Beruflich Dokumente
Kultur Dokumente
Background and PurposeTo develop and validate a risk score (acute ischemic stroke-associated pneumonia score [AIS-
APS]) for predicting in-hospital stroke-associated pneumonia (SAP) after AIS.
MethodsThe AIS-APS was developed based on the China National Stroke Registry, in which eligible patients were
randomly classified into derivation (60%) and internal validation cohort (40%). External validation was performed using
the prospective Chinese Intracranial Atherosclerosis Study. Independent predictors of in-hospital SAP after AIS were
obtained using multivariable logistic regression, and -coefficients were used to generate point scoring system of the
AIS-APS. The area under the receiver operating characteristic curve and the HosmerLemeshow goodness-of-fit test
Downloaded from http://stroke.ahajournals.org/ by guest on September 11, 2017
Key Words: acute ischemic stroke China National Stroke Registry stroke-associated pneumonia
1
2StrokeMay 2013
hospitals in China are classified into 3 grades: I (community hospitals); Statistical Analysis
II (hospitals that serve several communities); and III (central hospitals Model building was performed exclusively in the derivation cohort.
for a certain district or city). In total, 242 potential sites, including 114 In univariate analysis, 2 test was used to compare categorical vari-
grade III, 71 grade II, and 57 grade I hospitals, from both urban and ables, and MannWhitney test was used to compare continuous vari-
rural area, were initially identified by soliciting application. The CNSR ables. Univariate and multivariable logistic regression was performed
steering committee evaluated the research capability and commitment to determine the independent predictors of SAP after AIS in the
to the registry of each hospital with preliminary survey. Finally, a total derivation cohort. Candidate variables were those with biologically
of 132 hospitals including 100 grade III and 32 grade II were selected, plausible link to SAP on the basis of prior publication and those as-
which cover 27 provinces and 4 municipalities across China. Trained sociated with SAP in univariate analysis (P<0.2). On multivariable
research coordinators at each institute reviewed medical records daily analysis, stepwise backward estimation was used to remove nonsig-
to identify, consent, and enroll consecutively eligible patients. To be nificant variables from the model. To test for collinearity between the
eligible for this study, subjects had to meet the following criteria: covariates of the final multivariable model, the tolerance and variance
(1) age 18 years; (2) hospitalized with a primary diagnosis of AIS inflation factor of each covariate was calculated. The -coefficients
according to World Health Organization criteria30; (3) stroke confirmed from the final model were used to generate point scoring system of
by head computerized tomography or brain MRI; (4) direct admission the AIS-APS, as in previous studies.34 The resulting AIS-APS was
to hospital from a physicians clinic or emergency department. The then validated by assessing model discrimination and calibration,35
eligible patients from the CNSR were randomly classified into in the internal and external validation cohorts. Discrimination was
derivation cohort (60%) and internal validation cohort (40%). assessed by calculating the area under the receiver operating charac-
teristic curve (AUROC). Calibration was assessed by performing the
External Validation Cohort HosmerLemeshow goodness-of-fit test and was graphically depict-
The external validation cohort was derived from the cohort of ed in the plot of observed versus predicted SAP risk according to 10
Downloaded from http://stroke.ahajournals.org/ by guest on September 11, 2017
Chinese Intracranial Atherosclerosis Study (CICAS). CICAS was a deciles of predicted risk. Furthermore, we compared the discrimina-
multicenter and prospective study aiming at investigating prevalence, tion of the AIS-APS and 3 prior scores,2527 with regard to in-hospital
risk factors, and impact of intracranial atherosclerosis among patients SAP after AIS. The AUROC and maximum Youden index were used
with ischemic stroke.31 The CICAS inclusion criteria were as follows: to assess the discrimination of these scores for in-hospital SAP after
(1) age between 18 and 80 years; (2) onset of symptom within 7 days; AIS. AUROC was compared using Delong method,36 and sensitivity,
and (3) hospitalized with a primary diagnosis of AIS or transient specificity, positive predict value, and negative predictive value were
ischemic attack. Exclusion criteria were as follows: (1) preadmission calculated at each scores maximum Youden index.
modified Rankin Scale score >3; and (2) inability to undergo MRI All tests were 2-tailed, and statistical significance was determined
for cerebral vascular imaging. For this study, patients diagnosed of at level of 0.05. Statistical analysis was performed using SAS 9.1
transient ischemic attack were excluded. (SAS Institute, Cary, NC), SPSS 17.0 (SPSS Inc, Chicago, IL), and
Medcalc software 12.3 (MedCalc, Ostend, Belgium).
