Article

Estimation of the Period Prevalence of Inflammatory Bowel Disease Among Nine Health Plans Using Computerized Diagnoses and Outpatient Pharmacy Dispensings

Harvard University, Cambridge, Massachusetts, United States
Inflammatory Bowel Diseases (Impact Factor: 4.46). 04/2007; 13(4):451-61. DOI: 10.1002/ibd.20021
Source: PubMed
ABSTRACT
There are few contemporary estimates of prevalence rates for inflammatory bowel disease (IBD) in diverse North American communities.
We estimated the period prevalence of IBD for January 1, 1999, through June 30, 2001, among 1.8 million randomly sampled members of nine integrated healthcare organizations in the US using computerized diagnoses and outpatient pharmaceutical dispensing. We also assessed the positive predictive value (PPV) and sensitivities of 1) the case-finding algorithm, and 2) the 30-month sampling period using medical chart review and linkage to a 78-month dataset, respectively.
The PPV of the case-finding algorithm was 81% (95% confidence interval [CI], 78-87) and 84% (95% CI, 79-89) in two different organizations. In both, the sensitivity of the optimal algorithm, compared with the most inclusive, exceeded 90%. The sensitivity of the 30-month sampling period compared with 78 months was 61% (95% CI, 57-64) in one organization. Applying a slightly more sensitive case-finding algorithm, the average period prevalence of IBD across the nine organizations, standardized to the age- and gender-distribution of the US population, 2000 census, was 388 cases (95% CI, 378-397) per 100,000 persons (range 209-784 per 100,000; average follow-up 26 months). The prevalence of Crohn's disease, ulcerative colitis, and unspecified IBD was 129, 191, and 69 per 100,000, respectively.
The observed average prevalence was similar to prevalence proportions reported for other North American populations (369-408 per 100,000). Additional research is needed to understand differences in the occurrence of IBD among diverse populations as well as practice variation in diagnosis and treatment of IBD.

Full-text

Available from: Jennifer Elston Lafata, Sep 15, 2014
ORIGINAL ARTICLE
Estimation of the Period Prevalence of Inflammatory Bowel
Disease Among Nine Health Plans Using Computerized
Diagnoses and Outpatient Pharmacy Dispensings
Lisa J. Herrinton, PhD,* Liyan Liu, MD,* Jennifer Elston Lafata, PhD,† James E. Allison, MD,‡
Susan E. Andrade, ScD,§ Eli J. Korner, PharmD,P K. Arnold Chan, ScD, MD,¶ Richard Platt, MD,¶
6
Deborah Hiatt, RN, MPH,** and Siobha´n O’Connor, MD, MPH**
Background: There are few contemporary estimates of prevalence
rates for inflammatory bowel disease (IBD) in diverse North Amer-
ican communities.
Methods: We estimated the period prevalence of IBD for January
1, 1999, through June 30, 2001, among 1.8 million randomly sam-
pled members of nine integrated healthcare organizations in the US
using computerized diagnoses and outpatient pharmaceutical dis-
pensing. We also assessed the positive predictive value (PPV) and
sensitivities of 1) the case-finding algorithm, and 2) the 30-month
sampling period using medical chart review and linkage to a 78-
month dataset, respectively.
Results: The PPV of the case-finding algorithm was 81% (95%
confidence interval [CI], 78 87) and 84% (95% CI, 79 89) in two
different organizations. In both, the sensitivity of the optimal algo-
rithm, compared with the most inclusive, exceeded 90%. The sen-
sitivity of the 30-month sampling period compared with 78 months
was 61% (95% CI, 57– 64) in one organization. Applying a slightly
more sensitive case-finding algorithm, the average period prevalence
of IBD across the nine organizations, standardized to the age- and
gender-distribution of the US population, 2000 census, was 388
cases (95% CI, 378–397) per 100,000 persons (range 209 –784 per
100,000; average follow-up 26 months). The prevalence of Crohn’s
disease, ulcerative colitis, and unspecified IBD was 129, 191, and 69
per 100,000, respectively.
Conclusions: The observed average prevalence was similar to
prevalence proportions reported for other North American popula-
tions (369 408 per 100,000). Additional research is needed to
understand differences in the occurrence of IBD among diverse
populations as well as practice variation in diagnosis and treatment
of IBD.
(Inflamm Bowel Dis 2007;13:451– 461)
Key Words: inflammatory bowel disease, epidemiology, predictive
values, sensitivity, prevalence
P
revalence estimates of inflammatory bowel disease (IBD)
in North American populations are urgently needed.
IBD— comprising Crohn’s disease (CD), ulcerative colitis
(UC), and indeterminate colitis— constitutes a major chronic
disease associated with substantial morbidity. These disor-
ders have onsets early in life and have a chronic course, often
requiring long-term and often expensive health care. They
affect both quality of life and work productivity. Evidence
suggests that the prevalence varies across geographic loca-
tions, time periods, and racial and ethnic populations.
1
Yet
prevalence rates have been published for only a small number
of North American areas. Several of these estimates are
outdated, while others were obtained for populations that
were quite homogeneous.
We estimated the period prevalence for IBD among
nine integrated healthcare organizations across the US for
the 30-month period January 1, 1999, through June 30,
2001. The case-finding algorithm that detected prevalent
IBD cases used computerized diagnoses and outpatient
pharmacy dispensings for mesalamine, olsalazine, or bal-
salazide. The sensitivity and positive predictive value
(PPV) of the case-finding algorithm were assessed using
linkage and medical chart review.
Received for publication January 27, 2006; accepted October 4, 2006.
From the *Division of Research, Kaiser Permanente Northern California,
Oakland, CA; HMO Research Network CERT; †Center for Health Services
Research, Henry Ford Health System, Detroit, MI; HMO Research Network
CERT; ‡Division of Gastroenterology, Department of Internal Medicine,
University of California, San Francisco, CA; §Meyers Primary Care Institute
(University of Massachusetts Medical School and Fallon Foundation); HMO
Research Network CERT; PClinical Research Unit, Kaiser Permanente
Colorado, Aurora, CO; School of Pharmacy, University of Colorado Health
Sciences Center, Denver, CO; HMO Research Network CERT; ¶Department
of Ambulatory Care and Prevention, Harvard Pilgrim Health Care and
Harvard Medical School; Channing Laboratory, Brigham and Women’s
Hospital, Harvard Medical School; HMO Research Network CERT; **Cen-
ters for Disease Control and Prevention, National Center for Infectious
Diseases, Atlanta, GA.
Reprints: Lisa J. Herrinton, PhD, Division of Research, Kaiser Permanente
Northern California, 2000 Broadway, Oakland, CA, 94612 (e-mail:
Lisa.Herrinton@kp.org).
Copyright © 2007 Crohn’s & Colitis Foundation of America, Inc.
