Associations Between Practice-Reported Medical
Homeness and Health Care Utilization Among
Publicly Insured Children
Anna L. Christensen, PhD; Joseph S. Zickafoose, MD, MS; Brenda Natzke, MPP;
Stacey McMorrow, PhD; Henry T. Ireys, PhD
From the Mathematica Policy Research, Washington, DC (Dr Christensen, Ms Natzke, and Dr Ireys); Mathematica Policy Research, Ann
Arbor, Mich (Dr Zickafoose); and The Urban Institute, Washington, DC (Dr McMorrow)
The authors declare that they have no conflict of interest.
Address correspondence to Anna L. Christensen, PhD, Mathematica Policy Research, 1100 1st St NE, 12th Floor, Washington, DC 20002
(e-mail: achristensen@mathematica-mpr.com).
Received for publication June 12, 2014; accepted December 8, 2014.
ABSTRACT
BACKGROUND: The patient-centered medical home (PCMH)
is widely promoted as a model to improve the quality of
primary care and lead to more efficient use of health care
services. Few studies have examined the relationship between
PCMH implementation at the practice level and health care uti-
lization by children. Existing studies show mixed results.
METHODS: Using practice-reported PCMH assessments and
Medicaid claims from child-serving practices in 3 states partici-
pating in the Children’s Health Insurance Program Reauthoriza-
tion Act of 2009 Quality Demonstration Grant Program, this
study estimates the association between medical homeness (ter-
tiles) and receipt of well-child care and nonurgent, preventable,
or avoidable emergency department (ED) use. Multilevel logis-
tic regression models are estimated on data from 32 practices in
Illinois (IL) completing the National Committee for Quality As-
surance’s (NCQA) medical home self-assessment and 32 prac-
tices in North Carolina (NC) and South Carolina (SC)
completing the Medical Home Index (MHI) or Medical Home
Index—Revised Short Form (MHI-RSF).
RESULTS: Medical homeness was not associated with receipt of
age-appropriate well-child visits in either sample. Associations be-
WHAT’S NEW
Children in practices with high medical homeness were
less likely to have a nonurgent, preventable, or avoid-
able emergency department visit than those with low
or medium medical homeness, but only among prac-
tices in Illinois scored using the NCQA self-
assessment, not among a second sample.
THE PATIENT-CENTERED MEDICAL home (PCMH) is
widely promoted as a model to improve quality of primary
care and lead to more efficient use of health care services.1–
3
Programs promoting PCMH adoption typically measure
fidelity to the model, or medical homeness, through
primary care practices’ responses to inventories of
practice infrastructure and care processes.
In pediatric primary care, some of the most commonly
used practice-level PCMH measures include the National
ACADEMIC PEDIATRICS
Copyright ª 2015 by Academic Pediatric Association
Downloaded for Siti Fadhilah (fadhilahhazhiyah@gmail.com) at Universitas Muslim Indonesia from ClinicalKey.com by Elsevier on May 05, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
tween nonurgent, preventable, or avoidable ED visits and medical
homeness varied. No association was seen among practices in NC
and SC that completed the MHI/MHI-RSF. Children in practices
in IL with the highest tertile NCQA self-assessment scores were
less likely to have a nonurgent, preventable, or avoidable ED visit
than children in practices with low (odds ratio 0.65; 95% confidence
interval 0.47–0.92; P <.05) and marginally less likely to have such a
visit compared with children in practices with medium tertile scores
(odds ratio 0.72, 95% confidence interval 0.52–1.01; P ¼ .06).
CONCLUSIONS: Higher levels of medical homeness may be
associated with lower nonurgent, preventable, or avoidable
ED use by publicly insured children. Robust longitudinal
studies using multiple measures of medical homeness are
needed to confirm this observation.
KEYWORDS: child health services; health care quality, access,
and evaluation; medical home; patient-centered care; primary
health care
ACADEMIC PEDIATRICS 2015;15:267–274
Committee for Quality Assurance’s (NCQA) PCMH recog-
nition program4 and the Medical Home Index (MHI), often
used for internal quality improvement and practice transfor-
mation.5 However, little research has examined the rela-
tionship between PCMH implementation, as reflected in
these practice-level PCMH measures, and children’s use
of health services. Studies that examine parent-reported
medical homeness and health care utilization are more
common and are often based on data from a few large na-
tional surveys.6 These studies have had mixed findings. In
the general pediatric population, parent-reported care in a
medical home has been associated with increased receipt
of well-child care,7–9 decreased emergency department
(ED) use,7,8 and no differences in hospitalizations.8 In chil-
dren with chronic conditions, parent-reported care in a
medical home has been associated with no differences in
receipt of preventive care,10,11 mixed results for ED
visits,10,12,13 and no differences in hospitalizations.10
Volume 15, Number 3
267 May–June 2015
268 CHRISTENSEN ET AL
Two published studies that have examined the relation-
ship between practice-reported PCMH measures and
health care utilization by children also show mixed results.
In a study of children with chronic conditions in 43 primary
care practices in multiple states, Cooley et al14 found that
children were less likely to be hospitalized if their primary
care practice scored higher on the MHI, but there was no
reduction in ED use, a key target of many PCMH efforts.5
In a study of 296 primary care practices in Michigan, Paus-
tian et al15 found that children were more likely to receive
preventive services but had no differences in overall costs
when receiving care in practices that scored higher on an
insurer-specific practice-level PCMH measure.
The aim of the present analysis was to assess the rela-
tionship between practice-reported medical homeness
and health service use by children enrolled in Medicaid