Data Collection and Definitions
The scientific use of data obtained with informed consent and entered Results
in the CNSR and CICAS registries was approved by the local ethi-
cal committees. In these registries, a standardized case report form Clinical Characteristics
was used for data collection in the CNSR and CICAS network. The The clinical characteristics of the derivation, internal and
relevant data were extracted from the medical records by trained re- external validation cohort were shown in Table 1. From
search coordinators. Data from each case report form were manually
checked for completeness, correct coding, and proper application of
September 2007 to August 2008, a total of 14702 patients
diagnostic algorithm by a research specialist from an independent with AIS were enrolled in the CNSR network. The median age
contract research organization. For the present study, the following was 66 (interquartile range [IQR], 5675), and 61.9% were
candidate variables were analyzed: (1) demographics (age and sex); men. The median length of hospital stay was 14 days (IQR,
(2) stroke risk factors: hypertension (history of hypertension or anti- 1020), and a total of 1669 (11.4%) patients had SAP dur-
hypertensive medication use), diabetes mellitus (history of diabetes
mellitus or antidiabetic medication use), dyslipidemia (history of dys- ing hospitalization. The eligible patients from the CNSR were
lipidemia or lipid-lowering medication use), atrial fibrillation (history randomly classified into derivation cohort (60%, n=8820) and
of atrial fibrillation or documentation of atrial fibrillation at admis- internal validation cohort (40%, n=5882), which were well
sion), coronary heart disease, history of stroke/transient ischemic at- matched with respect to patient characteristics and overall rate
tack, current smoking, and excess alcohol consumption (2 standard of in-hospital SAP (Table 1).
alcohol beverages per day); (3) pre-existing comorbidities: congestive
heart failure, valvular heart disease, peripheral artery disease, chronic From October 2007 to June 2009, a total of 3580 patients of
obstructive pulmonary disease, hepatic cirrhosis, peptic ulcer, pre- AIS were registered in 22 participating hospitals in the CICAS
vious gastrointestinal bleeding, renal failure, arthritis, Alzheimers network. Of these, we excluded 329 patients with transient
disease/dementia, and cancer; (4) prestroke dependence (modified ischemic attack (9.1%; median age, IQR: 62, 5471; median
Rankin Scale 3); (5) admission stroke severity based on National
admission NIHSS, IQR: 0, 01) and an additional 214 (6.0%;
Institutes of Health Stroke Scale score (NIHSS) and Glasgow Coma
Scale score; (7) symptom of dysphagia; (8) stroke subtype: according median age, IQR: 64, 5573; median admission NIHSS, IQR:
to the Oxfordshire Community Stroke Project criteria,32 AIS was clas- 3, 06) who had missing data of 1 covariates in the AIS-
sified into partial anterior circulation infarct, total anterior circulation APS. The median length of hospital stay was 14 days (IQR,
infarct, lacunar infarction, and posterior circulation infarct; (9) admis- 1018), and a total of 222 (7.3%) patients had SAP during
sion blood glucose (mmol/L); and (10) length of hospital stay (days).
In this study, SAP was diagnosed by treating physician accord-
hospitalization (Table 1).