DOI 10.1002/ibd.20021
Published online 11 January 2007 in Wiley InterScience (www.
interscience.wiley.com).
Inflamm Bowel Dis
Volume 13, Number 4, April 2007 451
Page 1
MATERIALS AND METHODS
Overview
The study was conducted using a database that was
developed by the Health Maintenance Organization Research
Network Centers for Education and Research in Therapeutics
(HMORN CERT). The HMORN CERT consists of integrated
healthcare delivery systems; their research programs, enrollee
populations, and databases; and the procedures and trained
personnel to use the databases efficiently. It aims to conduct
research and provide education to advance the optimal use of
drugs, medical devices, and biological products.
A preexisting dataset was used for this study.
2
The
dataset, referred to here as the HMORN CERT core dataset,
provides a study population of about 1.8 million health-plan
members that were randomly selected from a base population
of about 6 million members from the nine health plans. At
each of the health plans, the HMORN CERT core dataset
contains a representative sample of about 200,000 health plan
members who had prescription coverage during the 30-month
period from January 1, 1999, through June 30, 2001. Nine
health plans (referred to here as Plans B through J) partici-
pated in this study, with two (Plans C and H) contributing
information to assess the sensitivity and PPV of the case-
finding algorithm. The nine plans vary with respect to the
number of members served (several hundred thousand to
several million), the organization of the providers (network,
staff, and group models), and the payment for medical ser-
vices (capitated versus fee-for-service).
The HMORN CERT core dataset is a distributed data-
set, i.e., comprised of datasets residing at each of the health
plans. The datasets have nearly identical formats and data
elements, and data requests are coordinated through the Data
Coordinating Center at the Channing Laboratory, Harvard
Medical School. The dataset contains computerized diag-
noses and outpatient dispensings for the entire 30-month
period, including the following: membership, demographics,
all outpatient prescription details, inpatient diagnoses and
procedures, and outpatient diagnoses and procedures. Inpa-
tient dispensings were not available.
With respect to covariates, the source dataset contains
information on age and gender, but not race and ethnicity, as
most health plans do not routinely capture this information
into computerized databases. At each of the nine health plans,
two ecological block-group variables—percent with a high
school education and percent at or above poverty level—were
determined by linking members’ addresses to their census
block-group. The number of comorbid conditions was com-
puted using the Deyo modification to the Charlson comor-
bidity index.
3
Length of follow-up was computed as the total
number of months that each health-plan member was enrolled
during the 30-month observation period.
Development and testing of the case-finding algorithm
and estimation of the period prevalence comprised five steps,
as described in the following five sections.
Evaluation of Diagnostic and Dispensing Codes and
Development of the Preliminary Case-finding
Algorithm
Clinicians with expertise in gastroenterology and med-
ical epidemiology assembled a list of relevant ICD-9 diag-
nostic codes and drug names to identify preliminary cases of
IBD, so that patients with relevant diagnostic and dispensing
codes would be selected from computerized encounter and
claims data (Table 1). In addition to codes for IBD, the list
included three codes for nonspecific gastrointestinal condi-
tions whose clinical symptoms may be difficult to distinguish
from those of IBD or that might be assigned to early IBD. It
also included four codes for possible complications of IBD or
comorbid conditions. The drugs included mesalamine, olsala-
zine, and balsalazide; sulfasalazine; immune modulators (6-
mercaptopurine, azathioprine, methotrexate); infliximab; and
three antibiotics (metronidazole, rifampin, and rifabutin). Al-
though commonly used to treat IBD, the study did not exam-
ine dispensings of corticosteroids because they are used for a
wide variety of other diseases.
These diagnostic and drug codes were evaluated among
the 200,000 members of one plan (Plan C) who were sampled
into the HMORN CERT core dataset. For the comorbid
TABLE 1. Relevant ICD-9 Diagnostic Codes and Drug Names
Identified by an Expert Panel for Detecting Preliminary Cases
of IBD
Diagnosis or Drug
Codes for inflammatory bowel disease
Crohn’s disease (555)
Ulcerative colitis (556)
Codes for nonspecific gastrointestinal conditions
Other noninfectious gastroenteritis and colitis (558.9)
Irritable bowel syndrome (564.1)
Functional diarrhea (564.5)
Codes for IBD complications and comorbid conditions
Fistula of the intestine (569.81)
Ulceration of the intestine (569.82)
Cholangitis (576.1)
Ankylosing spondylitis (720.0)
Drugs
Mesalamine, olsalazine, or balsalazide
Sulfasalazine
Immune modulator (6-mercaptopurine, azathioprine,
methotrexate)
Infliximab
Antibiotic (metronidazole, rifampin, rifabutin)
IBD, inflammatory bowel disease.
Herrinton et al
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
452
Page 2
conditions of IBD (fistula of the intestine, ulceration of the
intestine, cholangitis, and ankylosing spondylitis), we re-
viewed the complete computerized record of diagnostic and
dispensing information. We also reviewed the computerized
record for persons who filled a prescription for mesalamine,
olsalazine, or balsalazide but did not have a diagnosis code of
555 (CD) or 556 (UC). Although infliximab is dispensed in
the infusion clinic and could not be ascertained consistently
across the plans using outpatient pharmacy dispensings, we
nonetheless examined whether some cases would be detected
only through an infliximab dispensing.
Following the evaluation at Plan C, we then defined a
preliminary case-finding algorithm composed of nine mutu-
ally exclusive combinations of diagnosis codes and dispens-
ings (Table 2); we refer to these combinations as “subsets”.
For purposes of this study, the nine algorithm subsets defined
three IBD categories: 1) Crohn’s disease (CD), 2) ulcerative
colitis (UC), and 3) unspecified IBD. For the first two cate-
gories—CD and UC—the highest level of diagnostic cer-
tainty was defined as two or more recorded diagnosis codes of
555 and 556, respectively, during the 30-month period of the
HMORN CERT core dataset; a second level of certainty was
defined as one recorded diagnosis code of 555 or 556 plus an
outpatient dispensing of mesalamine, olsalazine, or balsala-
zide; while a third level of certainty was defined as a single
recorded diagnosis code of 555 or 556 without an outpatient
dispensing of these drugs. Diagnostic certainty was lowest for
the third category, “unspecified IBD.” The three subsets that
made up the unspecified IBD category included the follow-
ing: a) patients with mixed 555 (CD) and 556 (UC) diagnosis
codes; b) patients with an outpatient dispensing of me-
salamine, olsalazine, or balsalazide alone; and c) patients who
received the diagnosis codes 558.9 (other noninfectious gas-
troenteritis and colitis) or 564.1 (irritable bowel syndrome)
with an outpatient dispensing of mesalamine, olsalazine, or
balsalazide, or sulfasalazine, or an immune modulator (6-
mercaptopurine, azathioprine, methotrexate, cyclosporine). In
the remainder of this report we refer to the last subset as
“non-IBD diagnostic codes.”