in 3 states. We hypothesized that children would be more
likely to attend preventive care visits and less likely to
use the ED for nonurgent, potentially avoidable, and pri-
mary care–treatable conditions when receiving care in pri-
mary care practices with higher levels of medical
homeness. A secondary aim was to assess whether the as-
sociation between medical homeness and health care use
differed for children with chronic conditions and disabil-
ities versus all other children. Given the widespread pro-
motion and implementation of the PCMH model, more
information is needed to understand the association be-
tween practice-level measures of the PCMH and health
care utilization for children in a variety of populations.
This analysis helps fill the gap in the current literature.
METHODS
DATA
We performed a cross-sectional analysis assessing the
relationship between practice-reported medical homeness
and health service use by children enrolled in Medicaid
in 64 practices in 3 states participating in the Children’s
Health Insurance Program Reauthorization Act (CHIPRA)
Quality Demonstration Grant Program: Illinois (IL), North
Carolina (NC), and South Carolina (SC).16 The 2009 CHI-
PRA legislation authorized $100 million in grants to states
to test various approaches to improve the quality of health
care for children in the United States, particularly publicly
insured children. As part of this effort, several states pur-
sued projects to promote the implementation of the
PCMH model in primary care practices serving children.
The 3 states in this analysis have varied histories with pri-
mary care transformation and represent different state
Medicaid contexts, including managed care penetration.
Intervention practices in all 3 states volunteered to partic-
ipate with the aim of primary care practice transformation.
NC and SC also recruited comparison practices. This anal-
ysis uses baseline data that were collected for the national
evaluation of the program, funded by the Agency for
Healthcare Research and Quality (AHRQ).16
Medical homeness was assessed via standardized
practice-reported surveys. Each state selected the medical
home assessment that was most appropriate for its demon-
Downloaded for Siti Fadhilah (fadhilahhazhiyah@gmail.com) at Universitas Muslim Indonesia from ClinicalKey.com by Elsevier on May 05, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
ACADEMIC PEDIATRICS
stration project.17 NC’s practices completed the MHI, and
SC’s completed the Medical Home Index—Revised Short
Form (MHI-RSF).5,17 We have shown that the MHI and
MHI-RSF rank practices similarly17; consequently, we
combined data from NC and SC. NC practices were sur-
veyed in mid- to late 2011, at the beginning of the interven-
tion. SC practices were surveyed in early 2012, more than 1
year after the intervention began. Practices in IL completed
the NCQA 2011 medical home self-assessment in late
2011 to early 2012, less than a year before the intervention
began.4 The self-assessment does not require practices to
submit documentation to NCQA.
Patient characteristics and health care utilization were
assessed via Medicaid enrollment and claims data from
the calendar year before medical home assessment: 2010
in NC and IL, and 2011 in SC. Because the SC PCMH
intervention began in 2011, we were concerned that the
MHI-RSF score collected in early 2012 would not reflect
the level of medical homeness corresponding to utilization.
Thus, we exclude the SC intervention practices and include
only the comparison practices in this baseline analysis. Uti-
lization data included fee for service (FFS) or primary care
case management (PCCM) claims only and excluded tradi-
tional managed care encounters. Depending on the state,
managed care encounters were not available to researchers
or could not be linked to practices participating in the
demonstration due to the variation in provider identifica-
tion variables across managed care organizations.
The New England institutional review board deemed
this phase of the evaluation as exempt from review because
it used existing administrative data and practice-level data
(45 CFR 46.101(b)(4)).
STUDY POPULATION
The study population included children aged 0 to 18
years continuously enrolled for 12 months (with no more
than a 1-month gap in coverage) in FFS or PCCM
Medicaid who were attributed to primary care practices
enrolled in the demonstration in the 3 states. Children
were excluded if they received partial Medicaid benefits;
received benefits through a Medicaid waiver program;
had any other source of insurance coverage; or were insti-
tutionalized. Children were attributed to a primary care
practice using a staged, claims-based approach. First, chil-
dren were attributed to the practice where they received the
majority of their well-child visits (WCV) in the calendar
year (of children who could be attributed, 72% of children
in IL, 67% of children in NC, and 43% of children in SC
were attributed in this stage). Children without a WCV
were attributed to the practice where they received the ma-
jority of their other ambulatory care visits (24% of children
in IL, 30% of children in NC, and 50% of children in SC
were attributed in this stage). In the remaining cases where
there was no majority practice for WCV or other ambula-
tory visits (4% of children in IL, 3% of children in NC,
and 7% of children in SC), children were attributed to
the practice they attended most recently.
Practices participating in the demonstration were identi-
fied by billing and/or service provider identification
ACADEMIC PEDIATRICS
numbers supplied by the states. There is no variable to link
children to their primary care provider other than the billing
provider on a claim, so children without any WCVor ambu-
latory claims could not be attributed to a primary care
practice and were excluded from this analysis. Further in-
formation on our attribution algorithm may be obtained
upon request. Practices with fewer than 30 Medicaid
