Auditing Access to Specialty Care for Children with
Public Insurance
Joanna Bisgaier, M.S.W., and Karin V. Rhodes, M.D.
New England Journal of Medicine
June 16, 2011
http://www.nejm.org/doi/full/10.1056/NEJMsa1013285#t=articleTop

Expansions of Medicaid and the Children's Health
Insurance Program (CHIP) are designed to extend access
to high-quality medical care to all U.S. children.1-3
However, evidence suggests that the 37 million children
covered by Medicaid-CHIP4,5 are less likely to receive
specialty care than children covered by commercial
insurance.6-13 Children covered by Medicaid-CHIP may
face greater barriers to specialist care as a result of
fewer resources within their families, including lower
levels of income, education, language proficiency, and
health literacy.14 Another possible explanation for
disparities is that specialists choose not to accept
public insurance.15 In contrast to patient-related or
family-related barriers, which are less malleable to
change, provider-related barriers are potentially
modifiable through health care policies.16 To date,
research on children's access to specialty care has not
adequately distinguished between provider-related
barriers and patient-related ones.

Unraveling the contributions of clinical need and
patient-related versus provider-related barriers is a
vital first step in constructing effective policies that
improve children's access to specialty care. Given the
association between socioeconomic disadvantage and poor
health status, children covered by Medicaid-CHIP may
have a greater need for specialty care.17 However, most
studies to date have been unable to directly control for
children's clinical need for specialty services.6,18
Audit methodology, traditionally used for detecting
"real life" discriminatory behavior in housing and labor
markets, can be used to assess insurance-related
disparities in health care access.19 Using this approach
in a 1994 study, the Medicaid Access Study Group found
that adult patients with Medicaid had poor access to
outpatient care.20 Subsequent studies in which this
approach was used did not sufficiently examine
physicians' willingness to provide needed specialty care
for publicly insured children.7,13,21,22 In light of the
pending expansions of public insurance programs, we
sought to identify whether - and if so, to what extent -
provider acceptance of Medicaid-CHIP coverage is an
independent barrier to outpatient specialty care for
children in the current health care market, while
controlling for patient factors and the clinical urgency
of the referral.

Methods

Data Collection and Study Design

We designed an audit study in which research assistants
posing as mothers made paired calls to the same clinic
and attempted to schedule an appointment for a child
needing specialty care. The calls were separated by 1
month and varied only by insurance status (private vs.
Medicaid-CHIP insurance). Data were gathered by the
University of Chicago Survey Laboratory, where trained
and supervised graduate students made calls to specialty
clinics with the use of a central-computer-assisted
telephone interview. (Post-call evaluation forms and the
protocol flow chart for audit calls are available in the
Supplementary Appendix, available with the full text of
this article at NEJM.org.) Our study was conducted in
Cook County, Illinois, the second most populous U.S.
county (5,194,675 residents),23 where the ratio of
specialists to population is 218 to 100,000; the
national median is 32 to 100,000.24 Although Illinois
Medicaid has historically provided care through a fee-
for-service structure, it began implementing a primary
care case-management program in July 2006, which serves
approximately 67% of publicly insured children in Cook
County.25 The remaining children are served in a fee-
for-service structure (16%) or voluntary commercial
managed-care organizations (18%). Illinois is among 27
states that implement CHIP and Medicaid as a combined
program (i.e., identical program name [All Kids] and
reimbursements).26

Sampling Methods

We constructed an exhaustive list of providers, using
state-provided physician-licensure data, cross-
referenced with lists of physicians submitting specialty
claims for children in Cook County and lists of
specialists provided by children's hospitals and the
American Academy of Pediatrics. The final sample
included all specialists for whom there was any evidence
that they provided care to children (0 to 18 years of
age) residing in Cook County. Because several
specialists may practice at the same clinic and some
specialists practice at several clinics, we did not
sample providers; rather, we sampled clinics, defined by
unique (unduplicated) telephone numbers used for
scheduling appointments. Random samples of 40 clinics
per health-condition scenario were stratified according
to two key variables (provider licensure reporting
acceptance vs. nonacceptance of Medicaid-CHIP and urban
vs. suburban location) with the use of a computer
algorithm. During the study, physicians' licensure data
regarding Medicaid-CHIP acceptance were not publicly
available.

