	1/5	6/8
Eye examination	4/5	6/8
Nephropathy monitoring	4/5	6/8
On-site registry out of target range for:
Laboratory values	3/5	2/8
Physical findings (BP, BMI)	3/5	2/8
On-site registry for patients overdue for:
Screening services	5/5	6/8
Diabetes services	3/5	4/8
Other chronic disease services	0/4	3/8
Shared communication:
HbA1c testing	2/4	2/8
Cholesterol testing	2/4	2/8
Eye examination	2/4	2/8
Nephropathy monitoring	2/4	1/8

 

Source: Van der Wees PJ, Friedberg MW, Alcala E, Ayanian JZ, Rodriguez HP. Comparing the implementation of team approaches for improving diabetes care in community health centers. BMC Health Services Research.  In press. 

Slide 16

Patient Survey Respondent Characteristics: Education and Language (n=907)

Image: Bar graph shows the following Patient Survey respondent characteristics:

• Education:
  • 8th grade or less: 44%.
  • Some high school but did not graduate: 16%.
  • High school graduate or GED: 26%.
  • 4-year college graduate: 9%.
  • More than 4-year college graduate: 5%.
• Language:
  • Asian (predominantly Chinese): 32%.
  • English: 34%.
  • Spanish: 34%.

Slide 17

Methods-Change Over Time

• Analytic sample definition: 6,111 adult diabetic patients with 2+ visits in pre-intervention year (2011) and 1+ visit in the intervention year (2012).
• Cluster randomization of clinics did not result in balanced patient characteristics.  Exact matching was used to improve causal inference.
• Age (in 10 year bands), gender, race/ethnicity, language preference, and insurance type were used as matching variables.

Slide 18

Improving the Balance of Patients across the Study Arms

Image: Bar graph compares languages spoken by Community Health Workers and Medical Assistants. For both the Community Health Workers Arm (n=686) and Control Group (Weighted), over 60 % are Latino-Spanish speaking and approximately 20% are English-Spanish speaking; the remainder are Asian-English speaking (~10%) or Other English speaking (~10%), with only a tiny number Asian-Asian language. For both the Medical Assistant Arm (n=644) and Control Group (Weighted), more than 70% are Asian-Asian language speaking, ~10% Asian-English speaking, less than 10% Latino-Spanish speaking and ~5% Latino-Spanish speaking or Other English speaking.

Slide 19

Clinic Setting	Provider Organization	Health coach	Team composition	Panel size (Overall diabetes at clinic)	Main patient population	Workflow
1. Urban	2 clinics with ~5 clinicians serving low-income families	MA	Teams of 2 clinicians and 2 MA	119 (139)	Latino	MA panel management based on the Teamlet Model. No home visits. Combining regular MA work with health coaching. MA sees 4 patients per day for health coaching on alternate days.
2. Urban	7 clinics with ~50 clinicians serving low-income families	MA	Teams of 6 clinicians and 4 MA	NS (367)	Recent Chinese immigrants	MA works on weekly rotating schedule as health coach. Sees ~12 patients per day typically in post-visits to clinicians. No home visits.
3. Small community	7 clinics with ~40 clinicians serving low-income families	CHW	Teams of 2 clinicians and 1 CHW	118 (334)	Latino	CHW works mainly office-based via panel management in Teamlet Model. Sees 6-8 patients per day.
4. Small community	2 clinics with ~5 clinicians serving low-income families	CHW	Teams of 3 clinicians and 2 CHW	137 (143)	Latino	CHW does office-based visits and post-visits based on Teamlet Model. Started small-scale home visits, planning 3-4 joint visits per day by 2 CHW.
5. Suburban	7 clinics with ~40 clinicians serving low-income families	CHW	Teams of 3 clinicians and 1 CHW	84 (377)	Latino	CHW works community-based with home visits of 25-30 minutes during 4 days per week. One day office-based for follow-up phone calls. Separate from clinic workflow.

 

Source: Van der Wees PJ, Friedberg MW, Alcala E, Ayanian JZ, Rodriguez HP. Comparing the implementation of team approaches for improving diabetes care in community health centers. BMC Health Services Research.  In press.  

Slide 20

HbA1c controlled below 8.0%

Image: Bar graph compares HbA1c controlled below 8.0% for Community Health Workers and Medical Assistants in 2011 and 2012.

Community Health Workers Arm:

• 2011: 86.2.
• 2012: 88.5.

Community Health Workers Control Group Arm:

• 2011: 86.3.
• 2012: 88.7.

Medical Assistant Arm:

• 2011: 84.2.
• 2012: 86.3.

Medical Assistant Control Group Arm:

• 2011: 90.2.
• 2012: 91.6.

Note: Adjusted analyses control for patient age, gender, race/ethnicity, and insurance status.  Patient and clinic random effects are used to account for the clustering of time within patients and patients within clinics, respectively.

Slide 21

LDL-Cholesterol Control (≤100 mg/dL) Changes Over Time

Image: Bar graph compares LDL-cholesterol control changes over time for Community Health Workers and Medical Assistants in 2011 and 2012.