children attributed to them were excluded from the analysis
(6 practices in IL; no practices in NC or SC) to ensure a
sufficient sample size to estimate within-practice variation.
MEASURES
Medical homeness was measured using the MHI/MHI-
RSF or the NCQA self-assessment. We used the NCQA
self-assessment total score rather than recognition level
because we were interested in the overall relationship be-
tween medical homeness and service use and because no
practices were NCQA certified. Scores for all tools range
from 0 to 100; practices were categorized into tertiles, rep-
resenting low, medium, and high levels of medical home-
ness. Tertile cutoffs were set separately for practices
completing the MHI/MHI-RSF (NC, SC) and the NCQA
2011 self-assessment (IL).
We generated 2 measures of health care utilization:
receipt of the age-appropriate number of WCV, and use
of ED services for nonurgent, potentially avoidable, or pri-
mary care–treatable conditions. The WCV outcome is a bi-
nary measure of whether the child received at least 75% of
the recommended number of WCV for their age during the
measurement year, based on Bright Futures, the American
Academy of Pediatrics’ guideline and generally recognized
standard for preventive care of children.18 For children un-
der 36 months of age, the number of recommended visits
was determined for their age in months. All children 36
months and older are recommended to have one WCV
per year. We identified WCV using procedure codes from
the Centers for Medicare and Medicaid Services Core Set
of Children’s Health Care Quality Measures.19
The ED use outcome is a binary measure of whether the
child had any nonurgent, primary care–treatable, or poten-
tially avoidable ED visits in the measurement year. These
are ED visits that could hypothetically be reduced or
avoided when children receive continuous and appropriate
care in a medical home. The measure uses ED procedure
and diagnosis codes aligned with the Centers for Medicare
and Medicaid Services Core Set of Children’s Health Care
Quality Measures.19 We then categorized ED visits using a
modified version of the NYU algorithm, a claims-based
method for classifying the severity of ED visits based on
ICD-9 diagnosis codes.20–22 In the present study, visits
were considered to be nonurgent, primary care–treatable,
or potentially avoidable if 1) the NYU algorithm
classified them as having a 0.75 or higher probability of
being nonemergent plus emergent but primary care–
treatable plus emergent but preventable or avoidable, or
2) if they contain a diagnosis code consistent with a
pediatric-specific nonurgent ED visit, as identified by an
expert panel in a prior study.23 We refer to this group
Downloaded for Siti Fadhilah (fadhilahhazhiyah@gmail.com) at Universitas Muslim Indonesia from ClinicalKey.com by Elsevier on May 05, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
PRACTICE-REPORTED MEDICAL HOMENESS 269
of visits as nonurgent, preventable, or avoidable ED visits.
We elected not to assess primary care–sensitive hospitali-
zations because they were rare in this sample (0.4% prev-
alence), with limited variation across practices.
Covariates included in the multivariate analyses come
from the Medicaid enrollment files in all states (child age,
race/ethnicity) and a practice-level survey in NC/SC (prac-
tice geography and number of providers, including physi-
cians, nurse practitioners, and physician assistants), which
was not available for practices in IL. Additionally, children
were classified as having a chronic condition or disability
if the claims-based Pediatric Medical Complexity Algorithm
identified them has having a complex chronic condition or a
noncomplex chronic condition,24 or if their Medicaid eligi-
bility category indicated they are disabled. We refer to the re-
maining group as healthy children.
ANALYSES
We examined the number and characteristics of children
attributed to practices in each state overall and by medical
homeness level. We also summarized practice-level char-
acteristics. To measure the association between medical
homeness and children’s health care utilization, we esti-
mated multilevel logistic regression models using SAS/
STAT software, version 9.3, PROC GLIMMIX (SAS Insti-
tute, Cary, NC). The models account for clustering of chil-
dren’s outcomes within practices. Separate models were
estimated for practices using the MHI/MHI-RSF in NC/
SC and those using the NCQA 2011 self-assessment in
IL. Adjusted models include covariates that are thought
to be associated with medical home level and health care
utilization, based on theory or prior literature. All models
control for child age, race/ethnicity, and chronic condi-
tions/disability status. For NC/SC analyses, where
practice-level characteristics were available, models also
control for number of providers in the practice and practice
geography (urban/suburban, rural). We also include a
dummy variable for state.
To assess whether associations between medical home-
ness and service use differ for children with and without
chronic conditions or disabilities (ie, effect modification),
we tested stratified models and models with an interaction
term between medical homeness and health status.
Additionally, for sensitivity testing of the medical home
variable, we estimated the final models for each sample us-
ing the continuous medical home score, as well as with cat-
egorical cutoffs at the 25th and 75th percentile (rather than
33rd and 66th percentile) for medical home level; these
models resulted in similar inferences as the final models
presented in the results section.
RESULTS
The IL sample included 33,895 publicly insured children
attributed to 32 practices (Table 1). The NC/SC sample
included 57,553 children in 32 practices, primarily in
NC. In both samples, the number of children in FFS or
PCCM Medicaid per practice ranged widely, from 34 to
over 10,700. Over half of the children in both samples
270 CHRISTENSEN ET AL ACADEMIC PEDIATRICS