Specialty Conditions and Protocol

From January through May 2010, we investigated eight
specialties (allergy-immunology, pulmonary diseases,
dermatology, endocrinology, neurology, orthopedics,
otolaryngology, and psychiatry) in which providers treat
seven pediatric specialty health conditions (Table
1Table 1Specialties and Health-Condition Scenarios
Included in the Study.). Allergists-immunologists and
pulmonary disease specialists were audited together and
sampled in proportion to their representation in the
population, because both treat persistent, uncontrolled
asthma. Clinical scenarios (involving a diagnosis and
symptoms in a patient of a specified age) were chosen by
pediatric primary care providers (PCPs) and specialist
consultants with the use of an iterative review process
to identify conditions that affect a large number of
children, warrant timely outpatient specialty evaluation
and treatment to achieve optimal health outcomes, are
urgent situations but not emergencies, and have a known
effective treatment. A pilot study of these scripts with
standardized responses to possible questions was
conducted between November 2009 and January 2010.
(Scripts are available in the Supplementary Appendix.)

Every caller reported having a referral from the child's
PCP; three scenarios also involved referral by an
emergency department. To avoid geographic
discrimination, we geocoded all specialty clinics and
generated fake patient and PCP addresses that were in
the vicinity of (but more than 1.6 km [1 mi] from) each
clinic with the use of ArcGIS software (version 9.3). If
asked, callers reported an emergency department located
in the general area, cross-checked against specialists'
hospital affiliations (from licensure data) to avoid the
potential for shared electronic medical records.

We obtained dummy Medicaid-CHIP identification numbers
from the state that would appear in the online system as
"active" and that were linked to the demographic
characteristics (e.g., name, sex, and race or ethnic
group) corresponding to each caller's identity. If asked
for the PCP's name, callers gave 1 of the top 10
physician surnames from Medicaid-CHIP claims data for
fiscal year 2008. For questions that the caller was
unable to answer (e.g., Social Security number or
private insurance number), standardized "work-arounds"
were developed. To control for the racial or ethnic
characteristics of a caller's name and voice, all
samples were randomly assigned to one of three groups of
callers (black, white, or Hispanic) with the use of a
computer algorithm. Clinics were deemed "out of scope"
if they reported that they did not provide care for the
clinical condition or for children of the reported age
(before knowing the child's insurance status). Out-of-
scope clinics and nonfunctional telephone numbers were
replaced with the next randomly selected clinic
providing care for the condition. After three calls
without reaching a live person, callers left a voice-
mail message with their assigned name, telephone number,
and insurance type. If voice mail was not returned,
callers placed six additional calls, leaving voice-mail
messages.

The same caller called the same clinic twice. The order
of reported insurance type, the only variable differing
between the two calls, was randomly assigned. If asked,
there were minor variations in the patient's and
caller's names, the patient's address and date of birth,
and the PCP's name and address. For private insurance,
callers reported Blue Cross Blue Shield coverage because
it has the largest market share in Illinois.27 Callers
did not volunteer their insurance status, but if an
appointment was granted without a request for insurance
status, callers confirmed the acceptance of their
assigned insurance. All calls were kept as short as
possible, and all appointments were canceled at the end
of the call. Prepaid cell phones allowed callers to
provide telephone numbers, leave voice-mail messages,
and receive returned calls. Outcomes were the percentage
of callers according to insurance status who
successfully scheduled an appointment and the wait time
(number of days) between the call and the scheduled
appointment date. Descriptive data about medical and
insurance-related questions asked were collected.

