comparison of hospital mortality and long-term

7/26/2019 Comparison of Hospital Mortality and Long-term

1/8

618 Original Research [ 1 4 7 #3 CHEST MARCH 2015 ]

Comparison o Hospital Mortality and Long-term

Survival in Patients With Acute Lung Injury/ARDSvs Cardiogenic Pulmonary Edema

Christopher N. Schmickl, MD, MPH; Michelle Biehl, MD; Gregory A. Wilson, RRT; and Ognjen Gajic, MD, FCCP

BACKGROUND: Early differential diagnosis o acute lung injury (ALI) vs cardiogenic pulmo-

nary edema (CPE) is important or selecting the most appropriate therapy, but the prognostic

implications o this distinction have not been studied. Accurate prognostic inormation is

essential or providing appropriate inormed consent prior to initiation o mechanical

ventilation.

METHODS: Tis is a long-term ollow-up study o a previously established population-based

cohort o critically ill adult patients with acute pulmonary edema admitted at a tertiary-care

center during 2006 to 2009, in which post hoc expert review had established ALI vs CPE diag-

nosis. Using logistic and Cox regression, hospital mortality and long-term survival were com-

pared in patients with ALI vs patients with CPE.

RESULTS: O 328 patients (ALI5155, CPE5173), 240 patients (73%) died during a median

ollow-up o 160 days. Afer adjusting or conounders, patients with ALI were signicantly

more likely to die in the hospital (OR54.2, 95% CI52.3-7.8, n5325, P,.001), but among

hospital survivors the risk o death during ollow-up was the same in both groups (hazard

ratio5 1.13, 95% CI5 0.79-1.62, n5 229, P5 .50). Independent predictors o mortality

included age and APACHE (Acute Physiology and Chronic Health Evaluation) III score. Results

were similar when restricting patients with ALI to the subset with ARDS (Berlin denition). In

post hoc analyses, the mortality rate in hospital survivors compared with the general US popu-

lation was signicantly higher during the rst 2 years but essentially converged by year ve.

CONCLUSIONS: Although hospital mortality is higher in patients with ALI/ARDS compared

with patients with CPE, long-term survival is similar in hospital survivors rom both groups.

CHEST 2015; 147(3):618-625

[ Original Research Critical Care ]

Manuscript received June 9, 2014; revision accepted November 5, 2014;originally published Online First December 4, 2014.

ABBREVIATIONS: ALI5acute lung injury; APACHE5Acute Physiologyand Chronic Health Evaluation; CAD5coronary artery disease; CPE5cardiogenic pulmonary edema; HR5hazard ratio; PEEP5positiveend-expiratory pressure

AFFILIATIONS: From M.E..R.I.C. (Multidisciplinary Epidemiologyand ranslational Research in Intensive Care) (Drs Schmickl, Biehl,and Gajic and Mr Wilson), Division o Pulmonary and Critical CareMedicine, and the Department o Pulmonary and Critical Care Medicine(Drs Biehl and Gajic), Mayo Clinic, Rochester, MN; and the UniversityWitten-Herdecke (Dr Schmickl), Witten, Germany.

FUNDING/SUPPORT:he authors have reported to CHEST that nounding was received or this study.

CORRESPONDENCE TO: Christopher N. Schmickl, MD, MPH,Schnieglingerstrasse 225, 90427 Nuremberg, Germany; e-mail:[email protected]

2015 AMERICAN COLLEGE OF CHEST PHYSICIANS.Reproduction othis article is prohibited without written permission rom the AmericanCollege o Chest Physicians. See online or more details.

