The Comprehensive Care Physician (CCP) Program & Comprehensive Care Learning Collaborative (CCLC)David Meltzer, MD, PhD
Emily Perish, MPP
Overview of Problem
• Need to control U.S. health care spending growth and improve value• Small fraction of the population accounts
for a large fraction of overall health care spending• A large portion of this spending is on
hospitalizations• Fragmented medical care exacerbates this
problem
• Hospitals increasingly accountable for care of populations • Care coordination programs a central
strategy
Challenges
Solutions
Care coordination models have been developed and tested with little impact on outcomes or total cost of care
Ambulatory
Ambulatory
Hospital
HospitalCare
Coordination
Is it possible to better coordinate care without incurring large costs that are hard to recover?
Hospitalists
• Change from traditional model of primary care physicians (PCPs) who care for patients in and out of the hospital• Hoped to improve care, lower costs• Advantages: Inpatient expertise, presence• Disadvantages: Discontinuities, loss of Dr.-Pt.
Relationship• Net Effect: Modest
• Why did hospitalists grow?• Belief improve hospital care• Needs of primary care• Declining hospital vs. ambulatory volumes discourage
traditional PCP• Declining hospital use with shift from hospitalization to
ambulatory care• Increased ambulatory use with growth of preventive care• Organization of physicians into groups facilitated
specialization
Ambulatory Economics Theory of Hospitalist Growth(Meltzer, Chung, NBER Working Paper, 2010)
• Compare time costs of two models:• Traditional model:
• Internist time to see patients in hospital, clinic, transport• Hospitalist/PCP model
• Hospitalist time to see patient in hospital, communicate with PCP• PCP time to see patient in clinic, communicate with hospitalist
• Cost of PCP/Hospitalist vs. traditional model driven by per capita communication
• costs relative to transport costs for a traditional internist
• Cost of PCP/Hospitalist Model vs. Traditional Model falls when:• Admissions (p) fall relative to ambulatory visits• Communication costs (tc) decline• Transport costs (tT) rise • Physician work hours (TI) decline
• Confirm with data on PCP use of hospitalists from Community Tracking Study
∆"#$%&"&/(#$)*%+,*$% -$. /0+1*%*#2+, = 45%" − %/%7 + 5%(/9 − %/
= 45%" −%/:97
What is the Value of the Doctor-Patient Relationship for the Hospital Setting? And for Whom does it Matter?• Rich literature on the value of the doctor-patient relationship
• Trust, interpersonal relationship, communication btw. doctor/patient, knowledge of the patient
• Patients value seeing their own doctor in the hospital• But willingness to pay is not so high
• Observational studies show lower costs, better outcomes with continuity of care• Care by PCP for > 10 years: 15% lower Medicare costs (Weiss et al AJPH 1996)
• Lung CA patients cared for by own doctor in terminal hospitalization have 25% lower (OR=0.74, p<0.01) odds ICU use (Sharma et al, Annals, 2009)
• One experimental study• Wasson et al (JAMA, 1984) randomized 776 complex VA patients to see same physician vs.
different physician in each primary care visit. Continuous care group:• 49% lower emergent hospitalizations (20% vs. 39%, p<0.002)• 38% lower hospital days (6.6 vs. 9.1, p<0.02)• 74% lower ICU days (0.4 vs. 1.4, p<0.01)
à Discontinuity harmful/costly, esp. for complex, frequently hospitalized patientsà Better coordination of in/outpatient care may improve outcomes, but can we do it w/o
offsetting any savings?
CCP Approach to General Medical Care• Advantages?
• Most frequently hospitalized patients get own doctor in both settings. Continuity:• Is valued by patients• Decreases unneeded
testing/treatment, errors• Lowers doctor costs (travel, history
taking)• All hospitalized patients get doctors with
significant hospital experience and presence• Physicians can be specialists
• Patient choice restored• CCP model can work for physician• Patient-centered medical home / bundling
/ readmission penalties• Smaller primary care base can fill hospital
• Challenges?• Are enough patients willing to switch?• Will doctors let patients switch?• Will doctors do this job?• Can it be economically viable?
