Partial and Incremental PCMH Practice Transformation: Implications for Quality and Costs Michael L. Paustian, Jeffrey A. Alexander, Darline K. El Reda, Chris G. Wise, Lee A. Green, and Michael D. Fetters Mich bcbs july 2013

Health Services Research
© Health Research and Educational Trust
DOI: 10.1111/1475-6773.12085
RESEARCH ARTICLE

Partial and Incremental PCMH Practice
Transformation: Implications for Quality
and Costs
Michael L. Paustian, Jeffrey A. Alexander, Darline K. El Reda,
Chris G. Wise, Lee A. Green, and Michael D. Fetters
Objective. To examine the associations between partial and incremental implementation of the Patient Centered Medical Home (PCMH) model and measures of cost and
quality of care.
Data Source. We combined validated, self-reported PCMH capabilities data with
administrative claims data for a diverse statewide population of 2,432 primary care
practices in Michigan. These data were supplemented with contextual data from the
Area Resource File.
Study Design. We measured medical home capabilities in place as of June 2009 and
change in medical home capabilities implemented between July 2009 and June 2010.
Generalized estimating equations were used to estimate the mean effect of these
PCMH measures on total medical costs and quality of care delivered in physician practices between July 2009 and June 2010, while controlling for potential practice, patient
cohort, physician organization, and practice environment confounders.
Principal Findings. Based on the observed relationships for partial implementation,
full implementation of the PCMH model is associated with a 3.5 percent higher quality
composite score, a 5.1 percent higher preventive composite score, and $26.37 lower
per member per month medical costs for adults. Full PCMH implementation is also
associated with a 12.2 percent higher preventive composite score, but no reductions in
costs for pediatric populations. Incremental improvements in PCMH model implementation yielded similar positive effects on quality of care for both adult and pediatric
populations but were not associated with cost savings for either population.
Conclusions. Estimated effects of the PCMH model on quality and cost of care
appear to improve with the degree of PCMH implementation achieved and with incremental improvements in implementation.
Key Words. PCMH, medical home, cost, quality

Policy makers and payers have begun to engage with primary care providers
in testing an alternative model of practice organization and orientation under
the rubric of the “Patient Centered Medical Home (PCMH).” The PCMH is
1

2

HSR: Health Services Research

deﬁned as a holistic and integrated model of primary care designed to
improve the processes and outcomes of health care, including increasing
value, improving responsiveness to patients, and improving outcomes in the
areas of access, timeliness, patient centeredness, safety, equity, and efﬁciency
(Patient-Centered Primary Care Collaborative 2007). The recently passed federal health care reform bill, the Patient Protection and Affordable Care Act
(HR3590), includes federal PCMH demonstration programs, and PCMH
implementation is under way in a wide variety of practice settings across the
country (Patient-Centered Primary Care Collaborative 2009; Fields, Leshen,
and Patel 2010; The Commonwealth Fund 2011). Despite enthusiasm to rapidly pilot and scale up the PCMH model, including efforts to incorporate
PCMH into state Medicaid programs, attempts to empirically examine the
effects of PCMH on quality and cost-related outcomes are still in the early
stages and are often focused on small and self-selected samples of physician
practices.
A central problem in PCMH-effectiveness studies is determining when,
or to what degree, PCMH has been implemented (Alexander and Bae 2012;
Hoff, Weller, and DePuccio 2012). In some studies implementation is assumed
but not measured, whereas in other cases, it is measured in terms of degree but
only in “check the box” fashion. Because of the high potential for measurement error, both are problematic for studying the effects of PCMH. In our
study we assume that the complex, multicomponent nature of PCMH
increases the probability that components of the PCMH model will be implemented incrementally rather than at a single point in time. We further assume
that the “path” to full implementation of PCMH will likely vary among physician practices. For example, some practices might initially implement
extended access and then add individual care management, whereas others
might start with developing a patient registry followed by e-prescribing, etc.
As the PCMH model assumes its constituent components will function as a
system of care, it is not clear how cost and quality outcomes will be affected if

Address correspondence to Michael L. Paustian, Ph.D., Blue Cross Blue Shield of Michigan,
Department of Clinical Epidemiology and Biostatistics, 2311 Green Rd, Ann Arbor, MI 48105;
e-mail: mpaustian@bcbsm.com. Jeffrey A. Alexander, Ph.D., is with the School of Public Health,
Department of Health Management and Policy, University of Michigan, Ann Arbor, MI. Darline
K. El Reda, Dr.P.H., is also with the Department of Clinical Epidemiology and Biostatistics, Blue
Cross Blue Shield of Michigan, Ann Arbor, MI. Chris G. Wise, Ph.D., is with the Practice
Improvement Consultant, Ann Arbor, MI. Lee A. Green, M.D., is with the University of Alberta,
Edmonton, Edmonton, AB, Canada. Michael D. Fetters, M.D., is with the University of Michigan
Health System, Ann Arbor, MI.

Partial and Incremental PCMH Practice Transformation

3

complete implementation is not achieved, or if different paths to implementation will result in similar effects on quality and cost outcomes.
Our study attempts to advance PCMH research by examining the following research questions: (1) Is partial implementation of the PCMH model
associated with outcomes related to primary care cost and quality of care? (2)
Are incremental changes in implementation of PCMH associated with outcomes related to primary care cost and quality of care? and (3) Do different
approaches to PCMH implementation show similar associations with outcomes related to primary care cost and quality of care?