Table 1. Sample Selection, Sample Size, and Attribution of Children to Practices

No. of Children
No. of Children Attributed to No. of Children No. of Demonstration Sample Size per
(0–18 y) Demonstration in Analytic Practices With Demonstration Practice,
State in Medicaid Practices Sample* Complete Data Median (Range)
Illinois 1,599,800 40,353 33,895 32 469 (34–9,803)
North Carolina/ 1,632,819 92,882 57,553 32 846 (36–10,714)
South Carolina
North Carolina 1,055,162 73,715 46,632 18 1,819 (261–10,714)
South Carolina 577,657 19,167 10,921 14 583 (36–2,596)
*Children were included in the sample if they were continuously enrolled in fee for service or primary care case management Medicaid, had
no other source of coverage, had full benefits not through a waiver program, were not institutionalized, and the practice they were attributed to
completed a medical home assessment.

were age 5 or younger (Table 2). About 31% of the sample
in IL and 34% of the sample in NC/SC had a chronic con-
dition or disability.
Seventy-six percent of the sample in IL and 64% in
NC/SC received the appropriate number of WCV. More
than one-quarter of each sample had a nonurgent, pre-
ventable, or avoidable ED visit (30% in IL; 28% in
NC/SC). By design, practices were equally distributed
into categories of medical homeness (tertiles; Table 3).
Table 4 shows how sample characteristics vary by medi-
cal homeness level. In the IL sample, a higher proportion
of children in the high medical homeness group were
black compared with children in the other medical
home levels. The percentage of children with a chronic
condition/disability was similar across medical home
levels (range 29%–31%). In the NC/SC sample, a higher
proportion of children in the high medical homeness
group were from practices in NC and in rural practices.
Again, the percentage of children with a chronic condi-
tion/disability was fairly similar across medical home
levels (range 31%–37%).
In multilevel logistic regression models, the estimated
associations between medical home level and children’s
health care utilization differed for practices completing
the NCQA 2011 self-assessment in IL and the MHI/
MHI-RSF in NC/SC. Inferences about the association be-
tween medical homeness and outcomes were the same in
the unadjusted and adjusted models; here we present re-
sults from adjusted models (Table 5).
For practices completing the NCQA tool in IL, there was
no statistically significant association between medical
homeness and receipt of WCV. However, children who
received care in practices with high medical homeness
were less likely to have a nonurgent, preventable, or avoid-
able ED visit than children in practices with low medical
homeness (odds ratio 0.65; 95% confidence interval
0.47–0.92). They were marginally less likely to have a
nonurgent, preventable, or avoidable ED visit than children
in practices with medium medical homeness (odds ratio
0.72, 95% confidence interval 0.52–1.01; P ¼ .06; not
shown in Table 5). There was no difference in nonurgent,
preventable, or avoidable ED use between the middle and
low medical home tertiles (odds ratio 0.90; 95% confidence
interval 0.65–1.27). Among practices completing the MHI/
MHI-RSF in NC/SC, there was no statistically significant
association between medical homeness and receipt of
WCV or nonurgent, preventable, or avoidable ED visits
(Table 5).

Table 2. Child-Level Sample Characteristics

Illinois North Carolina/South Carolina North Carolina South Carolina
Characteristic n % n % n % n %
Total 33,895 100 57,553 100 46,632 100 10,921 100
Age group
0–5 y 18,081 53 32,510 56 26,810 57 5,700 52
6–12 y 10,498 31 17,362 30 13,936 30 3,426 31
13–18 y 5,316 16 7,681 14 5,886 13 1,795 16
Sex
Male 17,197 51 29,584 51 23,943 51 5,641 52
Female 16,698 49 27,969 49 22,689 49 5,280 48
Race/ethnicity
Black 15,204 45 18,886 33 14,316 31 4,570 42
White 10,644 31 26,059 45 21,643 46 4,416 40
Other 8,047 24 12,608 22 10,673 23 1,935 18
Chronic condition or disability (yes) 10,384 31 19,360 34 15,695 34 3,665 34
Health care utilization
Appropriate receipt of well-child 25,659 76 36,455 64 29,831 65 6,624 62
visits
Any nonurgent, preventable, or 10,103 30 15,914 28 12,808 27 3,106 28
avoidable ED visit

ED indicates emergency department.

