COPD: Journal of Chronic Obstructive Pulmonary Disease

ISSN: 1541-2555 (Print) 1541-2563 (Online) Journal homepage: https://www.tandfonline.com/loi/icop20

Hospitalizations for Acute Exacerbations of

Chronic Obstructive Pulmonary Disease: How You
Count Matters

Brian D. Stein, Jeffery T. Charbeneau, Todd A. Lee, Glen T. Schumock, Peter K.

Lindenauer, Adriana Bautista, Diane S. Lauderdale, Edward T. Naureckas &
Jerry A. Krishnan

To cite this article: Brian D. Stein, Jeffery T. Charbeneau, Todd A. Lee, Glen T. Schumock,
Peter K. Lindenauer, Adriana Bautista, Diane S. Lauderdale, Edward T. Naureckas & Jerry A.
Krishnan (2010) Hospitalizations for Acute Exacerbations of Chronic Obstructive Pulmonary
Disease: How You Count Matters, COPD: Journal of Chronic Obstructive Pulmonary Disease,
7:3, 164-171, DOI: 10.3109/15412555.2010.481696

To link to this article: https://doi.org/10.3109/15412555.2010.481696

Published online: 20 May 2010.

Submit your article to this journal

Article views: 486

View related articles

Citing articles: 7 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=icop20
COPD: Journal of Chronic Obstructive Pulmonary Disease, 7:164–171
ISSN: 1541-2555 print / 1541-2563 online
Copyright

c 2010 Informa Healthcare USA, Inc.
DOI: 10.3109/15412555.2010.481696

ORIGINAL RESEARCH

Hospitalizations for Acute Exacerbations of Chronic

Obstructive Pulmonary Disease: How You Count
Matters
Brian D. Stein1 (brian stein@rush.edu), Jeffery T. Charbeneau2 (jcharbeneau@health.bsd.uchicago.edu), Todd A. Lee3,4
(todd.lee@va.gov), Glen T. Schumock3 (schumock@uic.edu), Peter K. Lindenauer5 (peter.lindenauer@baystatehealth.org),
Adriana Bautista3 (adrianitabautista@hotmail.com), Diane S. Lauderdale2 (dlauderd@uchicago.edu), Edward T.
Naureckas6 (tnaureka@medicine.bsd.uchicago.edu), and Jerry A. Krishnan2,6 (jkrishna@medicine.bsd.uchicago.edu)
1 Rush University Medical Center, Section of Pulmonary and Critical Care Medicine, Chicago, Illinois, USA
2 University of Chicago, Department of Health Studies, Chicago, Illinois, USA
3 University of Illinois at Chicago, Center for Pharmacoeconomics Research, Chicago, Illinois, USA
4 Center for Management of Complex Chronic Care, Hines VA Hospital, Hines, Illinois, USA
5 Center for Quality of Care Research, Baystate Medical Center, Springfield, Massachusetts, USA
6 University of Chicago, Section of Pulmonary & Critical Care Medicine, Chicago, Illinois, USA

ABSTRACT
ICD-9-CM diagnosis codes are increasingly used to estimate the burden of disease, as well
as to evaluate the quality of care and outcomes of various conditions. Acute exacerbations of
COPD (AE-COPD) are common and associated with substantial health and financial burden in
the U.S. Whether published algorithms that employ different combinations of ICD-9-CM codes to
identify patients hospitalized for AE-COPD yield similar or different estimates of disease burden
is unclear. In this study, the Nationwide Inpatient Sample from years 2000–2006 was used to
identify and compare the number of hospitalizations, healthcare utilization, and outcomes for
patients hospitalized for AE-COPD in the U.S. AE-COPD was identified using five different
published ICD-9-CM algorithms. Estimates of the annual number of hospitalizations for AE-
COPD in the U.S. varied more than 2-fold (e.g., 421,000 to 870,000 in 2006). Outcomes and
healthcare utilization of patients hospitalized for AE-COPD varied substantially, depending
on the algorithm used (e.g., in-hospital mortality 2.0% to 5.1%, total hospital days 2.0 to 5.1
million in 2006). Observed trends in the number of hospitalizations over the 7-year period
varied depending on which algorithm was used. In conclusion, the estimated health burden
and trends in hospitalizations for AE-COPD in the United States differ, depending on which
ICD-9-CM algorithm is used. To improve our understanding of the burden of AE-COPD and to
ensure that quality of care initiatives are not misdirected, a validated approach to identifying
patients hospitalized for AE-COPD is needed.

