Charlson Index Is Associated with One-year Mortality in Emergency Department Patients with Suspected Infection

Article (PDF Available)inAcademic Emergency Medicine 13(5):530-6 · May 2006with413 Reads
DOI: 10.1197/j.aem.2005.11.084 · Source: PubMed
Abstract
A patient's baseline health status may affect the ability to survive an acute illness. Emergency medicine research requires tools to adjust for confounders such as comorbid illnesses. The Charlson Comorbidity Index has been validated in many settings but not extensively in the emergency department (ED). The purpose of this study was to examine the utility of the Charlson Index as a predictor of one-year mortality in a population of ED patients with suspected infection. The comorbid illness components of the Charlson Index were prospectively abstracted from the medical records of adult (age older than 18 years) ED patients at risk for infection (indicated by the clinical decision to obtain a blood culture) and weighted. Charlson scores were grouped into four previously established indices: 0 points (none), 1-2 points (low), 3-4 points (moderate), and > or =5 points (high). The primary outcome was one-year mortality assessed using the National Death Index and medical records. Cox proportional-hazards ratios were calculated, adjusting for age, gender, and markers of 28-day in-hospital mortality. Between February 1, 2000, and February 1, 2001, 3,102 unique patients (96% of eligible patients) were enrolled at an urban teaching hospital. Overall one-year mortality was 22% (667/3,102). Mortality rates increased with increasing Charlson scores: none, 7% (95% confidence interval [CI] = 5.4% to 8.5%); low, 22% (95% CI = 19% to 24%); moderate, 31% (95% CI = 27% to 35%); and high, 40% (95% CI = 36% to 44%). Controlling for age, gender, and factors associated with 28-day mortality, and using the "none" group as a reference group, the Charlson Index predicted mortality as follows: low, odds ratio of 2.0; moderate, odds ratio of 2.5; and high, odds ratio of 4.7. This study suggests that the Charlson Index predicts one-year mortality among ED patients with suspected infection.

Figures

Charlson Index Is Associated with One-year
Mortality in Emergency Department Patients
with Suspected Infection
Scott B. Murray, MD, David W. Bates, MD, MSc, Long Ngo, PhD, Jacob W. Ufberg, MD,
Nathan I. Shapiro, MD, MPH
Abstract
Objectives: A patient’s baseline health status may affect the ability to survive an acute illness. Emergency
medicine research requires tools to adjust for confounders such as comorbid illnesses. The Charlson
Comorbidity Index has been validated in many settings but not extensively in the emergency department
(ED). The purpose of this study was to examine the utility of the Charlson Index as a predictor of one-
year mortality in a population of ED patients with suspected infection.
Methods: The comorbid illness components of the Charlson Index were prospectively abstracted from the
medical records of adult (age older than 18 years) ED patients at risk for infection (indicated by the clinical
decision to obtain a blood culture) and weighted. Charlson scores were grouped into four previously estab-
lished indices: 0 points (none), 1–2 points (low), 3–4 points (moderate), and R5 points (high). The primary
outcome was one-year mortality assessed using the National Death Index and medical records. Cox
proportional-hazards ratios were calculated, adjusting for age, gender, and markers of 28-day in-hospital
mortality.
Results: Between February 1, 2000, and February 1, 2001, 3,102 unique patients (96% of eligible patients)
were enrolled at an urban teaching hospital. Overall one-year mortality was 22% (667/3,102). Mortality
rates increased with increasing Charlson scores: none, 7% (95% confidence interval [CI] = 5.4% to
8.5%); low, 22% (95% CI = 19% to 24%); moderate, 31% (95% CI = 27% to 35%); and high, 40% (95%
CI = 36% to 44%). Controlling for age, gender, and factors associated with 28-day mortality, and using
the ‘‘none’’ group as a reference group, the Charlson Index predicted mortality as follows: low, odds ratio
of 2.0; moderate, odds ratio of 2.5; and high, odds ratio of 4.7.
Conclusions: This study suggests that the Charlson Index predicts one-year mortality among ED patients
with suspected infection.
ACADEMIC EMERGENCY MEDICINE 2006; 13:530–536 ª 2006 by the Society for Academic Emergency
Medicine
Keywords: mortality prediction, risk adjustment, emergency medicine, sepsis
I
t is intuitive that a patient’s baseline health status
plays a role in a patient’s ability to survive after sus-
taining an acute injury or illness.
1,2
However, quanti-
fying a patient’s comorbid burden can be difficult. The
challenge lies in determining which comorbid conditions
affect patient mortality and to what degree. Charlson
et al. developed a scoring system for comorbid illness
that predicted one-year mortality in medical inpatients
and later validated the model in patients with breast can-
cer.
3
The Charlson score is based on the presence of clin-
ical conditions obtained from manual chart abstraction.
Each comorbid condition was assigned a whole number
integer that was proportional to the relative risk of death
(at one year) associated with that disease (Table 1). The
sum of the integers makes up the Charlson score, and
the scores are grouped to form the Charlson Index.
The importance of accurately measuring comorbidity
is apparent when researchers attempt to describe the
baseline illness severity in a patient population. A stan-
dardized tool that calculates a patient’s baseline disease
severity would be useful for researchers, because it
From the Department of Emergency Medicine, Beth Israel
Deaconess Medical Center (SBM, LN, NIS), Boston, MA; Division
of General Medicine, Department of Medicine, Brigham and
Women’s Hospital (DWB), Boston, MA; and Department of Emer-
gency Medicine, Temple University Hospital (JWU), Philadelphia,
PA.