Downloaded for Siti Fadhilah (fadhilahhazhiyah@gmail.com) at Universitas Muslim Indonesia from ClinicalKey.com by Elsevier on May 05, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
ACADEMIC PEDIATRICS PRACTICE-REPORTED MEDICAL HOMENESS 271

Table 3. Practice-Level Sample Characteristics and Medical Table 4. Child-Level Sample Characteristics by Level of MH
Homeness
Column % for:
Characteristic n %
Characteristic Low MH Medium MH High MH
Illinois (NCQA self-assessment)
Illinois (NCQA
Total 32 100
Self-Assessment) n ¼ 7,507 n ¼ 8,917 n ¼ 17,471
NCQA 2011 self-assessment medical home tertile
High (score range: 66.8 to 83.3) 11 34 Age group
Medium (score range: 29.0 to 64.0) 11 34 0–5 y 58 53 51
Low (score range: 10.3 to 27.8) 10 31 6–12 y 30 30 32
North Carolina/South Carolina (MHI/MHI-RSF) 13–18 y 12 16 17
Total 32 100 Sex
Median (range) no. of providers* 4.5 (1–21) NA Male 52 51 50
Geography* Female 48 49 50
Urban/suburban 11 34 Race/ethnicity
Rural 21 66 Black 27 21 65
MHI/MHI-RSF medical home tertile White 49 50 14
High (score range: 59.6 to 89.3) 10 31 Other 24 29 21
Medium (score range: 49.1 to 58.9) 11 34 Chronic condition or 29 30 31
Low (score range: 24.1 to 46.4) 11 34 disability (yes)

NCQA indicates National Committee for Quality Assurance; MHI/ North Carolina/South n ¼ 10,687 n ¼ 15,055 n ¼ 31,811
MHI-RSF, Medical Home Index/Medical Home Index—Revised Carolina (MHI/MHI-RSF)
Short Form; NA, not applicable. Age group
*Numbers of providers and geography were not available for 0–5 y 53 55 58
practices in Illinois. 6–12 y 31 31 29
13–18 y 15 14 12
Sex
An interaction term between medical home level and Male 52 51 51
chronic condition/disability status was not statistically sig- Female 48 49 49
nificant in the NC/SC or IL analyses for either outcome, Race/ethnicity
indicating that the association between medical homeness Black 34 27 35
White 40 57 42
and outcomes did not vary significantly by health status.
Other 26 16 23
Cross-tabulations and regression models stratified by Chronic condition or 37 31 34
chronic condition/disability status confirmed the interpre- disability (yes)
tation of the interaction term. This term was excluded Practice geography*
from final models presented in Table 5. Urban/suburban 53 14 23
Rural 47 86 77
State
North Carolina 40 82 95
DISCUSSION South Carolina 60 18 5
This study contributes to the small but growing body of MH indicates medical homeness; NCQA, National Committee for
evidence on the association between medical homeness Quality Assurance; MHI/MHI-RSF, Medical Home Index/Medical
and children’s health care utilization. In this cross- Home Index—Revised Short Form.
sectional baseline analysis of practices participating in *Practice geography was not available for practices in Illinois.
the CHIPRA Quality Demonstration Grant Program, med-
ical home level was not associated with age-appropriate low and medium medical home scores. Among practices
receipt of WCV among practices completing either the that completed the MHI/MHI-RSF in NC/SC, there was
MHI/MHI-RSF or the NCQA self-assessment. These re- no association with nonurgent, preventable, or avoidable
sults differ from findings by Paustian et al.15 There may ED visits, a finding similar to that reported by Cooley
be a methodological explanation for null results: children et al,14 who examined medical claims for children in prac-
are only included in the sample if they had a well-child tices that completed the MHI.
or other ambulatory visit during the measurement year. Several factors could explain the variation in our ED
Although we are able to attribute approximately 70% to visit results. First, the differences between the 2 PCMH
80% of children in FFS or PCCM Medicaid, depending tools may explain the variation in results. Previous unpub-
on the state, it is possible that if we were able to attribute lished research has shown that scores on the MHI and the
children with no visits to practices there might be signifi- NCQA self-assessment are not highly correlated.25,26 It is
cant variation across medical home levels. plausible that the NCQA self-assessment better captures
The association between nonurgent, preventable, or domains of medical homeness that could be expected to
avoidable ED visits and medical homeness varied depend- prevent nonurgent, preventable, or avoidable ED visits.
ing on the sample and PCMH measure. Among practices in For example, access to primary care has been linked to
IL that completed the NCQA 2011 medical home self- decreased ED use among children.27–30 The NCQA self-
assessment, children who received care in practices with assessment contains 7 items that measure “Access and
high medical home scores were less likely to have a nonur- Continuity,” including access during office hours, after-
gent, preventable, or avoidable ED visit than practices with hours access, and electronic access, whereas the

Downloaded for Siti Fadhilah (fadhilahhazhiyah@gmail.com) at Universitas Muslim Indonesia from ClinicalKey.com by Elsevier on May 05, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
272 CHRISTENSEN ET AL ACADEMIC PEDIATRICS

