CHIP/MEDICAID

Medicaid and CHIP Retention Among Children

in 12 States
Susmita Pati, MD, MPH; Angie T. Wong, MHS; Rose E. Calixte, PhD; Justin Ludwig, MA;
Ashley Zeigler, BA; A. Russell Localio, PhD; JeanHee Moon, PhD, MPH;
Jeffrey H. Silber, MD, PhD
From the Division of Primary Care Pediatrics, Stony Brook University School of Medicine, Stony Brook, NY (Dr Pati, Ms Wong, Dr Calixte);
Pediatric Quality Measures Program Center of Excellence (Dr Pati, Ms Wong, Dr Calixte, Mr Ludwig, Ms Zeigler, Dr Localio, Dr Moon, Dr
Silber), Center for Outcomes Research (Mr Ludwig, Ms Zeigler, Dr Silber), Division of General Pediatrics (Dr Moon), Childrens Hospital of
Philadelphia, Philadelphia, Pa; Department of Biostatistics and Epidemiology (Dr Localio), Leonard Davis Institute of Health Economics (Dr
Silber), and Department of Pediatrics (Dr Silber), University of Pennsylvania Perelman School of Medicine, Philadelphia, Pa
The authors declare that they have no conflict of interest.
Address correspondence to Susmita Pati, MD, MPH, Stony Brook University & Stony Brook Long Island Childrens Hospital, Health Sciences
Center, Level 11, Room 020, Stony Brook, NY 11794-8111 (e-mail: susmita.pati@stonybrook.edu).
Received for publication October 18, 2013; accepted September 28, 2014.

ABSTRACT
OBJECTIVE: Though stable insurance is important to support
optimal child health, the reproducibility of metrics to assess
child health insurance retention at the state and county level
has not been examined. We sought to determine reproducibility
of public insurance retention rates for children using 3 different
metrics at the state and county level.
METHODS: Public health insurance retention for children was
assessed using 3 different metrics calculated from 20062009
Medicaid Analytic Extract data from 12 selected states. The
metrics were: 1) Duration: a prospective metric that quantifies
the number of newly enrolled children continuously enrolled
in public insurance 6, 12, and 18 months after initial enrollment
during a selected period; (2) Infant Duration: assesses Duration
only among infants born during a selected period; (3) Coverage:
a prospective metric that quantifies the average percentage of
time a selected population is enrolled over an 18-month interval.

Reproducibility of the metrics was assessed using a range of

sample sizes with resampling and determining changes in relative rankings of states/counties by retention rate.
RESULTS: All 3 metrics demonstrated reproducible estimates
at the state level with sample sizes of 2000, 5000, and 10,000.
Reproducibility of relative rankings for child health insurance
retention of counties within states were sensitive to county child
population size and the amount of variability in retention rates
within the county and at the state level.
CONCLUSIONS: As health care reform unfolds, the complete
set of these 3 reproducible metrics can be used to evaluate multipronged and multilevel strategies to retain eligible children in
public health insurance.

KEYWORDS: CHIP; insurance; Medicaid; retention

ACADEMIC PEDIATRICS 2015;15:249257

WHATS NEW

health and well-being, coordination and adjustment of

management plans and treatments to maximize health,
and prevention of exacerbations of ambulatory caresensitive conditions that might lead to hospitalization.57
Children with public health insurance are disproportionately burdened with poor health outcomes, including hospitalization for ambulatory caresensitive conditions, than
children with private health insurance.814 Most children
enrolled in public health insurance programs remain
eligible for these programs for much of their childhood,
despite the churning that occurs within and between
different public health insurance programs.1517 In 2009,
Congress passed the Childrens Health Insurance Program
Reauthorization Act18 with the explicit goal of supporting
states in identifying, enrolling, and retaining public health
insurance coverage for eligible but uninsured children.
The 2010 Affordable Care Act19 expanded Medicaid eligibility and coverage and simplified the enrollment process

Three different metrics designed to assess public health

insurance retention demonstrate high reproducibility by
state and county for children. As health care reform unfolds, the complete set of these 3 metrics can inform
multipronged and multilevel strategies to retain eligible
children.

STABLE

INSURANCE COVERAGE is important for

health, especially for children and adolescents, because
continuous coverage permits children and adolescents
access to a regular source of care during critical phases
of development.13 A regular source of care allows for
treatment of chronic health conditions, routine preventive
care, and management of acute and urgent problems.4 A
regular source of preventive care also supports continuity
of care that facilitates regular monitoring of child

ACADEMIC PEDIATRICS
Copyright 2015 by Academic Pediatric Association

249

Volume 15, Number 3

MayJune 2015

250

PATI ET AL

ACADEMIC PEDIATRICS

for Medicaid and Childrens Health Insurance Program

(CHIP). With accountability requirements, limited budgets,
and increasing Medicaid enrollment due to the recent economic recession,20 states need robust metrics to assess child
health insurance retention rates.
The Pediatric Quality Measures Program Center of
Excellence at the Childrens Hospital of Philadelphia (Center) examined 3 insurance retention metrics: the Duration
of First Observed Enrollment metric adapted from a measure originally proposed by the Centers for Medicare and
Medicaid Services (CMS) and the Office of Management
and Budgets (OMB),21 and 2 complementary metrics
developed by the Center: the Duration of Infants First
Enrollment metric and the Coverage metric. Here we
assess the reproducibility of these 3 metrics in 12 states.