Received July 24, 2005; revision received November 14, 2005;
accepted November 14, 2005.
Address for correspondence and reprints: Nathan I. Shapiro,
MD, MPH, West Clinical Center 2–Department of Emergency
Medicine, Beth Israel Deaconess Medical Center, One Deaconess
Way, Boston, MA 02215. Fax: 617-754-2350; e-mail: nshapiro@
bidmc.harvard.edu.
ISSN 1069-6563 ª 2006 by the Society for Academic Emergency Medicine
PII ISSN 1069-6563583 doi: 10.1197/j.aem.2005.11.084530
would allow readers to better understand the comorbid
burden in the study population and to assess the general-
izability of the study to the reader’s patient population.
It could also be useful to analyze comorbid burden as a
potential confounder to study results, similar to what
would occur when discrepancies in age, race, or gender
are found between cohorts. Such a tool would also be of
value in small case series and nonrandomized studies.
Researchers could use this as a tool to describe popula-
tions when comparing data between different institu-
tions or time periods. This tool could also be used to
gauge the success of randomization.
The Charlson Index is a well-validated research metho-
dology that has been used in numerous populations.
4–16
Despite its prevalence in the medical literature, the utility
of the Charlson Index has not been well studied in an
emergency department (ED) population.
17
Therefore,
we sought to determine how well the Charlson Index
would predict one-year mortality in a population of ED
patients with suspected infection.
METHODS
Study Design
We performed an observational study of consecutive ED
patients with suspected infection seen over a one-year
period to assess the ability of the Charlson Comorbidity
Index to predict mortality. The primary outcome mea-
sured was one-year mortality. This study was approved
by the investigational review board at Beth Israel
Deaconess Medical Center.
Study Setting and Population
Subjects were patients seen at an urban, academic, ter-
tiary care hospital ED in Boston with an annual census
of 50,000 visits. The study period began on February 1,
2000, and ended on February 1, 2001. All patients (18
years or older) who had blood cultures drawn in the
ED (or within three hours of admission to an inpatient
bed) were eligible for the study.
18
Patients who had
more than one ED visit during the study period were
enrolled only once at the first visit. The clinical decision
to obtain a blood culture was used to identify patients
deemed at risk for infection, because clinicians com-
monly obtain blood cultures on patients who are
suspected to have bacteremia.
Study Protocol and Measurements
Emergency department charts were reviewed by trained
data abstractors without subsequent knowledge of the
patient’s hospital course or the study hypothesis. Data
collected included demographic information and the
presence of any of the Charlson comorbidities. Original
definitions and integer weights of comorbid conditions,
as published by Charlson et al., were used (Table 1).
3
A
Charlson score was calculated for each patient by adding
the integers (Table 1) assigned to each disease diagnosed
in that patient (e.g., congestive heart failure [1] + lym-
phoma [2] = Charlson score [3]). The Charlson score
was prospectively consolidated into four previously
defined groups known as the Charlson Index: 0 points
(none), 1–2 points (low), 3–4 points (moderate), and R5
points (high) as originally described.
3
One-year mortality (measured from date of index visit)
was used as the primary mortality outcome, because the
Charlson Index was originally derived for this time inter-
val. Patients were classified as dead (22%) if they were
found to be on the national death registry within one
year from the index visit or if they had a discharge diag-
nosis of death during the one-year period. Patients were
confirmed to be alive (53%) if they were found to have
had a hospital encounter (laboratory tests, office visits,
and so on) more than one year after the index visit. Pa-
tients were presumed to be alive (25%) if they were not
located in the national death registry (no evidence of
death) and did not have any hospital encounters (no evi-
dence of life). Both confirmed and presumed alive were
classified as alive for the purposes of analysis.
Data Analysis
Descriptive statistics were used to summarize the demo-
graphics and in-hospital clinical characteristics of 3,102
patients eligible for the analysis. Categorical variables
were reported as percentages. The distributions of con-
tinuous variables were described using the mean, stan-
dard deviation, and median with interquartile ranges.
One-year mortality rates with 95% confidence intervals
(CIs) were calculated for each Charlson Index group.
To determine independent association of the Charlson
Index with one-year mortality, we built a Cox propor-
tional-hazards model incorporating routinely reported
variables (age, gender) and significant predictors of
28-day in-hospital mortality identified in prior studies.
18
The Cox proportional-hazards model was used to com-
pare the adjusted one-year mortality hazard ratios for
the Charlson Index using the lowest comorbidity (none)
as the reference group. Adjusted relative hazard ratio
estimates with corresponding 95% CIs were reported
for the three Charlson Comorbidity Index groups (low,
moderate, and high). Our new model showing adjusted
relative hazard ratios includes covariates found to be
independent predictors of one-year mortality (p < 0.05).