Stratify Patients by Expected Hospital Use
Low Expected Hospital Use
Ambulatory-based Primary Care Physicianand Hospitalist
High Expected Hospital Use
Comprehensive Care Physician
Study Background and Aims
• Began in 2012 with funding from the Center for Medicare and Medicaid Innovation (CMMI)• Established CCP Program at University of Chicago Medicine, which provides general medical care
to socioeconomically vulnerable population on Chicago’s South Side• From November 2012 to June 2016, randomly assigned 2,000 Medicare and dual-eligible patients
at increased risk of hospitalization to CCP or to ‘standard care’ by different physicians in and out of the hospital• If patients in standard care did not have a PCP or wanted a new PCP, we offered help to find
one
• Primary Aims: To determine whether providing Medicare patients at increased risk of hospitalization with access to care from a CCP compared to standard care by different physicians in and out of the hospital affects patient outcomes over 1 year, including
• Patient experience with health care (satisfaction with provider)• Health outcomes (self-rated general health status and mental health status) • Resource utilization (patient-reported hospitalization rate, Medicare claims)
• Present analysis of patient reported outcomes from quarterly surveys at 1 year. Further follow-up and outcomes (e.g., costs, hospitalization from Medicare claims) in progress
Key CMMI Design Elements
Lessons from Literature/Theory Program ElementFocus on patients at increased risk of hospitalization
Patients expected to spend >10 days in hospital in next year; up to 40% of general medicine days, annual Medicare costs $50,000- $100,000 per year; diverse recruitment sources, including resident clinics
Maximize Direct Interaction with CCP/PCH Panel size: 200. AM on wards. Midday buffer. PM in clinic.Build Interdisciplinary Team 5 CCPs = 1000 patients. Organize CCP, RN, LPN, LCSW, clinic coordinator
around common patient medical and psychosocial needsMinimize costs (esp. coordination costs) Small, well-connected teams, provider continuity, daily multidisciplinary
roundsFocus on care transitions Post-discharge calls, Health ITFinancial incentives Prepare for shared savings (randomized internal controls)
Sustainable roles and training for care team
Support the team members (group to spread weekend coverage, night coverage, psychosocial support, relevant clinical training (e.g., communication, palliative care), academic development, recognition).
Rapid cycle innovation Frequent, data-driven meetings that seek to engage relevant leadersRigorous evaluation 2,000 person RCT, Triple Aim (Better Care Better Health, Lower Costs),
survey and Medicare claims data, external and internal evaluators
Data Collection and Analysis
• Patient reported data from in person baseline survey prior to randomization and quarterly surveys by phone of patients in CCP and standard care arms• Patient demographics• Patient-reported outcomes:
• Physician ratings (Hospital and Consumer Assessment of Health Plans rating (1 (worst possible) to 10 (best possible))
• Self-rated general health status and mental health status (1 (poor) – 5(excellent))• Self-reported number of hospitalization in past quarter
• Analysis• Comparison of baseline demographics and health status measures• Longitudinal mixed effect models with logit for physician ratings (best possible) and health status
variables (excellent or very good) and zero-inflated Poisson model for number of hospitalizations• Random intercepts to account for repeated measures within subjects• Covariates: gender, age categories (50-64,65-74,75-84,85+), dual-eligible status, Hispanic,
number of hospitalizations at baseline• Pattern-mixture models (Little, JASA 1993) to address missing data and deaths
• ~89% 1-year follow-up rate (91% CCP, 87% SC), 12.5% 1-year mortality rate
10
Data Subjects
Characteristic CCPN=996
SCN=996
P-value Characteristic
CCPN=996
SCN=996
P-value
Female, % 62 62 0.58 Health Outcomes
Dual, % 46 43 0.14 Provider rank, best possible, % 39 34 0.14
Black, % 88 86 0.14