Determining the Sensitivities of the 30-Month
Observation Period and the Algorithm Subsets
IBD is episodic, and remissions can last for years.
Therefore, the following sources of under-ascertainment of
prevalent IBD were considered: 1) cases with remitting IBD
who did not receive medical attention for their IBD during the
study period, and 2) cases that received medical attention for
active or remitting IBD, or as-yet undiagnosed IBD, but did
TABLE 2. Preliminary and Final Case-finding Algorithms
Assigned IBD Category
Preliminary Case-finding Algorithm
(Nine Subsets)
Final Case-finding Algorithm
(Four Subsets)
CD (1) 2 codes of 555; or
(2) 1 code of 555 and 1 dispensing for mesalamine,
olsalazine, or balsalazide; or
(1) 1 code of 555
(3) 1 code of 555
UC (4) 2 codes of 556; or
(5) 1 code of 556 and 1 dispensing for mesalamine,
olsalazine, or balsalazide; or
(2) 1 code of 556
(6) 1 code of 556
Unspecified IBD
Mixed CDUC
diagnostic codes
(7) 1 code 555 and 1 code 556
(3) 1 code 555 and 1 code 556
Mesalamine, olsalazine,
or balsalazide alone
(8) 1 dispensing of mesalamine, olsalazine, or
balsalazide without a diagnosis code of 555, 556,
558.9, or 564.1 or
(4) 1 dispensing of mesalamine, olsalazine, or
balsalazide without a diagnosis code of 555
or 556
Non-IBD diagnostic
codes
(9) Code 558.9 or 564.1 and 1 dispensing for
mesalamine, olsalazine, or balsalazide;
sulfasalazine; or immune modulator (6-
mercaptopurine, azathioprine, methotrexate)
IBD, inflammatory bowel disease; CD, Crohn’s disease; UC, ulcerative colitis.
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
Prevalence of IBD
453
Page 3
not receive an explicit diagnosis in the medical record. We
sought to separate these sources of under-ascertainment by
using two measures of sensitivity: i) sensitivity of the 30-
month observation period, and ii) the sensitivity of the algo-
rithm. We computed the sensitivity of the 30-month obser-
vation period at Plan C by comparing the occurrence of cases
detected in the 30-month dataset with their occurrence in a
dataset that contained the same health plan members and the
same data elements during an overlapping but longer 78-
month period from January 1, 1995, through June 30, 2001.
In other words, the 30-month HMORN CERT core dataset
was a subset of the 78-month dataset. We computed the
sensitivity of the algorithm at two health plans, Plans C and
H, by comparing the number of cases detected in each subset
of the algorithm against the number in the complete algo-
rithm. The 95% confidence intervals (CIs) were computed
using the exact binomial estimate.
4
Determining the PPV of the 30-Month Dataset Against
Chart Review
For two health plans, Plans C and H, medical chart
review was performed to assess the PPV of overall IBD, as
well as CD, UC, and unspecified IBD. The PPV was
defined as the proportion of cases in the 30-month dataset
whose clinical diagnoses were confirmed through chart
review. For the chart review, we sought to abstract infor-
mation for a total of 400 patients: 200 from each of the two
health plans. We used a stratified, random sampling pro-
cedure to sample preliminary cases into each of the nine
algorithm subsets defined by the preliminary case-finding
algorithm. We sampled more preliminary cases into the
algorithm subsets likely to demonstrate the lowest PPV.
However, for Plan H we did not abstract information for
persons who used mesalamine, olsalazine, or balsalazide
alone because of a programming error that was not iden-
tified until the study had been completed.
Charts were abstracted for the period from January 1,
1997, through June 30, 2001. Three trained medical record
abstractors reviewed all available chart materials. If abstrac-
tors found two or more gastroenterology or internal medicine
notes recording the definite diagnosis of IBD, and these notes
were separated by 6 months or more, they completed a 1-page
abstraction form that included 1–2 lines of narrative support-
ing their view; 316 charts met these criteria. If there was less
support for the diagnosis (N 82), they used an 8-page
instrument to record the diagnoses as well as the symptoms,
procedures, and findings from endoscopy, pathology, surgery,
and radiology, as well as symptoms and diagnoses from
progress notes. A gastroenterologist (J.A.) reviewed every
completed chart-review form, asking for additional support-
ing documents or the entire chart when needed. His opinions
were taken as final, and those are the diagnoses tallied in the
remainder of this report.
Because the numbers of patients sampled from each of
the nine algorithm subsets were fixed by the investigators, it
was necessary to weight the overall PPV. We used the SAS
(Cary, NC) procedure SURVEYMEANS to estimate the
weighted PPV and its 95% CI. The procedure takes into
account that the sampling design was stratified. For three
algorithm subsets (mixed CDUC diagnostic codes; dispens-
ing of mesalamine, olsalazine, or balsalazide alone; and non-
IBD diagnostic codes), we computed stratum-specific but not
weighted PPVs, because they appeared to perform quite dif-
ferently and we judged it inappropriate to summarize across
them.
5
Selecting the Final Case-finding Algorithm
We sought a final case-finding algorithm that maxi-
mized the sensitivities and the PPV for overall IBD (CD plus
UC plus unspecified IBD) rather than for each of the three
IBD categories. To determine which of the nine subsets
should compose the final case-finding algorithm, we esti-
mated the algorithm sensitivities and PPV for increasingly
more specific case-finding algorithms by excluding the fol-
lowing combinations of algorithm subsets that were less
likely to detect true cases of overall IBD: 1) non-IBD diag-
nostic codes; 2) non-IBD diagnostic codes plus mesalamine,
olsalazine, or balsalazide; 3) non-IBD diagnostic codes plus
mixed CDUC codes; and 4) both (2) and (3). We then
selected a final case-finding algorithm, which contained eight
of the nine subsets in their entirety, as well as a portion of
cases from the ninth subset (those with mesalamine, olsala-
zine, or balsalazide and non-IBD diagnostic codes) (Table 2).
The final case-finding algorithm was simplified into four
subsets: i) 1 diagnostic code of 555 (CD), ii) 1 diagnostic
codes of 556 (UC), iii) mixed CDUC codes, or iv) dispens-
ing for mesalamine, olsalazine, or balsalazide without a di-
agnosis code indicating CD or UC. “Overall IBD” included
cases detected by any of the four criteria.
Estimating the Period Prevalence
To identify prevalent cases, the final case-finding algo-
rithm was applied to computerized records for all persons in the
30-month HMORN CERT core dataset across the nine health
plans. The period prevalence was standardized to the age- and
gender-distribution of the US population, 2000 census.