METHODS
DATA SOURCE
The 20062009 Medicaid Analytic Extract (MAX) obtained from CMS was the data source for this study. These
data are commonly used by policy analysts interested in
public insurance usage because they include person-level
data about Medicaid eligibility, service utilization (ie,
claims), and payments. Eligibility data include the enrollment and termination information for each individual that
is needed to compute insurance retention rates for all 3 metrics. All states report Medicaid eligibility data to CMS in
MAX, and some states report Medicaid expansionChildrens Health Insurance Program (M-CHIP) and/or separate
Childrens Health Insurance Program (S-CHIP) eligibility
data. Medicaid and M-CHIP eligibility data in MAX include
the number of days per month that a child was enrolled. A
child was considered covered for the entire month in this
study if the child had >14 days of enrollment in that month.
S-CHIP eligibility data in MAX indicates whether a child
was enrolled each month using the CHIP FLAG variable.
A child was considered covered for a given month if the
child was marked as yes (ie, enrolled) for this variable.
Claims data include dates of visits (ie, encounter data) and
diagnostic codes (ie, ICD-9 codes), which are needed to
identify children with chronic conditions. Unlike fee-forservice plans, some managed care plans do not require
providers to submit claims for patient encounters because
providers receive per capita reimbursements for enrolled individuals. Therefore, states with high managed care penetration may not have complete encounter data for ambulatory
visits.22 Although here we focus the reproducibility of the
metrics at the state and county level, the Center was also
tasked with determining reproducibility of the metrics
among vulnerable groups, such as those with chronic conditions. This influenced our decision to include states that
report claims data in the MAX data set and/or have low
managed care penetration in this study. To achieve the Centers overarching goals, this study included states that report
all eligibility data for Medicaid and M-CHIP, S-CHIP, or
both (combination of Medicaid expansion and separate
child health program), and states that report claims data in
the MAX data set.23,24 The following 12 states were

selected for this study: Arizona, Idaho, Illinois, Indiana,

Louisiana, Montana, North Carolina, New Hampshire,
New Jersey, New Mexico, Virginia, and Wisconsin.
INSURANCE RETENTION METRICS
We studied 3 metrics designed to assess insurance retention. For the purposes of this study, transitions between
Medicaid and M-CHIP are not classified as gaps in
coverage if the child is insured for the majority of a given
month, while transitions between Medicaid and S-CHIP
are not classified as gaps in coverage if the child is indicated to have S-CHIP for a given month.
The Duration of First Observed Enrollment (hereafter
referred to as Duration) measure is adapted from the CMS
and OMBs Duration Measure of Selected Children.21 It is
a prospective metric that quantifies the proportion of newly
enrolled children, including newborns, who are continuously enrolled in public coverage at 6 months, 12 months,
and 18 months after initial observed enrollment, and who
would otherwise not age out of the program during the
selected study time period (Fig. 1A). Specifically, newly
enrolled children are defined by CMS and OMB as all children under age 16 years and 5 months: 1) enrolled for the
first time in Medicaid and/or CHIP, for states with joint
Medicaid and CHIP programs (Fig. 1A, rows 1 and 2); or
2) those who have a continuous 30-day period with no
Medicaid or CHIP coverage before the initial observed
enrollment during the selected enrollment period (Fig. 1A,
rows 3 and 4). As a result, the Duration metric censors children who are not new enrollees; this is termed left-censoring
(Fig. 1A, row 6). For this project, the enrollment period was
defined as January 1, 2007, to December 31, 2007; any child
insured by Medicaid and/or CHIP during this time period
was eligible for inclusion in the analytic sample. Subsequently, any newly enrolled children (as previously defined)
were observed for an 18-month period.
The Duration of Infants First Enrollment measure, or
Infant Duration, reports the duration of first observed
enrollment for infant enrollees only. Infant enrollees are
defined as infants born and enrolled in Medicaid (or
CHIP for states with joint Medicaid and CHIP programs)
at any time during the enrollment period (Fig. 1B). For
this project, the enrollment period was defined as January
1, 2007, to December 31, 2007.
The Coverage metric quantifies the percentage of time,
over a selected time interval (eg, 18 months), that an
eligible child is enrolled in public health insurance. Any
child who is publicly insured during the 18-month lookback interval before the start of the observation period is
presumed eligible. To calculate this metric, the sum of
the number of insured months is divided by the number
of eligible months for each child during a selected 18month observation period to produce an individual proportion; then the average of these individual proportions for
sampled children within a prespecified geographic area
(eg, state, county) is computed to produce the Coverage
metric (Fig. 1C). For this project, the 18-month observation
period was defined as January 1, 2008, to June 30, 2009.
For children who are new enrollees during the observation