Table 5. Effect of Medical Homeness on Health Care Utilization

Appropriate Receipt of Any Nonurgent, Preventable, or Avoidable
Well-Child Visits Emergency Department Visit
Unadjusted Regression Adjusted Unadjusted Regression Adjusted
Characteristic % Adjusted %† OR† 95% CI % Adjusted %† OR† 95% CI
Illinois (NCQA self-assessment)
MH tertile
High 75 77 1.02 0.71–1.46 30 23 0.65* 0.47–0.92
Medium 78 74 0.91 0.64–1.30 27 26 0.90 0.65–1.27
Low 74 76 (ref) 31 29 (ref)
North/South Carolina (MHI/MHI-RSF)
MH tertile (pooled)
High 64 63 0.77 0.47–1.26 25 27 0.92 0.70–1.20
Medium 66 69 1.03 0.68–1.55 31 29 0.99 0.79–1.25
Low 64 69 (ref) 29 29 (ref)
MH tertile (NC)
High 64 63 0.62 0.34–1.15 25 28 0.86 0.59–1.25
Medium 66 73 0.99 0.51–1.91 33 29 0.91 0.61–1.36
Low 70 73 (ref) 29 31 (ref)
MH tertile (SC)
High 72 75 1.88 0.75–4.74 27 25 0.96 0.55–1.66
Medium 63 60 0.95 0.58–1.57 26 26 1.04 0.77–1.41
Low 59 61 (ref) 30 26 (ref)

OR indicates odds ratio; CI, confidence interval; NCQA, National Committee for Quality Assurance; MH, medical home; MHI/MHI-RSF,
Medical Home Index/Medical Home Index—Revised Short Form.
*P < .05.
†All models adjusted for: child age, race/ethnicity, and chronic conditions/disability status. NC/SC models further adjusted for number of
providers in the practice, practice geography (urban/suburban, rural), and state.

MHI-RSF includes only one item for “Communication/Ac-
cess,” which does not specifically reference after-hours ac-
cess or electronic access. It is possible that variation in
access to care was not well measured in the NC/SC sample.
Alternatively, given that these tools were used in
different states, the state Medicaid context and prior incen-
tives for medical home implementation may also explain
the differences. For example, the racial composition of
children in the high medical homeness group varies across
the NC/SC and the IL samples. Although we controlled for
race in our models, these differences may be indicative of
other unobserved variations that could contribute to our re-
sults.
Variations in regression modeling across samples are un-
likely to account for the difference in inferences. Although
we controlled for practice characteristics in the NC/SC
model but not the IL model (due to lack of data), the infer-
ences regarding medical homeness did not change when we
removed practice characteristics from the NC/SC model in
sensitivity testing.
Several limitations of our study are important to note.
Medical homeness scores were self-reported by practices
and not verified against documentation. Practices volun-
teered to participate in the demonstration program and
may not be representative of child-serving practices state-
wide. Like many studies of Medicaid service use, our find-
ings might not be generalizable to publicly insured children
who are enrolled in managed care (other than PCCM)
because we were unable to include those children in our an-
alyses. In 2010, 71% of children with Medicaid in SC were
enrolled in traditional managed care (vs 0% of children in
NC and 9% of children in IL), making generalizability
more of a concern in SC than other states.31 However,
SC represents less than a fifth of the NC/SC sample, mini-
mizing this limitation.
Like Cooley et al,14 we analyzed Medicaid claims from
the calendar year before the year medical homeness was as-
sessed. If practices were increasing their medical homeness
during the claims year and quality of care improved after
the claims year, our models may attenuate any existing as-
sociations between higher medical home level and higher-
quality service use patterns. We have no information on
practices’ medical homeness or transformation activities
before the baseline data collection for the demonstration;
however, we know that practices began participating in
PCMH practice transformation activities under the CHI-
PRA Quality Demonstration Grant program immediately
after the baseline data collection, so we did not use claims
from the calendar year after medical homeness was as-
sessed.
Finally, the NYU algorithm assigns each ED visit a prob-
ability of falling into the nonurgent, preventable, and pri-
mary care–treatable categories and was not designed to
classify individual visits into mutually exclusive cate-
gories. Thus, there is likely to be measurement error in
our indicator for nonurgent, preventable, or avoidable ED
use. However, that error is unlikely to systematically
vary by medical homeness tertile and therefore is unlikely
to bias our estimates of the association between medical
homeness and ED utilization.
Despite these limitations, this study contributes to the
narrow knowledge base on the associations between pedi-
atric PCMH implementation and health service use in
several ways. First, it uses standardized, practice-reported