RESULTS
Evaluation of Diagnostic and Dispensing Codes and
Development of a Preliminary Case-finding
Algorithm
Evaluation at Plan C of the diagnostic codes that had
been selected by the expert panel revealed that no cases were
coded with 564.5 (functional diarrhea), and this code was
removed from further consideration. The other codes together
detected 11,816 persons, the vast majority (11,245) with
codes 558.9 (other noninfectious gastroenteritis and colitis)
Herrinton et al
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
454
Page 4
and 564.1 (irritable bowel syndrome). The number detected
by virtue of these two codes remained large (2738) even
when combined with relevant drug codes (mesalamine, ol-
salazine, or balsalazide; sulfasalazine; immune modulators;
or antibiotics). We therefore removed antibiotics and required
persons with these two non-IBD diagnostic codes to have
been dispensed mesalamine, olsalazine, or balsalazide; sul-
fasalazine; or immune modulator. No cases were detected
only through an infliximab dispensing; these cases all had
IBD diagnosis codes as well.
We reviewed the complete computerized record of di-
agnostic and dispensing information for patients without a
diagnosis code of 555 or 556 but with a diagnosis code for
fistula of the intestine (n 30), ulceration of the intestine (n
10), ankylosing spondylitis (n 77), or cholangitis (n
43). Additional computerized information was consistent
with IBD for only two of these patients who had fistula with
unknown etiology. In contrast, 46% of encounters coded as
fistula, 71% coded as ulceration, 59% coded as ankylosing
spondylitis, and 46% coded as cholangitis occurred in con-
junction with diagnosis codes for non-IBD conditions that
could explain the condition. Because of the relatively low
specificity of these codes for finding IBD, we therefore re-
moved them from further consideration. Among the 107
persons who filled a prescription for mesalamine, olsalazine,
or balsalazide but did not have a diagnosis code of 555 (CD)
or 556 (UC), review of all computerized data suggested that
some might have IBD but for others no evidence supported a
diagnosis of IBD during the study period: 46% had no re-
corded medical encounters (other than pharmacy) while 19%
had computerized information consistent with an alternative
therapeutic use of the queried medications. The study re-
tained subjects having drug dispensings of mesalamine, ol-
salazine, or balsalazide but no relevant diagnosis code be-
cause the available information could not dismiss IBD in over
50% of the cases.
Determining the Sensitivities of the 30-Month
Observation Period and the Algorithm Subsets
When we compared the 30-month data to the 78-month
data for Plan C, we observed the results shown in Table 3.
Among the 186 cases of CD observed in the 78-month
dataset, 124 (67%) occurred in the 30-month data, 111 with
1 diagnosis code of 555, two with a dispensing of me-
salamine, olsalazine, or balsalazide alone, and four with non-
IBD codes (not shown). Among the 394 UC cases, 230 (58%)
occurred in the 30-month dataset, 217 detected by 1 one
occurrence of 556, 20 with a dispensing alone, and 15 with
non-IBD codes. In the 78-month dataset, there were 60 per-
sons with mixed CD and UC codes and 45 with a dispensing
of mesalamine, olsalazine, or balsalazide alone; this com-
pares with the 30-month dataset, in which the corresponding
numbers were 13 and 49.
Table 4 provides the total number of cases identified by
the nine algorithm subsets in the 30-month dataset for Plans
C and H. For Plan C, the total number of preliminary cases
identified was 526, with 24% being identified with code 555
(CD), 44% with code 556 (UC), 3% with mixed CDUC
codes, 9% with mesalamine, olsalazine, or balsalazide alone,
and 21% with non-IBD codes. For Plan H, 1020 preliminary
cases were identified: 30% with code 555 (CD), 43% with
code 556 (UC), 7% with mixed CDUC codes, 8% with
mesalamine, olsalazine, or balsalazide alone, and 13% with
non-IBD codes.
Determining the PPV of the 30-Month Dataset
Against Chart Review
The numbers of patients whose charts were reviewed
for Plans C and H were 200 and 198, respectively, represent-
ing 38% and 19% of the total number of persons with these
diagnoses codes at these sites. For Plan C, 32 patient charts
were not accessible because they were being used in a clinic.
For Plan H, charts for two patients were not available.
TABLE 3. Proportion of Cases in the 78-Month Dataset (1/1/95-6/30/01) That Were Also Detected in the 30-Month Dataset (1/1/
99-6/30/01), Plan C
Assigned IBD Category
Number of Cases
Detected in the
30-Month Dataset
a
Number of Cases
Detected in the
78-Month Dataset
Proportion Detected in the
30-Month Dataset, %
CD 124 186 67
UC 230 394 58
Mixed CDUC diagnostic codes 13 60 22
Mesalamine, olsalazine, or balsalazide alone 49 45 109
Non-IBD diagnostic codes
b
90 267 33
IBD, inflammatory bowel disease; CD, Crohn’s disease; UC, ulcerative colitis.
a
Excludes 20 persons who were not members during the entire 78-month period.
b
An encounter coded 558.9 (other noninfectious gastroenteritis and colitis) or 564.1 (irritable bowel syndrome) with a relevant dispensing of mesalamine,
olsalazine, or balsalazide; sulfasalazine; or immune modulator (6-mercaptopurine, azathioprine, or methotrexate).
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
Prevalence of IBD
455
Page 5
TABLE 4. Comparison of the 30-Month CERT Core Dataset IBD Case Classifications, as Generated by the Preliminary Algorithm, Against Chart Review Diagnoses
in Sampled Cases, Plans C and H, 1/1/99-6/30/01
CERT Core
Dataset (30
Months)
Classification on Chart Review
Plan C Plan H
Total
Number
Identified
(n 526)
Total
Number
Abstracted
CD, Row
%
(95% CI)
UC, Row
%
(95% CI)
Indeterminate
IBD, Row %
Not IBD,
Row %
Total
Number
Identified
(n 1,020)
Total
Number
Abstracted
CD, Row
%
(95% CI)
UC, Row
%
(95% CI)
Indeterminate
IBD, Row %
Not IBD,
Row %
CD: 2
codes
(555) 85 22 91 0 0 9 261 36 89 0 6 5
CD: 1 code
dispensing 14 14 86 7 7 0 6 6 67 0 0 33
CD: 1 code 26 24 67 4 4 25 37 33 18 3 0 79
Overall CD 125 60 80 (77–83) 3 3 13 304 75 56 (52–60) 1 3 40
UC: 2
codes
(556) 116 19 0 84 0 16 358 28 0 93 0 7
UC: 1 code
dispensing 76 21 0 71 5 24 22 19 5 37 5 53
UC: 1 code 38 30 10 63 10 17 54 33 0 36 0 64
Overall UC 230 70 4 71 (66–77) 5 20 434 80 1 56 (50–62) 1 41
Mixed
CDUC
codes 15 14 29 36 14 21 76 28 43 39 4 14
Mesalamine,
olsalazine,
or
balsalazide
alone 47 14 7 43 14 36 81 0
Non-IBD
diagnostic
codes
a
109 42 7 14 10 69 125 15 7 0 0 93
IBD, inflammatory bowel disease; CD, Crohn’s disease; UC, ulcerative colitis; CERT, HMO Research Network Centers for Education and Research in Therapeutics. Dispensing: mesalamine,
olsalazine, or balsalzide.
a
An encounter coded 558.9 (other noninfectious gastroenteritis and colitis) or 564.1 (irritable bowel syndrome) with a relevant dispensing of mesalamine, olsalazine, or balsalazide; sulfasalazine;
or immune modulator (6-mercaptopurine, azathioprine, or methotrexate).