ABBREVIATIONS and Medicare Services; COPD: Chronic Obstructive Pulmonary

Disease; HEDIS: Health Effectiveness Data and Information
AE-COPD: Acute exacerbation of chronic obstructive pul- Set; ICD-9-CM: International Classification of Diseases, Ninth
monary disease; AHRQ: Agency for Healthcare Research and Revision, Clinical Modification; IHD: Ischemic Heard Disease;
Quality; CI: Confidence interval; CMS: Centers for Medicaid NHLBI: National Heart, Lung, and Blood Institute; NIS: Na-
tionwide Inpatient Sample; PVD: Pulmonary Vascular Disease;
RF: Respiratory Failure;
Correspondence to:
Jerry A. Krishnan, MD, PhD
The University of Chicago INTRODUCTION
5841 S. Maryland Ave, MC 6076
Chicago, IL 60637 USA Health services research and epidemiologic studies fre-
Phone: 773-702-6790 quently rely on administrative data to identify sufficiently large
email: jkrishna@medicine.bsd.uchicago.edu and representative populations. Administrative data include

164 June 2010 COPD: Journal of Chronic Obstructive Pulmonary Disease

diagnoses at healthcare encounters in the form of International
Classification of Diseases, Ninth Revision, Clinical Modifica-
tion (ICD-9-CM) codes. Although originally developed for epi-
demiologic purposes (e.g., to estimate and track changes in dis-
ease burden), ICD-9-CM codes have become critical for billing
and for health services research (1). ICD-9-CM codes also pro-
vide the basis for interventions to improve healthcare delivery,
a topic of substantial interest in current efforts to reform U.S.
healthcare (2). For example, ICD-9-CM codes have been used to
identify patients eligible for quality improvement initiatives, in-
cluding the Centers for Medicaid and Medicare (CMS) Report-
ing Hospital Quality Data for Annual Payment Update program
(3).
The validity of ICD-9-CM codes for identifying patients with
specific medical conditions, however, appears to vary depending
on both the condition examined and the ICD-9-CM algorithm
used (4–6). Chronic obstructive pulmonary disease (COPD) is
a leading cause of morbidity, mortality, and healthcare expen-
ditures and affects an estimated 24 million people in the United
States alone (7, 8). It is currently the fourth-leading cause of
death in the United States and is predicted to become the third
leading cause of death by 2020 (9). Hospitalizations for acute
exacerbations of COPD (AE-COPD) are common, associated
with substantial morbidity and mortality, and costly; therefore,
there is considerable interest in understanding and improving
the care and outcomes of these patients (10–12).
Despite the enormous health and financial burden of hospi-
talizations for AE-COPD, there is no consensus regarding which
ICD-9-CM code or group of codes is most appropriate to iden-
tify such hospitalizations. Some algorithms are based on a single
ICD-9-CM code for COPD exacerbation (12), whereas others
(7, 8, 10, 13–15) employ multiple codes for COPD (regardless
of whether such codes are specific to an exacerbation) and/or
include codes for respiratory failure combined with codes for
COPD (11, 16–18); the various algorithms also include different
age cut-offs to identify patients with COPD.
The effects of these different approaches on national esti-
mates of healthcare utilization and outcomes attributed to hos-
pitalizations for AE-COPD (e.g., number of hospitalizations,
hospital days, use of mechanical ventilation, mortality) are un-
clear. A better understanding about the variability in these esti-
mates arising from the use of different algorithms is critical in
planning and interpreting health services research and epidemi-
ologic studies about AE-COPD, as well as quality improvement
initiatives in this population.
We therefore conducted a study to compare estimates of
the number and outcomes of hospitalizations for AE-COPD
when different published algorithms are applied to a nation-
ally representative sample of all hospitalizations in the United
States. We also examined how our understanding regarding
trends in the health and financial burden of hospitalizations
for AE-COPD would vary depending on which algorithm was
employed. We hypothesized that algorithms based on different
ICD-9-CM codes would identify not only different numbers of
hospitalizations for AE-COPD but also patients with differing
characteristics and outcomes.
COPD: Journal of Chronic Obstructive Pulmonary Disease
MATERIALS AND METHODS
Data source
We used the Agency for Healthcare Research and Quality
(AHRQ)–sponsored Nationwide Inpatient Sample (NIS) be-
cause it is the largest all-payer source of data on hospitalized
patients in the United States. The NIS contains data on all ad-
missions from a 20% stratified probability sample of all acute
care non-governmental hospitals, providing the opportunity to
identify, track, and analyze national trends in hospitalizations
(12, 19, 20). For the years 2000 to 2006, approximately 1,000
hospitals located in 28 to 38 states contributed a total of nearly 8
million records per year to this database. This study was exempt
from institutional review board approval.
Study design
The number of hospitalizations for AE-COPD was estimated
based on 5 previously published (7, 10, 12, 14, 17) ICD-9-CM
algorithms (Table 1, Table 2). Our analysis was limited to these
5 algorithms for simplicity, as they represent the spectrum of
potential ICD-9-CM algorithms used to identify AE-COPD. For
ease of interpretation we refer to the algorithms by number and