Study Oversight

The study was approved, with a waiver of the requirement
for informed consent, by institutional review boards at
two institutions, with the caveat that debriefing
letters be sent to all clinics in the entire sampling
frame at the conclusion of the study. The deceptive
design was considered necessary to accomplish the
primary objective of the study: to identify the
existence and extent of any disparities in children's
access to specialty care according to insurance status
by measuring the real-life behavior of specialty
practices contacted for outpatient appointments. The
debriefing letters clearly stated that the purpose of
the study was to monitor the system rather than
individual providers, that individual clinics may or may
not have been randomly selected to be studied, and that
the identity of those selected will never be disclosed.

Statistical Analysis

For all calls, we calculated the relative risk that
children with Medicaid-CHIP coverage, as compared with
those who had commercial insurance, would not receive a
specialty care appointment. For paired calls, we
calculated the log-odds probability of a scheduled
appointment, using McNemar's test to assess the symmetry
of discordant pairs (i.e., pairs of calls in which
public and private insurance were not treated equally),
holding constant all other patient and clinical
characteristics. For subanalyses according to specialty
type, we anticipated extreme splits on the dependent
variable and used exact conditional (fixed-effects)
logistic regression, which is a generalization of
McNemar's test. Sample-size calculations for McNemar's
test before the study were based on previous data from
audit studies.21 We calculated that a sample of 20
clinics would provide 80% power to detect a 34%
difference and that 32 clinics would be needed to detect
a 20% difference in the rate of clinics accepting public
versus private insurance, at an alpha level of 0.05.

For specialty clinics that scheduled appointments for
both insurance types, we calculated the difference
between appointment wait times (in number of days) with
the use of paired t-tests. We did not test the
significance of wait-time disparities by specialty type
because of the small number of clinics that scheduled
appointments for both insurance types. All tests were
two-sided, and P values of less than 0.05 were
considered to indicate statistical significance. All
statistical analyses were performed with the use of
Stata/SE software (version 11.0).

Results

Clinics

During the 5-month study period, the survey center
attempted to contact 577 specialty clinics. As shown in
Figure 1Figure 1Clinics Included in the Study Sample.,
149 clinics (26%) did not treat patients with the given
age or clinical condition, and 151 clinics (26%) were
excluded because of nonfunctional telephone numbers. For
the 277 clinics in the final sample, callers were unable
to complete the study protocol with 4 clinics (1%),
which required more medical documentation than we could
provide. Two completed calls were made to each of the
remaining 273 clinics (546 total calls). Because of the
low number of endocrinology and neurology clinics with
evidence of providers seeing pediatric patients (30 and
66, respectively), we randomly sampled from the broader
pool of specialty clinics (68 endocrinology clinics and
99 neurology clinics) in an attempt to identify
additional specialists willing to see children.

Outcomes

Of the 546 calls to clinics, 297 (54%) involved a
request for information about the child's insurance type
before the caller was told whether an appointment could
be scheduled. For 153 (52%) of these 297 calls, the type
of insurance coverage was the first question asked.
Figure 2Figure 2Clinics Scheduling Specialty Care
Appointments for Children, According to Type of
Insurance. shows the proportions of specialty clinics
that scheduled appointments for children with public
insurance and for those with private insurance,
according to type of specialty. As shown in Table 2Table
2Likelihood of Being Denied a Scheduled Specialty Care
Appointment According to Type of Insurance., 66% (179)
of the callers reporting Medicaid-CHIP coverage were
denied an appointment for specialty care, as compared
with 11% (29) of the callers reporting Blue Cross Blue
Shield insurance (relative risk, 6.2; 95% confidence
interval [CI], 4.3 to 8.8; P<0.001). When calls to the
same clinic were analyzed as matched pairs, there were 5
discordant pairs (2%) in which children with Medicaid-
CHIP obtained an appointment but those with private
insurance did not, and 155 discordant pairs (57%) in
which the clinic accepted privately insured children but
not Medicaid-CHIP enrollees (odds ratio for appointment
denial with public insurance, 31.0; 95% CI, 13.0 to
96.8). All relative risks (when calculable) and exact
conditional logistic-regression analyses showed that,
across all tested specialties, children with Medicaid-
CHIP were significantly more likely to be denied an
appointment than privately insured children. Among 173
clinics with any providers whose license indicated
acceptance of Medicaid-CHIP, 43% scheduled Medicaid-CHIP
appointments. Of 100 clinics without licensure-reported
Medicaid-CHIP acceptance, 19% granted these
appointments.