DOI:10.1378/chest.14-1371

wnloaded From: http://journal.publications.chestnet.org/ by Deddy Herman on 03/18/2015
mailto:[email protected]:[email protected]


2/8

journal.publications.chestnet.org 619

Acute hypoxic respiratory ailure with pulmonary

edema is a common syndrome in the ICU,1but

distinguishing between the main dierential diag-

noses, noncardiogenic (ARDS,2previously known as

acute lung injury [ALI]3) and cardiogenic pulmonary

edema (CPE), is oten challenging in the early stages

o illness.4Recently, several eorts have been made to

aid with the early distinction o ALI/ARDS vs CPE,5-9

which may acilitate early enrollment into research

studies and improve clinical management o these

patients.10-13

Accurate prognostic inormation is essential or clin-

ical decision-making and or providing appropriate

inormed consent beore initiation o aggressive measures

o resuscitation, such as invasive mechanical ventilation.

Although several studies evaluated mortality in patients

with acute respiratory ailure rom any cause1and in

patients with ALI/ARDS,14-19the outcomes o patients

with CPE have been less well characterized20and, to our

knowledge, have not yet been compared with patients

with ARDS.

o assess i early dierential diagnosis o ALI/ARDSvs CPE has prognostic implications, our primary aim

was to compare mortality (overall, during hospitali-

zation, and ater hospital discharge) between patients

with ALI/ARDS and patients with CPE. Secondary

aims were to identiy (1) independent predictors o

mortality and (2) subgroups among patients with

ALI/ARDS and patients with CPE whose survival

diers.

Materials and MethodsFor this retrospective observational study we used a previouslydescribed population-based cohort o 332 adult patients with acutepulmonary edema admitted to the ICU o a tertiary-care centerbetween 2006 and 2009, who had been classied as having ALI vs CPEvs ALI1CPE based on post hoc expert review (gold standard diagnosis),described in detail elsewhere.7In brie, two critical care experts reviewedpatients records afer hospital discharge or death and classied them asCPE or ALI (American-European Consensus Conerence denition3)based on the presence or absence o lef atrial hypertension, respec-tively. Markers o lef atrial hypertension included clinical (jugularvenous pressure), echocardiographic (eject ion ract ion, E/E9ratio),ECG (signs o ischemia), radiographic (cardiothoracic ratio, vascularpedicle width), and laboratory (brain natriuretic peptide, troponin)

indings as well as invasive measures o illing pressures (centralvenous and pulmonary wedge pressure). Experts also took into accountresponse to therapy (eg, diuretics). Patients with mixed results wereclassied as both (ALI1CPE). Disagreements were resolved by a thirdexpert.

Patients were ollowed rom onset o acute pulmonary edema(deined as the time when patients irst had a Pa2/F2,300 onarterial blood gas analysis and pulmonary edema or bilateral inl-trates on the chest radiograph as read by radiologists, both withina 24-h period)7up until January 1, 2014. Te only exclusion crite-ria were withdrawal o research authorization (since initial recruit-ment) and immediate loss to ollow-up ater hospital discharge(ie, zero ollow-up time). Tis study was approved by the Mayo ClinicInstitutional Review Board, waiving the need or inormed consent(IRB# 08-003560).

Using a previously validated database, which eeds rom hospital-wide data,21 we obtained patients date o death or last date oollow-up (or right-censored patients), age, sex, BMI, APACHE(Acute Physiology and Chronic Health Evaluation) III score onadmission, level o positive end-expiratory pressure (PEEP),ICU/hospital length o stay, and year o admission. ALI/CPE riskactors, smoking, and alcohol abuse were assessed by manual chartreview.7

Statistical Analysis

Patient characteristics were summarized and compared using theWilcoxon rank-sum test and the x2test or continuous and categorical

variables, respectively. Patients with ALI1CPE were treated as ALI inall analyses unless noted otherwise.