BACKGROUND
The current body of research concerning implementation of PCMH among
physician practices is limited, although several recent studies have examined
the adoption of components of the PCMH model (Rittenhouse et al. 2008,
2011). These studies suggest that PCMH elements such as whole-person orientation and continuity of care through use of personal physicians have been
more readily adopted (Goldberg and Kuzel 2009). However, processes related
to care coordination and integration, enhanced access, team-based care, and
support from appropriate information systems have not been adopted as
broadly (Audet, Davis, and Schoenbaum 2006; Rittenhouse et al. 2008;
Friedberg et al. 2009).
Several studies have linked medical homes with improved health indicators. In general, these studies suggest that medical homes are associated with
increased patient satisfaction (Palfrey et al. 2004; Gill et al. 2005; DeVoe et al.
2008; Reid et al. 2009) and decreased hospitalizations and emergency room
visits (Martin et al. 2007; Cooley et al. 2009; Rankin et al. 2009). Studies
have also identified improvements in quality (Rankin et al. 2009; Reid et al.
2009) and reduced clinician burnout (Reid et al. 2009). Findings have been
more mixed regarding PCMH impact on reducing medical costs (McBurney,
Simpson, and Darden 2004; Damiano et al. 2006).
Although promising, conclusions from these PCMH-effectiveness evaluations warrant further discussion, given differences in PCMH measurement
approaches. Some studies directly measured the degree of “medical homeness,” whereas others simply assumed the presence of a medical home (i.e.,
Cooley et al. 2009). Other studies measured PCMH by using questions from
the Consumer Assessment of Healthcare Providers and Systems Survey or by
measuring elements of PCMH defined by the specific study. The variation

4


and limitations in current PCMH measurement approaches, coupled with the
relatively small sample size of many PCMH studies, increase the difficulty of
cross-study comparisons and drawing conclusions about PCMH effects. Additional investigation using larger samples and more granular PCMH measurement is necessary to assess how the degree of PCMH implementation affects
patient care outcomes, including whether this care delivery model will lead to
better processes and outcomes across a variety of practice types.
In this study we address the question of whether PCMH is an “all or
nothing” form of care delivery by examining the hypothesis that more extensive implementation of the PCMH model will be positively associated with
indicators of lower cost and higher quality of care. Underlying this hypothesis
are the assumptions that partial benefit of PCMH on patient outcomes will be
obtained even if total implementation of all components of the model is not in
place, and that the approach (or sequencing) to implementing components of
the PCMH model can be tailored to the circumstances of each primary care
practice. We also test the proposition that incremental progress in the implementation (regardless of the initial level of implementation) of the PCMH
model will be associated with lower cost and higher quality of care.

M ETHODS
Sample
The physician practice sample in this study comes from the population of
4,192 practices in Michigan with at least one physician affiliated with a physician organization participating in the Blue Cross Blue Shield of Michigan
(BCBSM) Physician Group Incentive Program (PGIP) in 2010. Within PGIP,
primary care practices are eligible for two PCMH-related incentives: (1) partial reimbursement for PCMH capability implementation and (2) 10 percent
office visit and preventive visit fee enhancements for practices that achieve significant PCMH capability implementation in combination with delivering
high quality of care (Share and Mason 2012). Among these practices, 2,494
self-reported at least one primary care physician (PCP). These practices
account for 5,750 primary care physicians or 65 percent of all primary care
physicians practicing in Michigan (Michigan Department of Community
Health 2010), and operate in 82 of Michigan’s 83 counties. Because our
research questions focus on the effects of PCMH transformation on primary
care outcomes, a practice needed at least one primary care physician and must
have reported medical home capabilities in both June 2009 and June 2010 to


5

be included. Practices were excluded if more than one half of the physicians in
the practice were nonprimary care specialists. Application of inclusion and
exclusion criteria resulted in a study group of 2,432 practices providing primary care to approximately 1.5 million BCBSM members.
Data Sources
Data for this research were obtained from (1) the BCBSM Self-Reported Database (SRD); (2) BCBSM member enrollment data files; (3) BCBSM claims
data; and (4) the Health Resources and Services Administration (HRSA) Area
Resource File 2009–2010 (ARF 2010). Physician organizations report practice
and physician information to BCBSM through the SRD semiannually. This
database includes physician demographic information, physician organization
and practice affiliation (i.e., which physicians make up each practice), a primary care physician indicator, practice PCMH capabilities based on interpretive guideline case definitions (Blue Cross Blue Shield of Michigan 2010a),
and approximate date of implementation for each capability. BCBSM field
staff trained in process improvement conducted 235 practice site visits
between June 2009 and June 2010 to validate capability reporting and provide
educational guidance. To address potential overreporting that may be encouraged through the incentive design, practices must demonstrate functional use
of capabilities during the site visit, such as a populated patient registry usable
at the point of care.
We used enrollment information to obtain demographic data on members who received care at these practices and administrative claims data for
the services received. An administrative claims-based retrospective primary
care member attribution algorithm was applied to determine the BCBSM
commercial member cohort, denominator population, for each practice.
Members were assigned based on 24 months of outpatient and office-based
claims history, beginning July 1, 2008 and ending June 30, 2010, with attribution assigned to the physician who had the most evaluation and management
office or preventive visits with that member; ties were resolved by assigning
members to the physician with the most recent evaluation and management
office or preventive visit with that member or to the physician who rendered
the most total services for that member.
We incorporated items from HRSA’s Area Resource File to address
potential confounding from market area socioeconomic or demographic
sources not available through the SRD or through BCBSM administrative
data.

6


M EASURES
PCMH Implementation Score
We developed a practice-level PCMH implementation score to reflect the
degree of PCMH implementation across 13 domains of function (Table 1) by
combining SRD PCMH capability information in a multistage process. In the
first stage, capabilities reported as “fully in place” were assigned a value of 1,
whereas capabilities reported as “not in place” were assigned a value of 0.
When capabilities had multiple gradients, the capability score was calculated
as a proportion of the maximum gradient. For example, the Patient–Provider
Agreement domain (Appendix A) asked respondents to identify the percentage of their patient population who had established documented patient–
provider agreements from the following options: 10, 30, 50, 60, 80, or 90
percent. A response of 30 percent implementation on this patient–provider
agreement communication was assigned a value of 0.33 (0.3/0.9). Domainspecific scores were calculated in the second stage, by summing all capability
scores within the domain and dividing by the maximum capability score
possible, that is, distinct capabilities within that domain. Lastly, we calculated
the overall PCMH implementation score as the mean of all 13 domain-specific
scores. Thus, a one unit change in the continuous overall PCMH implementation score corresponds to the difference between no implementation (0) and
full implementation (1), although almost all practices in our study group fall
along the continuum between these two endpoints. This method intentionally
gives equal scoring weight to each PCMH domain to reflect the unknown relative importance of each domain, and thereby avoids giving greater weight to
domains with a greater number of capabilities.
Baseline PCMH Implementation and Change in PCMH Implementation
Although the PCMH program began in January 2008, practices needed adequate time to establish enough capabilities before assessing their impact. The
PCMH implementation score was calculated for both the June 2009 and June
2010 reporting cycles. Capability implementation dates from the June 2010
reporting cycle were used to correct for any overreporting of capabilities in
the June 2009 time period due to differences in interpretation of capability
case definitions from the interpretive guidelines and to account for any interpretive guideline changes made between reporting time periods. This correction creates greater comparability of the two measurement time points for the