Herrinton et al
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
456
Page 6
Comparing the 30-month dataset to chart review, the
weighted PPVs for detecting overall IBD (either CD or UC)
with the preliminary algorithm were 81% (95% CI, 78 87)
and 84% (95% CI, 79 89) for Plans C and H, respectively.
For Plan C, it was 80% (95% CI, 77–83) for CD and 71%
(66 –77) for UC (Table 4). For Plan H, it was 56% for each
(95% CI: CD, 52– 60; UC, 50 62). In addition, 13 cases for
Plan C and 11 for Plan H were classified by the gastroenter-
ologist as “indeterminate IBD.” There were striking differ-
ences between Plans C and H in the PPVs for CD and UC. A
single code of 555, indicating CD, yielded a PPV of 67% for
Plan C but 18% for Plan H. For UC, the respective PPVs were
63% and 36%. Also for Plan H, the PPV for one diagnosis
code plus a pharmacy dispensing was higher for CD (67%)
than for UC (37%).
The stratum-specific PPV for the “mixed CDUC codes”
was 79% (95% CI, 49–95) for Plan C (comprising 29% CD,
36% UC, and 14% indeterminate IBD) and 86% (95% CI,
67–96) for Plan H (comprising 43% CD, 39% UC, and 4%
indeterminate IBD). For a dispensing of mesalamine, olsalazine,
or balsalazide alone, it was 64% (95% CI, 35– 87) for Plan C
(comprising 7% CD, 43% UC, and 14% indeterminate IBD).
For “non-IBD codes,” it was 31% (95% CI, 18 47) for Plan C
(comprising 7% CD, 14% UC, and 10% indeterminate IBD).
For Plan H, only one individual (7%) with non-IBD codes had
IBD, specifically CD, by chart review. Persons detected by the
preliminary algorithm who did not have IBD on chart review
had diverticulitis, irritable bowel syndrome, ischemic colitis, and
gastroenteritis, among less common diagnoses.
Selecting the Final Case-finding Algorithm
The results shown in Table 5 were used to select the
final case-finding algorithm. The table shows the sensitivity
of the algorithm in Plans C and H, the sensitivity of the
30-month observation period in Plan C, and the PPV for
various combinations of subsets that were considered for the
final case-finding algorithm. Compared with including all
nine subsets (Plan C, algorithm sensitivity, 100% by defini-
tion, PPV, 71%; Plan H, algorithm sensitivity, 100% by
definition, PPV, 76%), reducing the number to eight by
excluding persons who received non-IBD codes by definition
decreased the sensitivity of the preliminary algorithm some-
what (Plan C, 91%; Plan H, 99%), but increased the PPV for
each plan (Plan C, 81%, 95% CI, 75%–87%; Plan H, 84%,
95% CI, 79%– 89%); the sensitivity of the 30-month obser-
vation period in Plan C was 61% (95% CI, 57%– 64%).
Exclusion of mixed CD and UC codes and dispensings of
mesalamine, olsalazine, or balsalazide alone did not further
improve the algorithm.
TABLE 5. Comparison of the Sensitivities and PPV of the Preliminary Case-finding Algorithm for Identifying Overall IBD Using
Various Subsets of Codes, 30-Month CERT Core Dataset, Plans C and H, 1/1/99-6/30/01
No. of Subsets
in the
Algorithm
Specific Subsets Excluded
(Refer to Table 2)
Plan C Plan H
n
Sensitivity
of the
Algorithm
a
,
%
Sensitivity of the
30-Month
Observation
Period
b
,%
PPV
for
Overall
IBD,
% n
Sensitivity
of the
Algorithm
a
,
%
PPV
for
Overall
IBD,
%
Nine None 526 100
a
71 1,020 100
c
76
Eight Non-IBD diagnostic codes
d
(9)
417 91 61 (95% CI, 57–64) 81 895 99 86
Seven
Non-IBD diagnostic codes (9);
and Mesalamine, olsalazine,
or balsalazide alone (8) 370 83 84 814 88 84
Seven
Non-IBD diagnostic codes (9);
and Mixed CD and UC
codes (7) 402 88 82 819 90 86
Six
Non-IBD diagnostic codes (9);
and Mesalamine, olsalazine,
or balsalazide alone (8); and
Mixed CDUC codes (7) 355 80 84 738 80 84
IBD, inflammatory bowel disease; CD, Crohn’s disease; UC, ulcerative colitis; CERT, HMO Research Network Centers for Education and Research in
Therapeutics; PPV, positive predictive value.
a
Using subsets of the preliminary case-finding algorithm and comparing algorithms within the 30-month dataset.
b
Using the final case-finding algorithm and comparing the 30-month data set to the 78-month data set in Plan C.
c
By definition.
d
An encounter coded 558.9 (other noninfectious gastroenteritis and colitis) or 564.1 (irritable bowel syndrome) with a relevant dispensing of mesalamine,
olsalazine, or balsalazide; sulfasalazine; or immune modulator (6-mercaptopurine, azathioprine, or methotrexate).
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
Prevalence of IBD
457
Page 7
Estimating the Period Prevalence
The final case-finding algorithm identified 6649 per-
sons with IBD from a representative sample of about 1.8
million health members from nine health plans. Characteris-
tics of members identified with IBD are summarized in Table
6. The number of patients ascertained with IBD using the
algorithm ranged 3-fold from 416 for Plan C to 1,463 for Plan
E, with Plan E having an appreciably higher number of cases
than the other plans.
The proportion of patients ascertained with IBD sub-
types was as follows: CD, 34% (SD, 3%); UC, 48% (SD,
8%); both CD and UC diagnosis codes, 6% (SD, 3%); and an
outpatient dispensing of mesalamine, olsalazine, or balsala-
zide alone, 12% (SD, 11%). The nine health plans generally
were similar in these proportions except that dispensings of
mesalamine, olsalazine, or balsalazide alone ranged from 4%
for Plan E, where the case count was greatest, to 39% for Plan
G, which ranked second in case count.