ACADEMIC PEDIATRICS

MEDICAID AND CHIP RETENTION

251

Figure 1. Duration, Infant Duration, and Coverage metrics illustrating different scenarios of enrollment patterns; these scenarios were
created by the authors and are not taken directly from any single childs individual enrollment experience as recorded in MAX. (A) The Duration
metric quantifies the number of newly enrolled children (as defined by CMS and OMB) continuously enrolled in public insurance at 6, 12, and
18 months after enrollment. (B) The Infant Duration metric calculates duration for the subset of infants born during the enrollment period. In this
study, the enrollment observation period for the Duration metrics was January 1, 2007, to December 31, 2007. (C) The Coverage metric measures the population average of months enrolled in public insurance. In this study, the look-back period for the Coverage metrics was July 1,
2006, to December 31, 2007, and the observation period was January 1, 2008, to June 30, 2009.

period (ie, no evidence of enrollment in the 18-month lookback interval of July 1, 2006, to December 31, 2007), the
denominator begins with the first instance of enrollment
in the observation period (Fig. 1C, row 4). For children
who are not new enrollees (ie, evidence of enrollment in
the 18-month look-back interval), the denominator is the
complete 18-month observation period (Fig. 1C, rows 1,
2, 3, and 5). The total percentage of enrollment may reflect
a single enrollment spell or the sum of noncontiguous

spells of enrollment occurring within the selected observation period.

REPRODUCIBILITY
For the purposes of this report, we define the reproducibility of a metric as its ability to produce consistent as
well as precise results under similar conditions. Specifically, we determined whether the retention rates measured
by each metric and the rankings based on these rates were

252

PATI ET AL

ACADEMIC PEDIATRICS

consistent upon repeated sampling from the population of

beneficiaries. Reproducibility is relevant for states that
may decide to compare retention rates across counties
for example, to identify counties that have retention rates
in the top and bottom 5% within the state. If the metrics
are reproducible, then the retention rates and rankings
measured using random samplings should be consistent
when repeated.
To test reproducibility at the state level, we implemented
random sampling of sizes 2000, 5000, and 10,000 stratified
by county. We stratified by county to ensure the sample was
representative of the population and avoid the possible bias
from simple random sampling because counties might
differ markedly in patient characteristics and the outcome
retention metrics. For example, if the eligible population
for a metric in a given state is 50% in county A, 20% in
county B, and 30% in county C, then the sample composition when resampling using 2000 observations comprises
1000 (ie, 2000*0.5) observations from county A, 400 (ie,
2000*0.2) observations from county B, and 600 (ie,
2000*0.3) observations from county C. Three samples of
each size were then used to calculate the metric and assess
similarities across samples within a state. In New Hampshire, we sampled using only sizes of 2000 and 5000
because the total infant population was less than 10,000.
To test reproducibility at the county level, we used samples of 2000 for counties with more than 2000 children,
samples of 200 for counties with more than 200 but fewer
than 2000 children, and all observations without resampling for any county with fewer than 200 children. We
did not determine the reproducibility of the Infant Duration
metric at the county level because most states do not have
more than 2000 infants in each of its counties. To further
explore the stability of the metrics at the county level, we
ranked the counties according to their retention rates. We
hypothesized that the variation in each metric calculated
across the 3 samples within counties would be far less
than the actual variation in the metrics across counties.
Given that each of the samples within each state and
county are independent, we used Greenwoods method
for variances (and confidence bounds) of differences.25,26
With multiple samples, we estimated pairwise differences
among these samples and the width of the resulting
95% confidence interval of these differences. Power to
demonstrate equivalence of proportions (or zero
differences in proportions) requires relatively large
samples.27 As an example, assuming that the duration proportion for children at 12 months is 0.875 (87.5%) and that
the desired bounds for equivalence of the estimated proportions from independent samples is plus or minus 2.5 percentage points, and that the true difference in proportions is zero,
then a sample size of 3000 children in each of 2 samples will
have 80% power to achieve this level of equivalence. Power
calculations were performed using PASS 11 software.28