Table 1
Charlson Comorbid Conditions
Comorbid Condition
Charlson Integer
Weight
Acquired immunodeficiency syndrome 6
Metastatic solid tumor 6
Moderate or severe liver disease 3
Diabetes with end organ damage 2
Hemiplegia, paraplegia 2
Nonmetastatic solid tumor, leukemia,
lymphoma
2
Renal disease 2
Cerebrovascular accident 1
Chronic pulmonary disease 1
Congestive heart failure 1
Connective tissue disease 1
Dementia 1
Diabetes without organ damage 1
Mild liver disease 1
Myocardial infarction 1
Peptic ulcer disease 1
Peripheral vascular disease 1
ACAD EMERG MED
May 2006, Vol. 13, No. 5
www.aemj.org 531
The Kaplan–Meier estimation method was used to obtain
the survival distribution estimates. All analyses were per-
formed using SAS software version 9 (SAS Institute, Inc.,
Cary, NC).
19
RESULTS
Over a one-year period, 4,044 patient visits occurred
during which blood cultures were drawn. Repeat visits
occurred in 824 cases, and only the first visit was eligible
for inclusion. Of the 3,220 eligible patient visits, 3,102
were enrolled (96% of eligible patients). Using medical
records, 53% were found to be alive one year later, while
22% had died as determined by medical records or
the national death index. Twenty-five percent of the study
group was presumed to be alive because no evidence of
death was found in the national death index. The average
age was 59.9 years (median, 61 years). Female patients
made up a slightly larger proportion than male patients
(Table 2). The median length of stay in the hospital was
three days, and 11.6% were admitted to the intensive
care unit (median length of stay in the intensive care
unit, two days). The most frequent suspected sources of
infection, as documented by the treating physician,
were lower respiratory infections and skin/soft tissue
infections (Table 2).
Table 2 also shows the prevalence of baseline comor-
bid conditions, as defined by the Charlson Comorbidity
Index. The categories of nonmetastatic solid tumor, con-
gestive heart failure, renal disease, and chronic pulmo-
nary diseases make up slightly more than half of all the
comorbidities listed. All comorbidity index groups were
represented: none, 32%; low, 34%; moderate, 17%; and
high, 17%.
The one-year unadjusted mortality rate increased as
Charlson Index severity increased: none, 7% (95% CI =
5.4% to 8.5%); low, 22% (95% CI = 19% to 24%); moder-
ate, 31% (95% CI = 26.8% to 34.7%); and high, 40%
(95% CI = 35.5% to 43.8%) (Table 3). Survival over time
(Kaplan–Meier method) decreased with increasing Charl-
son Index severity (Figure 1). Table 3 shows the observed
incidence of death at one year stratified by the Charlson
Comorbidity Index, gender, and age. There is not a sig-
nificant difference in mortality between the two genders.
The adjusted survival analysis model shows the indepen-
dent effect of the Charlson Comorbidity Index on mor-
tality after controlling for gender, age, and factors that
were found to be predictive of inpatient 28-day mortality
(nursing home residents, lower respiratory infection,
respiratory rate >20 breaths/min or pulse oximetry
<90%, altered mental status, and platelet count <150,000/
mL). Age was strongly correlated with mortality risk, but
even after adjusting for age the Charlson Index remained
strongly predictive of mortality. Using the Charlson
Index group with the lowest comorbidity (none) as the
reference group (odds ratio of 1), the adjusted odds of
death for each group were as follows: low, odds ratio
of 2.0 (95% CI = 1.6 to 2.7), moderate, odds ratio of 2.5
(95% CI = 1.9 to 3.4), and high, odds ratio of 4.7 (95%
CI = 3.6 to 6.2). Adjusted odds ratios were statistically dif-
ferent using the Cox proportional-hazards model. Each
index group was an independent predictor of death,
with risk progressively increasing as Charlson Index
severity increased.
Table 2
Descriptive Statistics of Patient Population
Demographic and admission
Percent female (n) 55 (1,706)
Median (IQR) age (yr) 61 (43–77)
Median (IQR) length of stay (days) 3.0 (1–6)
ICU admission rate (n) 11.6 (360)
Median (IQR) ICU length of stay (days) 2.0 (1–5)
Indication for blood cultures (% [n])
Lower respiratory infection 23 (719)
Skin or soft tissue infection 19 (590)
Intra-abdominal infection 14 (424)
Fever without source 13 (414)
Urologic infection 10 (296)
Line infection 4 (117)
Central nervous system infection 2 (74)
Rule out endocarditis 2 (65)
Neutropenic fever 2 (61)
Charlson comorbid conditions (% [n])
Nonmetastatic solid tumor, leukemia, lymphoma 19 (605)
Congestive heart failure 13 (425)
Renal disease 10 (366)
Chronic pulmonary disease 11 (360)
Diabetes without organ damage 12 (351)
Diabetes with end organ damage 10 (341)
Myocardial infarction 9 (288)
Cerebrovascular accident 8 (257)
Dementia 8 (257)
Peripheral vascular disease 7 (233)
Acquired immunodeficiency syndrome 6 (217)
Peptic ulcer disease 5 (174)
Moderate or severe liver disease 4 (158)
Metastatic solid tumor 4 (152)
Mild liver disease 2 (65)
Connective tissue disease 2 (62)
Hemiplegia, paraplegia 1 (34)
Patients may have more than one comorbidity, so the totals may be
>100%.