an author or organization who has published using the specified
algorithm. The first algorithm, (algorithm 1 [Patil] (12)) selects
hospitalizations with primary diagnosis of COPD (defined as a
single ICD-9-CM code for COPD exacerbation: 491.21 obstruc-
tive chronic bronchitis with acute exacerbation) and represents
one extreme approach (use of a ICD-9-CM code for acute ex-
acerbation).
Algorithm 2, perhaps the most commonly used, is recom-
mended by the National Committee for Quality Assurance
Health Effectiveness Data and Information Set (HEDIS) to
measure the quality of care in patients hospitalized for AE-
COPD (14). It selects hospitalizations with a primary diagnosis
of COPD (defined as one of several ICD-9-CM codes, even
those that are not specific for an exacerbation). Algorithm 3,
used by Holguin et al. (10), has been employed by health ser-
vices researchers to identify patients hospitalized for AE-COPD
and includes the codes recommended by HEDIS to identify pa-
tients with a primary diagnosis of COPD, as well as a primary
diagnosis of ICD-9-CM code 490 (bronchitis, not specified as
acute or chronic). Algorithm 4 is used by the National Heart,
Lung, and Blood Institute (NHLBI) for official estimates of dis-
ease burden, morbidity and mortality for patients with COPD
(7).
The NHLBI algorithm includes all the codes in algorithm 3
plus codes for bronchiectasis and allergic alveolitis to establish
a primary diagnosis of COPD. Algorithm 5, used by Lee et al.
(17), is representative of algorithms that include hospitalizations
with a primary diagnosis of COPD (even codes that are not
specific for an exacerbation) or a secondary diagnosis of COPD
when combined with a primary diagnosis of respiratory failure
(ICD-9-CM codes 518.81, 518.82 and 518.84). Algorithms 1,
2, and 5 use a lower age limit of 40 years; however, algorithm 3
June 2010 165
 Table 1. ICD-9-CM codes used in algorithms to identify uses a lower age limit of 25 years and algorithm 4 has no lower
AE-COPD age limit.
Among patients hospitalized for AE-COPD, characteristics
ICD-9-CM examined were age, sex, and comorbidity. We did not exam-
code Description
ine race because race data were unavailable in 24 to 29% of
490 Bronchitis, not specified as acute or chronic records in the NIS data sets from 2000 to 2006. Variables for
491.0 Simple chronic bronchitis
age and sex were missing in less than 1% of all hospitalizations
491.1 Mucopurulent chronic bronchitis
491.20 Obstructive chronic bronchitis without exacerbation for AE-COPD. We calculated a weighted Charlson Comorbid-
491.21 Obstructive chronic bronchitis with acute ity Index, a validated measure for use with administrative data
exacerbation that correlates with in-hospital morbidity and mortality (21,
491.22 Obstructive chronic bronchitis with acute bronchitis 22). Additionally, we examined the prevalence of specific co-
491.8 Other chronic bronchitis
morbid conditions that were associated with an increased risk
491.9 Unspecified chronic bronchitis
492.0 Emphysematous bleb of in-hospital mortality (10) in a previous study of hospitalized
492.8 Other emphysema patients with COPD (hypertension, heart failure, pneumonia,
493.22 Chronic obstructive asthma with acute ischemic heart disease, diabetes, pulmonary vascular disease,
exacerbation and thoracic malignancies); the Clinical Classifications Soft-
494.0 Bronchiectasis without acute exacerbation
ware (CCS) developed by AHRQ was used for the identification
494.1 Bronchiectasis with acute exacerbation
495.0 Farmers’ lung of comorbid conditions (23).
495.1 Bagassosis Outcomes and measures of healthcare utilization included
495.2 Bird-fanciers’ lung median length of stay (number of days from hospital admis-
495.3 Suberosis sion to discharge) and total hospital days (sum of hospital days
495.4 Malt workers’ lung
for all patients for that calendar year), median and total hospital
495.5 Mushroom workers’ lung
495.6 Maple bark-strippers’ lung charges (sum of hospital charges for all patients for that calendar
495.7 Ventilation” pneumonitis year), the percentage of patients who received mechanical venti-
495.8 Other specified allergic alveolitis and pneumonitis lation during hospitalization (ICD-9-CM procedure codes 96.04
495.9 Unspecified allergic alveolitis and pneumonitis insertion of endotracheal tube, 93.90 continuous positive airway
496 Chronic airway obstruction, not elsewhere
pressure, or 96.7x continuous mechanical ventilation) (12, 20),
classified
and the percentage of patients who died prior to discharge (in-
518.81 Acute respiratory failure hospital mortality). National estimates of these outcomes and
518.82 Other pulmonary insufficiency not elsewhere measures were developed for years 2000 to 2006 using each
classified algorithm.
518.84 Acute and chronic respiratory failure
Analysis
The table above includes all ICD-9-CM codes that have been used
in five published algorithms employed to identify patients Hospital discharge weights included in the NIS were used
hospitalized for acute exacerbations of COPD (AE-COPD). to develop national estimates for the number of hospitalizations
Abbreviations: ICD-9-CM, International Classification of Disease,
Ninth Edition, Clinical Modification.
for AE-COPD as well as patient characteristics and outcomes
(24). Hospital charges for years 2000 to 2005 were converted to
2006 U.S. dollars using the Consumer Price Index for medical