ing to the Centers for Disease Control and Prevention criteria for
hospital-acquired pneumonia,33 on a basis of clinical and laboratory Predictors of In-hospital SAP
indices of respiratory tract infection (fever, cough, auscultatory re- The univariate analysis for potential predictors of in-hospital
spiratory crackles, new purulent sputum, or positive sputum culture),
and supported by typical chest X-ray findings. Only hospital-acquired SAP after AIS in the derivation cohort was shown in Table
pneumonia was documented and pneumonia before stroke was not I in the online-only Data Supplement, and the multivariable
considered. Data on in-hospital SAP was prospectively collected. predictors were listed in Table 2. Age, present history of
Ji et al A Risk Score to Predict SAP After AIS 3
Congestive heart failure 169 (1.9) 121 (2.1) 0.55 25 (0.8) <0.001
Valvular heart disease 213 (2.4) 139 (2.4) 0.83 38 (1.3) <0.001
Peripheral artery disease 64 (0.7) 29 (0.5) 0.08 26 (0.9) 0.48
COPD 98 (1.1) 64 (1.1) 0.89 13 (0.4) 0.001
Hepatic cirrhosis 29 (0.3) 21 (0.4) 0.78 7 (0.2) 0.40
Peptic ulcer or previous GIB 283 (3.2) 195 (3.3) 0.72 76 (2.5) 0.05
Renal failure 7 (0.1) 4 (0.1) 0.78 4 (0.1) 0.50
Arthritis 266 (3.0) 176 (3.0) <0.001 46 (1.5) <0.001
Dementia 113 (1.3) 82 (1.4) 0.56 22 (0.7) 0.01
Cancer 150 (1.7) 109 (1.9) 0.50 52 (1.7) 1.00
Prestroke dependence (mRS>3), n, % 809 (9.2) 535 (9.1) 0.87 0 (0.0)
Admission NIHSS score, median (IQR) 5 (29) 5 (29) 0.18 4 (27) <0.001
Admission GCS score, median (IQR) 15 (1415) 15 (1415) 0.36 15 (1515) 0.12
Symptom of dysphagia, n, % 840 (9.5) 529 (9.0) 0.28 236 (7.8) 0.004
OCSP subtype, n, % 0.22 <0.001
Partial anterior circulation infarct 4834 (54.8) 3327 (56.6) 1869 (61.5)
Total anterior circulation infarct 811 (9.2) 519 (8.8) 177 (5.8)
Lacunar infarction 1667 (18.9) 1074 (18.3) 249 (8.2)
Posterior circulation infarct 1508 (17.1) 962 (18.4) 742 (24.4)
Admission blood glucose, median (IQR), 6.2 (5.47.0) 6.2 (5.57.0) 0.28 5.5 (4.96.9) <0.001
mmol/L
Length of hospital stay, median (IQR), d 14 (1020) 14 (1020) 0.96 15 (1220) <0.001
In-hospital SAP, n, % 1007 (11.4) 662 (11.3) 0.76 222 (7.3) <0.001
COPD indicates chronic obstructive pulmonary disease; GCS, Glasgow Coma Scale; GIB, gastrointestinal bleeding; IQR, interquartile range; mRS, modified Rankin
Scale; NIHSS, National Institutes of Health Stroke Scale; OCSP, Oxfordshire Community Stroke Project; SAP, stroke-associated pneumonia; and TIA, transient ischemic
attack.
P 1 denotes significant test between the derivation and internal validation cohort; P2 denotes significant test between the derivation and external validation cohort.
atrial fibrillation, congestive heart failure, chronic obstructive Derivation of the AIS-APS
pulmonary disease, and current smoking, prestroke The -coefficients from the multivariable regression model
dependence (modified Rankin Scale>3), admission NIHSS were used to generate point scoring system of the AIS-APS.
score, Glasgow Coma Scale score, symptom of dysphagia, To derive an integer value for each predictor, the -coefficient
Oxfordshire Community Stroke Project subtype of total of current smoking was used as reference and the value was
anterior circulation infarct and posterior circulation infarct, rounded to the closest integer. The point scoring system of the
and blood glucose were significantly predictive of in-hospital AIS-APS was shown in Figure 1. The median AIS-APS score
SAP after AIS. The tolerance of covariates in the final was 8 (range, 032) in the derivation cohort. The risk catego-
multivariable model ranged between 0.66 and 0.99; the mean ries were assigned in 7-point increments, and the magnitude
variance inflation factor was 1.16 (range, 1.021.68). of the score had prognostic implication (Figure 2).
4StrokeMay 2013
Table 2. Multivariable Predictors of In-hospital SAP After AIS in the Derivation Cohort (n=8820)
Variables -Coefficients SE Adjusted OR* 95% CI P Value
Model intercept 6.873
Age (1 yr increase) 0.049 0.003 1.05 1.041.06 <0.001
Atrial fibrillation (yes) 0.199 0.098 1.22 1.031.50 0.03
Congestive heart failure (yes) 0.764 0.162 2.15 1.562.95 <0.001
COPD (yes) 0.510 0.217 1.67 1.092.55 0.02
Current smoking (yes) 0.238 0.063 1.27 1.121.44 <0.001
Prestroke dependence (mRS>3) (yes) 0.313 0.086 1.37 1.161.62 <0.001
Admission NIHSS score (1 increase) 0.077 0.005 1.08 1.071.09 <0.001
Admission GCS score (1 decrease) 0.043 0.015 1.04 1.011.08 0.005
Dysphagia (yes) 0.642 0.083 1.90 1.622.24 <0.001
OCSP subtype
Lacunar infarction) 1.00
Partial anterior circulation infarct 0.001 0.088 1.00 0.831.19 0.99
Downloaded from http://stroke.ahajournals.org/ by guest on September 11, 2017
Internal Validation of the AIS-APS and external validation cohort, the AIS-APS demonstrated
The performance of the AIS-APS (AUROC) in the deriva- the highest AUROC. The difference in AUROC between AIS-
tion (n=8820) and internal validation cohort (n=5882) was APS and 3 compared scores was statistically significant (all
0.797 (95% confidence interval, 0.7820.811) and 0.785 P<0.0001). The AIS-APS had the highest maximum Youden
(95% confidence interval, 0.7660.803), respectively. The index and associated sensitivity, specificity, positive predict
HosmerLemeshow test was not significant (P=0.22), and the value and negative predictive value (Table 3).