7

Table 1: Patient-Centered Medical Home (PCMH) Domains of Function,
Number of Total Capabilities, and Distinct Capabilities by Domain

Domain
Patient–provider
agreement
Patient registry
Performance reporting

Individual care
management

Extended access

Test results tracking
and follow-up

E-prescribing

Preventive services

Linkage to community
services

Description
Practice has developed and is using
PCMH-related communication tools
An all-payer registry is used to manage
established patients in the practice
Performance reports are generated that
allow tracking and comparison of
results for the established population
of patients in the practice
Practice has ability to deliver
coordinated care management
services with an integrated team of
multidisciplinary providers and a
systematic approach is in place to
deliver comprehensive care that
addresses patients’ full range of health
care needs
Patients have 24-hour access to a
clinical decision maker by phone, and
clinical decision maker has a feedback
loop within 24 hours or next business
day to the patient’s PCMH
Practice has test tracking process
documented and in place which
requires tracking and follow-up for all
tests and results, with identiﬁed
timeframes for notifying patients of
results
Practice has adopted and uses
electronic prescribing and clinical
decision support tools to improve the
safety, quality, and cost effectiveness
of the prescription process
Primary prevention program is in place
that focuses on identifying and
educating patients about personal
health behaviors to reduce their risk of
disease and injury
A comprehensive review of, and
linkage to, community resources has
been completed

No. of
Capabilities

Maximum
Distinct
Capabilities

8

3

17

17

12

12

15

15

9

7

9

9

2

1

8

8

8

8

continued

8


Table 1. Continued

Domain
Self-management support

Patient web portal

Coordination of care

Specialist referral process

Description
A systematic approach to empowering
patients to understand their central
role in effectively managing their
illness, making informed decisions
about care, and engaging in healthy
behaviors is in place
A patient web portal is in use by the
practice to allow for electronic
communication between patients and
physicians, and provide greater access
to medical information and technical
tools
For patients with selected chronic
conditions, a mechanism is established
for being notiﬁed of each patient
admit and discharge or other type of
encounter, and appropriate transition
plans are in place
Procedures are in place to guide each
phase of the specialist referral process

Total capabilities

No. of
Capabilities

Maximum
Distinct
Capabilities

8

8

12

12

9

7

9

9

126

116

degree of PCMH implementation. Change in PCMH implementation, or
incremental implementation, was recorded as the difference between the June
2010 PCMH implementation score and the June 2009 PCMH implementation score. Thus, the estimated value for incremental implementation is
dependent on the baseline level of PCMH implementation previously
attained. The maximum combined value of the partial implementation (baseline) and the incremental implementation (change) is 1.
Cost and Quality Outcomes
For this study, we examined one practice-level measure of cost and four composite measures of quality of care for the July 2009 to June 2010 time period
( Jaen et al. 2010; Higgins et al. 2011; The Commonwealth Fund 2012). Total
combined medical and surgical allowed cost per member per month (PMPM)
was calculated separately for the adult and pediatric primary care–attributed


9

member population. Composite measures of quality were selected over individual quality measures because few physicians have sufficient numbers of
patients on any individual quality measure for comparative purposes (Scholle
et al. 2009) and because of concerns about heterogeneity in physician performance across individual quality measures (Parkerton et al. 2003). Four overall
percentage (Reeves et al. 2007) composite quality and preventive measures
were generated from HEDIS-defined and BCBSM-defined individual quality
and preventive measures (Blue Cross Blue Shield of Michigan 2010b): adult
quality, adult preventive, pediatric quality, and pediatric preventive (See
Appendix B). The pediatric quality composite measure was log transformed
after assessment of its normality.
Practice Characteristics
Four practice characteristics were used in the analysis: practice size, primary
care focus, BCBSM market share, and inpatient service provision. Practice
size was evaluated categorically based on total number of physicians, including specialists, in the practice as reported in the SRD. We used the primary
care physician indicator in the SRD to classify practices as “primary care
focus” if only primary care physicians were present and “mixed specialty” if
both primary care and nonprimary care specialty physicians were present.
Total BCBSM-paid services per PCP delivered between July 2009 and June
2010 were calculated for each practice as a proxy for BCBSM volume within
the practice. We also calculated the proportion of these services provided
within an inpatient setting to ensure comparability of physician practice patterns across practices.
Patient Cohort Characteristics
We used the primary care–attributed member cohort for each practice and
their member enrollment information to estimate the following five practicelevel patient characteristics: proportion of members under 18 years, proportion of members over 64 years, proportion of members who were female, and
mean Ingenixâ Symmetry version 6 prospective risk score (Ingenixâ 2008)
for both adult and pediatric members in the practice. The prospective risk
score employs a large national database of aggregated claims and membership
information to derive a numerical, diagnosis-based episode assessment used
to predict future medical costs.

10


Market Characteristics
We examined five county-level market characteristics and one physician organization characteristic to address additional sources of variation that might
influence cost and quality outcomes. County-level data from the ARF 2010
were used for median household income, percent of residents who were nonwhite or Hispanic, total primary care physicians per 1,000 population, and
total county population. We also calculated county-level BCBSM market
share based on member subscriber addresses from BCBSM member enrollment information and total estimated county population from the ARF 2010.
County-level measures were weighted for each practice to account for the proportion of their care provided to members residing in each county. The
weighted total population estimate for the practice was converted into a categorical measure of urbanicity based on 1990 census metropolitan statistical
area classifications. Physician organization size was measured as the total number of practices with at least one primary care physician.
Analytic Approach
The unit of analysis for this study was the primary care physician practice, the
functional unit at which the PCMH program was targeted. We analyzed relationships between level of PCMH implementation, change in PCMH implementation, and our outcome measures in multivariable models using both
generalized estimating equations and random intercept generalized linearmixed models. Generalized estimating equations were calculated to account
for clustering of practices within physician organizations and to estimate the
mean population-level practice effects of PCMH implementation on
measures of cost and quality (Hubbard et al. 2010). Random intercept linearmixed models were fitted to address potential heterogeneity in outcomes conditional on the physician organization of the practice (Singer 1998). To assess
the effects of previously implemented PCMH capabilities in combination
with newly implemented PCMH capabilities, the models for each of the six
outcomes included both the level of partial PCMH implementation measured
at baseline and the incremental PCMH implementation that occurred during
the study time period. All analyses were performed in SASâ version 9.2 (SAS
Institute Inc 2011). Outcome-specific exclusion criteria were applied prior to
constructing multivariable models. Pediatric practices (N = 296), defined as
practices with at least 80 percent of attributed members below 18 years, were
excluded from adult outcome analyses. Practices that failed to meet the