The mean period prevalence, standardized to the age-
and gender-distribution of the 2000 US census, is shown in
Table 7, with plan-specific estimates shown in Table 8. We
observed a period prevalence of 388 cases of IBD (95% CI,
378 –397) per 100,000 persons during the average 26-month
period of observation. For CD, the period prevalence was 129
per 100,000 persons; for UC, 191 per 100,000; and for
unspecified IBD, 69 per 100,000. The period prevalence of
overall IBD increased with age, as expected, and among those
age 65 years and older was somewhat higher in men than
women. Mirroring the case-counts, there was 3-fold variation
in period prevalence among the nine health plans (range,
209 –784 per 100,000).
DISCUSSION
We estimated the period prevalence of IBD among a
random sample of 1.8 million persons representing more than
6 million members of nine health plans across the US. Av-
eraging across the nine health plans for overall IBD (sum of
CD, UC, and unspecified IBD), we observed a period prev-
alence of 388 IBD cases (95% CI, 378 –397) per 100,000
persons during an average 26-month period of observation.
As expected with a lifelong disease, the period prevalence
increased with age. It was generally similar in men and
women, although among those age 65 and older it was
greatest among men.
Our study was novel in using information from nine
health plans across the US. The 3-fold variation in prevalence
among the health plans suggests that the validity of case-
finding algorithms that use computerized data is dependent on
the context in which they are used, including the care delivery
system and its data systems and practices. Three specific
differences among health plans merit discussion: 1) variation
in clinical practice with respect to classifying and treating
IBD, 2) differences in local coding practices, and 3) genuine
differences in true prevalence proportions among the popu-
lations served.
Doubtless there are differences among the physicians
with respect to the weight of evidence needed to assign a
diagnosis code of 555 or 556 for confirmed or suspected CD
or UC, respectively, and these differences may cluster at the
level of the plan. Some physicians may be more inclined to
code 555 or 556 while a definitive work-up is still in process,
particularly when a condition has an unusual presentation and
is difficult to diagnose. However, the life-long duration of
most IBD and short observation period of the study likely
resulted in the great majority of patients in this population
having prevalent, not incident disease. Or, for example, there
may be great differences in the frequency of coding IBD that
is in remission, or UC that has been “cured” through colec-
tomy. The organization of care may be important in this
regard. The plans included in the study vary in structure from
networks of preferred providers to independent physician
associations and staff/group models. There is variation in the
organization of IBD-related services, e.g., referral to special-
ist gastroenterologists and in-plan versus out-of-plan phar-
macy utilization that could impact diagnosis and treatment.
Understanding these organizational factors would require fur-
ther detailed study.
Another factor that might contribute to variability in
clinical practice is the lack of consensus regarding evidence-
based guidelines in diagnosis and therapeutic management.
6
Considering variation in diagnosis, there are several condi-
tions that may be coded as IBD among some providers but
not others, including microscopic colitis, collagenous colitis,
ischemic colitis, and undiagnosed infectious colitis. In addi-
tion, there could be variation in the diagnosis of indetermi-
nate IBD, which is not clearly CD or UC. Variation in
therapeutic management could result in physician and health-
plan differences in the use of pharmaceuticals to maintain
remission, for example. Similarly, in some settings, diagnoses
recorded in urgent care, primary care, and other non-GI
specialty settings, often based on an oral history of suspected
IBD and past summaries of evaluations for possible IBD or
irritable bowel syndrome, could produce records of multiple
encounters inappropriately coded by 555 or 556.
We strongly suspect that the incentive for coding, be it
reimbursement, quality assessment, or planning, is a key
factor in coding practices, with reimbursement providing the
strongest incentive. Indeed, providers practicing in Plan H see
both fee-for-service and capitated patients, and thus routinely
code all encounters for reimbursement. In contrast, those in
Plan C see capitated patients only. The level of undercoding
and miscoding of computerized encounter and claims data
can be large enough to be of concern for purposes of under-
standing epidemiologic patterns and planning health care
expenditures.
7,8
Recall that outpatient encounter, hospitalization, and
Herrinton et al
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
458
Page 8
TABLE 6. Characteristics of 6649 Members Identified With IBD Diagnostic Codes or Outpatient Pharmacy Dispensings of
Mesalamine, Olsalazine, or Balsalazide Drugs Among 1.8 Million Members of Nine Health Plans (B Through J), 1/1/99-6/30/01
% IBD Cases Detected per Plan (Combined CD, UC, and Unspecified IBD)
B
(n 843)
C
(n 416)
D
(n 682)
E
(n 1463)
F
(n 481)
G
(n 460)
H
(n 983)
I
(n 581)
J
(n 751)
Final algorithm subset
1 code 555, Crohn’s disease 33 30 36 37 35 30 28 36 34
1 code 556, ulcerative colitis 49 55 43 49 48 29 53 56 50
Codes for both 555 and 556 8 3 6 10 4 2 11 4 10
Dispensing of mesalamine,
olsalazine, or balsalazide alone 10 12 15 4 12 39 8 5 6
Source of ascertainment
Inpatient diagnosis, with or
without outpatient diagnosis 17 21 18 20 13 28 20 19 20
2 outpatient diagnoses, no
inpatient diagnosis 49 49 46 52 46 9 52 50 49
1 outpatient diagnosis, no
inpatient diagnosis 24 18 21 23 28 24 21 27 25
Dispensing only 10 12 15 4 12 39 8 5 6
Length of follow-up, mo
12 8 6 7 6 11 8 6 12 6
12–23 20 12 14 20 21 14 12 15 15
24–29 11 6 5 21 1488810
30 (maximum) 62 77 73 53 54 69 73 65 69
Mean 25 27 26 25 24 26 27 25 26
Gender
Male 48 44 44 47 47 45 47 46 48
Female 52 56 56 53 53 55 53 54 52
Age, years
0–18 8 4 5 5 10575 6
19–44 38 41 36 40 52 27 32 40 42
45–64 36 35 39 35 31 37 33 39 39
65 18 21 20 20 6 31 28 17 13
Blockgroup education and income
% with high school education 83 86 89 88 86 89 85 86 91
% income at or above poverty
level 90 90 91 91 92 93 91 88 93
Comorbid conditions
Myocardial infarct 2 2 2 3 1732 2
Congestive heart failure 3 2 3 5 2742 3
Cerebrovascular disease 2 2 1 3 1731 2
Chronic pulmonary disease 6 5 5 6 2 13 6 5 5
Ulcer disease 2 2 1 2 1200 1
Diabetes 2 4 3 5 3843 3
Any tumor 3 3 2 3 2432 2
Charlson Comorbidity Index
0 88 88 89 84 91738589 86
1 5657410556
2 3 2 3 4 2653 3
3 2 2 1 2 1321 1
4 2 2 3 3 2822 3
IBD, inflammatory bowel disease; CD, Crohn’s disease; UC, ulcerative colitis.