predicted and observed risk of in-hospital SAP after AIS were
in close agreement according to 10 deciles of predicted risk in
Discussion
the internal validation cohort (Figure I in the online-only Data
In the present study, we derived and validated a risk score for
Supplement).
predicting in-hospital SAP after AIS. A 34-point AIS-APS
was developed from the set of independent predictors. The
External Validation of the AIS-APS AIS-APS showed good discrimination and calibration in the
In the external validation cohort (n=3037), the AIS-APS derivation, internal and external validation cohort. When com-
showed good discrimination with an AUROC of 0.792 (95% pared with 3 prior scores, the AIS-APS showed significantly
confidence interval, 0.7610.823). The HosmerLemeshow better discrimination with regard to in-hospital SAP after AIS.
test was not significant (P=0.30). The plot of observed To preserve the clinical use of the model for decision-mak-
versus predicted risk of in-hospital SAP after AIS showed ing during acute hospitalization, we used only patient char-
high correlation between observed and predicted risk in the acteristics available at presentation. We chose not to include
external validation cohort (Figure I in the online-only Data variables related to in-hospital management and those not
Supplement). routinely collected, such as quality of care,37 mechanical ven-
tilation, and swallowing test,11,38 despite the fact that these fac-
Sensitivity Analysis tors might influence the development of SAP after AIS. This
We completed prespecified subgroup analyses by age, sex, model therefore predicts the expected risk of in-hospital SAP
time delay from onset to hospital arrival, and length of hospital at presentation, and as such, the predictions could be used to
stay. Similar good discrimination was seen in these subgroups guide subsequent in-hospital management.
(AUROC range, 0.7400.837) (Table III in the online-only Several risk factors for SAP have been identified. Consistent
Data Supplement). with these studies, we confirmed that in-hospital SAP was sig-
nificantly associated with older age, present history of atrial
Comparison of SAP Scores fibrillation, congestive heart failure, chronic obstructive pul-
Table 3 showed the discrimination of the AIS-APS and 3 com- monary disease and smoking, prestroke dependence, admis-
pared prior scores for in-hospital SAP after AIS. Although sion NIHSS score, Glasgow Coma Scale score, dysphagia, and
Chumbler and Hoffmann scores consistently showed good stroke subtypes. In addition, our study showed that admission
discrimination for in-hospital SAP in the derivation, internal blood glucose was significantly associated with in-hospital
Ji et al A Risk Score to Predict SAP After AIS 5
Figure 2. The proportion of in-hospital stroke-associated pneumonia (SAP) after acute ischemic stroke (AIS) according to the AISpneu-
monia score (PS) in the derivation, internal and external validation cohort. The potential risk of in-hospital SAP after AIS increased steadily
with higher AIS-PS score. Error bars indicated 95% confidence interval for the proportion of in-hospital SAP after AIS in each category.