RESULTS
The 3 metrics are highly reproducible, with retention
rates remaining stable across 3 samples at each sample

size (2000, 5000, 10,000) in all 12 states. Naturally the

variability between samples decreases as the sample size
increases, but even at the smallest sample size, the difference in measurements provided by all the 3 metrics between samples is less than 5%. The largest differences
observed for the Duration metric at 6, 12, and 18 months
after enrollment are 1.4% across samples of 10,000 and
3.6% for samples of 2000 (Fig. 2). For the Infant metric,
the largest differences observed at 6, 12, and 18 months after enrollment are 1.4% across samples of 10,000 and 3.9%
for samples of 2000 (Fig. 3). Similarly, the largest differences observed for the Coverage metric ranged from
1.3% across samples of 10,000 to 3.2% across samples of
2000 (Fig. 4).
The Duration and Coverage metrics are also highly
reproducible across counties within each state (Online
Appendixes A and B). The number of counties varies in
each state; thus, the population size in each county could
range from 100 to up to 100,000. The ranking of the
counties in the top 5% and bottom 5% within each state remained relatively stable for counties with a sample size over
2000 when using the Duration metric and the Coverage
metric (Tables 1 and 2; Online Appendixes C and D). The
ranking of the counties within states became unstable in
some cases where the sample size is less than 2000.
The demographic information about the states for all 3
metrics is presented in Online Appendix E. We found
that child health insurance retention rates vary over time,
between states, and when using different metrics (Online
Appendix F). Using the Duration metric, retention rates
dropped as the time period after the start of the observation
period lengthened such that the rates at 12 and 18 months
were lower than those at 6 months (Online Appendix F,
panel A). At 6 months after enrollment, children who remained continuously enrolled ranged from 80.2% in Arizona to 96.1% in Idaho. At 12 months, children who
remained continuously enrolled ranged from 60.6% in Arizona to 81.9% in Louisiana. At 18 months, children who
remained continuously enrolled ranged from 41.0% in Arizona to 72.8% in Louisiana. For infants in most states, the
6- and 12-month retention rates remained relatively stable,
followed by a sharp drop at 18 months (Online Appendix F,
panel B). When compared to the Duration metric for all
newly enrolled children, the retention rates for infants at
all 3 time points were generally higher. At 6 months, infants that remained continuously enrolled ranged from
80.6% in New Jersey to 96.3% in Idaho. At 12 months, infants that remained continuously enrolled ranged from
64.0% in New Jersey to 91.9% in Indiana. At 18 months,
infants that remained continuously enrolled ranged from
39.7% in Arizona to 85.1% in Illinois. Coverage rates
also varied by state, ranging from 72.6% for a child in Arizona to 84.0% in Illinois (Online Appendix F, panel C).

DISCUSSION
All 3 insurance retention metrics were highly reproducibleeven with sample sizes of 2000thus facilitating
potential use by governmental agencies and other

ACADEMIC PEDIATRICS

MEDICAID AND CHIP RETENTION

253

A
% of children continuously enrolled

100.0%

AZ
ID

95.0%

IL
IN

90.0%

LA
85.0%

MT
NC

80.0%

NH
NJ

75.0%

NM

VA

70.0%
N=2000,
Sample 1

% of children continuously enrolled

N=2000,
Sample 2

N=2000,
Sample3

N=5000,
Sample 1

N=5000,
Sample 2

N=5000,
Sample 3

N=10000,
Sample 1

N=10000,
Sample 2

WI

N=10000,
Sample3

85.0%

AZ
ID

80.0%

IL
IN

75.0%

LA
70.0%

MT
NC

65.0%

NH
NJ

60.0%

NM
VA

55.0%

N=2000,
Sample 1

% of children continuously enrolled

N=2000,
Sample 2

N=2000,
Sample3

N=5000,
Sample 1

N=5000,
Sample 2

N=5000,
Sample 3

N=10000,
Sample 1

N=10000,
Sample 2

WI

N=10000,
Sample3

80.0%

AZ

75.0%

ID

70.0%

IL

65.0%

IN
LA

60.0%

MT

55.0%

NC

50.0%

NH

45.0%

NJ

40.0%

NM
VA

35.0%
N=2000,
Sample 1

N=2000,
Sample 2

N=2000,
Sample3

N=5000,
Sample 1

N=5000,
Sample 2

N=5000,
Sample 3

N=10000,
Sample 1

N=10000,
Sample 2

N=10000,
Sample3

WI

Figure 2. Duration metric reproducibility for 12 states across 3 random samples using 3 sample sizes. Reproducibility of the (A) 6-month, (B)
12-month, and (C) 18-month Duration metric. The Duration metric quantifies the number of newly enrolled children (as defined by CMS and
OMB) continuously enrolled in public insurance at 6, 12, 18 months after enrollment.

