Académique Documents
Professionnel Documents
Culture Documents
T
he obesity epidemic is rapidly becoming a public MATERIAL AND METHODS
health concern; more than 30% of Americans
The data and in-kind database development support and guid-
may be classified as obese (with a body mass index
ance were provided by the Blue Cross and Blue Shield (BCBS)
[BMI] ⱖ30).1 Obesity is associated with multiple medical Plans, BCBS of Tennessee, BCBS of Hawaii, BCBS of Michi-
comorbidities, including type 2 diabetes, heart disease, gan, BCBS of North Carolina, Highmark, Inc., of Pennsylvania,
hypertension, pregnancy complications, and sleep apnea. Independence Blue Cross of Pennsylvania, Wellmark BCBS of
Obesity has also been associated with urological conditions Iowa, and Wellmark BCBS of South Dakota. All individuals
such as male and female sexual dysfunction, infertility, in- with one of these 7 plans as their primary insurer were eligible
continence, genitourinary malignancy, and nephrolithia- for inclusion in the dataset. The claims data used in this study
sis.2-5 A similar relationship between obesity and uri- were de-identified in accordance with the Health Insurance
Portability and Accountability Act of 1996 definition of a
nary tract infection (UTI) may exist but it has not
limited dataset, and were used in accordance with federal stan-
been as well defined. Therefore, we performed a study dards for protecting confidentiality of the personal health in-
to characterize the association between obesity and the formation of the enrollee. The Institutional Review Board of
incidence of UTI and pyelonephritis and stratified the the Johns Hopkins University found this analysis to be exempt
results by BMI subcategories. from the requirement for review.
The dataset includes approximately 3.7 million insured lives
over a 5-year period (2002-2006) and includes information on
enrollee age and sex, enrollment dates, and claims for reim-
Funding Support: This publication was made possible by grant number T32DK07552
bursement for billable health care services. Included in these
from the NIH-NIDDK and its contents are solely the responsibility of the authors and
do not necessarily represent the official views of the NIH-NIDDK. data are patient diagnoses as identified by the International
From the Departments of Urology, Surgery, and Medicine, the Johns Hopkins Classification of Diseases, Ninth Revision (ICD-9) codes and
University School of Medicine, and the Departments of Health Policy and Management, diagnosis-related groups (DRGs), and medical procedures clas-
Johns Hopkins School of Public Health, Baltimore, MD sified by current procedural terminology (CPT) codes and
Reprint requests: Brian R. Matlaga, M.D., M.P.H., James Buchanan Brady
Urological Institute, the Johns Hopkins University School of Medicine, 600 North
ICD-9 procedure codes. A subset of patients (n ⫽ 95,598) also
Wolfe Street, Baltimore, MD 21287. E-mail: bmatlaga@jhmi.edu completed a Health Risk Assessment (HRA) that included
Submitted: June 8, 2011, accepted (with revisions): September 28, 2011 their body mass index (BMI); these provided the data used for
this study. A substantial proportion of patients (30.6%) com- Women were 4.2 times more likely to be diagnosed with
pleted HRAs on more than one occasion over the 5-year period. a UTI (19.3% vs 4.6%) and 3.6 times more likely to be
Subjects who did not complete an HRA document were not diagnosed with pyelonephritis (1.22% vs 0.34%) than
included in further analysis. were men. At all stratifications of obesity, the obese were
UTI and pyelonephritis were defined by ICD-9 coding, as
significantly more likely to be diagnosed with a UTI than
described in Table 1. These codes have been previously used by
other large-scale investigations, such as the Urologic Diseases in
the nonobese (Table 2). With regard to pyelonephritis,
America project and Agency for Healthcare Research and the obese were significantly more likely to be diagnosed
Quality metrics.6-8 Specifically, for the purposes of this manu- than the nonobese, for certain BMI ranges, but not for all
script, the terminology “UTI” includes diagnosis codes for both strata (Table 3). These findings were significant, as con-
cystitis as well as UTI, and the terminology “pyelonephritis” is firmed by logistic regression. The confidence intervals for
restricted to codes for acute or chronic pyelonephritis. For those the odds ratios overlap among the obese cohorts, suggest-
who had a diagnosis of or were treated for UTI or pyelonephri- ing that there are no significant differences among the
tis, the BMI closest in time to their diagnosis or treatment was different strata of obesity.
used. Otherwise, the highest BMI reading for a given patient
was used. Age was calculated as the date of the HRA that was
used to obtain the BMI reading. We excluded female subjects COMMENT
with a code indicating pregnancy the year before, the year of, or We found that obesity is a significant risk factor for being
the year after an obese BMI value. diagnosed with UTI as well as pyelonephritis. Within the
For statistical analysis, BMI strata were set at less than obese population, approximately 20% of females and 8%
30, 30.00-34.99, 35.00-39.99, 40.00-44.99, 45.00-49.99, and of males were diagnosed with a UTI. As a group, the
greater ⱖ50. The main dependent variables in this study were obese were up to 2.5 times more likely to be diagnosed
a diagnosis or treatment of UTI or pyelonephritis. Simple with a UTI than were the nonobese. In particular, males
descriptive statistics were obtained and logistic regression was
demonstrated a higher risk than females. With regard to
performed to model the odds of a diagnosis or procedure by age,
gender, and BMI. SAS version 9.13 was used for all analyses
pyelonephritis, the obese were nearly 5 times more likely
(SAS Institute, Cary, NC). to be diagnosed than were the nonobese; females were at
particularly higher risk.