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
Prevalence of IBD
459
Page 9
pharmacy data were used for case ascertainment in the
present study. We expected that the number of inpatient
diagnoses would be consistent across the plans, and that there
would be more extensive differences in coding outpatient
encounter and claims forms. However, we observed differ-
ences in the prevalence of cases with inpatient diagnoses,
giving evidence that differences in period prevalence among
the plans may reflect true differences in risk in the underlying
populations served by the plans. Alternatively, it could indi-
cate differences in hospitalization of potential IBD cases for
evaluation and diagnosis, or in physician coding of symptom-
atic cases compared with those in remission.
The preliminary case-finding algorithm demonstrated a
sensitivity exceeding 90% and a PPV exceeding 80% for
overall IBD, with the final algorithm being slightly more
sensitive. The PPV for the IBD subtypes, CD and UC, were
lower, largely because of cross-classification of CD with UC
and vice versa. For Plan C, we estimated the sensitivity of the
30-month study period (compared with a 78-month study
period) to be 61%. It should be noted that Plan C had the
lowest prevalence of all the plans, with coding being unre-
lated to reimbursement. For these reasons, we suspect that
61% may be lower than the sensitivity of the 30-month
sampling period for the other plans.
In this study we obtained only limited sociodemo-
graphic information. Our populations are somewhat better
educated and higher income than the general US population
(2000 Census, attained high school education, 80%; at or
above the poverty level, 88%), but have a similar age
structure.
9
At the same time, we observed rather larger
variations in prevalence from plan to plan that likely were
related to data quality. Considering these limitations, we
believe that our study was not adequate to allow inferences
about population characteristics and regional variation in
TABLE 7. Period Prevalence (per 100,000) of IBD by Gender
and Age Among 1.8 Million Members of Nine Health Plans in
the US, 1999 to June, 2001, 26-Month Average Enrollment
CD UC
CD
UC
Mesalamine,
olsalazine, or
balsalazide
All IBD
Period
prevalence 95% CI
All 129 191 28 41 388 378–397
Male
0–18 40 28 5 7 79 68–92
19–44 118 159 25 31 332 309–356
45–64 185 326 38 56 604 573–636
65 248 452 52 119 871 835–909
Female
0–18 27 29 9 4 68 58–80
19–44 134 174 27 38 374 349–399
45–64 198 303 54 69 623 591–655
65 232 372 44 91 740 706–775
IBD, inflammatory bowel disease; CD, Crohn’s disease; UC, ulcerative
colitis; CI, confidence interval.
a
Age, sex standardized to 2000 US population. Age group used for standard
-
ization: 0 –24, 25–44, 45–54, 55– 64, 65–74, 75.
TABLE 8. Period Prevalence (per 100,000) for 1.8 Million Members of Nine Health Plans in the US, 1999 to June, 2001, 26-Month
Average Enrollment
All
Plans B C D E F G H I J
Average prevalence 388 434 219 334 785 301 230 476 297 420
95% CI 378–397 406–463 199–240 309–360 366–420 278–326 210–252 447–506 274–321 393–449
Male
0–18 79 100 27 77 152 57 27 136 55 68
19–44 332 456 161 283 648 277 151 384 254 374
45–64 604 697 339 497 1084 543 348 750 488 683
65 871 913 474 733 2212 671 610 1008 515 790
Female
0–18 68 106 25 34 131 69 26 113 33 85
19–44 374 433 264 351 707 310 182 464 276 346
45–64 623 663 319 582 11124 495 378 769 546 703
65 740 614 459 539 1773 337 589 870 566 931
IBD, inflammatory bowel disease; CD, Crohn’s disease; UC, ulcerative colitis; CI, confidence interval.
a
Age, sex standardized to 2000 US population. Age group used for standardization: 0–24, 25– 44, 45–54, 55–64, 65–74, 75.
Herrinton et al
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
460
Page 10
prevalence. Further research is needed to understand these
relationships.
Our results can be compared with other North Ameri-
can population-based studies that have been conducted in
northern Alberta and Manitoba (Canada), California, and
Olmsted County (Minnesota) as early as 1977. In Alberta
during the period 1977–1981, counts of patients with one or
more hospital discharge diagnoses or relevant visits to gas-
troenterology practices during the preceding 5-year period
were used to determine the point prevalence of IBD on
December 31, 1981.
10
The number of cases per 100,000 was
44 for CD and 37 for UC, far lower than observed in the
present study or other published North American studies but
reflecting only cases with multiple encounters coded as IBD.
Studies in other geographic areas have yielded estimates of
170 –246 per 100,000 for UC and 144–199 per 100,000 for
CD using collated diagnostic indices with medical-record
review or computerized databases.
11–13
The combined prev-
alence of UC and CD in these studies ranged from 369 408
per 100,000, although they did not report numbers for un-
specified or indeterminate colitis. Our mean estimate of the
period prevalence across the nine health plans for all IBD
(including UC, CD, and unspecified IBD) was comparable, at
388 per 100,000. (Although the authors of the earlier studies
refer to their estimates as “point prevalence,” their time
intervals for case ascertainment were similar to that used in
the present study. For a life-long, chronic disease like IBD,
there is little difference between point prevalence and period
prevalence.)
In conclusion, we found the period prevalence of over-
all IBD to be 388 per 100,000 health plan enrollees (95% CI,
378 –397), suggesting a current total of 1.12 million cases of
IBD in the US. We found substantial variation in prevalence
rates across the nine participating health plans; this variation
in prevalence at the health-plan level may have resulted from
differences in clinical practice with respect to classifying and
treating IBD, differences in local coding practices, and gen-
uine differences in true prevalence proportions among the
populations served. IBD represents a significant public health
burden across the US. Further studies are needed to under-
stand population differences in the occurrence of IBD and
practice variation in diagnosing and managing the disease.
ACKNOWLEDGMENTS
This research was supported by grants from the Crohn’s
and Colitis Foundation (CCFA), the Agency for Healthcare
Research and Quality, and the Kaiser Foundation Research
Institute and with funding from the Centers for Disease
Control and Prevention (CDC). The authors thank Matthew
Zack and Benjamin Gold for technical input.
REFERENCES
1. Loftus EV Jr. Clinical epidemiology of inflammatory bowel disease:
incidence, prevalence, and environmental influences. Gastroenterology.
2004;126:1504 –1517.
2. Chan KA, for the HMO Research Network CERTs Patient Safety Study
Investigators. Development of a multipurpose dataset to evaluate poten-
tial medication errors in ambulatory settings. In: Henriksen K, Battles
JB, Marks ES, Lewin DI, editors. Advances in Patient Safety: From
Research to Implementation, Vol 2. Concepts and Methodology. AHRQ
Publication No. 05-0021-2. Rockville, MD: Agency for Healthcare
Research and Quality; 2005:225–238.
3. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index
for use with ICD-9-CM administrative databases. J Clin Epidemiol.
1992;45:613– 639.