organizations interested in monitoring retention. The insurance retention metrics were less reproducible with sample
sizes of fewer than 2000. Therefore, we recommend that
states with small counties group counties with similar population characteristics together to attain sample sizes of at
least 2000.
Examining child enrollment and retention rates is of significant interest to state public insurance program leaders,
and this study builds on ongoing work to this end.29 Each of
the 3 metrics provides different information about the status of public insurance retention among children in a given
state. The Duration metric provides an accurate estimate of
the retention rates of new enrollees, defined by the CMS
and OMB as those without insurance in the 30 days before
the initial observed enrollment during a selected enrollment period. As previously noted, the Duration metric
thus censors children who are not new enrollees (ie, leftcensoring), excluding up to 75% of children enrolled in
Medicaid/CHIP from the analysis. Many of these censored
children are the most successful, with the highest rates of
insurance retention. The Infant Duration metric does not
suffer from this form of left-censoring because all infants
born during the observation period and enrolled in public
insurance are included in the analysis. However, infants
generally have higher retention rates than the entire child

population as a result of more liberal income-eligibility

thresholds for younger versus older children and greater
opportunities for continuity of coverage provided by
maternal prenatal insurance.3032 Thus, results for infants
have limited generalizability to older children. The
Coverage metric uses the entire child population and
therefore provides population-level information. On the
other hand, this metric does not capture the precise timing
or number of insurance gaps that children may encounter
and presumes that all children in the look-back period
are eligible for public insurance, thus providing an upper
bound of the true estimate of population-level coverage
rates. In addition to the different properties of the 3 metrics,
the observation periods for the Duration metrics are
different from the Coverage metric, which makes the estimates not directly comparable but provides a more complete picture of the entire study period.
Child health insurance retention rates also vary greatly
across states. One explanation for the high retention rates
in some states might be related to the states eligibility,
enrollment, and renewal policies. In 2006, states implemented simplification policies such as no interview at
application and/or renewal, presumptive eligibility, and
12-month continuous eligibility.30 Newer policies such as
Express Lane Eligibility, which uses data matches with

254

PATI ET AL

100.0%

% of infants continuously enrolled

ACADEMIC PEDIATRICS

AZ

ID

95.0%

IL
IN

90.0%

LA
MT

85.0%

NC
NH

80.0%

NJ
NM

75.0%
N=2000,
Sample 1

N=2000,
Sample 2

N=2000,
Sample3

N=5000,
Sample 1

N=5000,
Sample 2

N=5000,
Sample 3

N=10000,
Sample 1

N=10000,
Sample 2

N=10000,
Sample3

VA

95.0%

% of infants continuously enrolled

AZ
90.0%

ID
IL

85.0%

IN
LA

80.0%

MT
75.0%

NC
NH

70.0%

NJ

65.0%

NM
VA

60.0%
N=2000,
Sample 1

% of infants continuously enrolled

N=2000,
Sample 2

N=2000,
Sample3

N=5000,
Sample 1

N=5000,
Sample 2

N=5000,
Sample 3

N=10000,
Sample 1

N=10000,
Sample 2

N=10000,
Sample3

90.0%

WI

AZ
ID

80.0%

IL

IN

70.0%

LA
60.0%

MT
NC

50.0%

NH
NJ

40.0%

NM

VA

30.0%
N=2000,
Sample 1

N=2000,
Sample 2

N=2000,
Sample3

N=5000,
Sample 1

N=5000,
Sample 2

N=5000,
Sample 3

N=10000,
Sample 1

N=10000,
Sample 2

N=10000,
Sample3

WI

Figure 3. Infant Duration metric reproducibility for 12 states across 3 random samples using 3 sample sizes. Reproducibility of the (A) 6month, (B) 12-month, (C) and 18-month Infant Duration metric. The Infant Duration metric quantifies the number of infants born during the
enrollment period continuously enrolled in public insurance at 6, 12, and 18 months after enrollment. Resampling using a size of 10,000
for the Infants Duration metric in New Hampshire was not possible because there were fewer than 10,000 infants.

Population average % of months covered in

18 months

the Supplemental Nutrition Assistance Program to find

already verified low-income families and give them
Medicaid cards, has also been shown to increase enrollment and retention among eligible children.33
Our findings also demonstrate that child health insurance retention rates vary with the duration of measurement such that, in most states, retention rates at 18
months were lower than those at 6 and 12 months. The
drop in retention rates over time is consistent with
other studies exploring the disenrollment patterns from

Medicaid and CHIP.3437 Fairbrother and colleagues37

found a sharp drop after 12 months among children
newly enrolled in Ohio Medicaid. Dick and colleagues35 looked specifically at CHIP renewal policies
and found that states without passive reenrollment
resulted in about half of the children dropping out of
CHIP at every recertification point. As most of states
examined in our study require renewal annually (ie, 12
months after initial enrollment), these findings contribute
to the evidence that children are particularly vulnerable to

88.0%
AZ

86.0%

ID

84.0%

IL

82.0%

IN

80.0%

LA
MT

78.0%

NC

76.0%

NH

74.0%

NJ

72.0%

NM
VA

70.0%
N=2000,
Sample 1

N=2000,
Sample 2

N=2000,
Sample3

N=5000,
Sample 1

N=5000,
Sample 2

N=5000,
Sample 3

N=10000,
Sample 1

N=10000,
Sample 2

N=10000,
Sample3

WI

Figure 4. Coverage metric reproducibility for 12 states across 3 random samples using 3 sample sizes. The Coverage metric measures the
population average of months enrolled in public insurance.