Among the 89 specialty clinics that scheduled
appointments for both Medicaid-CHIP enrollees and
privately insured children, children with Medicaid-CHIP
had greater delays in obtaining needed specialty care
(Table 3Table 3Wait Times for Appointments for Children
with Public versus Private Insurance among Clinics
Accepting Both Insurance Types.). On average, children
with public insurance waited 42 days for an appointment
with a specialist, whereas privately insured children
waited 20 days (mean difference, 22.1 days; 95% CI, 6.8
to 37.5; P=0.005).

Discussion

With the use of an experimental study design involving
simulated requests for specialty care, we measured real-
world scheduling behavior in an urban area with a high
density of medical specialists.24 The results showed
significant disparities in children's access to needed
outpatient specialty care, attributable to specialists'
reluctance to accept public health insurance. These
results held across all audited specialties. Moreover,
even when children with Medicaid-CHIP were not denied
appointments outright, the appointments were, on
average, 22 days later than those obtained for privately
insured children with identical health conditions.
Notably, even callers claiming to have a privately
insured child faced an average wait time of 20 days when
urgently requesting an appointment. These findings
signal a need to consider refining specialty care
delivery processes to more efficiently use the
specialist workforce.28,29

Two previous audit studies of pediatric specialty care
have shown even lower Medicaid acceptance rates: 4%13
and 8%.7 However, both studies investigated only one
specialty type (orthopedics), and both had weaknesses in
their sampling strategies that may have biased their
results, including failure to exclude ineligible
providers,7 sampling at the physician level rather than
the clinic level (i.e., possibly calling the same clinic
multiple times),7 and the exclusion of physicians
practicing at tertiary pediatric referral centers,13
which are key sources of outpatient orthopedic care.30

A recent population-based survey by Kogan et al. showed
that parents whose children had Medicaid-CHIP coverage
were more likely to report that insurance did not allow
their child to see needed providers.31 Our results
corroborate and add to this important finding by
measuring the real-life experience of attempting to
schedule an appointment when all other factors besides
insurance status (e.g., parental persistence or savvy
and the child's clinical symptoms) are held constant.
The strength of the current study stems from its ability
to isolate the effect of one dimension of access. Our
results indicate that increasing the number of providers
who accept public insurance will increase access
opportunities. Without correcting this dimension, it is
unlikely that disparities in access between public and
private insurance can be fully eliminated, even if all
other barriers to access (e.g., out-of-pocket costs,
referral requirement, and need for language proficiency,
transportation, and health literacy) could be
addressed.15,16

The Affordable Care Act represents an opportunity to
remold health care delivery processes in the United
States.32,33 It is well established that reimbursement
levels influence providers' decisions about whether to
accept public insurance.8,34-36 In Illinois, an office
consultation visit for a problem of moderate severity
(Healthcare Common Procedure Coding System code 99243)
is reimbursed at $99.86 by Medicaid-CHIP,37 whereas the
average reimbursement for the same code by a commercial
preferred-provider organization is approximately $160.
Although disparities in insurance-reimbursement rates
are important, the literature indicates that additional
variables affect physicians' decisions about whether to
accept public insurance, such as delays in payment and
hassles of payment procedures,35,36 personal
characteristics of providers (e.g., credentials or
experience,34,38,39 race or ethnic group,34,38-41 and
underlying attitudes or prejudices39,42), and structural
features of the system in which they provide care (e.g.,
institutional affiliations,34,43,44 location,34,38,41
and practice size or type22,34,38,44). Further research
on the multiple underlying variables associated with
provider behavior in our current system can help with
workforce planning and inform innovations in service
delivery.