11

minimum sample threshold of 50 members for cost measures or 30 events for
composite quality and preventive outcomes were excluded from their respective analyses. Practices whose cost outcome measure exceeded three interquartile range (IQR) units from the median or whose composite outcome
measures exceeded two IQR units from the median were excluded from their
respective analyses to minimize the impact of the tail of the distributions on
parameter estimates. Measures of standard influence on predicted value, a
standard regression diagnostic, were calculated to identify “influential observations” that might have excessive influence and bias parameter estimates
(Fox 1991). After applying exclusion criteria, we compared excluded practices
to retained practices to assess differences that might impact interpretation or
generalizability of the results.
We examined potential colinearity of predictor variables using Pearson
correlation coefficients and then evaluated variable colinearity in the multivariable models. Percentage of members over 64 years and percent of services
provided in the inpatient setting were dropped due to colinearity with the
mean risk score.

RESULTS
As of June 2009, 1,647 study practices (67.7 percent of the study group)
had implemented at least one PCMH capability. The mean baseline
PCMH implementation score across study practices was 0.11 (SD: 0.13),
whereas the mean PCMH implementation score among practices with at
least one capability was 0.17 (SD: 0.13). By June 2010, 2,219 study practices (91.2 percent) had implemented at least one PCMH capability. The
mean PCMH implementation score had risen to 0.34 (SD: 0.24) and was
0.37 (SD: 0.22) among practices with at least one capability. The mean
change in PCMH implementation between the two time periods in study
practices was 0.22 (SD: 0.20).
Table 2 illustrates the member, practice, market characteristics, and
outcome distributions of the study practice population. On average, solo
physician practices had 334 BCBSM members, practices with two or three
physicians had 771 members, practices with four or five physicians had
1,482 members, and practices with six or more physicians had 2,122 members. Solo physician practices accounted for 1,411 (58.0 percent) study
practices but were more likely to be excluded in each analysis due to sample size criteria.

12


Table 2: Practice, Patient Cohort, and Market Characteristics of Physician
Practices with at Least One Primary Care Physician Participating in the
BCBSM Physician Group Incentive Program, July 2009 to June 2010
Adult and Family Practices
(N = 2,136)

Pediatric Practices
(N = 296)

Median
Outcomes
Adult preventive composite
Adult quality composite
Pediatric preventive composite
Pediatric quality composite
Adult cost PMPM
Pediatric cost PMPM
Continuous variables
PCMH score June 2009
PCMH change to June 2010
Median household income
Total practices in PO with a PCP
Services per PCP
PCP’s per 1,000 population
Mean prospective risk score (adult)
Mean prospective risk score
(pediatric)
Percent of nonwhite
attributed members
Percent of female attributed members
Percent BCBSM market share

IQR

Median

IQR

74.3%
70.4%
33.7%
76.2%
$296.67
$80.19

66.5–80.0%
63.6–76.6%
21.2–46.7%
50.0–100.0%
$252.40–$357.56
$56.50–$105.17

NA
NA
64.0%
86.8%
NA
$96.82

NA
NA
56.6–70.6%
78.7–92.5%
NA
$81.44–$114.42

0.06
0.19
$48,363
111
1,979
0.98
1.60
0.45

0–0.19
0.05–0.35
$44,843–$58,332
59–710
1,132–3,209
0.71–1.26
1.41–1.87
0.38–0.54

0.06
0.23
$50,666
104
3,054
1.04
0.67
0.44

0–0.15
0.08–0.38
$43,929–$55,321
55–177
1,870–5,071
0.77–1.40
0.58–0.77
0.40–0.50

Categorical Variables
Practice size
Solo physician practice
2–3 physicians
4–5 physicians
6 or more physicians
Practice specialty
Mixed
Primary care only
Metropolitan statistical area status
Metropolitan: 1,000,000 or more persons
Metropolitan: 250,000–999,999 persons
Metropolitan: 100,000–249,999 persons
Metropolitan: below 100,000 persons
Micropolitan
Rural

20.5%

12.1–26.8%

21.3%

14.0–25.9%

50.8%
31.1%

45.9–58.2%
25.7–34.4%

48.6%
31.3%

46.7–50.7%
26.0–34.7%

N

%

N

%

1,274
500
189
173

59.6
23.4
8.8
8.1

137
87
43
29

46.3
29.4
14.5
9.8

83
2,053

3.9
96.1

13
283

4.4
95.6

754
514
406
69
208
177

35.3
24.1
19.0
3.2
9.7
8.3

121
84
49
5
23
14

40.9
28.4
16.6
1.7
7.8
4.7

Note. IQR, interquartile range; NA, not applicable; PCMH, patient-centered medical home;
PMPM, per member per month.