Downloaded for Siti Fadhilah (fadhilahhazhiyah@gmail.com) at Universitas Muslim Indonesia from ClinicalKey.com by Elsevier on May 05, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
ACADEMIC PEDIATRICS
PCMH assessment tools and medical claims rather than
parent reports of medical homeness and service use. We
are aware of only 2 other published studies that use these
types of data to examine the relationship between medical
homeness and children’s health care utilization. As noted,
these studies yield mixed findings. Second, the multilevel
models account for the clustering of patients within prac-
tices, which has not been done in most other studies
(including adult studies) of PCMH outcomes. Failing to ac-
count for clustering can result in biased estimates.32
Finally, by presenting results using 2 different PCMH
tools, this study also suggests that differences in the defini-
tion and measurement of medical homeness may
contribute to the mixed findings in the existing literature.
Continued emphasis by insurers and states on PCMH in-
terventions underscores the value of determining whether
medical homes in fact improve care for children. The
mixed findings to date reflect the complexity of this issue.
To determine the effects in various populations and policy
contexts with different measures, researchers must design
studies that can identify whether mixed medical home out-
comes are due to variations in the measures of medical
homeness used, actual variations in medical home imple-
mentation fidelity, or variations in other factors not yet
measured. Longitudinal studies that examine changes in
multiple measures of medical homeness and health care
service use within practices over time would contribute
greatly to the evidence base.
CONCLUSIONS
With the CHIPRA Quality Demonstration Grant Pro-
gram and other PCMH demonstrations currently under-
way, the evidence base on pediatric PCMH impacts will
expand over the next several years. Although our findings
suggest a potential association between medical homeness
and nonurgent, preventable, or avoidable ED use, the
mixed results of this and previous studies imply a need to
recognize the effects that differences in PCMH measure-
ment tools and state policy context may have on the
research findings.
ACKNOWLEDGMENTS
All phases of this study were supported by a contract with the US
Department of Health and Human Services, AHRQ
(HHSA290200900019I/HHSA29032004T). The authors wish to
acknowledge Carl Cooley, MD, co-chair of the National Center for Med-
ical Home Implementation; Sarah Scholle, PhD, of the National Commit-
tee for Quality Assurance; Genevieve Kenney, PhD, of the Urban Institute;
Catherine McLaughlin, PhD, of Mathematica Policy Research; and Cindy
Brach, MPP, and Linda Bergofsky, MSW, MBA, of the AHRQ for
providing valuable comments. We also thank the demonstration project
staff and state Medicaid agency staff who provided data and guidance.
Preliminary results from this study were presented at the Academy Health
Annual Research Meeting, June 23, 2013, in Baltimore, Md.
REFERENCES
1. Patient-Centered Primary Care Collaborative. Joint principles of
the patient-centered medical home. Available at: http://www.aafp.
org/dam/AAFP/documents/practice_management/pcmh/initiatives/
PCMHJoint.pdf. Accessed October 14, 2013.
Downloaded for Siti Fadhilah (fadhilahhazhiyah@gmail.com) at Universitas Muslim Indonesia from ClinicalKey.com by Elsevier on May 05, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
PRACTICE-REPORTED MEDICAL HOMENESS 273
2. Bitton A, Martin C, Landon BE. A nationwide survey of patient
centered medical home demonstration projects. J Gen Intern Med.
2010;25:584–592.
3. Grundy P, Hagan KR, Hansen JC, et al. The multi-stakeholder move-
ment for primary care renewal and reform. Health Aff (Millwood).
2010;29:791–798.
4. National Committee for Quality Assurance. Standards and guidelines
for NCQA’s patient-centered medical home (PCMH). Available at:
http://www.ncqa.org/PublicationsProducts/RecognitionProducts/
PCMHPublications.aspx; 2011. Accessed October 14, 2013.
5. Cooley WC, McAllister JW, Sherrieb K, et al. The Medical Home In-
dex: development and validation of a new practice-level measure of
implementation of the medical home model. Ambul Pediatr. 2003;3:
173–180.
6. Hadland SE, Long WE. A systematic review of the medical home for
children without special health care needs. Matern Child Health J.
2014;18:891–898.
7. Long WE, Bauchner H, Sege RD, et al. The value of the medical home
for children without special health care needs. Pediatrics. 2012;129:
87–98.
8. Romaire MA, Bell JF, Grossman DC. Health care use and expendi-
tures associated with access to the medical home for children and
youth. Med Care. 2012;50:262–269.
9. Strickland BB, Jones JR, Ghandour RM, et al. The medical home:
health care access and impact for children and youth in the United
States. Pediatrics. 2011;127:604–611.
10. Kieckhefer GM, Greek AA, Joesch JM, et al. Presence and character-
istics of medical home and health services utilization among children