ACADEMIC PEDIATRICS

Table 1. Counties Ranked in Top or Bottom 5% Using Duration Metric Within Each State Across 3 Samples

State
AZ
ID
IL
IN
LA
MT
NC
NH
NJ
NM
VA
WI

No. of Counties
in State

No. of Counties Ever

Ranked in Top or
Bottom 5% Across 3
Samples*

15
44
102
92
64
115
100
10
21
33
134
72

6
10
31
33
18
34
44
5
8
3
41
22

6 Months

12 Months

No. of Counties With

Large Rank Changes

Population Sizes
Among Counties With
Large Rank Changes

3
0
3
8
3
6
10
0
1
0
2
5

145736,833
.
11791622
3411865
4731195
4101179
6411999
.
1430
.
7511171
3861411

18 Months

No. of Counties With

Large Rank Changes

Population Sizes
Among Counties With
Large Rank Changes

No. of Counties With

Large Rank Changes

Population Sizes
Among Counties With
Large Rank Changes

0
0
4
5
3
2
6
0
1
0
2
4

.
.
4661622
4281908
4731956
10611094
6271931
.
1396
.
424725
6891411

0
0
3
5
2
2
7
0
2
0
2
3

.
.
3351466
5921754
473889
11331424
3731999
.
13021396
.
7251171
3391411

MEDICAID AND CHIP RETENTION

The Duration metric quantifies the proportion of newly enrolled children (as defined by the Centers for Medicare and Medicaid Services and the Office of Management and Budget) who are continuously
enrolled in public insurance at 6, 12, and 18 months after enrollment.
*In each state, 3 samples were used to calculate the Duration metric for each county. The counties were then ranked by the percentage of children continuously enrolled at 6, 12, and 18 months (ie, the
county with highest percentage of children continuously enrolled will be ranked as 1). For counties with 2000 or more children, 3 random samples of 2000 were used. For counties with 200 to <2000 children,
3 random samples of 200 were used. For counties with fewer than 200 children, all observations were used without resampling.
Counties with large rank changes are defined as counties that changed ranking by more than half of the total number of counties in that state across 3 random samples. For example, county 9 in Arizona
was ranked 11 in sample 1 and ranked 1 in samples 2 and 3. The ranking for county 9 across the 3 samples changed by more than 7.5 (ie, 15/2), and thus this county was included in the total count for Arizona
in this column.

255

256

PATI ET AL

ACADEMIC PEDIATRICS

Table 2. Counties Ranked in Top or Bottom 5% Using the Coverage

Metric Within Each State Across 3 Samples

State
AZ
ID
IL
IN
LA
MT
NC
NH
NJ
NM
VA
WI

No. of
Counties
in State

No. of Counties
Ranked in
Top or Bottom
5% Across 3
Samples*

No. of
Counties
With Large
Rank
Changes

Population
Sizes in
Counties
With Large
Rank Changes

15
44
102
92
64
115
100
10
21
33
134
72

2
8
17
18
8
17
15
3
2
3
23
12

0
0
2
3
0
0
1
0
0
0
1
2

.
.
11951970
11341943
.
.
1621
.
.
.
1664
14231549

The Coverage metric measures the population average of months

enrolled in public insurance.
*In each state, 3 samples were used to calculate the Coverage
metric for each county. The counties were then ranked by the
average of months children were enrolled in public insurance (ie,
the county with highest average will be ranked as 1). For counties
with 2000 or more children, 3 random samples of 2000 were used.
For counties with 200 to <2000 children, 3 random samples of
200 were used. For counties with fewer than 200 children, all observations were used without resampling.
Counties with large rank changes are defined as counties that
changed ranking by more than half of the total number of counties
in that state across 3 random samples. For example, county 79 in
Illinois was ranked 78 in sample 1, 2 in sample 2, and 9 in sample
3. The ranking for county 79 across the 3 samples changed by
more than 51 (ie, 102/2), and thus this county was included in the total count for Illinois in this column.

in public health insurance or remain uninsured than try to

obtain private insurance.37,38 A further limitation of this
study is that 2009 MAX data were the most recent
national data available at time of study, thus limiting the
timeliness of the results. As a result, the impact of some
recent state efforts to streamline Medicaid and CHIP
enrollment and renewal processes are not yet reflected in
our data. Despite this limitation, this study demonstrates
the utility of using the comprehensive set of these 3
insurance retention measures. Lastly, details of state and
local implementation of renewal procedures are difficult
to ascertain. If states implement these metrics, additional
information about local implementation policies and
processes (ie, lessons learned) could provide valuable
insights to share between states to improve retention rates.
During the current period of health care reform and
budgetary constraints, federal and state policy makers need
to regularly assess the status of Medicaid and CHIP child
health insurance retention. Utilizing the 3 insurance retention
metrics described in this study offers policy makers a robust
picture of the status of new enrollees, infants, and the broader
population of Medicaid/CHIP-eligible children to better
inform decision making. Further validation should test the
reproducibility of these 3 metrics among subgroups of children, such as those with chronic conditions and racial/ethnic
minorities. The recent success experienced by some states in
streamlining enrollment and renewal processes while cutting
costs, suggests that other states may be able to implement
similar policies to ensure eligible children benefit from
continuous public health insurance.