4. Newcombe RG. Two sided confidence intervals for the single propor-
tion: comparison of seven methods. Stat Med. 1998;17:857– 872.
5. Weiss NS. Clinical Epidemiology: The Study of the Outcome of Illness.
New York: Oxford University Press; 1986:19.
6. Tremaine WJ, Sandborn WJ, Loftus EV, et al. A prospective cohort
study of pratice guidelines in inflammatory bowel disease. Am J Gas-
troenterol. 2001;96:2401–2406.
7. Harrold LR, Yood RA, Andrade SE, et al. Evaluating the predictive
value of osteoarthritis diagnoses in an administrative database. Arthritis
Rheum. 2000;43:1881–1885.
8. Andrade SE, Gurwitz JH, Chan KA, et al. Validation of diagnoses of
peptic ulcers and bleeding from administrative databases: a multi-
health maintenance organization study. J Clin Epidemiol. 2002;55:
310 –313.
9. Wagner E, Greene SM, Hart G, et al. Building a research consortium of
large health systems: the Cancer Research Network. J Natl Cancer Inst
Monogr. 2005:3–11.
10. Pinchbeck BR, Kirdeikis J, Thomson AB. Inflammatory bowel disease
in northern Alberta. An epidemiologic study. J Clin Gastroenterol.
1988;10:505–515.
11. Loftus EV Jr, Silverstein MD, Sandborn WJ, et al. Crohn’s disease in
Olmsted County, Minnesota 1940-1993: incidence, prevalence, and sur-
vival. Gastroenterology. 1998;114:1161–1168.
12. Loftus EV Jr, Silverstein MD, Sandborn WJ, et al. Ulcerative colitis in
Olmsted County, Minnesota 1940-1993: incidence, prevalence, and sur-
vival. Gut. 2000;46:336 –343.
13. Bernstein CN, Blanchard JF, Rawsthorne P, at al. Epidemiology of
Crohn’s disease and ulcerative colitis in a central Canadian province: a
population-based study. Am J Epidemiol. 1999;149:916 –924.
Inflamm Bowel Dis
Volume 13, Number 4, April 2007
Prevalence of IBD
461
Page 11
  • Source
    • "Epidemiological studies of IBD in the USA are necessary to quantify the public health burden of disease and inform policy regarding the allocation of resources and provision of health services for affected individuals. Because IBD is not a reportable condition in the USA and comprehensive, nationwide registries for IBD surveillance have not been established, published studies on the epidemiology of IBD in the USA [1–6] are limited and primarily include studies which have sampled small, geographically restricted populations. Furthermore, no studies of IBD prevalence have been published using data from the last 5 years and, therefore, current time trends remain unknown. "
    [Show abstract] [Hide abstract] ABSTRACT: Purpose Most US inflammatory bowel disease (IBD) epidemiology studies conducted to date have sampled small, geographically restricted populations and have not examined time trends. The aim of our study was to determine the prevalence of Crohn’s disease (CD) and ulcerative colitis (UC) in a commercially insured US population and compare prevalences across sociodemographic characteristics and time. Methods Using claims data from approximately 12 million Americans, we performed three consecutive 2-year cross-sectional studies. Cases of CD and UC were identified using a previously described algorithm. Prevalence was estimated by dividing cases by individuals in the source population. Logistic regression was used to compare prevalences by region, age, and sex. Results In 2009, the prevalences of CD and UC in children were 58 [95 % confidence interval (CI) 55–60] and 34 (95 % CI 32–36), respectively. In adults, the respective prevalences were 241 (95 % CI 238–245) and 263 (95 % CI 260–266). Data analysis revealed that IBD prevalences have slightly increased over time. Based on census data, an estimated 1,171,000 Americans have IBD (565,000 CD and 593,000 UC). Conclusions Analysis of the epidemiological data revealed an increasing burden of IBD in recent years, which may be used to inform policy.
    Preview · Article · Aug 2012 · Digestive Diseases and Sciences
  • Source
    • "Validation studies for chronic diseases have also been pub- lished [16,17]. These generally, though implicitly, are studies of disease prevalence rather than incidence. "
    [Show abstract] [Hide abstract] ABSTRACT: Osteonecrosis (ON) is a rare disease associated with alcohol and glucocorticoid use. Identifying additional risk factors is difficult as the number of cases at any single center is small. We investigated whether data available in large health care databases can be used to identify incident ON cases. Using data from the Boston Veterans Affairs Healthcare system, we identified potential cases of ON. These records, including available radiographs and reports, were reviewed. Using published criteria, we evaluated whether the subjects had confirmed ON (radiographs/reports met criteria), incident ON (onset of symptoms within 6 months of first code), or prevalent ON (onset more than 6 months prior to first code or onset could not be determined). We tested different definitions for incident ON using information derived from administrative data. These were compared to the 'gold standard' (record review) and positive predictive values (PPVs) were derived. Since PPVs for incident cases were low, we found the number of incident cases expected for every 1,000 potential cases identified, using the definitions as an initial screening tool to reduce the number of medical records that required examination. We identified 87 potential cases. No case of jaw ON was identified. Only 15 (17%) incident cases of ON were identified. PPVs never exceed 50% for incident ON. However, if we used the definition '(at least 1 inpatient ON code) and (no prior codes for osteoarthritis)' as an initial screen, then for every 1,000 records, we would need to review only 150 to find 69 incident cases. Though the precise PPVs we found may not be generalizable to other databases, we believe that administrative data alone should not be used to identify incident cases of ON without confirming the diagnosis through a review of medical records. By applying the above definition, the number of records requiring review can be markedly reduced. This method can be used to find cases for valid case-control studies of risk factors for ON.
    Full-text · Article · Jul 2009 · Arthritis research & therapy
  • Source
    [Show abstract] [Hide abstract] ABSTRACT: Lamotrigine is an important new addition to the drugs used to treat people with seizure disorders, but disconcerting are reports of a higher than expected incidence of severe skin reaction among children. Using automated data from three HMOs, we conducted a retrospective investigation of children (<15 years) exposed to lamotrigine from 1 January 1995 to 30 June 1997. The outcome of interest was hospitalization for a severe skin reaction (e.g. erythema multiforme). Lamotrigine was dispensed to 124 children (56% female, mean age 8.7 years); the mean number of dispensings per person was 10. Of those exposed, 59 (47%) were hospitalized at least once during the study period, mainly for convulsions and epilepsy. There were no hospitalizations for or with a diagnosis of severe skin reactions. Our investigation revealed no evidence to support a causal relationship between lamotrigine and severe skin reactions. However, because our sample size was small we had power to detect only a very strong association between lamotrigine and severe skin disease. Taken alone, our study does not establish the risks of lamotrigine. These results should be viewed as a contribution to the totality of evidence that will be used to assess the safety of lamotrigine.
    Preview · Article · Nov 1998 · Pharmacoepidemiology and Drug Safety
Show more