Table 2. Different ICD-9-CM algorithms used to identify hospitalizations for AE-COPD

Algorithm Age Primary ICD-9-CM Diagnosis Secondary ICD-9-CM Diagnosis∗

1 [Patil] (12) ≥40 yrs 491.21 None
2 [HEDIS] (14) ≥40 yrs 491.x, 492.x, or 496 None
3 [Holguin] (10) ≥25 yrs 490, 491.x,492.x, or 496 None
4 [NHLBI] (7) All yrs 490, 491.x, 492.x, 494.x, 495.x, or 496 None
5 [Lee] (17) ≥40 yrs 491.0, 491.1, 491.21, 491.22, 491.8, 491.9, None
492.0, 492.8, 493.22, or 496∗∗
OR
518.81, 518.82 or 518.84 AND 491.0, 491.1, 491.21, 491.22, 491.8, 491.9, 492.0,
492.8, 493.22, or 496∗∗

Abbreviations: ICD-9-CM, International Classification of Disease, Ninth Edition, Clinical Modification; AE-COPD, acute exacerbation of chronic
obstructive pulmonary disease.
∗ Secondary ICD-9-CM Diagnosis may be in any position after the first.
∗∗ Excluded if 491.20 (obstructive chronic bronchitis without exacerbation) in any diagnostic field.