ACKNOWLEDGMENTS
disenrollment at renewal time points and during the
renewal process.
One limitation of the study is the potential selection bias
in the MAX data set, which includes only children enrolled
in Medicaid and CHIP. Unfortunately, a comprehensive
child health insurance database inclusive of private and
public insurance information is not available. We could
not capture whether there is a gap when a child transitions
from one public insurance program to another because
Medicaid and CHIP programs record and report eligibility
data differently. Additionally, MAX does not track the
reason a child disenrolls and this prevents us from knowing
whether a child was not properly reenrolled, or whether the
child became ineligible due to an increase in household
income or change in employment. Another limitation of
these metrics is that data must be collected for 12 to 18
months after a policy change occurs to determine its
impact. Prospective monitoring of reasons for disenrollment from a sample of disenrollees may better serve the
goal of understanding causes of disenrollment and transitions to other insurance programs, but this approach
requires substantial resources and additional work for staff
as eligible children have to be identified and tracked. Nonetheless, the literature documents that children disenrolled
from Medicaid and CHIP are more likely to try to reenroll

This work was funded by a Pediatric Quality Measures Program grant

U18HS20408 from the Agency for Healthcare Research and Quality. The
content is solely the responsibility of the authors and does not necessarily
represent the official views of the Agency for Healthcare Research and
Quality. We also thank the Medicaid staff from Illinois and Louisiana
for discussing these results with the research team.

SUPPLEMENTARY DATA
Supplementary data related to this article can be found
online at http://dx.doi.org/10.1016/j.acap.2014.09.012.

REFERENCES
1. Aiken KD, Freed GL, Davis MM. When insurance status is not static:
insurance transitions of low-income children and implications for
health and health care. Ambul Pediatr. 2004;4:237243.
2. Holl JL, Szilagyi PG, Rodewald LE, et al. Evaluation of New York
States Child Health Plus: access, utilization, quality of health care,
and health status. Pediatrics. 2000;105(3 suppl E):711718.
3. Schoen C, DesRoches C. Uninsured and unstably insured: the importance of continuous insurance coverage. Health Serv Res. 2000;35(1
pt 2):187206.
4. Olson LM, Tang SF, Newacheck PW. Children in the United States
with discontinuous health insurance coverage. N Engl J Med. 2005;
353:382391.
5. Fairbrother G, Jain A, Park HL, et al. Churning in Medicaid managed
care and its effect on accountability. J Health Care Poor Underserved.
2004;15:3041.

ACADEMIC PEDIATRICS
6. Haggerty JL, Reid RJ, Freeman GK, et al. Continuity of care: a multidisciplinary review. BMJ. 2003;327(7425):12191221.
7. Weissman JS, Gatsonis C, Epstein AM. Rates of avoidable hospitalization by insurance status in Massachusetts and Maryland. JAMA.
1992;268:23882394.
8. Yu SM, Bellamy HA, Kogan MD, et al. Factors that influence receipt
of recommended preventive pediatric health and dental care. Pediatrics. 2002;110:e73.
9. Eisert S, Gabow P. Effect of Child Health Insurance Plan enrollment
on the utilization of health care services by children using a public
safety net system. Pediatrics. 2002;110:940945.
10. Bethell CD, Kogan MD, Strickland BB, et al. A national and state profile of leading health problems and health care quality for US children: key insurance disparities and across-state variations. Acad
Pediatr. 2011;11(3 suppl):S22S33.
11. Halfon N, Newacheck PW, Wood DL, St Peter RF. Routine emergency department use for sick care by children in the United States.
Pediatrics. 1996;98:2834.
12. Shi LY, Samuels ME, Pease M, et al. Patient characteristics associated
with hospitalizations for ambulatory care sensitive conditions in
South Carolina. South Med J. 1999;92:989998.
13. Suruda A, Burns TJ, Knight S, Dean JM. Health insurance, neighborhood income, and emergency department usage by Utah children,
19961998. BMC Health Serv Res. 2005;5:29.
14. Orr ST, Charney E, Straus J, Bloom B. Emergency room use by lowincome children with a regular source of health-care. Med Care. 1991;
29:283286.
15. Simon AE, Driscoll A, Gorina Y, et al. A longitudinal view of child
enrollment in Medicaid. Pediatrics. 2013;132:656662.
16. Sommers BD. From Medicaid to uninsured: drop-out among children
in public insurance programs. Health Serv Res. 2005;40:5978.
17. Sommers BD. Enrolling eligible children in Medicaid and CHIP: a
research update. Health Aff (Millwood). 2010;29:13501355.
18. 111 th Congress. Childrens Health Insurance Program Reauthorization Act of 2009. HR 2. Available at: http://www.govtrack.us/
congress/bills/111/hr. Accessed February 22, 2013.
19. 111 th Congress. Patient Protection and Affordable Care Act. HR
3590. Available at: http://www.govtrack.us/congress/bills/111/hr3.
Accessed February 22, 2013.
20. Kaiser Commission on Medicaid and the Uninsured. Hoping for Economic Recovery, Preparing for Health Reform: A Look at Medicaid
Spending, Coverage and Policy TrendsResults from a 50-State
Medicaid Budget Survey for State Fiscal Years 2010 and 2011. Washington, DC: Kaiser Commission on Medicaid and the Uninsured;
2010.
21. General Services Administration. CHIP Annual Report Template System (CARTs). Available at: http://www.reginfo.gov/public/do/
PRAViewIC?ref_nbr201111-0938-009&icID200547. Accessed
May 10, 2013.
22. Byrd V, Nysenbaum J, Lipson D. Encounter Data Toolkit. Washington, DC: Mathematica Policy Research; 2013.