The Urologic Diseases in America (UDA) Project has
RESULTS previously reported that use of outpatient hospital and
Of the 95,598 subjects identified for evaluation, 54,572 physician office care for male subjects with UTI have
(57.1%) were female and 41,026 (42.9%) were male. In increased in the last decade, and health care expenditure
the overall study cohort, the diagnosis of a UTI or pye- for privately insured men with UTI was twice that for
lonephritis occurred in 13% and 0.84%, respectively. men without UTI.6 In addition, women have a reported
⬎50% lifetime risk of UTI.7 The UDA investigators also tionately affected by such policy. Risk factors such as
explored the costly nature of UTI, because overall ex- obesity should be accounted for as these paradigm-shift-
penditures for men with this condition in the United ing reimbursement programs are applied, so that such
States were approximately $1 billion in 2000 and for negative incentives do not affect the way in which health
women were $2.5 billion.7 care is delivered to the obese—a potentially vulnerable
Because the treatment of UTIs poses a significant population. Such perverse incentives that penalize sur-
economic burden on society, both in terms of health care geons for caring for obese patients could disincline pro-
resource usage as well as absolute economic cost, a better viders to serve this population. Metrics of performance
epidemiologic characterization of this condition is impor- should adjust for the influence of obesity in contributing
tant. At present, however, there is a paucity of literature to these types of infections.
defining the relationship between obesity and urinary It is a cross-sectional study and thus only elucidates a
infection. What literature that does exist, though, is statistical association between obesity and urinary infec-
consistent with our findings. Ribera et al evaluated risk tion and one cannot infer a cause and effect relationship.
factors associated with UTI in a series of diabetic patients Because obesity is associated with significant medical
with and without asymptomatic bacteriuria, and found comorbidities, the extremely obese may have an in-
obesity to be a significant risk factor for development of creased use of health care resources and encounters. Such
symptomatic UTI in males.9 Geerlings and associates usage may lead to a detection bias, resulting in an in-
have also demonstrated a relationship between BMI and crease in incidental, asymptomatic urinary infections
asymptomatic bacteriuria although their study did not that do not require treatment. In addition, the database
describe any relationship between obesity and symptom- we examined identifies infection by ICD-9 code so they
atic UTI.10 In 2006, Olumuyiqa and associates reported a are not necessarily culture-proven but rely on physicians
retrospective analysis in which they evaluated postoper- to properly code. Given the nature of an administrative
ative complications in obese vs nonobese patients, and claims database, we also did not have access to certain
found that the obese had a significantly increased risk for clinical data, including culture data, laboratory studies, or
UTI.11 diagnostic circumstances, and the database may contain
Our present work is the first study dedicated to describe erroneous and incomplete coding (ie, failing to list codes
the relationship between obesity and urinary infection. for all diagnoses relevant to a given admission, or
We have also subcategorized lower UTI and upper UTI missing claims). That is to say, the diagnosis of UTI
so these 2 unique conditions may be examined separately. was identified by coding data but was not confirmed
With all data taken together, this study provides a robust with clinical studies or evaluations. Although this is a
description of the prevalence of cystitis and pyelonephri- limitation, these codes have been used previously by
tis among all categories of obesity; to our knowledge, other large-scale investigations such as the UDA proj-
there has been no similar report stratifying the effect of ect and Agency for Healthcare Research and Quality
increasing degrees of obesity, particularly among those metrics.6-8 Finally, the database is from a private in-
with extreme obesity, on urinary infection. surance company and may not be representative of the
Our work presented herein may have implications on general population. Further study to confirm the asso-
future health care policies and highlight the need for ciation in other populations should be done. In addi-
circumspection with large-scale efforts directed at reduc- tion, studies to confirm cause and effect need to be
ing, or even eliminating, reimbursement for the care of carried out to demonstrate that weight loss among the
hospital-acquired UTI. Because we have demonstrated obese will subsequently reduce the likelihood of being
that obesity is a significant risk factor for the diagnosis of diagnosed with UTI and pyelonephritis.
UTI, one might expect that hospitals providing care to a Although these limitations only allow us to draw an
greater proportion of obese subjects may be dispropor- association between obesity and urinary infection and we