6StrokeMay 2013
Table 3. Discrimination of AIS-APS and 3 Compared Prior Scores With Regard to In-hospital SAP After AIS
AUROC 95% CI AUROC* P Value Youden Index Cutoff Sensitivity Specificity PPV NPV
In the derivation cohort (n=8850)
Pneumonia score (Kwon et al. 2006)27 0.713 0.7040.723 0.084 <0.0001 0.342 2 0.762 0.580 0.189 0.950
Chumblers score (Chumbler et al. 2010)25 0.752 0.7430.761 0.045 <0.0001 0.401 13 0.605 0.796 0.276 0.940
A2DS2 score (Hoffmann, et al. 2012)26 0.745 0.7360.754 0.052 <0.0001 0.373 5 0.545 0.828 0.291 0.934
AIS-APS score (2012) 0.797 0.7820.811 Ref 0.435 8 0.793 0.769 0.306 0.967
In the internal validation cohort (n=5882)
Pneumonia score (Kwon et al. 2006)27 0.694 0.6820.706 0.091 <0.0001 0.313 2 0.730 0.583 0.182 0.945
Chumblers score (Chumbler et al. 2010)25 0.737 0.7260.749 0.048 <0.0001 0.393 14 0.577 0.816 0.285 0.938
A DS score (Hoffmann et al. 2012)
2 2 26
0.728 0.7160.739 0.057 <0.0001 0.364 5 0.527 0.837 0.290 0.933
AIS-APS (2012) 0.785 0.7660.803 Ref 0.435 8 0.781 0.764 0.300 0.965
In the external validation cohort-A (n=3037)
Pneumonia score (Kwon et al. 2006)27 0.676 0.6590.693 0.116 <0.0001 0.294 2 0.662 0.632 0.124 0.960
Chumblers score (Chumbler et al. 2010) 25
0.752 0.7260.787 0.040 <0.0001 0.418 11 0.658 0.760 0.178 0.966
Downloaded from http://stroke.ahajournals.org/ by guest on September 11, 2017
A2DS2 score (Hoffmann et al. 2012)26 0.759 0.7440.774 0.033 <0.0001 0.392 5 0.725 0.667 0.146 0.969
AIS-APS score (2012) 0.792 0.7610.823 Ref 0.530 8 0.728 0.802 0.225 0.973
AIS-APS indicates acute ischemic stroke-associated pneumonia score; AUROC, area under the receiver operating characteristic curve; CI, confidence interval; PPV,
positive predictive value; NPV, negative predictive value; and SAP, stroke-associated pneumonia.
*AUROC denotes the difference in AUROC between the AIS-APS and compared scores with regard to in-hospital SAP after AIS.
P value of comparing paired AUROC with Delong method.
a score (A2DS2) to predict pneumonia after AIS in a large Our study had some limitations that deserve comment. First,
derivation and validation sample. The model showed good as all observational studies, we cannot rule out the possibility
discrimination and calibration properties; however, it was not that additional baseline variable (unmeasured confounders)
externally validated in geographically, culturally, and socio- might have some impact on the development of in-hospital
economically distinct population. SAP after AIS, such as use of angiotensin-converting enzyme
With several SAP-related risk stratification and prognos- inhibitors or angiotensin receptor blockers.44 Second, the time
tic models available, identification of the most accurate and course of SAP after AIS is unclear. Because we only have
reliable grading scale(s) would be of great value to patients, information on new-onset SAP during hospitalization without
clinicians, and researchers. In this study, we compared the dis- documentation of the exact date, our data allow no conclu-
crimination of the AIS-APS and 3 prior scores with regard to sion as to whether patients with a longer length of stay per se
in-hospital SAP after AIS. When using 2 measures to assess are more likely to develop pneumonia or whether diagnosis of
model discrimination (AUROC and maximum Youden index), pneumonia leads to a longer hospitalization. However, by sen-
the AIS-APS consistently showed to be superior at predicting sitive analysis, the AIS-APS was proven to be valid and signif-
in-hospital SAP after AIS than 3 compared scores in our popu- icant, regardless of length of hospital stay. Third, the limited
lation. In addition, similar results were verified in the deriva- information on time course of pneumonia also does not allow
tion, internal and external validation cohort. It was noteworthy for a safe interpretation of causality between mechanical ven-
that all scales had higher negative predictive value than posi- tilation and pneumonia. We therefore abstained from includ-
tive predict value for in-hospital SAP after AIS, which sug- ing mechanical ventilation in our predictive score. In addition,
gested that lower values more consistently predict patients the primary aim of our study was to develop a predictive score
without in-hospital SAP than higher values predicting those for SAP after AIS at presentation. Fourth, the study included
developing in-hospital SAP after AIS. Development of future only hospitalized patients with stroke, and those patients died
prognostic models might benefit from attempts to make them in the emergency department, shortly after admission, or
more balanced in this regard, with discriminative use distrib- treated in outpatient clinics were not included. Meanwhile,
uted more evenly among higher and lower values. the results of current analysis apply only to patients of AIS
Prior studies have shown that SAP was an important risk and cannot be extrapolated to patients of hemorrhagic stroke.
factor of mortality and morbidity after stroke; however, a sys- Finally, the AIS-APS need to be further validated in additional
tematic review on efficacy of early antibiotics prophylaxis populations.
after stroke failed to show benefit in patients outcome.43 This In summary, the AIS-APS is a valid clinical grading scale for
might be attributable to inclusion of patients with low risk of predicting in-hospital SAP after AIS at presentation. Further
developing SAP in these studies. The AIS-APS and other SAP studies on effect of AIS-APS on stroke outcomes are needed.
predictive tools could be used to identify patients who are at
high risk of developing SAP after stroke. It is expected that Sources of Funding
randomized controlled trials on efficacy of antibiotics pro- The China National Stroke Registry is funded by the Ministry of
phylaxis on stroke functional outcome with stratification of Science and Technology (2006BA101A11) and the Ministry of
patients potential risk of developing SAP. Health of the Peoples Republic of China (2009CB521905).