ICD-9.x represents all possible ICD-9-CM sub-codes following the initial 3 digit ICD-9 code.

166 June 2010 COPD: Journal of Chronic Obstructive Pulmonary Disease

care (25). Trends in the estimated number of hospitalizations for
AE-COPD, total hospital days, and total hospital charges across
all years were evaluated for each algorithm using weighted least
squares regression using the inverse variance of the estimate as
the analytic weight (26). Trends in mechanical ventilation use,
in-hospital mortality, and comorbidity were evaluated with lo-
gistic regression using year as the independent variable. Formal
statistical tests comparing these parameters across algorithms
were not performed as there is partial overlap in the hospital-
ization identified by the various algorithms (i.e., more than one
algorithm may identify the same hospitalization).
Because algorithm 5 [Lee] is comprised of both patients
identified by primary ICD-9-CM codes for COPD as well as
patients identified by primary ICD-9-CM codes for respiratory
failure (when combined with a secondary diagnosis of COPD),
we also reported the prevalence of selected co-morbid conditions
in each of these 2 subsets of patients in algorithm 5 to explain
possible differences between this algorithm and algorithms 1-4
(which all relied on a primary diagnosis of COPD). All reported
p-values are two sided, and p-values of <0.05 were considered
statistically significant. Analyses were performed using STATA
software package, release 10.0 (Stata Corp Inc., College Station,
Texas).
RESULTS
Hospitalizations for AE-COPD
For the years 2000 to 2006, there were a total of 36.4 to 39.5
million hospitalizations per year (19). The estimated number of
hospitalizations for AE-COPD in 2006 varied more than 2-fold
(421,000 [algorithm 1; 1.1% of hospitalizations] to 870,000 [al-
gorithm 5; 2.2% of hospitalizations]; Table 3). Algorithms 2-4
produced very similar estimates of the number of hospitaliza-
tions for COPD in 2006. In 2006, the addition of the code 490
(bronchitis, not specified as acute or chronic) added approxi-
mately 23,000 additional hospitalizations in both algorithms 3
[Holguin] and 4 [NHLBI] compared to algorithm 2 [HEDIS].
Codes for allergic alveolitis and bronchiectasis in algorithm 4
[NHLBI] added another 12,000 hospitalizations compared to
algorithm 2 [HEDIS]. Including patients with primary diagno-
sis of respiratory failure with a secondary diagnosis of COPD
added 201,000 hospitalizations to algorithm 5 compared to al-
gorithm 2; including patients with a code for chronic obstructive
asthma (493.22) added 118,000 hospitalizations to algorithm 5
compared to algorithm 2.
The annual number of hospitalizations for AE-COPD ap-
peared to decrease from 2000 to 2006 when using algorithm 1,

Table 3. Hospitalizations, patient characteristics, and outcomes in 2006

Algorithm 2 Algorithm 3 Algorithm 4

Algorithm 1 [Patil] [HEDIS] [Holguin] [NHLBI] Algorithm 5 [Lee]
Number (99% CI) Patient 421,000 (395,000 557,000 582,000 (547,000 599,000 (563,000 870,000
Characteristics – 447,000) (524,000–591,000) – 617,000) – 634,000) (817,000–922,000)
Age, yrs Mean (99% CI) 69.4 (69.1, 69.7) 69.5 (69.2, 69.8) 69.1 (68.8, 69.4) 68.5 (68.1, 68.9) 68.8 (68.5,69.1)
< 40 years,% 0.0 0.0 1.1 2.2 0.0
Female,% (99% CI) 55 (55, 56) 56 (55,56) 56 (55, 56) 56 (55, 57) 57 (57, 58)
Charlson Index,%
1 42 43 43 44 39
2 30 30 30 30 30
3 16 15 15 15 16
>4 12 12 12 12 14
Outcomes
Hospital days,
Median per hospitalization 4 (2,6) 4 (2,6) 4 (2,6) 4 (2,6) 4 (3,7)
(IQR)
Total for all hospitalizations x 1 2.0 (1.9,2.1) 2.7 (2.5, 2.8) 2.7 (2.6, 2.9) 2.9 (2.7, 3.0) 5.1 (4.8, 5.4)
million (99% CI)
Hospital charges, $
Median per hospitalization x 12.6 (7.6,21.7) 12.7 (7.7,21.8) 12.5 (7.6,21.5) 12.6 (7.6,21.8) 15.2 (8.7, 29.2)
1,000 (IQR)
Total for all Hospitalizations x 1 8.0 (7.4, 8.7) 10.7 (9.8,11.6) 11.1 (10.1, 12.0) 11.5 (10.6, 12.5) 23.8 (21.8,25.8)
billion (99% CI)
Mechanical ventilation,% (99% 5.2 (4.7, 5.7) 4.7 (4.3, 5.2) 4.6 (4.2, 5.0) 4.6 (4.2, 5.0) 16.5 (15.8, 17.3)
CI)
In-hospital mortality,% (99% CI) 2.1 (1.9,2.2) 2.1 (2.0, 2.3) 2.0 (1.9,2.2) 2.0 (1.9,2.2) 5.1 (4.8, 5.3)