Patients with ALI vs patients with CPE were compared in terms o(1) long-term survival (rom onset o acute pulmonary edema until endo ollow-up) using Kaplan-Meier log-rank test, (2) hospital mortalityusing logistic regression, and (3) long-term survival (rom hospi-tal discharge until end o ollow-up) among hospital survivors basedon Cox regression and Kaplan-Meier methodology. o adjust orpotential conounding and to identiy independent predictors o mor-tality, regression models were built in a stepwise ashion (Pvalue toenter5 .1; P value to leave5 .25) using clinical judgment. Potentialconounders/predictors were chosen on clinical grounds, includingpatient demographics, ALI/CPE risk actors, APACHE III score, andhospital length o stay (or long-term survival analysis). For all regres-

sion models, we assessed the need or nonlinear modeling o con-tinuous covariates (as restricted cubic splines with three knots) andtested i the eect o diagnosis on mortality was modiied by any othe candidate variables (by adding interaction terms and main effectsas appropriate). Te validity o the proportional hazards assumptionwas checked with the Cox proportional hazards test based on weightedresiduals (using P,.01 to judge signicant violation).22Patients withmissing data on any o the covariates were omitted rom regressionanalyses.

o test robustness o results, a propensity score or ALI vs CPE diag-nosis was created accounting or all initially considered potential con-ounders and entered into regression models as a covariate. Furthersensitivity analyses were perormed comparing (1) patients with ALIexposed to a PEEP5 cm H2O within 12 to 72 h o onset o acute pul-monary edema (ie, patients meeting criteria or ARDS as per Berlindenition)2vs patients with ALI and vs patients with CPE, respectively;and (2) ALI1CPE against patients with ALI and patients with CPE,respectively.

In a post hoc analysis, the probability o death was estimated or eacho the rst 5 years afer hospital discharge or all hospital survivors(irrespective o diagnosis) and compared against the general US popu-lation (matched by age and sex) using exact binomial tests.23

Analyses were perormed in JMP (version 9.0.3; SAS Institute Inc)and R (version 3.0.3; Te R Foundation or Statistical Computing).wo-sided Pvalues,.05 were used to judge statistical signiicanceunless noted otherwise.

http://journal.publications.chestnet.org/http://journal.publications.chestnet.org/http://journal.publications.chestnet.org/


3/8

620 Original Research [ 1 4 7#3 CHEST MARCH 2015 ]

Results

Te nal cohort comprised 328 patients (ALI5155

including 25 ALI1CPE; CPE5173) (Fig 1).7Patients

with ALI were signicantly younger, sicker (higher

APACHE III score), and more likely to abuse alcohol

and to be admitted at earlier stages o the study and

had a lower BMI than patients with CPE. As expected,ALI and CPE risk actors were signicantly associated

with ALI and CPE, respectively. Further, compared

with patients with CPE, those with ALI stayed longer

in the ICU and hospital, and their hospital mortality

was higher. Data on covariates were near complete

(able 1).

wo hundred orty patients (73%) died during the

ollow-up period (median ollow-up5160 days, max-

imum ollow-up57.8 years). Long-term survival (afer

onset o acute pulmonary edema) was signicantly lower

in patients with ALI vs patients with CPE (P5 .01), butvisually this difference was ully explained by a higher

mortality rate among patients with ALI at the beginning

(approximately 50-100 days) o the ollow-up period

(Fig 2).

Hospital Mortality

During hospitalization, 95 patients (29%) with acute

pulmonary edema died. Crude hospital mortality

was signicantly higher in patients with ALI than

patients with CPE (OR5 3.7, 95% CI5 2.2-6.2,

P,.001) and did not change when adjusting orpotential conounders (OR54.2, 95% CI5 2.3-7.8,

P,.001) (able 2). Age (P5 .02) and APACHE III

score (P,.001) were independent predictors o mor-

tality, with a trend toward higher mortality in patients

who had a history o coronary artery disease (CAD)

(P5 .15) or received chemotherapy (P5 .06) in the

6 months prior (able 3).

Tere was no evidence that continuous covariates

required nonlinear modeling or that the effect o diag-

nosis on hospital mortality was modied by any o the

candidate variables. Results were similar when adjustingor potential conounding by the propensity score and

when comparing ARDS and ALI1CPE vs CPE, respec-

tively (able 4). Hospital mortality in ARDS and

ALI1CPE subgroups was higher than in other patients

with ALI, respectively, but without reaching statistical

signicance (able 4).