13

Table 3 shows the PCMH effect estimates from the multivariable models adjusting for member, practice, and market characteristics. Full multivariable model results are available in appendices C–E.
Preventive Care Measures
The multivariable model for the adult prevention composite measure
included 1,636 study practices after practice exclusions: insufficient sample
size (303), missing predictors (42), and exceeding IQR thresholds (155). Practices included in the model had a median of 194 prevention opportunities per
practice, and the mean adult preventive composite score was 74.8 percent,
ranging from 38.7 to 94.5 percent. After multivariable adjustment, a practice
that achieved full PCMH implementation would have a 5.1 percent higher
adult preventive composite score compared with a practice that never
achieved any PCMH implementation (p = .0316). A practice without preexisting PCMH infrastructure that implemented all PCMH capabilities during
the study time period would have a 3.3 percent higher adult preventive composite score compared with that same practice with no incremental PCMH
implementation (p = .0028). Effect estimates were slightly higher for baseline
implementation (6.3 percent) and incremental implementation (3.8 percent)
when practices exceeding the IQR thresholds were included in the model as a
sensitivity test.
The pediatric preventive composite measure multivariable model
included 1,218 study practices after practice exclusions: insufficient sample
size (1,151), missing predictors (39), and exceeding IQR thresholds (24). The
median number of preventive opportunities was 115 per practice included in
the model, and the mean pediatric preventive composite score was 44.5 percent, ranging between 10.1 and 85.2 percent. After multivariable adjustment,
a practice that achieved full PCMH implementation would have a 12.2 percent higher pediatric preventive composite score compared with a practice
that never achieved any PCMH implementation (p = .0008). A practice without preexisting PCMH infrastructure that implemented all PCMH capabilities during the study time period would have a 4.9 percent higher pediatric
preventive composite score compared with that same practice with no incremental PCMH implementation (p = .0260). While the random intercept
model resulted in an increased effect of full baseline PCMH implementation
from 12.2 to 15.7 percent, it did not improve model fit (AIC: À1,577) over the
model without a random intercept (AIC: À1,719) for the pediatric preventive

9.7%
19.3%

À0.4%
À0.05
À$53.08
À$13.13

3.5%
0.10

À$26.37
À$1.72
$0.33
$9.70

7.4%
0.25

0.5%
5.1%

5.1%
12.2%

.0529
.7682

.0806
.1745

.0316
.0008

À$3.08
$7.45

5.2%
0.05

3.3%
4.9%

À$18.07
À$1.33

2.6%
À0.07

1.1%
0.6%

$11.91
$16.24

7.8%
0.17

5.4%
9.3%

.6868
.0964

<.0001
.4113

.0028
.0260

Beta Estimate 95% CI (Lower) 95% CI (Upper) p-Value Beta Estimate 95% CI (Lower) 95% CI (Upper) p-Value

PCMH Change from June 2009 to June 2010

Note. Full multivariable model results are available in appendices C–E. Covariates included the following practice characteristics: mean prospective risk
score, percent female attributed members, paid services per PCP, practice size based on total physicians in the practice, practice as primary care only or
mixed primary and specialty care, and whether the practice was a pediatric practice or not; and market characteristics: total primary care practices in
the physician organization, median household income, PCPs per 1,000 population, BCBSM market share, percent nonwhite residents, and metropolitan statistical area status.
*Log transformed.
BCBSM, Blue Cross Blue Shield of Michigan; CI, conﬁdence interval; PCMH, patient-centered medical home.

Preventive composite scores
Adult preventive composite
Pediatric preventive composite
Quality composite scores
Adult quality composite
Pediatric quality composite*
Medical and surgical PMPM costs
Adult PMPM cost
Pediatric PMPM cost

Outcome Variable

Baseline PCMH Score – June 2009

Table 3: Multivariable Generalized Estimating Equation Results: Medical Home Implementation and Adult and Pediatric Composite Quality of Care Measures, Composite Preventive Measures, and Per Member Per Month (PMPM) Medical and Surgical Costs

14


15

composite measure. Random intercept models did not identify any noticeable
changes in parameters or model fit for any other outcome.
Quality of Care Measures
The multivariable model for the adult quality composite measure consisted of
1,590 study practices after practice exclusions: insufficient sample size (498),
missing predictors (43), and exceeding IQR thresholds (5). For practices
included in the model, the median number of quality opportunities per practice was 166, and the mean adult quality composite score was 70.2 percent,
ranging between 33.8 and 96.4 percent. After multivariable adjustment, a
practice that achieved full PCMH implementation would have a 3.5 percent
higher adult quality composite compared with a practice that never achieved
any PCMH implementation (p = .0806). A practice without preexisting
PCMH infrastructure that implemented all PCMH capabilities during the
study time period would have a 5.2 percent higher adult quality composite
score compared with that same practice with no incremental PCMH implementation (p < .0001).
A total of 337 study practices were included in the multivariable model
for the pediatric quality measure after practice exclusions: insufficient sample
size (2,077), missing predictors (7), and exceeding IQR thresholds (11). Practices included in the multivariable model had a median of 61 quality opportunities per practice, and the mean untransformed pediatric quality measure was
80.4 percent, ranging between 41.8 and 100 percent. Although positively associated, neither baseline nor incremental PCMH implementation was significantly associated with a higher pediatric quality score after multivariable
adjustment (p = .1745 and p = .4113, respectively).
Cost Measures
A total of 1,787 study practices were used for the adult PMPM cost multivariable model after the following practice exclusions: insufficient sample size
(286), missing predictors (45), exceeding IQR thresholds (15), and influential
observations (3). Practices included in the model had a median population of
303 adult members, and their overall mean adult PMPM cost was $310.79,
ranging between $102.72 and $1,053.45. After multivariable adjustment, a
practice that achieves full PCMH implementation would have a $26.37
PMPM lower adult PMPM cost compared with a practice that never achieved
any PCMH implementation (p = .0529). Based on a practice with median

16


population characteristics, this difference corresponds to a 7.7 percent (95 percent CI: À0.1 to 15.4 percent) lower adult PMPM cost. Incremental PCMH
implementation was associated with lower cost, albeit nonsignificant
(p = .6868).
The pediatric PMPM cost multivariable model included 956 practices
after practice exclusions: insufficient sample size (1,438), missing predictors
(28), exceeding IQR thresholds (5), and influential observations (5). Practices
included in the multivariable model had a median 172 pediatric members per
practice, and their mean pediatric PMPM cost was $91.40, ranging from
$21.59 to $236.14. After multivariable adjustment, baseline PCMH implementation was not associated with cost (p = .7682). However, a practice without preexisting PCMH infrastructure that implemented all PCMH
capabilities during the study time period would have a $7.45 higher pediatric
PMPM compared with that same practice with no incremental PCMH implementation (p = .0964).