Ji et al A Risk Score to Predict SAP After AIS 7
tors of early infection after an acute ischaemic stroke. Singapore Med J. Escalating levels of access to in-hospital care and stroke mortality.
2003;44:344346. Stroke. 2008;39:25222530.
8. Aslanyan S, Weir CJ, Diener HC, Kaste M, Lees KR; GAIN International 29. Wang Y, Cui L, Ji X, Dong Q, Zeng J, Wang Y, et al; China National
Steering Committee and Investigators. Pneumonia and urinary tract Stroke Registry Investigators. The China National Stroke Registry for
infection after acute ischaemic stroke: a tertiary analysis of the GAIN patients with acute cerebrovascular events: design, rationale, and base-
International trial. Eur J Neurol. 2004;11:4953. line patient characteristics. Int J Stroke. 2011;6:355361.
9. Ovbiagele B, Hills NK, Saver JL, Johnston SC; California Acute Stroke 30. Stroke--1989. Recommendations on stroke prevention, diagnosis, and
Prototype Registry Investigators. Frequency and determinants of pneu- therapy. Report of the WHO task force on stroke and other cerebrovascu-
monia and urinary tract infection during stroke hospitalization. J Stroke lar disorders. Stroke. 1989;20:14071431
Cerebrovasc Dis. 2006;15:209213. 31. Wang Y, Liu, L, Wong Y, Soo Y, Pu Y, Wong KL. The Chinese IntraCranial
10. Vargas M, Horcajada JP, Obach V, Revilla M, Cervera A, Torres F, et AtheroSclerosis (CICAS) Study Group. A multicenter study of the prev-
al. Clinical consequences of infection in patients with acute stroke: is it alence and outcomes of intracranial large artery atherosclerosis among
prime time for further antibiotic trials? Stroke. 2006;37:461465. stroke and TIA patients in China. Stroke. 2012;43:A120
11. Sellars C, Bowie L, Bagg J, Sweeney MP, Miller H, Tilston J, et al. Risk 32. Bamford J, Sandercock P, Dennis M, Burn J, Warlow C. Classification
factors for chest infection in acute stroke: a prospective cohort study. and natural history of clinically identifiable subtypes of cerebral infarc-
Stroke. 2007;38:22842291. tion. Lancet. 1991;337:15211526.
12. Walter U, Knoblich R, Steinhagen V, Donat M, Benecke R, Kloth A. 33. Garner JS, Jarvis WR, Emori TG, Horan TC, Hughes JM. CDC definitions
Predictors of pneumonia in acute stroke patients admitted to a neurologi- for nosocomial infections, 1988. Am J Infect Control. 1988;16:128140.
cal intensive care unit. J Neurol. 2007;254:13231329. 34. Sullivan LM, Massaro JM, DAgostino RB Sr. Presentation of multivari-
13. Vermeij FH, Scholte op Reimer WJ, de Man P, van Oostenbrugge RJ, ate data for clinical use: the Framingham Study risk score functions. Stat
Franke CL, de Jong G, et al; Netherlands Stroke Survey Investigators. Med. 2004;23:16311660.
Stroke-associated infection is an independent risk factor for poor out- 35. Cook NR. Use and misuse of the receiver operating characteristic curve
come after acute ischemic stroke: data from the Netherlands Stroke in risk prediction. Circulation. 2007;115:928935.
Survey. Cerebrovasc Dis. 2009;27:465471. 36. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas
14. Kumar S, Selim MH, Caplan LR. Medical complications after stroke. under two or more correlated receiver operating characteristic curves: a
Lancet Neurol. 2010;9:105118. nonparametric approach. Biometrics. 1988;44:837845.
15. Westendorp WF, Nederkoorn PJ, Vermeij JD, Dijkgraaf MG, van de 37. Ingeman A, Andersen G, Hundborg HH, Svendsen ML, Johnsen SP.
Beek D. Post-stroke infection: a systematic review and meta-analysis. Processes of care and medical complications in patients with stroke.