Long-term Survival Among Hospital Survivors

O 233 patients who survived the hospital stay, our

were immediately lost to ollow-up afer hospital dis-

charge. Tus, the long-term survival analyses amonghospital survivors were based on 229 patients (ALI588,

including 13 ALI1CPE; CPE5 141), o whom

145 patients (63.3%) died during the ollow-up period

(median ollow-up5790 days).

Ater hospital discharge, there was no dierence

between patients with ALI and patients with CPE

regarding overall survival (hazard ratio [HR] or ALI

vs CPE50.89, 95% CI50.64-1.25, P5 .50) (Fig 3).

Results did not change when adjusting or potential con-

ounders (HR or ALI vs CPE51.13, 95% CI50.79-1.62,

P5 .50) (able 2). Independently o diagnosis, risk odeath signicantly increased with age and was higher

among smokers and patients who previously received

chemotherapy, but without reaching statistical signi-

cance (able 3).

here was no evidence that continuous covariates

required nonlinear modeling, that the proportional haz-

ards assumption was violated, or that the effect o diag-

nosis on long-term survival was modied by any o the

candidate variables. Tere was no difference between

patients with ARDS or ALI1CPE compared with those

with ALI not meeting these criteria, respectively, and

results were not altered by any o the other sensitivity

analyses (able 4).

Among all hospital survivors (ie, in patients with ALI

and patients with CPE combined), median survival

time was 913 days (95% CI5641-1,198 days), and esti-

mated survival probability decreased rom 63% (95%

CI557%-69%) at 1 year to 35% (95% CI528%-41%)

at 5 years o ollow-up vs 97% and 79% in the general

population, respectively (e-Fig 1). Compared with theFigure 1 Study owchart. ALI5acute lung injury; CPE5 cardiogenic

pulmonary edema.



4/8


TABLE 1 ] Patient Characteristics

Characteristic ALI/ALI1CPE (n5155) CPE (n5173) PValue

During hospitalization (n5328)

Age, y 72 (58-81) 78 (67-86) ,.001

Female sex 47 (73) 56 (97) .11

BMI (n5327) 26 (23-31) 29 (24-33) .03

APACHE III score (n5325) 85 (68-102) 77 (65-94) .03

Year of admission .04

2006 45 (69) 36 (62)

2007 19 (29) 18 (31)

2008 24 (37) 21 (36)

2009 13 (20) 25 (44)

ALI risk factor

Sepsis or pancreatitis 52 (81) 15 (26) ,.001

Shock 35 (54) 20 (34) .002

Pneumonia 44 (68) 23 (39) ,.001

Gastric aspiration 14 (22) 7 (12) .03

Chemotherapy 10 (16) 1 (2) ,.001

CPE risk factor

History of heart failure 21 (32) 41 (70) ,.001

History of CAD 30 (46) 51 (88) ,.001

ST changes 9 (14) 22 (38) .001

Miscellaneous

Smoking (active/.20 pack-y) 46 (71) 44 (76) .73

Alcohol abuse (.2 drinks/d) 12 (19) 4 (7) .006

ARDS, Berlin definition

ALI with PEEP5 cm H2O within 12 h 61 (94) na

ALI with PEEP5 cm H2O within 72 h 80 (124) na

Outcomes

Hospital mortality 43 (66) 17 (29) ,.001

ICU length of stay, d ( n5325) 4.9 (2.1-8.3) 2.2 (1.3-4.4) ,.001

Hospital length of stay, d 11.3 (6.1-19) 8.1 (4.8-12.8) .002

After Hospitalization ( n5229) ALI/ALI1CPE (n588) CPE (n5141) PValue

Outcomes

Long-term mortality 60 (53) 65 (92) .44

Survival time, d 859 (81-1,801) 732 (127-1,677) .62

Data are presented as median (interquartile range) or % (No.). ALI5acute lung injury; APACHE5Acute Physiology and Chronic Health Evaluation;CAD5coronary artery disease; CPE5cardiogenic pulmonary edema; na5not applicable; PEEP5positive end-expiratory pressure.

general population, the rate o death in patients with

ALI/CPE was signicantly higher in the rst 2 years

afer hospital discharge but essentially converged by

5 years (able 5).