DISCUSSION
Our study demonstrated positive associations between baseline PCMH
implementation and composite measures of quality of care for both adults and
children, and we observed these associations for both primary and secondary
preventive care. PCMH implementation was also associated with lower overall medical and surgical costs for adults, but not for children. This cost finding
is consistent with the underlying PCMH principles of improved chronic care
management and care coordination due to the higher chronic disease burden
in adults.
Notably, these associations between PCMH and cost and quality measures were observed 2 years into the program, even though practices were still
expanding their PCMH infrastructure. These findings suggest that partial
implementation of the PCMH model may have quality and cost benefits well
before full PCMH implementation has been achieved, and these benefits
likely increase as practices progress toward full implementation. Although the
PCMH is intended to operate as a system of care, our findings suggest that its
constituent elements may have independent benefits and that substantial progress on all or most elements is not needed to significantly improve care.
Composite measures of quality and preventive care were also positively
associated with incremental PCMH implementation, controlling for baseline
levels of PCMH implementation. However, cost measures were not


17

associated with incremental PCMH implementation. These observations may
reflect the proximal impacts of recent implementation, whereby initial
improvements in quality are followed later by reductions in cost. Although
not statistically significant, the increased costs for pediatric members associated with incremental PCMH implementation may result from an increased
emphasis and use of preventive services reflected in the pediatric preventive
composite measure of well-child visits and immunizations.
The measures of association between PCMH implementation and quality and cost-related outcomes are intended to estimate the potential PCMH
effects when full implementation is eventually achieved. However, extrapolation of results observed at partial implementation to estimate full implementation effects should be interpreted cautiously, as the relationship between these
outcomes and PCMH implementation may change at higher levels of implementation. While the effect could diminish at higher levels, we believe that
synergistic interaction of PCMH elements will more likely result in greater
effects on outcomes as practices achieve higher levels of PCMH implementation. Although the measurement instrument developed for this program may
not capture all elements crucial to creating a fully functional PCMH, this
instrument was developed with significant physician input, incorporated the
PC-PCC guiding principles, and was specifically designed to capture partial
PCMH implementation and incremental PCMH changes under multiple
implementation scenarios (Alexander et al. 2012).
A major strength of this study is that these practices encompass nearly
two thirds of primary care physicians practicing in Michigan, span 82 of the
83 counties in Michigan, represent both small and large practices, urban and
rural practices, practices within integrated systems, and practices loosely affiliated in independent physician associations—all approaching PCMH implementation from a broad array of perspectives suited to the individual
practices. The external validity is furthered by evaluating the impact on the
total population rather than smaller chronic disease subpopulations that may
overstate the total population benefits of a PCMH. Although limited to a single payer’s data, BCBSM members represent a substantial portion of practice
populations statewide, but this raises questions as to whether the PCMH
effects we observed extend to other commercial or noncommercial payer populations. However, recent literature suggests a need to evaluate PCMH in a
broad array of practice settings with greater emphasis on the diversity of practices rather than the number of patients (Peikes et al. 2012).
Our PCMH evaluation was limited to a 1-year period, so we cannot
determine whether these results reflect true causal relationships or preexisting

18


practice patterns. For example, physicians motivated to improve quality and
reduce costs may be more likely to implement PCMH capabilities, and such
motivation may explain observed associations between PCMH and quality
and cost measures. The PCMH change variable may capture elements of physician motivation, as more motivated practices might expand their capabilities
more rapidly initially. However, such expansion may wane as fewer opportunities to expand PCMH capabilities are available and only relatively difficult
to implement capabilities remain. Additional limitations include the use of
county-level socioeconomic characteristics instead of individual socioeconomic characteristics and the lack of information on benefit design.
This study demonstrated relationships between the PCMH model,
higher quality of care, and reduced cost of care. Based on this initial assessment, the study supports continued investment in PCMH implementation.
However, further evaluation is needed to determine the sources of utilization
(e.g., emergency department, inpatient) affected by the PCMH that contribute
to lower costs, whether the association between PCMH and medical costs
extends to pharmacy costs, whether these effects span multiple payers, and
whether these effects persist under a longitudinal design. Additional efforts
should aim to understand which specific PCMH elements contribute to higher
quality care and lower cost of care, and potential for synergism between
PCMH elements with regard to these outcomes. Importantly, the associations
with quality we observed span a broad array of primary and secondary preventive measures rather than narrowly focused on any one individual measure. Further research is needed to determine whether these relationships
span other quality measures, are limited to specific subsets of quality measures, or extend into additional health areas such as preconception care where
chronic condition management may yield additional health benefits ( Johnson
et al. 2006). The diverse population of practices in this study population will
allow for subsequent examination of these areas and whether PCMH effects
are universal across practices or dependent on the practice setting.
Finally, although multiple statistically significant relationships between
PCMH and quality measures were observed in this study, our results do not
directly address the clinical significance of these relationships, an important
distinction given the potentially high costs associated with implementing and
maintaining a transformational change such as PCMH. Three points are relevant to this issue. First, many of our quality measures are based on established
clinical guidelines strongly linked to improved health outcomes in other studies, such as monitoring HbA1c levels in persons with diabetes (Larsen, Horder, and Mogensen 1990). Second, it may be premature to assess the cost


19

effectiveness of PCMH as many study practices are still in the early stages of
PCMH implementation and have not yet fully realized its clinical beneﬁts.
Third, even relatively small changes at the physician practice level may translate into important differences at the population level. The fact that our ﬁndings apply to a large study sample of over 2,000 practices in Michigan
partially supports this claim and may mitigate to some extent the costs of
implementation.
If the cost savings and quality improvement relationships observed in
this study are reinforced by additional evaluations of the PCMH model, further support for PCMH may be warranted. Implementing PCMH capabilities
presents a considerable challenge for many primary care practices, with significant investment of time and expense (Nutting et al. 2011). Requiring primary
care practices to shoulder this investment alone may severely limit PCMH
implementation. Payers, purchasers, and providers should consider methods
for sharing cost savings derived from PCMH implementation to provide
further incentives to support ongoing efforts to implement the PCMH model.
If cost savings and improved quality can indeed be derived at intermediate
stages of PCMH implementation, and sustained during more advanced stages
of implementation, the potential for a permanent program of shared savings
that support continuous improvements may well be viable.

ACKNOWLEDGMENTS
Joint Acknowledgment/Disclosure Statement: This study was funded by a grant
from the Agency for Healthcare Research and Quality: R18 RFA-HS-10-002.
Data were supplied by Blue Cross Blue Shield of Michigan. Michael Paustian
and Darline El Reda are employees of Blue Cross Blue Shield of Michigan.
Disclosures: None.
Disclaimers: None.