General health, excellent + very good, % 11 14 0.14
White, % 6.9 7.6 0.55
Mental health, excellent + very good, % 39 36 0.12
Hispanic, % 3.7 3.8 0.91
Hospitalizations in previous 12
months, %: 0.51
Age in years, mean (SD) 63 (16) 64 (16) 0.33 0 0.4 0.3
Age groups % 1 33 33
<50 22 21 0.55 2 or 3 27 28
50-64 25 24 0.64 4 or 5 6.7 7
65-74 30 30 0.71 5 < times <= 10 4.4 3.5
75-84 17 17 1.00 10 < times 28 27
85+ 7.2 8.6 0.25 Missing 0.7 1.6` Average per quarter (minimum est.) 1.13 1.10
20%
25%
30%
35%
40%
45%
50%
55%
60%
65%
70%
0 3 6 9 12 15 18 21 24
Perc
ent
Perc
ent
Bes
t Po
ssib
le (
10)
month
CCP
SC
Physician Rating (0 worst possible - 10 best possible)
Longitudinal ordinal mixed effects model: p=0.0001Longitudinal mean model: µCCP=9.30, µSC=9.07, D=0.23, p<0.0001
p=0.07
p=0.0007p<0.0001
p<0.0001 p<0.0001 p<0.0001p=0.0001 p=0.0001 p=0.0001
CCP 478 602 544 517 494 450 411 384 344
SC 481 436 390 351 366 347 323 294 297
10%
12%
14%
16%
18%
20%
22%
24%
0 3 6 9 12 15 18 21 24
Perc
ent
Exce
llen
t +
Ver
y G
oo
d
month
CCP
SC
General Health Rating
Longitudinal ordinal mixed effects model: p=0.86Longitudinal mean model: µCCP=3.26, µSC=3.28, Δ=-0.02, p=0.58
p=0.76
p=0.79
p=0.46p=0.51
p=0.47p=0.43
p=0.47
p=0.13
CCP 995 912 871 816 781 743 706 662 621
SC 994 899 847 799 746 710 678 636 614
p=0.36
25%
30%
35%
40%
45%
50%
0 3 6 9 12 15 18 21 24
Perc
ent E
xcel
lent
+ V
ery
Goo
d
month
CCP
SC
Mental Health Rating
Longitudinal ordinal mixed effects model: p<0.0001Longitudinal mean model: µCCP=3.21, µSC=3.41, Δ=0.20, p<0.0001
p=0.0004p=0.02
p=0.10
p<0.0001
p<0.0001p=0.08
p=0.10p=0.14
CCP 991 889 848 799 758 730 689 648 609
SC 991 867 831 778 727 695 667 625 604
p=0.11
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0 3 6 9 12 15 18 21 24
Ho
spit
aliz
atio
ns,
mea
n
month
CCP
SC
Hospitalization Rate
Longitudinal Zero-inflated Poisson mixed effects model: p=0.005Longitudinal mean model : µCCP=0.34, µSC=0.27, Δ=0.07, p<0.0001
p=0.24
p=0.01
p=0.0001
p=0.001
p=0.06
p=0.001
p=0.012p=0.13
p=0.10
CCP 996 911 867 815 771 730 686 608 517
SC 996 900 842 793 734 695 654 585 483
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0 3 6 9 12 15 18 21 24
Ho
spit
aliz
atio
ns,
mea
n
month
CCP
SC
Hospitalization Rate
Longitudinal Zero-inflated Poisson mixed effects model: p=0.005Longitudinal mean model : µCCP=0.34, µSC=0.27, Δ=0.07, p<0.0001
p=0.24
p=0.01
p=0.0001
p=0.001
p=0.06
p=0.001
p=0.012p=0.13
p=0.10
CCP 996 911 867 815 771 730 686 608 517
SC 996 900 842 793 734 695 654 585 483
Number Needed to Treat = 4 to prevent 1 hospitalization over 1 year$4,000/pt/year hospital costs, $3,000/pt/year total costsPrevent 250 Hospitalizations in 1000 patients in 1 year~ $4 million decrease in hospital costs, $3 million total costs
Conclusions and Limitations
• It was possible to implement a CCP program at UCM
• Positive patient outcomes, acceptable volumes for clinicians, acceptable ROI for
hospital
• CCP care improved patient experience and at least maintained patient outcomes while
reducing hospitalization by ~20% up to 1 year
• Number needed to treat = enroll 4 patients to prevent 1 hospitalization over 1 year
• Prevented about 250 hospitalizations in 1000 patients over year
• Implies ~ $4,000 lower hospital cost/patient/year if avg. cost of hospitalization ~$15,000
• Program savings of $4 million/year substantially greater than program costs
• Limitations
• Self-reported outcomes may be biased by patients and less than complete follow-up
• Medicare claims data not yet available to assess hospitalization and costs
• Dual eligibles more likely to drop out due to Illinois Medicare-Medicaid Alignment initiative, especially healthier ones
• New CCP program, one hospital, limited set of doctors, socioeconomically disadvantaged population
Next Steps
• Further analysis of CCP at UCM• Longer follow up (Donaghue), Costs, Duals/Non-Duals, hospitalization
risk, qualitative assessment of Dr.-Patient relationship in CCP, EOL care• Expansion efforts
• UCM• Migrating from fee for service to value based contracts• Expanding to affiliated Ingalls Community Hospital