Discussion

Te main ndings o this study are that (1) patients

with ALI have a signicantly higher risk o death

during hospitalization than those with CPE, but i dis-

charged alive the long-term survival in both patient

groups is similar; (2) the same applies to the subset o

patients with ALI meeting the criteria or ARDS based

on the Berlin denition, although the difference in

hospital mortality compared with patients with CPE

may be even more pronounced; (3) compared with the

general population, patients (ALI or CPE) discharged



5/8


rom the hospital are at high risk o death during the

rst 2 years, afer which their lie expectancy starts to

assimilate; (4) there seem to be no subgroups among

patients with ALI or patients with CPE in whom

hospital/long-term mortality differs (except or pos-

sibly the ARDS and ALI1CPE subgroups among

ALI, which may have a higher hospital mortality);

(5) independent predictors o hospital mortality are

age, severity o illness, and possibly history o CAD

and recent chemotherapy, whereas age and possibly

smoking and recent chemotherapy are independently

associated with higher long-term mortality.

Te difference in hospital mortality in ALI/ARDS vs CPE

is plausible, given the differences in underlying patho-

physiology (decreased vs intact capacity to reabsorb

uid rom the alveolar space in ALI/ARDS vs CPE,

respectively).4Further, our ndings are consistent with

other studies reporting a 40% to 60% hospital/short-

term mortality in patients with ALI/ARDS14-19and

12% hospital mortality along with a 40% 1-year mor-

tality among (predominantly) patients with CPE.20

However, to our knowledge this is the rst study explicitly

comparing outcomes between these two distinct patient

groups, especially regarding long-term prognosis.

One o the strengths o this study is that patients were

exclusively rom the well-characterized Olmsted County

population.24,25Tese patients receive virtually all o

their health care at our institution (or its local outposts),

which uses a hospital-wide unied electronic medical

record system linked to our ICU database.21Tis inra-

structure allowed us to closely ollow patients over an

extensive period o time, as demonstrated by a median

(interquartile range) ollow-up among long-term survi-

vors o 5.2 (4.3-6.6) years.

One o the limitations o this study is that generaliz-

ability may be limited, since Olmsted County patients

are less ethnically diverse, wealthier, and more highly

educated than the national average.25Further, Olmsted

County patients may also have access to high-quality

health care more readily than elsewhere. Another limi-

tation o this study is that the gold standard diagnosis

had been established based on the ALI denition3beore

it was replaced by the Berlin denition o ARDS.2Te

main difference is the newly introduced requirement o

a PEEP o5 cm H2O.2However, the Berlin denition

does not speciy the time rame within which this crite-

rion has to be met. In this study, the majority o patients

with ALI ullled this criterion within 12 to 72 h, and

results were similar when comparing these subsets o

patients with ALI against CPE (with the difference in

hospital mortality possibly being even more pro-

nounced), suggesting that the results apply equally to

patients with ARDS vs patients with CPE.

Although we addressed in our analysis many potential

conounding variables (using different approaches), as

Figure 2 Kaplan-Meier plot: long-term survival in patients with ALI(solid black line) vs patients with CPE (dotted gray line) afer the onset

o acute pulmonary edema. Visually, patients with ALI died aster thanpatients with CPE (diverging survival curves) during the rst 50 to 100 d,afer which the mortality rate appeared to be similar (parallel survivalcurves).

TABLE 2 ] Risk of Death in Patients With ALI vs Patients With CPE During and After Hospitalization

Hospital Mortality

Crude Adjusted

OR (95% CI) Deaths No. PValue ORa(95% CI) Deaths No. PValue

ALI vs CPE 3.7 (2.2-6.2) 95 328 ,.001 4.2 (2.3-7.8) 92 325 ,.001

Long-term Mortality HR (95% CI) Deaths No. PValue HRb(95% CI) Deaths No. PValue

ALI vs CPE 0.89 (0.64-1.25) 145 229 .50 1.13 (0.79-1.62) 145 229 .50

HR5hazard ratio. See Table 1 legend for expansion of other abbreviations.aAdjusted for age, APACHE III score, history of coronary artery disease, and chemotherapy.bAdjusted for age, smoking, and chemotherapy.