REFERENCES
Alexander, J. A., and D. Bae. 2012. Does the Patient-Centred Medical Home Work? A
Critical Synthesis of Research on Patient-Centred Medical Homes and PatientRelated Outcomes. Health Services Management Research 25 (2): 51–9.
Alexander, J. A., C. G. Wise, L. A. Green, M. D. Fetters, M. H. Mason, and D. K. El
Reda. 2013. Assessment and Measurement of Patient-Centered Medical Home

20


Implementation: the BCBSM Experience. Annals of Family Medicine 11 (Suppl 1):
S74–81.
Area Resource File (ARF). 2010. 2009–2010. Rockville, MD: US Department of
Health and Human Services, Health Resources and Services Administration,
Bureau of Health Professions.
Audet, A. M., K. Davis, and S. C. Schoenbaum. 2006. “Adoption of Patient-Centered
Care Practices by Physicians: Results from a National Survey.” Archives of Internal
Medicine 166 (7): 754–9.
Blue Cross Blue Shield of Michigan. September 2010a. BCBSM Physician Group Incentive Program Patient-Centered Medical Home Domains of Function: Interpretive Guidelines. Detroit, MI: Blue Cross Blue Shield of Michigan, Value Partnerships.
Blue Cross Blue Shield of Michigan. November 2010b. BCBSM Physician Group Incentive Program Evidence Based Care Report (EBCR) Measure Speciﬁcations for PGIP
2010 Program Year. Detroit, MI: Blue Cross Blue Shield of Michigan, Value Partnerships.
Cooley, W. C., J. W. McAllister, K. Sherrieb, and K. Kuhlthau. 2009. “Improved Outcomes Associated with Medical Home Implementation in Pediatric Primary
Care.” Pediatrics 124 (1): 358–64.
The Commonwealth Fund. 2011. Patient-Centered Coordinated Care Programs
[accessed on March 18, 2011]. Available at http://www.commonwealthfund.
org/Content/Program-Areas/Delivery-System-Innovation and-Improvement/
Patient-Centered-Coordinated-Care.aspx.
The Commonwealth Fund. 2012. Recommended Core Measures for Evaluating the
Patient-Centered Medical Home: Cost, Utilization, and Clinical Quality
[accessed on June 19, 2012]. Available at http://www.commonwealthfund.org/
Publications/Data-Briefs/2012/May/Measures-Medical-Home.aspx.
Damiano, P. C., E. T. Momany, M. C. Tyler, A. J. Penziner, and J. C. Lobas. 2006.
“Cost of Outpatient Medical Care for Children and Youth with Special Health
Care Needs: Investigating the Impact of the Medical Home.” Pediatrics 118 (4):
e1187–94.
DeVoe, J. E., L. S. Wallace, N. Pandhi, R. Solotaroff, and G. E. Fryer Jr. 2008. “Comprehending Care in a Medical Home: A Usual Source of Care and Patients Perceptions about Healthcare Communication.” Journal of the American Board of
Family Medicine 21 (5): 441–50.
Fields, D., E. Leshen, and K. Patel. 2010. “Driving Quality Gains and Costs Savings
through Adoption of Medical Homes.” Health Affairs 29 (5): 819–26.
Fox, J. 1991. Regression Diagnostics: An Introduction. Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-079. Newbury Park, CA: Sage.
Friedberg, M., D. Safran, K. Coltin, M. Dresser, and E. Schneider. 2009. “Readiness
for the Patient-Centered Medical Home: Structural Capabilities of Massachusetts Primary Care Practices.” Journal of General Internal Medicine 24 (2): 162–9.
Gill, J. M., H. Bittner Fagan, B. Townsend, and A. G. Mainous III. 2005. “Impact of
Providing a Medical Home to the Uninsured: Evaluation of a Statewide Program.” Journal of Health Care for the Poor and Underserved 16 (3): 515–35.


21

Goldberg, D. G., and A. J. Kuzel. 2009. “Elements of the Patient-Centered Medical
Home in Family Practices in Virginia.” Annals of Family Medicine 7 (4): 301–8.
Higgins, A., K. Stewart, K. Dawson, and C. Bocchino. 2011. “Early Lessons from
Accountable Care Models in the Private Sector: Partnerships between Health
Plans and Providers.” Health Affairs 30 (9): 1718–27.
Hoff, T., W. Weller, and M. DePuccio. 2012. “The Patient-Centered Medical Home: A
Review of Recent Research.” Published online before print May 29, 2012. Medical Care Research and Review 69 (6): 619–44.
Hubbard, A. E., J. Ahern, N. L. Fleischer, M. Van der Laan, S. A. Lippman, N. Jewell,
T. Bruckner, and W. A. Satariano. 2010. “To GEE or Not to GEE: Comparing
Population Average and Mixed Models for Estimating the Associations between
Neighborhood Risk Factors and Health.” Epidemiology 21 (4): 467–74.
Ingenixâ. 2008. Symmetry Episode Risk Groupsâ: A Successful Approach to Health Risk
Assessment. Eden Prairie, MN: Ingenix.
Jaen, C. R., R. L. Ferrer, W. L. Miller, R. F. Palmer, R. Wood, M. Davila, E. E. Stewart,
B. F. Crabtree, P. A. Nutting, and K. C. Stange. 2010. “Patient Outcomes at
26 Months in the Patient-Centered Medical Home National Demonstration
Project.” Annals of Family Medicine 8 (Suppl 1): S57–67.
Johnson, K., S. F. Posner, J. Biermann, J. F. Cordero, H. K. Atrash, C. S. Parker, S.
Boulet, and M. G. Curtis, CDC/ATSDR Preconception Care Work Group;
Select Panel on Preconception Care. 2006. Recommendations to Improve
Preconception Health and Health Care—United States. A Report of the CDC/
ATSDR Preconception Care Work Group and the Select Panel on Preconception Care. MMWR Recommended Reports, 55 (RR-6): 1–23.
Larsen, M. L., M. Horder, and E. F. Mogensen. 1990. “Effect of Long-Term Monitoring
of Glycosylated Hemoglobin Levels in Insulin-Dependent Diabetes Mellitus.”
New England Journal of Medicine 323 (15): 1021–5.
Martin, A. B., S. Crawford, J. C. Probst, G. Smith, R. P. Saunders, K. W. Watkins, and
K. Luchok. 2007. “Medical Homes for Children with Special Health Care
Needs.” Journal of Health Care for the Poor and Underserved 18 (4): 916–30.
McBurney, P. G., K. N. Simpson, and P. M. Darden. 2004. “Potential Cost Savings of
Decreased Emergency Department Visits through Increased Continuity in a
Pediatric Medical Home.” Ambulatory Pediatrics 4 (3): 204–8.
Michigan Department of Community Health. 2010. MDCH Survey of Physicians, Survey
Findings 2010. January 2011. Lansing, MI: Public Sector Consultants.
Nutting, P. A., B. F. Crabtree, W. L. Miller, K. C. Stange, E. Stewart, and C. Jaen. 2011.
“Transforming Physician Practices to Patient-Centered Medical Homes: Lessons
from the National Demonstration Project.” Health Affairs 30 (3): 439–45.
Palfrey, J. S., L. A. Soﬁs, E. J. Davidson, J. Liu, L. Freeman, and M. L. Ganz. 2004.
“The Pediatric Alliance for Coordinated Care: Evaluation of a Medical Home
Model.” Pediatrics 113 (5 Suppl): 1507–16.
Parkerton, P. H., D. G. Smith, T. R. Belin, and G. A. Feldbau. 2003. “Physician Performance Assessment: Nonequivalence of Primary Care Measures.” Medical Care
41 (9): 1034–47.