Table 3
One-year Mortality Stratified by Charlson Index, Gender, and
Age
One-year
Unadjusted
Mortality (%)
Adjusted Relative
Hazard Ratio
(95% CI)*
Charlson Comorbidity
Index
None (0) 7.0 1.0
Low (1–2) 22 2.0 (1.6, 2.7)
Moderate (3–4) 31 2.5 (1.9, 3.4)
High (R5) 40 4.7 (3.6, 6.2)
Gender
Female 21 1.0
Male 22 1.1 (0.9, 1.3)
Age (yr)
Younger than 65 11 1.0
Between 65 and 80 29 2.1 (1.7, 2.6)
Older than 80 42 2.8 (2.3, 3.5)
* Cox proportional-hazards model, adjusted for gender, age, and signifi-
cant covariates: nursing home residents, lower respiratory infection, res-
piratory rate >20 breaths/min or pulse oximetry <90%, altered mental
status, and platelet count <150,000/mL.
532 Murray et al.
CHARLSON INDEX AND ONE-YEAR MORTALITY IN ED PATIENTS
DISCUSSION
We found that the Charlson Index was a valid predictor
of one-year mortality in an ED patient population, inde-
pendent of age and gender. The index was able to quan-
tify the comorbid status of this population. These data
suggest that researchers can use this index to assess if
study cohorts have similar comorbid burdens or if in
fact there is excess comorbidity that may act as a con-
founder. This is particularly important when comparing
research between different studies or data obtained
from populations at different institutions. Investigators
can also verify if randomization has equally distributed
comorbidity between cohorts.
If cohorts are dissimilar, adjustment may be made to
account for baseline comorbidity. This tool may help
researchers assess population equality when performing
small nonrandomized studies or case series. In fact, the
Charlson Index has been used as a primary outcome,
an independent variable, and as a way to control for
comorbid illness for a variety outcomes besides mor-
tality.
5,6,20–31
Assuming that patients with more comorbid disease
will have higher mortality rates, the ability of the Charl-
son Index to measure comorbidity may be verified by
plotting mortality rates against the Charlson Index and
then observing a progressive increase in mortality. The
original investigation used to derive the Charlson Index
predicted one-year mortality in medical inpatients in
1984 as follows: 12%, none; 26%, low; 52%, moderate;
and 85%, high. The same index predicted different rates
of one-year mortality in the validation group of patients
with breast cancer between 1962 and 1969 as follows:
1%, none; 3%, low; 16%, moderate; and 31%, high.
3
Our work shows that the Charlson Index predicted
one-year mortality due to comorbid conditions in a dif-
ferent population, ED patients with suspected infection,
as follows: 7%, none; 22%, low; 31%, moderate; and
40%, high. It is inherent that the absolute mortality rates
will vary between populations presenting with different
acute diseases, and researchers should not apply the
mortality rates of patients at risk for sepsis to their
cohorts with different acute medical problems. However,
as this ED study and numerous studies from other
patient populations have demonstrated the ability of the
Charlson Index to predict one-year mortality, it is likely
that the model can assess comorbidity in other ED
patient cohorts, albeit with differing absolute mortality
rates.
3,4,32–53
Some have commented that age alone may be a suffi-
cient surrogate marker for comorbidity when trying to
assess cohort similarity.
54
In the only previous study
examining the Charlson Index in an ED population, the
Charlson Index was found to be equivalent to the modi-
fied Trauma and Injury Severity Score (modified using
age as a continuous variable) in predicting Australian
inpatient mortality after acute trauma.
17
We found age
to be an important predictor of mortality but that comor-
bidity is an independent and equally important predictor.
Instinctively, clinicians realize that despite advanced age,
a 72-year-old patient with a history of hypertension may
fare better than a 45-year-old patient with diabetes, renal
failure, prior myocardial infarction, and chronic obstruc-
tive pulmonary disease. Ignoring the effect of comorbid-
ity on mortality could potentially produce as much bias
as if age were ignored. To detect potential confounders,
researchers report the similarity between cohorts; data
on age, gender, and severity of acute illness are routinely
presented. Given the results of this study, researchers
Figure 1. Observed survival distributions estimated by Kaplan–Meier product limit method for each of the four categories of
the Charlson Index over time. The four categories of the Charlson Index display four distinct survival distributions that are
further apart toward the end of the one-year follow-up period, indicating stronger association between the index and mor-
tality through time. The log-rank test for differences between the groups is statistically significant (p < 0.001).
ACAD EMERG MED
May 2006, Vol. 13, No. 5
www.aemj.org 533
may wish to comment on comorbidity equality between
populations using the Charlson Index in future studies.
We used the original integer weights as initially de-
scribed by Charlson et al. When the weights for the score
are rederived based on data specific to a particular pop-
ulation, the index becomes more accurate when used
with that population.
4,35–37,54
This probably reflects the
impact a given comorbid condition has on a particular
disease cohort (e.g., congestive heart failure has an inte-
ger weight of 3 when the model is rederived in patients
undergoing coronary artery bypass graft surgery).
36
Rederivation also reflects regional differences that
make study populations different. The utility of using
the original Charlson weighting system is that it can be
(and has been successfully) applied to many populations,
but this generalizability diminishes its ‘goodness of fit’
to any specific population. Rederiving the weighting of
comorbid factors based on local patient demographic
and hospital characteristics would make the tool more
specific to the area where it was created. This is advanta-
geous for the researcher locally, but disadvantageous
when trying to create a more universal tool. Theoreti-
cally, researchers would have to collect large amounts
of data, rederive, and validate a model every time a
new population (new hospital, different presenting dis-
ease, and so on) was studied, which is an expensive,
time-consuming project.