MEDICAID AND CHIP RETENTION

257

23. Department of Health and Human Services. Medicaid Managed Care

Encounter Data: Collection and Use. Washington, DC: Office of
Inspector General; Dept of Health and Human Services; 2009.
24. Byrd VLH, Verdier J. Collecting, Using, and Reporting Medicaid
Encounter Data: A Primer for States. Washington, DC: Mathematica
Policy Research; 2011.
25. Greenwood M. A report on the natural duration of cancer. Rep Public
Health Med Subj (Lond). 1926;33:126.
26. Hosmer DW, Lemeshow S, May S. Applied Survival Analysis:
Regression Modeling of Time to Event Data. New York, NY: J. Wiley;
2011.
27. Farrington CP, Manning G. Test statistics and sample size formulae
for comparative binomial trials with null hypothesis of non-zero
risk difference or non-unity relative risk. Stat Med. 1990;9:
14471454.
28. PASS11 [computer program]. Kaysville, Utah: NCSS; 2011.
29. Trenholm C, Harrington M, Snyder A, Weiss AM. Using Data to
Drive State Improvement in Enrollment and Retention Performance.
Princeton, NJ: Robert Wood Johnson Foundation; 2011.
30. Kaiser Commission on Medicaid and the Uninsured. Resuming the
Path to Health Coverage for Children and Parents: A 50 State Update
on Eligibility Rules, Enrollment and Renewal Procedures, and CostSharing Practices in Medicaid and SCHIP in 2006. Washington, DC:
Henry J. Kaiser Family Foundation; 2007.
31. Kaiser Commission on Medicaid and the Uninsured. Health
Coverage for Children and Families in Medicaid and SCHIP: State
Efforts Face New Hurdles. Washington, DC: Henry J. Kaiser Family
Foundation; 2008.
32. Kaiser Commission on Medicaid and the Uninsured. Challenges of
Providing Health Coverage for Children and Parents in a Recession:
A 50 State Update on Eligibility Rules, Enrollment and Renewal Procedures, and Cost-Sharing Practices in Medicaid and SCHIP in 2009.
Washington, DC: Henry J. Kaiser Family Foundation; 2009.
33. Blavin F, Kenney GM, Huntress M. The effects of express lane eligibility on Medicaid and CHIP enrollment among children. Health Serv
Res. 2014;49:12681289.
34. Carrasquillo O, Himmelstein DU, Woolhandler S, Bor DH. Can
Medicaid managed care provide continuity of care to new Medicaid
enrollees? An analysis of tenure on Medicaid. Am J Public Health.
1998;88:464466.
35. Dick AW, Allison RA, Haber SG, et al. Consequences of states policies for SCHIP disenrollment. Health Care Financ Rev. 2002;23:
6588.
36. Sommers BD. Loss of health insurance among non-elderly adults in
Medicaid. J Gen Intern Med. 2009;24:17.
37. Fairbrother G, Madhavan G, Goudie A, et al. Reporting on continuity
of coverage for children in Medicaid and CHIP: what states can learn
from monitoring continuity and duration of coverage. Acad Pediatr.
2011;11:318325.
38. Sommers BD. Why millions of children eligible for Medicaid and
SCHIP are uninsured: poor retention versus poor take-up. Health
Aff (Millwood). 2007;26:w560w567.