• Chicagoland area: Ingalls Hospital, Rush, St. Anthony Hospital, Mt. Sinai• National: Vanderbilt, Kaiser Mid-Atlantic Region, Villages Health• International: National University Singapore, UK National Health Service,
Manipal University India• CMS Physician Focused Payment Model Technical Advisory Committee
(PTAC) recommended implementation of CCP payment model with PMPM care coordination fee
• GLPTN, CMS-funded Learning Collaborative
Comprehensive Care Learning Collaborative (CCLC) Overview
• Project ECHO web-based video platform
• Two Series: 1) Patient-level and 2) Program-level
• Topics focused on managing the care of patients with complex medical and social needs
• Sessions structure: Presentation/discussion and case-based discussion
• Participants: 15+ institutions, 40 participants
• Evaluated experience with pre/post survey
CCLC Series 1 Recap: Survey Questions about Topics Covered
• Please rate your knowledge, skills and/or competencies in the following areas, as they relate to patients at increased risk of hospitalization, which we refer to as "Patients" below. Please select a number to rate yourselves on a seven point scale (1 = "none or no skill at all", 2 = "vague knowledge, skills, or competence", 3 = "slight knowledge, skills or competence", 4 = "average among my peers", 5 = "competent", 6 = "very competent", 7 = "expert, teach others".)
• Establishing non-emergent primary care for Patients who have not established non-emergent primary care in the recent past
• Helping Patients reduce unneeded or unwanted medical care or polypharmacy
• Establishing multidisciplinary approaches to address chronic pain and opiate use in Patients
• Sustaining needed behavioral health services for Patients
• Assessing how Patients’ unmet social needs affect their medical care needs, and vice versa
• Decreasing social isolation among Patients
• Integrating trainees (e.g., medical students) into the care of Patients
• Using multidisciplinary teams to care for Patients
Pre-Survey Results (N=34) On average, participants report slightly more than “average competence among their peers” on each of the topics of this series, with at least one participant reporting “expert” knowledge on each topic
# report 1-7
AVG MEDIAN MIN MAX STD
Establishing non-emergent primary care for Patients who have not established non-
emergent primary care in the recent past
28 5.3 5 3 7 1.1
Helping Patients reduce unneeded or unwanted medical care or polypharmacy 31 5.0 6 1 7 1.5
Establishing multidisciplinary approaches to address chronic pain and opiate use in
Patients
30 4.2 4 1 7 1.6
Sustaining needed behavioral health services for Patients 30 4.3 4.5 1 7 1.5
Assessing how Patients' unmet social needs affect their medical care needs, and vice
versa
31 5.2 5 2 7 1.6
Decreasing social isolation among Patients 30 4.3 4 1 7 1.6Integrating trainees (e.g., medical students)
into the care of Patients 27 4.4 4 1 7 1.7Using multidisciplinary teams to care for
Patients 29 5.3 5 2 7 1.4
Pre-Survey Results (N=23 of 34 who answered the post-survey): The average pre-series competency of those who answered the post-series survey is similar to overall average
# report 1-7
AVG (total) AVG (pre & post)
MEDIAN MIN MAX STD
Establishing non-emergent primary care for Patients who have not established non-
emergent primary care in the recent past
21 5.3 5.3 5 3 7 1.1
Helping Patients reduce unneeded or unwanted medical care or polypharmacy 21 5.0 5.2 6 2 7 1.2
Establishing multidisciplinary approaches to address chronic pain and opiate use in Patients 20 4.2 4.3 4 2 7 1.3
Sustaining needed behavioral health services for Patients 21 4.3 4.5 5 2 7 1.2
Assessing how Patients' unmet social needs affect their medical care needs, and vice versa 21 5.2 5.2 5 2 7 1.5
Decreasing social isolation among Patients 21 4.3 4.3 4 1 7 1.6Integrating trainees (e.g., medical students) into
the care of Patients 18 4.4 4.2 4.5 1 7 1.9Using multidisciplinary teams to care for Patients 19 5.3 5.4 5 2 7 1.3