LIMITATIONS
This study has several limitations. Free-text ED charts
were used for data abstraction, and it is likely that
more medical history became available after admission
or that documentation inaccuracies occurred, creating
misclassification bias for the variables. The model may
have performed better had clinicians been required to
specify the presence or absence of each Charlson Index
component or had inpatient charts been used. However,
this limitation shows that researchers who only have ac-
cess to ED charts can assess comorbidity adequately with
the information at hand. Future work that includes inpa-
tient chart data may increase the accuracy of the Charl-
son score. Misclassification bias in assessing mortality
is possible; however, we confirmed life or death in 75%
of the study group and assumed the remaining 25% to
be alive because they were not listed in the national death
index. External validity was not addressed because data
were collected from only one hospital site.
Future research may include using inpatient records to
more accurately assess comorbid disease or rederiving
the integer weights. The Charlson Index has been previ-
ously adapted to ICD-9-CM inpatient databases.
32,33
The
utility of using computerized administrative databases to
determine comorbidity has not been examined in an ED
population.
CONCLUSIONS
We found that the Charlson Index, a well-established
method to assess for comorbid illness burden, predicted
one-year mortality for an ED patient population with
suspected infection. The Charlson Index may be used
by researchers to measure comorbidity and assess co-
hort similarity or to control for a patient’s comorbidity.
Further validation in other ED patient populations
should be performed before widespread implementation.
References
1. Jencks SF, Dobson A. Refining case-mix adjustment:
the research evidence. N Engl J Med. 1987; 317:
679–86.
2. Greenfield S, Aronow HV, Elashoff RM, Watanaba D.
Flaws in mortality data: the hazards of ignoring
comorbid disease. JAMA. 1988; 260:2253–5.
3. Charlson ME, Pompei P, Ales KL, MacKenzie RC. A
new method of classifying prognostic comorbidity
in longitudinal studies: development and validation.
J Chronic Dis. 1987; 40:373–83.
4. de Groot V, Beckerman H, Lankhorst GJ, Bouter LM.
How to measure comorbidity: a critical review of
available methods. J Clin Epidemiol. 2003; 56:221–9.
5. Navarro-Cano G, Del Rincon I, Pogosian S, Roldan
JF, Escalante A. Association of mortality with disease
severity in rheumatoid arthritis, independent of
comorbidity. Arthritis Rheum. 2003; 48:2425–33.
6. Sin DD, McAlister FA. The effects of beta-blockers in
morbidity and mortality in a population-based cohort
of 11,942 elderly patients with heart failure. Am J
Med. 2002; 113:650–6.
7. Yu DT, Black E, Sands KE, et al. Severe sepsis: varia-
tion in resource and therapeutic modality among
academic centers. Crit Care. 2003; 7(3):24–34.
8. Durairaj L, Will JG, Torner JC, Doebbleling BN. Prog-
nostic factors for mortality following interhospital
transfers to the medical intensive care unit of a ter-
tiary referral center. Crit Care Med. 2003; 31:1981–6.
9. Yu DT, Platt R, Lanken PN, et al. Relationship of pul-
monary artery catheter use to mortality and resource
utilization in patients with severe sepsis. Crit Care
Med. 2003; 31:2734–41.
10. Tu JV, Austin PC, Johnston KW. The influence of sur-
gical specialty training on the outcomes of elective
abdominal aortic aneurysm surgery. J Vasc Surg.
2002; 33:447–52.
11. Zacks SL, Sandler RS, Rutledge R, Brown RS. A pop-
ulation-based cohort study comparing laparoscopic
cholecystectomy and open cholecystectomy. Am J
Gastroenterol. 2002; 97:334–40.
12. Collins AJ, Weinhandl E, Snyder JJ, Chen SC, Gil-
bertson D. Comparison and survival of hemodialysis
and peritoneal dialysis in the elderly. Semin Dial.
2002; 15:98–102.
13. Rozental TD, Branas CC, Bozentka DL, Beredjklian PK.
Survival among elderly patients after fractures of the
distal radius. J Hand Surg [Am]. 2002; 27:948–52.
14. Cin MH, Goldman L. Gender differences in 1-year
survival and quality of life among patients admitted
with congestive heart failure. Med Care. 1998; 36:
1033–46.
15. Gabriel SE, Crowson CS, O’Fallon WM. Mortality in
rheumatoid arthritis: have we made an impact in 4
decades? J Rheumatol. 1999; 26:2529–33.
16. Covinsky KE, Kahana E, Chin MH, Palmer RM,
Fortinsky RH, Landefeld CS. Depressive symptoms
534 Murray et al.
CHARLSON INDEX AND ONE-YEAR MORTALITY IN ED PATIENTS
and 3-year mortality in older hospitalized medical
patients. Ann Intern Med. 1999; 130:563–9.
17. Gabbe BJ, Magtengaard K, Hannaford AP, Cameron
PA. Is the Charlson Index useful for predicting trauma
outcomes? Acad Emerg Med. 2005; 12:318–21.
18. Shapiro NI, Wolfe RE, Moore RB, Smith E, Burdick E,
Bates DW. Mortality in Emergency Department Sep-
sis (MEDS) score: a prospectively derived and vali-
dated clinical rule. Crit Care Med. 2003; 31:670–5.
19. SAS/STAT User’s Guide. Version 8. Cary, NC: SAS
Institute, 2000.