BMC Neurol. 2011;11:110. Stroke. 2011;42:167172.
16. Wang PL, Zhao XQ, Yang ZH, Wang AX, Wang CX, Liu LP, et al. Effect 38. Mann G, Hankey GJ, Cameron D. Swallowing function after stroke:
of in-hospital medical complications on case fatality post-acute isch- prognosis and prognostic factors at 6 months. Stroke. 1999;30:744748.
emic stroke: data from the China National Stroke Registry. Chin Med J. 39. Jeon CY, Furuya EY, Berman MF, Larson EL. The role of pre-operative
2012;125:24492454. and post-operative glucose control in surgical-site infections and mortal-
17. Ingeman A, Andersen G, Hundborg HH, Svendsen ML, Johnsen SP. ity. PLoS ONE. 2012;7:e45616.
In-hospital medical complications, length of stay, and mortality among 40. Ata A, Lee J, Bestle SL, Desemone J, Stain SC. Postoperative hypergly-
stroke unit patients. Stroke. 2011;42:32143218. cemia and surgical site infection in general surgery patients. Arch Surg.
18. Katzan IL, Dawson NV, Thomas CL, Votruba ME, Cebul RD. The cost 2010;145:858864.
of pneumonia after acute stroke. Neurology. 2007;68:19381943. 41. Delamaire M, Maugendre D, Moreno M, Le Goff MC, Allannic H,
19. Saposnik G, Hill MD, ODonnell M, Fang J, Hachinski V, Kapral MK; Genetet B. Impaired leucocyte functions in diabetic patients. Diabet
Registry of the Canadian Stroke Network for the Stroke Outcome Med. 1997;14:2934.
Research Canada (SORCan) Working Group. Variables associated 42. Rayfield EJ, Ault MJ, Keusch GT, Brothers MJ, Nechemias C, Smith
with 7-day, 30-day, and 1-year fatality after ischemic stroke. Stroke. H. Infection and diabetes: the case for glucose control. Am J Med.
2008;39:23182324. 1982;72:439450.
20. Vernino S, Brown RD Jr, Sejvar JJ, Sicks JD, Petty GW, OFallon WM. 43. van de Beek D, Wijdicks EF, Vermeij FH, de Haan RJ, Prins JM, Spanjaard
Cause-specific mortality after first cerebral infarction: a population- L, et al. Preventive antibiotics for infections in acute stroke: A systematic
based study. Stroke. 2003;34:18281832. review and meta-analysis. Archives of neurology. 2009;66:10761081
21. Kwan J, Hand P. Infection after acute stroke is associated with poor 44. Caldeira D, Alarco J, Vaz-Carneiro A, Costa J. Risk of pneumonia asso-
short-term outcome. Acta Neurol Scand. 2007;115:331338. ciated with use of angiotensin converting enzyme inhibitors and angio-
22. Heuschmann PU, Kolominsky-Rabas PL, Misselwitz B, Hermanek P, tensin receptor blockers: systematic review and meta-analysis. BMJ.
Leffmann C, Janzen RW, et al; German Stroke Registers Study Group. 2012;345:e4260.
Novel Risk Score to Predict Pneumonia After Acute Ischemic Stroke
Ruijun Ji, Haipeng Shen, Yuesong Pan, Panglian Wang, Gaifen Liu, Yilong Wang, Hao Li and
Yongjun Wang
Downloaded from http://stroke.ahajournals.org/ by guest on September 11, 2017
The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://stroke.ahajournals.org/content/early/2013/03/08/STROKEAHA.111.000598
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Stroke can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office.
Once the online version of the published article for which permission is being requested is located, click
Request Permissions in the middle column of the Web page under Services. Further information about this
process is available in the Permissions and Rights Question and Answer document.