with asthma. J Pediatr Health Care. 2005;19:285–292.
11. Knapp CA, Hinojosa M, Baron-Lee J, et al. Factors associated with a
medical home among children with attention-deficit hyperactivity
disorder. Matern Child Health J. 2012;16:1771–1778.
12. Diedhiou A, Probst JC, Hardin JW, et al. Relationship between pres-
ence of a reported medical home and emergency department use
among children with asthma. Med Care Res Rev. 2010;67:450–475.
13. Romaire MA, Bell JF, Grossman DC. Medical home access and health
care use and expenditures among children with special health care
needs. Arch Pediatr Adolesc Med. 2012;166:323–330.
14. Cooley WC, McAllister JW, Sherrieb K, et al. Improved outcomes
associated with medical home implementation in pediatric primary
care. Pediatrics. 2009;124:358–364.
15. Paustian ML, Alexander JA, El Reda DK, et al. Partial and incremen-
tal PCMH practice transformation: implications for quality and costs.
Health Serv Res. 2014;49:52–74.
16. Agency for Healthcare Research and Quality. National evaluation of
the CHIPRA quality demonstration grant program. Available at:
http://www.ahrq.gov/policymakers/chipra/demoeval/index.html. Ac-
cessed October 14, 2013.
17. McMorrow S, Christensen A, Natzke B, et al. How are states and evalu-
ators measuring medical homeness in the CHIPRA Quality Demonstra-
tion Grant Program? Available at: http://www.ahrq.gov/policymakers/
chipra/demoeval/what-we-learned/highlight02.pdf; 2013. Accessed
January 9, 2015.
18. American Academy of Pediatrics. Recommendations for preventive
pediatric health care. Available at: http://brightfutures.aap.org/pdfs/
AAP%20Bright%20Futures%20Periodicity%20Sched%20101107.
pdf. Accessed October 15, 2013.
19. Centers for Medicare and Medicaid Services. Core set of children’s
health care quality measures for Medicaid and CHIP (child core
set): technical specifications and resource manual for federal fiscal
year 2013 reporting. Available at: http://www.medicaid.gov/
Medicaid-CHIP-Program-Information/By-Topics/Quality-of-Care/
Downloads/Medicaid-and-CHIP-Child-Core-Set-Manual.pdf. Ac-
cessed October 16, 2013.
20. Ballard DW, Price M, Fung V, et al. Validation of an algorithm for cat-
egorizing the severity of hospital emergency department visits. Med
Care. 2010;48:58–63.
21. New York University. NYU ED background. Available at: http://
wagner.nyu.edu/faculty/billings/nyued-background. Accessed
October 15, 2013.
274 CHRISTENSEN ET AL
22. Billings J, Parikh N, Mijanovich T. Emergency Department Use: The
New York Story. New York, NY: Commonwealth Fund. Available at:
http://www.commonwealthfund.org/usr_doc/billings_nystory.pdf;
2010. Accessed September 24, 2014.
23. Ben-Isaac E, Schrager SM, Keefer M, et al. National profile of non-
emergent pediatric emergency department visits. Pediatrics. 2010;
125:454–459.
24. Simon TD, Cawthon ML, Stanford S, et al. Pediatric medical
complexity algorithm: a new method to stratify children by medical
complexity. Pediatrics. 2014;133:e1647–e1654.
25. Swigonski N, Nutakki K, Eckert G, et al. What is the measure of a
medical home? A comparison of the two leading instruments. Paper
presented at: Pediatric Academic Societies Annual Meeting; May
4–7, 2013; Washington, DC.
26. Reutland CP, Gallia C, Laurie A, et al. Becoming better medical
homes: improving systems and processes in the Tri-state Children’s
Health Improvement Consortium (T-CHIC). Poster presented at:
Academy Health Child Health Services Research Meeting; June 22,
2013; Baltimore, Md.
27. Brousseau DC, Gorelick MH, Hoffmann RG, et al. Primary care qual-
ity and subsequent emergency department utilization for children in
Wisconsin Medicaid. Acad Pediatr. 2009;9:33–39.
Downloaded for Siti Fadhilah (fadhilahhazhiyah@gmail.com) at Universitas Muslim Indonesia from ClinicalKey.com by Elsevier on May 05, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
ACADEMIC PEDIATRICS
28. Johnson WG, Rimsza ME. The effects of access to pediatric care and
insurance coverage on emergency department utilization. Pediatrics.
2004;113:483–487.
29. Zickafoose JS, DeCamp LR, Prosser LA. Association between
enhanced access services in pediatric primary care and utilization
of emergency departments: a national parent survey. J Pediatr.
2013;163:1389–1395.
30. Wang C, Villar ME, Mulligan DA, et al. Cost and utilization analysis
of a pediatric emergency department diversion project. Pediatrics.
2005;116:1075–1079.
31. Centers for Medicare and Medicaid Services. Medicaid Statistical Infor-
mation System (MSIS) state summary datamarts, 2010. Available at:
http://www.cms.gov/Research-Statistics-Data-and-Systems/Computer-
Data-and-Systems/MedicaidDataSourcesGenInfo/MSIS-Mart-Home.
html. Accessed October 16, 2013.
32. Peikes D, Dale S, Lundquist E, et al. Building the Evidence Base for
the Medical Home: What Sample and Sample Size Do Studies Need?.
Rockville, Md: Agency for Healthcare Research and Quality; 2011.
AHRQ Publication No. 11-0100-EF. White paper prepared by Math-
ematica Policy Research under contract HHSA290200900019I TO2.