6/8


in all observational studies we cannot exclude the pos-sibility o residual unmeasured conounding. Te main

limitation in this regard is that we could not control or

the number o previous episodes o acute pulmonary

edema. Particularly in patients with CPE, the duration

o the underlying heart ailure as indicated by the

number o previous episodes is likely an independent

risk actor or death. Given that or each enrolled

patient we only included the rst episode during the

accrual period, patients with CPE may have been on

average at an earlier stage o heart ailure, so that we

TABLE 3 ] Independent Predictors of Hospital andLong-term Mortality (in Hospital Survivors)

Predictors of Hospital

Mortalitya OR 95% CI PValue

Age, 1-y increase 1.02 1.01-1.05 .02

APACHE III score,

1-unit increase

1.04 1.02-1.05 ,.001

History of CAD 1.56 0.85-2.9 .15

Chemotherapy 3.0 0.95-9.8 .07

Predictors of Long-term

Mortalitya HR 95% CI PValue

Age, 1-y increase 1.04 1.03-1.06 ,.001

Smoking 1.38 0.98-1.9 .07

Chemotherapy 1.87 0.73-4.8 .19

See Table 1 and 2 legends for expansion of abbreviations.aAfter adjusting for ALI vs CPE diagnosis.

may have overestimated their survival time to some

extent.

Furthermore, or assessing the gold standard diagnosis

o ALI vs CPE, experts previously reviewed patients

medical records afer death or hospital discharge, also

taking into account patients clinical course (especially

response to therapy). Tereore, we cannot exclude thepossibility that experts were more likely to classiy

patients who died during hospitalization as having ALI

rather than CPE. However, assessment was based on

denitions including multiple components,7and our

results are consistent with those rom other reports,

thus making it unlikely that ascertainment bias was a

major issue.

Another limitation is that power to detect a difference in

long-term mortality between patients with ALI/ARDS

and patients with CPE may have been limited, but based

on the relatively small effect estimate (which is less

affected by sample size and event rate) it is unlikely that

there is a clinically signicant difference between these

two groups. Similarly, we did not identiy any sub-

groups o patients with ALI/ARDS and patients with

CPE in whom mortality differs, but power to detect such

interactions is generally low.26

Although some studies in the past indicated improve-

ments in incidence27and mortality19o these syndromes

over time, our results suggest that mortality in

TABLE 4 ] Sensitivity Analyses

In-Hospital Mortality Comment OR (95% CI) No. PValue

ALI vs CPE Adjusted for propensity score 3.2 (1.8-6.0) 324 ,.001

ARDS vs ALI ARDS5ALI with PEEP5 cm H2O within 12 h 1.8 (0.91-3.4) 155 .10

ARDS vs CPE ARDS5ALI with PEEP5 cm H2O within 12 h 4.6 (2.6-8.1) 267 ,.001


ARDS vs CPE ARDS5ALI with PEEP5 cm H2O within 72 h 4.1 (2.4-7) 297 ,.001

ALI1CPE vs ALI 1.30 (0.55-3.1) 155 .55

ALI1CPE vs CPE 4.6 (1.9-11) 198 .001

Long-term Mortality Comment HR (95%-CI) No. PValue

ALI vs CPE Adjusted for propensity score 0.98 (0.65-1.49) 228 .94


ARDS vs CPE ARDS5ALI with PEEP5 cm H2O within 12 h 0.86 (0.56-1.32) 190 .49


ARDS vs CPE ARDS5ALI with PEEP5 cm H2O within 72 h 0.82 (0.56-1.19) 209 .29

ALI1CPE vs ALI 0.95 (0.45-2.0) 88 .90

ALI1CPE vs CPE 0.85 (0.41-1.76) 154 .67

See Table 1 and 2 legends for expansion of abbreviations.