Abbreviations: CI, confidence interval; IQR, Inter-quartile range (interval of the 25th and 75th percentile); yrs, years.
Percentages may not sum to 100% due to rounding.
Missing values less than 1% for all variables.
Total hospital days is the sum of all hospital days for each algorithm in 2006.
Total hospital charges is the sum of all hospital charges for each algorithm in 2006.

COPD: Journal of Chronic Obstructive Pulmonary Disease June 2010 167


60
40
Percent
20
0
Systemic Heart Failure
Hypertension
Figure 1. Prevalence of selected comorbidities∗ across ICD-9-CM algorithms for AE-COPD.
to remain unchanged when using algorithms 2, 3 or 4, and to
increase when using algorithm 5 (Figure 2A).
Patient characteristics
Age varied little between algorithms 1-5 in 2006 despite
differences in age inclusion criteria (Table 3); very few hospi-
talizations included subjects <40 years in the various algorithms
(maximum 2.2% in algorithm 4). Slightly more than half of all
hospitalizations for COPD occurred in women, regardless of
algorithm. The Charlson comorbidity index was similar across
algorithms, with the majority of patients identified having at
least 1 additional clinically significant comorbid condition.
The prevalence of co-morbid conditions was remarkably con-
sistent across the algorithms, except algorithm 5, which had a
higher prevalence than others (Figure 1). The patients with a pri-
mary diagnosis of respiratory failure (and secondary diagnosis
Figure 2. Trends in the number of hospitalizations, total hospital days, and total charges for AE-COPD, 2000-2006
168 June 2010 COPD: Journal of Chronic Obstructive Pulmonary Disease
Selected Comorbidities
1 [Patil]
2 [HEDIS]
3 [Holguin]
4 [NHLBI]
5 [Lee]
5 [Lee, 1o COPD code]
5 [Lee, 1o RF code]
Pneumonia IHD Diabetes PVD
of COPD) in algorithm 5 [Lee] largely accounted for differences
between algorithm 5 and other algorithms.
Hospital days
Median length of stay for all algorithms was 4 days. In con-
trast, total hospital days for COPD in 2006 varied by more than
2-fold, depending on which algorithm was used (least using
algorithm 1 [Patil], 2.0 million days; most using algorithm 5
[Lee], 5.1 million days). Trends in total hospital days for AE-
COPD varied in both direction and magnitude from 2000 to
2006. There was a significant increase in the total number of
hospital days when algorithm 5 [Lee] was applied as compared
to a significant decrease in the total number of hospital days
when algorithms 1-3 were applied (Figure 2B). No significant
trend in the total number of hospital days for AE-COPD was
evident using algorithm 4 [NHLBI].
 Hospital charges
Hospital charges differed both between algorithms and across
time. Median hospital charges in 2006 varied by as much as
$2,700 (Table 3). The total yearly hospital charges were nearly
three-fold higher when algorithm 5 [Lee] was compared to al-
gorithm 1 ([Patil], $23.8 billion vs. $8.0 billion). The patients
with a primary diagnosis of respiratory failure in algorithm 5
[Lee] had higher hospital charges than patients identified by a
primary diagnosis of COPD (median charge per hospitalization:
$31,900 vs. 12,800, respectively).
Mechanical ventilation
The use of mechanical ventilation varied by more than three-
fold in 2006 (4.6% using algorithm 4 [NHLBI] vs. 16.5% using
algorithm 5 [Lee], Table 3). The percentage of patients ad-
mitted for AE-COPD who received mechanical ventilation did
not change significantly from 2000 to 2006 using algorithms
2 [HEDIS], 3 [Holguin], or 4 [NHLBI]. However, mechanical
ventilation use increased significantly when algorithms 1 [Patil]
and 5 [Lee] were applied (Figure 3A). The patients with a pri-
mary diagnosis of respiratory failure in algorithm 5 [Lee] had
higher rates of mechanical ventilation than patients identified by
a primary diagnosis of COPD (55.7% vs. 4.7%, respectively).
In-hospital mortality
In 2006, in-hospital mortality varied by more than 2-fold de-
pending on which algorithm was applied (2.0 % using algorithm
3 [Holguin] vs. 5.1% using algorithm 5 [Lee], Table 3). In those
identified by algorithm 5, mortality was 8 times greater in the
subset of patients with a primary diagnosis of respiratory failure
and a secondary diagnosis of COPD (16%) compared to those
with a primary diagnosis of COPD (2%), even though the same
codes for COPD were employed. In-hospital mortality appeared
to decrease for all algorithms across years 2000 to 2006 (Figure
3B).
Figure 3. Trends in mechanical ventilation use and in-hospital mortality for AE-COPD, 2000-2006
COPD: Journal of Chronic Obstructive Pulmonary Disease
DISCUSSION
We have demonstrated that estimates of the number of hospi-
talizations for AE-COPD in the United States vary more than 2-
fold, depending on which ICD-9-CM algorithm is used. We have
also demonstrated that annual estimates of outcomes and mea-
sures of healthcare utilization differ substantially, as do trends
in these estimates from 2000 to 2006, depending on which ICD-
9-CM algorithm is applied. To our knowledge, our study is
the first to quantify these discrepancies across commonly used
algorithms and to highlight the need for validation studies to
identify the optimal search strategy for hospitalizations due to
AE-COPD. Without such information, efforts to measure and
improve the quality of COPD care in hospitalized populations
may be misdirected.
Although a number of previous studies have used ICD-9-
CM codes to identify hospitalizations for AE-COPD, we are
not aware of studies that have directly compared the various ap-
proaches. We expected to see the number of hospitalizations for
AE-COPD increase in algorithms that include a larger number
of ICD-9-CM codes (e.g., algorithm 5 vs. algorithm 1). There
was, however, more overlap than we had expected in algorithms
2-4 due to the small numbers of additional hospitalizations iden-
tified by algorithms 3 or 4 compared to algorithm 2. Importantly,
our data indicate that different algorithms lead to very differ-
ent conclusions about trends in the health and economic burden
of AE-COPD (e.g. total number of hospital days, total hospital
charges, and in-hospital mortality). These differences are likely
to be attributable to differences in the characteristics of patients
captured by each algorithm.
Official U.S. estimates of the burden of AE-COPD are based