20. Liu JH, Zingmond D, Etzioni DA, et al. Characterizing
the performance and outcomes of obesity surgery in
California. Am Surg. 2003; 69:823–8.
21. Zhang Q, Safford M, Ottenweller J, et al. Performance
status of health care facilities changes with risk adjust-
ment of HbA1c. Diabetes Care. 2000; 23:919–27.
22. Wang PS, Solomon DH, Mogun H, Avorn J.
HMG-COA reductase inhibitors and the risk of hip
fractures in elderly patients. JAMA. 2000; 283:3211–6.
23. Monane M, Kanter DS, Glynn RJ, Avorn J. Variability
in length of hospitalization for stroke. The role of
managed care in an elderly population. Arch Neurol.
1996; 53:875–80.
24. Winkelmayer WC, Glynn RJ, Lebin R, Mittleman MA,
Pliskin JS, Avorn J. Late nephrologist referral and
access to renal transplantation. Transplantation.
2002; 73:1918–23.
25. Filit H, Hill JW, Futterman R. Health care utilization
and costs of Alzheimer’s disease: the role of co-morbid
conditions, disease stage, and pharmacotherapy.
Fam Med. 2002; 34:528–35.
26. Fischer LR, Defor TA, Cooper S, et al. Pharmaceutical
care and health care utilization in an HMO. Effect
Clin Pract. 2002; 5(2):49–57.
27. Saigal CS, Pashos CL, Henning JM, Litwin MS. Vari-
ations in use of imaging in a national sample of men
with early-stage prostate cancer. Urology. 2002; 59:
400–4.
28. Jaeger TM, Lohr RH, Pankratz VS. Symptom-trig-
gered therapy for alcohol withdrawal syndrome in
medical inpatients. Mayo Clin Proc. 2001; 76:695–701.
29. Martin BC, Ricci JF, Kotzan JA, Lang K, Menzin J.
The net cost of Alzheimer disease and associated dis-
orders. Alzheimer Dis Assoc Disord. 2000; 14:151–9.
30. Chaung KH, Covinsky KE, Sands LP, Fortinsky RH,
Palmer RM, Landefeld CS. Diagnosis-related-group-
adjusted hospital costs are higher in older medical
patients with lower functional status. J Am Geriatr
Soc. 2003; 51:1729–34.
31. Phillips Bute B, Matthew J, Blumenthal JA, et al.
Female gender is associated with impaired quality
of life 1 year after coronary artery bypass surgery.
Psychosom Med. 2003; 65:944–51.
32. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical
comorbidity index for use with ICD-9-CM adminis-
trative databases. J Clin Epidemiol. 1992; 45:613–9.
33. Romano PS, Roos LL, Jollis JG. Adapting a clinical
comorbidity index for use with ICD-9-CM adminis-
trative data: differing perspectives. J Clin Epidemiol.
1993; 46:1075–9.
34. Poses RM, McClish DK, Smith WR, Bekes C, Scott
WE. Prediction of survival of critically ill patients by
admission comorbidity. J Clin Epidemiol. 1996; 49:
743–7.
35. Romano PS, Roos LL, Jollis JG. Further evidence con-
cerning the use of a clinical comorbidity index with
ICD-9-CM administrative data. J Clin Epidemiol.
1993; 46:1085–90.
36. Ghali WA, Hall RE, Rosen AK, Ash AS, Moskowitz
MA. Searching for an improved clinical comorbidity
index for use in ICD-9-CM administrative data.
J Clin Epidemiol. 1996; 49:273–8.
37. van Doorn C, Bogardus ST, Willliams CS, Concato J,
Towle VR, Inouye SJ. Risk adjustment for older
hospitalized persons: a comparison of two methods
of data collection for the Charlson index. J Clin
Epidemiol. 2001; 54:694–701.
38. D’Hoore W, Bouckaert A, Tilquin C. Practical consid-
erations on the use of the Charlson comorbidity
index. J Clin Epidemiol. 1996; 49:1429–33.
39. Kieszak SM, Flanders WD, Kosinski AS, Ship CC,
Karp H. A comparison of the Charlson comorbidity
index derived from medical record data and admin-
istrative billing data. J Clin Epidemiol. 1999; 52:
137–42.
40. Cleves MA, Sanchez N, Draheim M. Evaluation of
two competing methods for calculating Charlson’s
comorbidity index when analyzing short-term
mortality using administrative data. J Clin Epidemiol.
1997; 50:903–8.
41. van Mansen JG, Korevaar JC, Dekker FW, Boescho-
ten EW, Bossuyt PM, Krediet RT. How to adjust for
comorbidity in survival studies in ESRD patients: a
comparison of different indices. Am J Kidney Dis.
2002; 40:82–9.
42. Librero J, Peiro S, Ordinana R. Chronic comorbidity
and outcomes of hospital care: length of stay, mor-
tality, and readmission at 30 and 365 days. J Clin
Epidemiol. 1999; 52:171–9.
43. D’Hoore W, Sicotte C, Tilquin C. Risk adjustment in
outcome assessment: the Charlson comorbidity
index. Methods Inf Med. 1993; 32:382–7.
44. Sachdev M, Sun JL, Tsiatis AA, Nelson CL, Mark DB,
Jollis JG. The prognostic importance of comorbidity
for mortality in patients with stable coronary artery
disease. J Am Coll Cardiol. 2004; 43:576–82.