7/8


patients with ALI/ARDS and patients with CPE is

still high and did not change substantially over the

past decade(s).14-20Future research should, thus, con-

tinue to ocus on prevention and treatment o these

syndromes.

Figure 3 Kaplan-Meier plot: long-term survival afer hospital dischargein patients with ALI (solid black line) vs patients with CPE (dotted gray

line). Log-rank test P5 .50.

TABLE 5 ] Probability of Death During the First 5 YAfter Hospital Discharge in the Cohortof All Hospital Survivors (Observed)Compared With the US Population Matchedby Age and Sex (Expected)

During Initial ly at Riska

Probability of Death, %

PValueObserved Expected

Year 1 229 36.90 3.44 ,.001

Year 2 141 8.64 3.78 .006

Year 3 119 7.50 4.18 .13

Year 4 99 6.88 4.63 .35

Year 5 81 5.45 5.11 .80

aDuring each year some patients were lost to follow-up.

Conclusions

Although hospital mortality is higher in patients with

ALI/ARDS compared with patients with CPE, long-term

survival is not signicantly different in hospital survivorsrom both groups. Age and severity o illness are inde-

pendent predictors o hospital mortality, whereas age is

the only signicant independent predictor o long-term

mortality among hospital survivors. Patients with either

syndrome who survive the rst couple o years afer hos-

pital discharge appear to have eventually a similar long-

term prognosis as the general population.

AcknowledgmentsAuthor contributions:C. N. S. serves as theguarantor o the paper as a whole. C. N. S.contributed to designing the study, obtainingand analyzing data, and drafing and criticallyrevising the manuscript; M. B. and O. G.contributed to designing the study and criti-cally revising the manuscript; and G. A. W.contributed to obtaining the data and criti-cally revising the manuscript.

Financial/nonnancial disclosures:Teauthors have reported to CHESTthat nopotential conicts o interest exist with anycompanies/organizations whose products orservices may be discussed in this article.

Other contributions:We thank L. Renro,PhD, who provided methodologic supportduring the initial phases o the study.

Additional information:Te e-Figure canbe ound in the Supplemental Materialssection o the online article.

Reerences1. Vincent JL, Aka S, De Mendona A,

et al; SOFA Working Group. Sequentialorgan ailure assessment. Te epidemi-ology o acute respiratory ailure in criti-cally ill patients. Chest. 2002;121(5):1602-1609.

2. Ranieri VM, Rubeneld GD, TompsonB, et al; ARDS Denition ask Force.

Acute respiratory distress syndrome: theBerlin Denition.JAMA. 2012;307(23):

2526-2533.3. Bernard GR, Artigas A, Brigham KL,

et al. Te American-European ConsensusConerence on ARDS. Denitions, mech-anisms, relevant outcomes, and clinicaltrial coordination.Am J Respir Crit CareMed. 1994;149(3 pt 1):818-824.

4. Ware LB, Matthay MA. Clinical practice.Acute pulmonary edema. N Engl J Med.2005;353(26):2788-2796.

5. Rana R, Vlahakis NE, Daniels CE, et al.B-type natriuretic peptide in the assess-ment o acute lung injury and cardio-genic pulmonary edema. Crit Care Med.2006;34(7):1941-1946.

6. Ware LB, Fremont RD, Bastarache JA,

Calee CS, Matthay MA. Determiningthe aetiology o pulmonary oedema bythe oedema uid-to-plasma protein ratio.Eur Respir J. 2010;35(2):331-337.

7. Schmickl CN, Shahjehan K, Li G, et al.Decision support tool or early differ-ential diagnosis o acute lung injuryand cardiogenic pulmonary edema inmedical critically ill patients. Chest.2012;141(1):43-50.

8. Lin Q, Fu F, Chen H, Zhu B. Copeptin inthe assessment o acute lung injury andcardiogenic pulmonary edema. RespirMed. 2012;106(9):1268-1277.