on the NHLBI algorithm (algorithm 4), which consists of a
primary diagnosis of chronic bronchitis, emphysema, chronic
airflow obstruction, or a number of other clinical conditions
(i.e.., bronchiectasis and disorders linked to allergic alveolitis).
Inclusion of patients hospitalized with a primary diagnosis of
bronchiectasis or allergic alveolitis will lead to a small overes-
timate of the health burden from COPD, since current clinical
June 2010 169
definitions of COPD do not include these disorders (27, 28).
The NHLBI algorithm (as well as algorithms 1–3) may also sub-
stantially underestimate the burden of AE-COPD, since more
severe exacerbations that lead to respiratory failure are likely
to be missed (i.e., patients with a primary diagnosis of respi-
ratory failure are excluded in these algorithms). These more
severe exacerbations are more likely to be captured in algorithm
5 [Lee].
We found that patients with a primary diagnosis of respira-
tory failure and a secondary diagnosis of COPD (identified by
algorithm 5) have a substantially higher prevalence of comor-
bid conditions linked with respiratory symptoms that can mimic
AE-COPD (e.g., heart failure, pneumonia) than patients with
a primary diagnosis of COPD (e.g., algorithms 1–4). A higher
prevalence of these comorbid conditions raises the possibility
that some patients identified by algorithm 5 were in fact admit-
ted for a condition other than AE-COPD (i.e., hospitalized for
respiratory failure due to heart failure, rather than AE-COPD).
Alternatively, algorithm 5 may have helped to identify the
subset of patients with more severe exacerbations (leading to
respiratory failure). Thus, while it is possible that the specificity
of algorithm 5 in hospitalized patients for identifying AE-COPD
may be lower than algorithms 1-4, the sensitivity of algorithm 5
is likely to be higher than other algorithms. Additional research
is needed to confirm this possibility.
Our study has two notable strengths. First, the data were de-
rived from the largest nationally representative sample of hos-
pital admissions; therefore our findings are likely to be broadly
applicable to patients in the United States. Second, we directly
examined the effects of applying five previously used algorithms
that ranged from simple to complex for hospitalizations due to
AE-COPD.
This study, however, also has some important limitations.
First, the assignment of ICD-9-CM discharge diagnoses for
hospitalized patients is done specifically for reimbursement pur-
poses. The assignment of these codes may be influenced by both
the clinical evidence in the medical record and by the relative
rate of reimbursement anticipated for each diagnosis (29, 30).
Because our analyses were restricted to the data available in
the NIS, we were unable to assess the validity of the individual
ICD-9-CM coding algorithms. As we were unaware of prior
validation studies of the algorithms to identify hospitalizations
for AE-COPD, we undertook a pilot validation study in two ur-
ban teaching hospitals, comparing different algorithms to chart
abstracted physician diagnosis (31). We compared the primary
physician diagnosis (as recorded in the physician notes) and
3 different ICD-9-CM algorithms for AE-COPD (algorithms 1
[Patil]; 2 [HEDIS]; and 5 [Lee]) in a sample of 200 hospitaliza-
tions.
The positive predictive values were 74% (algorithm 1), 62%
(algorithm 2), and 60% (algorithm 5), respectively, suggesting
that current algorithms for identifying AE-COPD may identify
a substantial number of patients admitted for alternative condi-
tions. Based on these data, we estimate that about 109,000 (26%
of 421,000 hospitalizations identified by algorithm 1), 212,000
(38% of 557,000 hospitalizations identified by algorithm 2),
170 June 2010 COPD: Journal of Chronic Obstructive Pulmonary Disease
and 348,000 (40% of 870,000 hospitalizations identified by al-
gorithm 5) hospital admissions were misclassified as AE-COPD
in 2006.
These findings suggest that the selection of an algorithm
should depend on its intended purpose. If, for example, the intent
is to identify patients for quality measurement, an algorithm
with the lowest false positive rate would be desirable. In this
instance, a more specific algorithm (e.g., one based on code
specifically for AE-COPD) may be preferrred (e.g., algorithm
1). By contrast, if the intent is to estimate the overall burden of
disease, then a more inclusive approach may be needed (e.g.,
algorithm 3).
Larger, multi-center validation studies are necessary, how-
ever, to develop more generalizable estimates regarding the
positive predictive values of these algorithms (1, 2, and 5) and
others. Second, outcomes assessed in the current study are lim-
ited to those occurring in-hospital. The NIS data are limited
to in-hospital mortality, and therefore we cannot discount the
effect of possible changes in discharge practices over time and
their influence on in-hospital outcomes if they varied across
patients identified by the different algorithms (e.g., increasing
use of home health care or home hospice). Last, our findings
are based on data in acute care non-governmental hospitals and
may not be generalizable to other healthcare settings.
In summary, the estimated health burden and trends in hos-
pitalizations for AE-COPD in the U.S. differ substantially de-
pending on which ICD-9-CM algorithm is used. To ensure an
accurate estimate of the burden of AE-COPD in the United
States and to ensure that quality of care initiatives target the
appropriate patient population, we recommend studies be con-
ducted to identify algorithms with the most suitable feasible
level of sensitivity and specificity.
ACKNOWLEDGMENTS
We would like to thank Dr. Fernando Holguin, Division of
Pulmonary, Allergy, and Critical Care Medicine, University of
Pittsburgh, for his input on a prior version of this manuscript.
This work was supported by the National Institutes of Health
[Grant HL07605, B. Stein] and AHRQ [Grant 5U18HS016967-
02, J. Krishnan, G. Schumock, T. Lee].
Declaration of interest
The authors report no conflicts of interest. The authors alone
are responsible for the content and writing of the paper.
REFERENCES
1. National Center for Health Statistics. International Classifica-
tion of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM),