Community Health Workers Arm:

• 2011: 66.
• 2012: 65.8.

Community Health Workers Control Group Arm:

• 2011: 60.9.
• 2012: 63.

Medical Assistant Arm:

• 2011: 47.6.
• 2012: 56*.

Medical Assistant Control Group Arm:

• 2011: 65.4.
• 2012: 68.5.

Note: Adjusted analyses control for patient age, gender, race/ethnicity, and insurance status.  Patient and clinic random effects are used to account for the clustering of time within patients and patients within clinics, respectively.

Medical Assistant intervention arm has 8.4% point improvement in LDL-C control.

Slide 22

Blood Pressure Control (≤140/90 mmHg) Over Time

Image: Bar graph compares blood pressure control over time for Community Health Workers and Medical Assistants in 2011 and 2012.

Community Health Workers Arm:

• 2011: 57.6.
• 2012: 63.7*.

Community Health Workers Control Group Arm:

• 2011: 63.1.
• 2012: 62.8.

Medical Assistant Arm:

• 2011: 65.1.
• 2012: 69.

Medical Assistant Control Group Arm:

• 2011: 64.2.
• 2012: 65.3.

Note: Adjusted analyses control for patient age, gender, race/ethnicity, and insurance status.  Patient and clinic random effects are used to account for the clustering of time within patients and patients within clinics, respectively.

Community Health Workers intervention arm has 6.1% point improvement (p=0.10).

Slide 23

Communication (CG-CAHPS)

Image: Bar graph compares Pre and Post Communication:

Community Health Workers Arm:

• Pre: 67.4.
• Post: 68.5.

Medical Assistant Panel Manager Arm:

• Pre: 74.1.
• Post: 76.0.

Control Arm:

• Pre: 71.0.
• Post: 70.0.

Note: Adjusted analyses control for patient age, gender, race/ethnicity, and insurance status.  Patient and clinic random effects are used to account for the clustering of time within patients and patients within clinics, respectively.

Slide 24

Patient Assessment of Chronic Illness Care (PACIC-11)

Image: Bar graph compares Pre and Post Patient Assessment of Chronic Illness Care:

Community Health Workers Arm:

• Pre: 50.4.
• Post: 53.4.

Medical Assistant Panel Manager Arm:

• Pre: 47.5.
• Post: 55.1*.

Control Arm:

• Pre: 48.0.
• Post: 51.3.

Note: Adjusted analyses control for patient age, gender, race/ethnicity, and insurance status. Patient and clinic random effects are used to account for the clustering of time within patients and patients within clinics, respectively.

Slide 25

Outcomes Summary

• Clinical Outcomes:
  • Improved LDL-C  control for the MA arm (8.4% points).
  • Improved blood pressure control for CHW arm (6.1% points).
• Patient Experience:
  • More improvement (7.6 points) in patients’ experiences of chronic illness care (PACIC-11) for MA arm.
  • No differential improvements in clinician-patient communication.

Slide 26

Implementation Insights

• Practice Modifications:
  • Changes to diabetes care management were concentrated in the five intervention sites.
  • Only one control clinic respondent indicated any changes to the management of diabetic patients during early or late intervention periods.
• Support of New Team Member Role Integration:
  • Perceived support of health coaching role of MA or CHW at all levels of the organization for the 5 intervention sites.
  • Dedicated time of MA and CHW crucial for implementation.
  • Rotating responsibilities for health coaching among staff impeded the learning process.

Slide 27

Implementation Insights II

• Structural capabilities (like registry use for diabetics) were perceived as foundational requirements for implementing CHW or MA team-based approaches.
• Cultural adaptations to the models were important:
  • Emphasizing physician-led teams for Chinese patients.
  • Emphasizing family roles and social support for Latino patients.
  • Gender seemed to play a role in the implementation of CHW home visits.

Slide 28

Key Conclusions

• Diabetic patients improved intermediate outcomes (Blood pressure for CHW; LDL-C for MA panel manager) in the short run (1 year)
• First multi-site intervention study to pool patient-level data across diverse CHC organizations serving different ethnic communities and link with patient experience surveys.
• Patient experiences of care quite low- need for improvement and appear to be difficult to change over time.
• Money and Facilitation Isn’t Enough!:  Even with implementation resources, extensive data management support and intervention technical assistance, intervention sites did not achieve breakthrough improvements.

Slide 29

Should We Be Spreading These Team-Based Approaches in the Safety Net?

1. The right thing to do for patients, but effect sizes are discouraging (compared to control)
2. Frontline experiences (key informant interviews) indicate that the study period (2011-2013) was turbulent for CHCs (EHR implementation, staff turnover).
3. Without supportive payment policies, implementation of MA and CHW models will not likely spread
4. More practice-based evidence to support future implementation?
5. Patient experience has got to be front and center of future efforts, as team-based models require patient acceptance.

Slide 30

Additional Questions?

• Hector Rodriguez
  • hrod@berkeley.edu, (510) 642-4578