45. Charlson M, Szatrowski TP, Peterson J, Gold J.
Validation of a combined comorbidity index. J Clin
Epidemiol. 1994; 47:1245–51.
46. Liu M, Domen K, Chino N. Comorbidity measures for
stroke outcome research: a preliminary study. Arch
Phys Med Rehabil. 1997; 78:166–72.
47. Newschaffer CJ, Bush TL, Penberthy LE, Bellatoni M,
Helzlsour K, Diener-West M. Does comorbid disease
interact with cancer? An epidemiologic analysis of
mortality in a cohort of elderly breast cancer patients.
J Gerontol. 1998; 53:372–8.
48. Beddhu S, Bruns FJ, Saul M, Seddon P, Zeidel ML. A
simple comorbidity scale predicts clinical outcomes
and costs in dialysis patients. Am J Med. 2000; 108:
609–13.
49. Beddhu S, Zeidel ML, Saul M, et al. The effects of
comorbid conditions on the outcomes of patients
undergoing peritoneal dialysis. Am J Med. 2002;
112:696–701.
ACAD EMERG MED
May 2006, Vol. 13, No. 5
www.aemj.org 535
50. Rochon PA, Katz JN, Morrow LA, et al. Comorbid ill-
ness is associated with survival and length of hospital
stay in patients with chronic disability. Med Care.
1996; 34:1093–101.
51. Gabriel SE, Crowson CS, O’Fallon WM. A compari-
son of two comorbidity instruments in arthritis.
J Clin Epidemiol. 1999; 52:137–42.
52. Birim O, Zuydendorp HM, Maat AP, Kappetein AP,
Eijkmans MJ, Bogers AJ. Lung resection for non-small
cell lung cancer in patients older than 70: mortality,
morbidity, and late survival compared with the
general population. Ann Thorac Surg. 2003; 76:
1796–801.
53. Skiest DJ, Rubinstein E, Carley N, Gioiella L, Lyons R.
The importance of comorbidity in HIV-infected
patients over 55: a retrospective case-control study.
Am J Med. 1996; 101:605–11.
54. Schneeweiss S, Maclure M. Use of comorbidity
scores for control of confounding in studies using
administrative data. Int J Epidemiol. 2000; 29:891–8.
Where to Find AEM Instructions for Authors
For complete instruct ions for authors, see the Jan uary or July issue of
Academic Emergency Medicine ; visit http://ees.else vier.com/acaeme/
default.asp and click on ‘Guide for Authors’’; or contact SAEM via
e-mail at aem@saem.org, via phone at 517-485-5484, or via fax at
517-485-0801.
536 Murray et al.
CHARLSON INDEX AND ONE-YEAR MORTALITY IN ED PATIENTS
    • "The patients' underlying diseases analyzed were: diabetes mellitus, liver cirrhosis, cancer, transplant recipient, HIV infection, chronic renal disease, obstructive pulmonary disease, trauma, and systemic arterial hypertension. We also analyzed the Charlson Comorbidity Score [14]. Bacteremias were classified as primary and secondary bloodstream infections. "
    [Show abstract] [Hide abstract] ABSTRACT: Background It has been challenging to determine the true clinical impact of Acinetobacter spp., due to the predilection of this pathogen to colonize and infect critically ill patients, who often have a poor prognosis. The aim of this study was to assess whether Acinetobacter spp. bacteremia is associated with lower survival compared with bacteremia caused by other pathogens in critically ill patients. Methods This study was performed at Hospital das Clínicas, University of São Paulo, Brazil. There are 12 intensive care units (ICUs) in the hospital: five Internal Medicine ICUs (emergency, nephrology, infectious diseases and respiratory critical care), three surgical ICU (for general surgery and liver transplantion), an Emergency Department ICU for trauma patients, an ICU for burned patients, a neurosurgical ICU and a post-operative ICU. A retrospective review of medical records was conducted for all patients admitted to any of the ICUs, who developed bacteremia from January 2010 through December 2011. Patients with Acinetobacter spp. were compared with those with other pathogens (Klebsiella pneumoniae, Staphylococcus aureus, Enterobacter spp., Enterococcus spp., Pseudomonas aeruginosa). We did a 30-day survival analysis. The Kaplan-Meier method and log-rank test were used to determine the overall survival. Potential prognostic factors were identified by bivariate and multivariate Cox regression analysis. ResultsOne hundred forty-one patients were evaluated. No differences between patients with Acinetobacter spp. and other pathogens were observed with regard to age, sex, APACHE II score, Charlson Comorbidity Score and type of infection. Initial inappropriate antimicrobial treatment was more frequent in Acinetobacter bacteremia (88 % vs 51 %). Bivariate analysis showed that age > 60 years, diabetes mellitus, and Acinetobacter spp. infection were significantly associated with a poor prognosis. Multivariate model showed that Acinetobacter spp. infection (HR = 1.93, 95 % CI: 1.25–2.97) and age > 60 years were independent prognostic factors. Conclusion Acinetobacter is associated with lower survival compared with other pathogens in critically ill patients with bacteremia, and is not merely a marker of disease severity.