Post-Survey Results (N=23 of 34) On average, respondents to the pre-and post-survey report higher competence post-series vs. pre-series, with few reports of “below average competence among my peers”
# report 1-7
AVG MEDIAN MIN MAX STD
Establishing non-emergent primary care for Patients who have not established non-
emergent primary care in the recent past
23 5.5 6 3 7 0.8
Helping Patients reduce unneeded or unwanted medical care or polypharmacy 22 5.5 6 4 7 0.9
Establishing multidisciplinary approaches to address chronic pain and opiate use in
Patients
22 5.1 5 4 7 0.9
Sustaining needed behavioral health services for Patients 23 5.4 6 4 7 0.9
Assessing how Patients' unmet social needs affect their medical care needs, and vice
versa
22 5.8 6 4 7 0.8
Decreasing social isolation among Patients 21 4.9 5 3 7 0.8Integrating trainees (e.g., medical students)
into the care of Patients 19 4.9 5 1 7 1.4Using multidisciplinary teams to care for
Patients 21 5.9 6 5 7 0.7
Post-Survey Results: participant competency improvementGreatest improvements in behavioral health sustainment, chronic pain management, decreasing social isolation, and assessing unmet social needs
# report 1-7 on PRE &
POST
AVG(PRE) AVG(POST) AVG (POST-
PRE)
P-value (paired t-
test)Establishing non-emergent primary
care for Patients who have not established non-emergent primary
care in the recent past
21 5.29 5.43 0.14 0.67
Helping Patients reduce unneeded or unwanted medical care or
polypharmacy
20 5.25 5.45 0.20 0.48
Establishing multidisciplinary approaches to address chronic pain
and opiate use in Patients
19 4.42 5.05 0.63 0.08
Sustaining needed behavioral health services for Patients
21 4.52 5.43 0.90 0.02Assessing how Patients' unmet social needs affect their medical care needs,
and vice versa
20 5.20 5.75 0.55 0.11
Decreasing social isolation among Patients
19 4.26 4.89 0.63 0.17Integrating trainees (e.g., medical students) into the care of Patients
14 4.21 4.79 0.57 0.28Using multidisciplinary teams to care
for Patients17 5.35 5.88 0.53 0.21
Post-Survey Results: participant competency improvementGreatest improvements in behavioral health sustainment, chronic pain management, decreasing social isolation, and assessing unmet social needs
# report 1-7 on PRE &
POST
AVG(PRE) AVG(POST) AVG (POST-
PRE)
P-value (paired t-
test)Establishing non-emergent primary
care for Patients who have not established non-emergent primary
care in the recent past
21 5.29 5.43 0.14 0.67
Helping Patients reduce unneeded or unwanted medical care or
polypharmacy
20 5.25 5.45 0.20 0.48
Establishing multidisciplinary approaches to address chronic pain
and opiate use in Patients
19 4.42 5.05 0.63 0.08
Sustaining needed behavioral health services for Patients
21 4.52 5.43 0.90 0.02Assessing how Patients' unmet social needs affect their medical care needs,
and vice versa
20 5.20 5.75 0.55 0.11
Decreasing social isolation among Patients
19 4.26 4.89 0.63 0.17Integrating trainees (e.g., medical students) into the care of Patients
14 4.21 4.79 0.57 0.28Using multidisciplinary teams to care
for Patients17 5.35 5.88 0.53 0.21
Series 1 Recap: Summary of Free-text Survey Responses
• Respondents participated in the series to learn new approaches to complex care and apply lessons within their own practices
• Respondents reported benefitting from hearing from others about challenges/successes, learning new techniques to care for complex patient populations and networking with other providers
• Respondents identified time as a challenge and enjoyed the case discussions, requesting expanding on this in future sessions to deepen engagement
Series 1 Recap: Summary of Free-text Survey Responses
• Respondents participated in the series to learn new approaches to complex care and apply lessons within their own practices
• Respondents reported benefitting from hearing from others about challenges/successes, learning new techniques to care for complex patient populations and networking with other providers
• Respondents identified time as a challenge and enjoyed the case discussions, requesting expanding on this in future sessions to deepen engagement
Next CCLC Series, April 2019!