9. Komiya K, Ishii H, eramoto S, et al.Diagnostic utility o C-reactive protein

combined with brain natriuretic peptidein acute pulmonary edema: a cross sec-

tional study. Respir Res. 2011;12:83.10. Herasevich V, Yilmaz M, Khan H,

Hubmayr RD, Gajic O. Validation oan electronic surveillance system oracute lung injury. Intensive Care Med.2009;35(6):1018-1023.

11. Gurin C, Reignier J, Richard JC, et al;PROSEVA Study Group. Prone positioningin severe acute respiratory distresssyndrome. N Engl J Med. 2013;368(23):2159-2168.

12. Papazian L, Forel JM, Gacouin A, et al;ACURASYS Study Investigators.Neuromuscular blockers in early acuterespiratory distress syndrome. N EnglJ Med. 2010;363(12):1107-1116.

13. Amato MB, Barbas CS, Medeiros DM,et al. Effect o a protective-ventilationstrategy on mortality in the acute respi-ratory distress syndrome. N Engl J Med.1998;338(6):347-354.

14. Rubeneld GD, Caldwell E, Peabody E,et al. Incidence and outcomes o acutelung injury. N Engl J Med. 2005;353(16):1685-1693.

15. Zilberberg MD, Epstein SK. Acute lunginjury in the medical ICU: comorbidconditions, age, etiology, and hospitaloutcome.Am J Respir Crit Care Med.1998;157(4 pt 1):1159-1164.

16. Doyle RL, Szaarski N, Modin GW,Wiener-Kronish JP, Matthay MA.



8/8


Identication o patients with acute lunginjury. Predictors o mortality.Am J RespirCrit Care Med. 1995;152(6 pt 1):1818-1824.

17. Sloane PJ, Gee MH, Gottlieb JE, et al.A multicenter registry o patients withacute respiratory distress syndrome.Physiology and outcome.Am Rev RespirDis. 1992;146(2):419-426.

18. Suchyta MR, Clemmer P, Elliott CG,

Orme JF Jr, Weaver LK. Te adultrespiratory distress syndrome. A reporto survival and modiying actors. Chest.1992;101(4):1074-1079.

19. Milberg JA, Davis DR, Steinberg KP,Hudson LD. Improved survival opatients with acute respiratory distresssyndrome (ARDS): 1983-1993.JAMA.1995;273(4):306-309.

20. Roguin A, Behar D, Ben Ami H, et al.Long-term prognosis o acute pulmo-

nary oedemaan ominous outcome.Eur J Heart Fail. 2000;2(2):137-144.

21. Herasevich V, Pickering BW, Dong Y,Peters SG, Gajic O. Inormatics inra-structure or syndrome surveillance,decision support, reporting, and mod-eling o critical illness.Mayo Clin Proc.2010;85(3):247-254.

22. Grambsch P, Terneau . Proportional

hazards tests and diagnostics basedon weighted residuals. Biometrika.1994;81(3):515-526.

23. Arias E. United States Lie Tables,2009. National Vital Statistics Reports.Hyattsville, MD: US Department oHealth and Human Services, Centers orDisease Control and Prevention NationalCenter or Health Statistics; 2014.

24. St Sauver JL, Grossardt BR, Yawn BP,Melton LJ III, Rocca WA. Use o a med-

ical records linkage system to enumeratea dynamic population over time: theRochester epidemiology project.Am JEpidemiol. 2011;173(9):1059-1068.

25. St Sauver JL, Grossardt BR, Leibson CL,Yawn BP, Melton LJ III, Rocca WA.Generalizability o epidemiologicalndings and public health decisions: anillustration rom the Rochester Epide-miology Project.Mayo Clin Proc. 2012;87(2):151-160.

26. Marshall SW. Power or tests o interac-tion: effect o raising the ype I error rate.Epidemiol Perspect Innov. 2007;4:4.

27. Li G, Malinchoc M, Cartin-Ceba R,et al. Eight-year trend o acute respiratorydistress syndrome: a population-basedstudy in Olmsted County, Minnesota.Am J Respir Crit Care Med. 2011;183(1):59-66.

comparison of hospital mortality and long-term

Documents