    Full-text · Article · Dec 2016
    • "In order to ensure early normocapnia was independently associated with good neurological outcome, we performed sensitivity analyses adjusting for candidate variables known to be strong predictors of poor outcome in post-cardiac arrest patients. We selected the following candidate variables for the regression models: (1) initial cardiac rhythm (asystole or pulseless electrical activity (PEA) versus ventricular fibrillation/ventricular tachycardia (VF/VT)), (2) prolonged duration of cardiopulmonary resuscitation (CPR duration > 20 minutes) [4,242526272829, (3) post-resuscitation shock (defined as systolic blood pressure < 100 mmHg or vasopressor support required to maintain systolic blood pressure > 100 mmHg during the first 24 hours after ROSC) [6,30,31], (4) metabolic acidosis (defined as one or more recorded base deficit ≤ -6 mmol/L during the first 24 hours after ROSC) [6,32], (5) age (decile), (6) pre-arrest comorbidities (that is Charlson comorbidities index) [33] , (7) pre-arrest pulmonary disease , and (8) initiation of therapeutic hypothermia. We also performed an additional multivariable logistic regression analysis to test the association between a narrower early PaCO 2 range (between 35 and 45 mmHg) and good neurological function at hospital discharge. "
    [Show abstract] [Hide abstract] ABSTRACT: Post-cardiac arrest hypocapnia/hypercapnia have been associated with poor neurological outcome. However, the impact of arterial carbon dioxide (CO2) derangements during the immediate post-resuscitation period following cardiac arrest remains uncertain. We sought to test the correlation between prescribed minute ventilation and post-resuscitation partial pressure of CO2 (PaCO2), and to test the association between early PaCO2 and neurological outcome. We retrospectively analyzed a prospectively compiled single-center cardiac arrest registry. We included adult (age >= 18 years) patients who experienced a non-traumatic cardiac arrest and required mechanical ventilation. We analyzed initial post-resuscitation ventilator settings and initial arterial blood gas analysis (ABG) after initiation of post-resuscitation ventilator settings. We calculated prescribed minute ventilation:MVmL/kg/min=tidalvolumeTV/idealbodyweightIBWxrespiratoryrateRRfor each patient. We then used Pearson's correlation to test the correlations between prescribed MV and PaCO2. We also determined whether patients had normocapnia (PaCO2 between 30 and 50 mmHg) on initial ABG and tested the association between normocapnia and good neurological function (Cerebral Performance Category 1 or 2) at hospital discharge using logistic regression analyses. Seventy-five patients were included. The majority of patients were in-hospital arrests (85%). Pulseless electrical activity/asystole was the initial rhythm in 75% of patients. The median (IQR) TV, RR, and MV were 7 (7 to 8) mL/kg, 14 (14 to 16) breaths/minute, and 106 (91 to 125) mL/kg/min, respectively. Hypocapnia, normocapnia, and hypercapnia were found in 15%, 62%, and 23% of patients, respectively. Good neurological function occurred in 32% of all patients, and 18%, 43%, and 12% of patients with hypocapnia, normocapnia, and hypercapnia respectively. We found prescribed MV had only a weak correlation with initial PaCO2, R = -0.40 (P < 0.001). Normocapnia was associated with good neurological function, odds ratio 4.44 (95%CI 1.33 to 14.85). We found initial prescribed MV had only a weak correlation with subsequent PaCO2 and that early Normocapnia was associated with good neurological outcome. These data provide rationale for future research to determine the impact of PaCO2 management during mechanical ventilation in post-cardiac arrest patients.
    Full-text · Article · Mar 2014
    • "The following data were collected for analysis. Clinical and demographic variables included age, sex, body mass index, alcohol, smoking, body temperature, respiratory rate, Charcot's triad, symptom to door time (time from symptom onset until arrival at the hospital), and Charlson comorbidity index score [18, 19]. Etiologic variables were malignant biliary obstruction, benign biliary stricture, and choledocholithiasis. "
    [Show abstract] [Hide abstract] ABSTRACT: Background Bacteremic cholangitis carries a high mortality rate of up to 10 % in relation to organ failure (OF), including septic shock. Aim The purpose of this study was to elucidate predictive factors for OF in bacteremic cholangitis. Methods A retrospective review of all patients diagnosed with acute cholangitis and proven bacteremia from 2003 to 2011 was performed. Comprehensive clinical and laboratory data of 211 patients were analyzed. Results There were 42 cases (19.9 %) of OF and 5 deaths (2.4 %). In the multivariate logistic regression analysis, significant predictive factors for OF were successful biliary decompression, presence of extended-spectrum beta-lactamase organism (ESBL), higher total bilirubin, and higher blood urea nitrogen (BUN) level at admission with odds ratios (ORs) of 0.129, 6.793, 1.148, and 1.089, respectively. Subgroup analysis of 165 patients who underwent biliary decompression before an event (with OF: 20, without OF: 145) was performed to elucidate the risk factors for organ failure even after successful biliary drainage. Variables significantly associated with OF included ESBL and BUN (OR = 4.123 and 1.177, respectively). We developed a scoring system with regression coefficient of each significant variable. The organ failure score was calculated using the following equation: (1.4 × ESBL) + (0.2 × BUN). This scoring system for predicting OF was highly sensitive (85.0 %) and specific (83.4 %). Conclusions Biliary decompression, ESBL, total bilirubin, and BUN are prognostic determinants in patients with bacteremic cholangitis. An organ failure scoring system may allow clinicians to identify groups with poor prognosis even after successful biliary decompression.
    Full-text · Article · Nov 2012
Show more