2009. Retrieved September 22, 2009, from http://www.cdc.gov/
nchs/icd.htm.
2. Institute of Medicine. Initial national priorities for comparative
effectiveness research. Washington, DC: Institute of Medicine,
2009.
 3. Centers for Medicare and Medicaid Services. Reporting Hos-
pital Quality Data for Annual Payment Update. Retrieved April 2,
2009, from http://www.cms.hhs.gov/HospitalQualityInits.
4. Aronsky D, Haug PJ, Lagor C, Dean NC. Accuracy of administra-
tive data for identifying patients with pneumonia. Am J Med Qual
Nov–Dec 2005; 20:319–328.
5. Fisher ES, Whaley FS, Krushat WM, Malenka DJ, Fleming C,
Baron JA, Hsia DC. The accuracy of Medicare’s hospital claims
data: progress has been made, but problems remain. Am J Public
Health Feb 1992; 82:243–248.
6. Ginde AA, Tsai CL, Blanc PG, Camargo CA, Jr. Positive pre-
dictive value of ICD-9-CM codes to detect acute exacerbation of
COPD in the emergency department. Jt Comm J Qual Patient Saf
Nov 2008; 34:678–680.
7. National Heart Lung and Blood Institute. 2007 NHLBI Morbid-
ity and Mortality Chart Book. Retrieved December 1, 2008, from
http://www.nhlbi.nih.gov/resources/docs/cht-book.htm.
8. Mannino DM, Homa DM, Akinbami LJ, Ford ES, Redd
SC. Chronic obstructive pulmonary disease surveillance—United
States, 1971–2000. MMWR Surveill Summ. Aug 2 2002; 51:1–16.
9. Jemal A, Ward E, Hao Y, Thun M. Trends in the leading causes
of death in the United States, 1970-2002. JAMA Sep 14 2005;
294:1255–1259.
10. Holguin F, Folch E, Redd SC, Mannino DM. Comorbidity and
mortality in COPD-related hospitalizations in the United States,
1979 to 2001. Chest Oct 2005; 128:2005–2011.
11. Lindenauer PK, Pekow P, Gao S, Crawford AS, Gutierrez B,
Benjamin EM. Quality of care for patients hospitalized for acute
exacerbations of chronic obstructive pulmonary disease. Ann In-
tern Med Jun 20 2006; 144:894–903.
12. Patil SP, Krishnan JA, Lechtzin N, Diette GB. In-hospital mor-
tality following acute exacerbations of chronic obstructive pul-
monary disease. Arch Intern Med May 26 2003; 163:1180–
1186.
13. McGhan R, Radcliff T, Fish R, Sutherland ER, Welsh C, Make
B. Predictors of rehospitalization and death after a severe exacer-
bation of COPD. Chest Dec 2007; 132:1748–1755.
14. National Committee on Quality Assurance. HEDIS 2009.
Retrieved April 2, 2009, from http://www.ncqa.org/tabid/
59/Default.aspx.
15. Fan VS, Bryson CL, Curtis JR, Fihn SD, Bridevaux PO, Mc-
Donell MB, Au DH. Inhaled corticosteroids in chronic obstruc-
tive pulmonary disease and risk of death and hospitalization:
time-dependent analysis. Am J Respir Crit Care Med Dec 15
2003;168:1488–1494.
16. Lee TA, Wilke C, Joo M, Stroupe KT, Krishnan JA, Schumock
GT, Pickard AS. Outcomes associated with tiotropium use in pa-
tients with chronic obstructive pulmonary disease. Arch Intern Med
Aug 10 2009; 169:1403–1410.
17. Lee TA, Schumock GT, Bartle B, Pickard AS. Mortality
risk in patients receiving drug regimens with theophylline for
chronic obstructive pulmonary disease. Pharmacotherapy Sep
2009;29:1039–1053.
18. Cannon KT, Sarrazin MV, Rosenthal GE, Curtis AE, Thomas
KW, Kaldjian LC. Use of mechanical and noninvasive ventilation
in black and white chronic obstructive pulmonary disease patients
COPD: Journal of Chronic Obstructive Pulmonary Disease
within the Veterans Administration health care system. Med Care
Jan 2009;47:129–133.
19. HCUP Nationwide Inpatient Sample (NIS). Healthcare Cost and
Utilization Project (HCUP) 2000–2006. Agency for Healthcare Re-
search and Quality, Rockville, MD. Retrieved December 1, 2008,
from www.hcup-us.ahrq.gov/nisoverview.jsp.
20. Krishnan V, Diette GB, Rand CS, Bilderback AL, Merriman B,
Hansel NN, Krishnan JA. Mortality in patients hospitalized for
asthma exacerbations in the United States. Am J Respir Crit Care
Med Sep 15 2006; 174:633–638.
21. Li B, Evans D, Faris P, Dean S, Quan H. Risk adjustment per-
formance of Charlson and Elixhauser comorbidities in ICD-9 and
ICD-10 administrative databases. BMC Health Serv Res 2008;
8:12.
22. Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi
JC, Saunders LD, Beck CA, Feasby TE, Ghali WA. Coding algo-
rithms for defining comorbidities in ICD-9-CM and ICD-10 admin-
istrative data. Med Care Nov 2005; 43:1130–1139.
23. Elixhauser, A., Steiner, C., Palmer, L. Clinical Classifi-
cations Software (CCS), 2004. February 6, 2004. U.S.
Agency for Healthcare Research and Quality. Retrieved
June 24, 2009, from http://www.hcup-us.ahrq.gov/reports/
2004CCSUserGuideFinal.pdf.
24. Houchens R, Elixhauser A. Final Report on Calculating Na-
tionwide Inpatient Sample (NIS) Variances, 2001. HCUP Method
Series Report # 2003-02. ONLINE June 2005 (revised June 6,
2005). U.S. Agency for Healthcare Research and Quality. Re-
trieved December 28, 2008, from http://www.hcup-us.ahrq.gov//
reports/CalculatingNISVariances 525200106092005.pdf.
25. U.S. Bureau of Labor Statistics Data. Consumer Price
Index—All Urban Consumers-Medical Care. US Bureau of La-
bor Statistics Web site. Retrieved February 19, 2009, from
http://www.bls.gov/cpi/. Accessed February 19, 2009.
26. Houchens RL, Elixhauser A. Using the HCUP Nationwide Inpa-
tient Sample to Estimate Trends. (Updated for 1988–2004). HCUP
Methods Series Report #2006-05 Online. Retrieved August 18,
2006, from http://www.hcup-us.ahrq.gov/reports/methods.jsp.
27. Celli BR, MacNee W. Standards for the diagnosis and treatment
of patients with COPD: A summary of the ATS/ERS position paper.
Eur Respir J Jun 2004; 23:932–946.
28. Rabe KF, Hurd S, Anzueto A, Barnes PJ, Buist SA, Calver-
ley P, Fukuchi Y, Jenkins C, Rodriguez-Roisin R, van Weel C,
Zielinski J. Global strategy for the diagnosis, management, and
prevention of chronic obstructive pulmonary disease: GOLD exec-
utive summary. Am J Respir Crit Care Med 2007; 176:532–535.
29. Hsia DC, Krushat WM, Fagan AB, Tebbutt JA, Kusserow
RP. Accuracy of diagnostic coding for Medicare patients under
the prospective-payment system. N Engl J Med. Feb 11 1988;
318:352–355.
30. O’Malley KJ, Cook KF, Price MD, Wildes KR, Hurdle JF, Ashton
CM. Measuring diagnoses: ICD code accuracy. Health Serv Res
Oct 2005;40(5 Pt 2):1620–1639.
31. Stein BD, Bautista A, Schumock GT, Lee TA, Shah LM, Meltzer
DO, Krishnan JA. Validity of ICD-9 discharge diagnoses in iden-
tifying patients hospitalized for COPD. Am J Respir Crit Care Med
2009;179:A2159.
June 2010 171