More work is needed to understand the benefits or
opportunity costs of potential policy changes. For
example, is it better to raise reimbursement rates
globally for all specialists or to provide targeted
incentives to specialists or medical centers located in
low-resource neighborhoods and committed to serving as
safety-net specialty providers? Do we need more
specialists or should we reorganize the manner in which
we provide specialty care? Such information is
fundamental to the formation of integrated delivery
systems and the configuration of payment methods that
can optimize access and decrease disparities.

Caution is needed in generalizing our results to
specialists other than those in the specific specialties
and region that were audited in this study. In
particular, there is no evidence that pediatric
specialists working in inpatient or rural settings are
unwilling to accept Medicaid-CHIP. Nonetheless, our
experimental design affords high internal validity
within the context of understanding specialist behavior
relative to our simulated children's insurance status,
with adequate controls for clinical urgency and other
patient-level factors. Our study only assessed access to
specialty care for publicly and privately insured
children, and it should be noted that access to
specialty care may be different for uninsured children
and for publicly insured or uninsured adults.

Our study was powered to measure appointment denials and
delays across a number of outpatient specialty types,
but it was not powered to identify the effect of
specific provider or clinic characteristics associated
with appointment denials or delays. In addition, we did
not identify the causes of interspecialty variation. Nor
did we assess whether acceptance of public insurance
varies between specialists who provide cognitive
consultations and procedural or surgical specialists,
who may be more dependent on their affiliated hospitals
to provide technologically advanced diagnostic and
surgical resources.29 Finally, although we used the
literature and experts in both primary and specialty
care to inform the urgency and importance of our
clinical scenarios, more work is needed to clarify
whether identified disparities are clinically meaningful
for children's long-term health and safety.

Overall, we found considerable disparities in access to
outpatient pediatric specialty care that were
attributable to providers' nonacceptance of public
insurance. These findings speak to the imperative for
policymakers to identify regulatory mechanisms and
incentives that target provider behavior and to explore
innovative models of specialty care delivery that have
the potential to increase access to specialty
expertise.45-47 As we encounter new opportunities for
restructuring the U.S. health care delivery system,
there is a need for empirical data on policy mechanisms
that can minimize disparities in access to care and
deliver on health care reform's commitment to the
provision of high-quality care for all Americans.

Supported by the state of Illinois, which provided
funding, detailed physician-licensure data, data
regarding Medicaid and state-employee health insurance
claims, and dummy Medicaid identification numbers as a
result of a court-ordered consent decree stemming from
class-action litigation on behalf of Cook County
children enrolled in Medicaid.

Disclosure forms provided by the authors are available
with the full text of this article at NEJM.org.

No potential conflict of interest relevant to this
article was reported.

We thank the attorneys from Heath and Disability
Advocates, the Sargent Shriver National Center on
Poverty Law, and Goldberg Kohn (particularly Frederick
Cohen, J.D.), for generating the impetus for this study;
the staff of the Illinois Department of Healthcare and
Family Services for their collaboration and review; and
Martha Van Haitsma, David Chearo, and Theresa Anasti
from the University of Chicago Survey Laboratory, as
well as Daniel Polsky, Paul Allison, A. Russell Localio,
and members of our expert review panel for their input
and methodologic advice.

Source Information

From the School of Social Policy and Practice (J.B.,
K.V.R.) and the Division of Emergency Care Policy
Research, Department of Emergency Medicine, School of
Medicine (K.V.R.) - both at the University of
Pennsylvania, Philadelphia.

Address reprint requests to Dr. Rhodes at the School of
Social Policy and Practice, University of Pennsylvania,
3815 Walnut St., Rm. 201, Philadelphia, PA 19104, or at
[log in to unmask]

___________________________________________

Portside aims to provide material of interest to people
on the left that will help them to interpret the world
and to change it.

Submit via email: [log in to unmask]

Submit via the Web: http://portside.org/submittous3

Frequently asked questions: http://portside.org/faq

Sub/Unsub: http://portside.org/subscribe-and-unsubscribe

Search Portside archives: http://portside.org/archive

Contribute to Portside: https://portside.org/donate