22


Patient-Centered Primary Care Collaborative. 2007. Joint Principles of the PatientCentered Medical Home [accessed May 7, 2013]. Available at http://www.
pcpcc.net/joint-principles.
Patient-Centered Primary Care Collaborative. 2009. Proof in Practice: A Compilation
of Patient Centered Medical Home Pilot and Demonstration Projects [accessed
on June 7, 2013]. Available at http://www.pcpcc.net/sites/default/ﬁles/medial/
PilotGuidePip.pdf.
Peikes, D., A. Zutshi, J. L. Genevro, M. L. Parchman, and D. S. Meyers. 2012. “Early
Evaluations of the Medical Home: Building on a Promising Start.” American
Journal of Managed Care 18 (2): 105–16.
Rankin, K. M., A. Cooper, K. Sanabria, H. J. Binns, and C. Onufer. 2009. “Illinois
Medical Home Project: Pilot Intervention and Evaluation.” American Journal of
Medical Quality 24 (4): 302–9.
Reeves, D., S. M. Campbell, J. Adams, P. G. Shekelle, E. Kontopantelis, and M. O.
Roland. 2007. “Combining Multiple Indicators of Clinical Quality: An Evaluation of Different Analytic Approaches.” Medical Care 45 (6): 489–96.
Reid, R. J., P. A. Fishman, O. Yu, T. R. Ross, J. T. Tufano, M. P. Soman, and E. B.
Larson. 2009. “Patient-Centered Medical Home Demonstration: A Prospective,
Quasi-Experimental, before and after Evaluation.” American Journal of Managed
Care 15 (9): 71–87.
Rittenhouse, D. R., L. P. Casalino, R. R. Gillies, S. M. Shortell, and B. Lau. 2008. “Measuring the Medical Home Infrastructure in Large Medical Groups.” Health
Affairs 27 (5): 1246–58.
Rittenhouse, D., L. Casalino, S. Shortell, and J. A. Alexander. 2011. “Small and Medium-Size Physician Practices Use Few Patient-Centered Medical Home
Processes.” Health Affairs 30 (8): 1575–85.
SAS Institute Inc. 2011. SASâ 9.2 for Windows. Cary, NC: SAS.
Scholle, S. H., J. Roski, D. L. Dunn, J. L. Adams, D. P. Dugan, L. G. Pawlson, and E. A.
Kerr. 2009. “Availability of Data for Measuring Physician Quality Performance.”
American Journal of Managed Care 15 (1): 67–72.
Share, D. A., and M. H. Mason. 2012. “Michigan’s Physician Group Incentive Program
Offers a Regional Model for Incremental ‘Fee for Value’ Payment Reform.”
Health Affairs 31 (9): 1993–2001.
Singer, J. D. 1998. “Using SAS Proc Mixed to Fit Multilevel Models, Hierarchical
Models and Individual Growth Models.” Journal of Educational and Behavioral Statistics 24 (4): 323–55.

S UPPORTING I NFORMATION
Additional supporting information may be found in the online version of this
article:
Appendix A: Example Showing Capability and Domain Scoring for the
Patient–Provider Agreement PCMH Functional Domain, June 2010.


23

Appendix B: Individual Component Measures of Adult and Pediatric
Quality and Preventive Composite Measures and Source Deﬁnition for Each
Individual Component Measure.
Appendix C: Multivariable Generalized Estimating Equation Model for
the Association between Medical Home Implementation and Adult and Pediatric Composite Quality of Care Measures in BCBSM PGIP Practices, July
2009 to June 2010.
Appendix D: Multivariable Generalized Estimating Equation Model for
the Association between Medical Home Implementation and Adult and Pediatric Composite Preventive Measures in BCBSM PGIP Practices, July 2009
to June 2010.
Appendix E: Multivariable Generalized Estimating Equation Model for
the Association between Medical Home Implementation and Adult and Pediatric Medical and Surgical per Member per Month Costs in BCBSM PGIP
Practices, July 2009 to June 2010.
Appendix SA1: Author Matrix.

Partial and Incremental PCMH Practice Transformation: Implications for Quality and Costs Michael L. Paustian, Jeffrey A. Alexander, Darline K. El Reda, Chris G. Wise, Lee A. Green, and Michael D. Fetters Mich bcbs july 2013

More Related Content

What's hot

Similar to Partial and Incremental PCMH Practice Transformation: Implications for Quality and Costs Michael L. Paustian, Jeffrey A. Alexander, Darline K. El Reda, Chris G. Wise, Lee A. Green, and Michael D. Fetters Mich bcbs july 2013

More from Paul Grundy

Recently uploaded

Partial and Incremental PCMH Practice Transformation: Implications for Quality and Costs Michael L. Paustian, Jeffrey A. Alexander, Darline K. El Reda, Chris G. Wise, Lee A. Green, and Michael D. Fetters Mich bcbs july 2013