Table e-1. Univariable predictor of SAP after AIS in the derivation cohort (n=8820)
Demographics
Age, year 1 year increase 1.05 1.05-1.07 <0.001
Gender female vs. male 1.18 1.06-1.31 0.002
Vascular risk factor
Hypertension yes vs. no 1.12 1.00-1.25 0.04
Diabetes mellitus yes vs. no 1.09 0.96-1.23 0.19
Dyslipidemia yes vs. no 0.76 0.64-0.91 0.003
Atrial fibrillation yes vs. no 2.91 2.51-3.38 <0.001
Coronary artery disease yes vs. no 1.73 1.53-1.98 <0.001
History of stroke/TIA yes vs. no 1.40 1.26-1.56 <0.001
Current smoking yes vs. no 0.88 0.79-0.98 0.02
Excess alcohol yes vs. no 0.84 0.73-0.98 0.02
Preexisting comorbidities
Congestive heart failure yes vs. no 3.84 2.98-4.95 <0.001
Valvular heart disease yes vs. no 2.29 1.77-2.96 <0.001
Peripheral artery disease yes vs. no 1.27 0.71-2.29 0.42
COPD yes vs. no 3.46 2.46-4.86 <0.001
Hepatic cirrhosis yes vs. no 1.72 0.83-3.54 0.14
Peptic ulcer or previous GIB yes vs. no 1.04 0.78-1.38 0.80
Renal failure yes vs. no 4.47 1.31-15.3 0.02
Arthritis yes vs. no 1.38 1.06-1.81 0.02
Alzheimers disease/dementia yes vs. no 3.14 2.29-4.31 <0.001
Cancer yes vs. no 1.75 1.27-2.41 0.001
Pre-stroke dependence (mRS>3) yes vs. no 2.56 2.22-2.93 <0.001
Admission NIHSS score 1 point increase 1.11 1.10-1.12 <0.001
Admission GCS score 1 point decrease 1.28 1.26-1.31 <0.001
Symptom of dysphagia yes vs. no 2.85 2.46-3.26 <0.001
OCSP subtype, n (%)
Lacunar infarction (LACI) 1.00
Partial anterior circulation infarct (PACI) PACI vs. LACI 1.16 1.00-1.35 0.05
Total anterior circulation infarct (TACI) TACI vs. LACI 3.14 2.61-3.78 <0.001
Posterior circulation infarct (POCI) POCI vs. LACI 1.50 1.26-1.80 <0.001
Admission blood glucose (mmol/L) 1mmol/L increase 1.08 1.06-1.10 <0.001
Length of hospital stay (days) 1 day increase 1.02 1.01-1.03 <0.001
Abbreviation: SAP, Stroke Associated Pneumonia; AIS, Acute Ischemic Stroke; OR, Odds Ratio; CI, Confidence
Interval; TIA, Transient Ischemic Attack; COPD, Chronic Obstructive Pulmonary Disease; GIB, Gastrointestinal
Bleeding; mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale score; GCS, Glasgow
Coma Scale; OCSP, Oxfordshire Community Stroke Project
Table e-2. Subgroup analysis of discrimination of the AIS-PS with regard to in-hospital SAP after AIS
Derivation cohort Internal validation cohort External validation cohort
(n=8820) (n=5882) (n=3037)
AUROC 95% CI AUROC 95% CI AUROC 95% CI
Overall cohort 0.797 0.782-0.811 0.785 0.766-0.803 0.792 0.761-0.823
Subgroups
Age
<59 0.802 0.764-0.840 0.774 0.718-0.830 0.769 0.701-0.840
>60 0.766 0.748-0.783 0.755 0.733-0.777 0.783 0.749-0.818
Gender
Male 0.806 0.789-0.823 0.803 0.779-0.826 0.798 0.759-0.838
Female 0.784 0.760-0.807 0.756 0.725-0.787 0.778 0.78-0.828
Time from onset to arrival (hours)
<6 0.802 0.776-0.829 0.744 0.705-0.784 0.792 0.761-0.823
6-12 0.812 0.771-0.854 0.803 0.751-0.854 0.837 0.749-0.927
12-24 0.803 0.772-0.835 0.780 0.736-0.823 0.740 0.700-0.820
>24 0.781 0.759-0.803 0.792 0.764-0.820 0.744 0.701-0.812
Length of hospital stay (days)
<7 0.830 0.800-0.860 0.809 0.765-0.853 0.782 0.742-0.816
8-13 0.786 0.758-0.815 0.774 0.734-0.813 0.814 0.736-0.893
>14 0.793 0.775-0.812 0.779 0.756-0.802 0.774 0.740-0.809
Abbreviation: SAP, Stroke Associated Pneumonia; AIS, Acute Ischemic Stroke; AUROC, Area Under the Receiver Operating Characteristic Curve; CI, Confidence Interval.
Figure e-1. Plot of observed versus predicted risk of SAP after AIS in the derivation and
validation cohorts
Figure e-1 legend
Plot of observed versus predicted risk of in-hospital SAP after AIS with 95% confidence interval
(C.I.) in the derivation and validation cohorts according to 10 deciles of predicted risk. Overall,
there was a very high correlation between observed and predicted risk in the derivation cohort (A)
(n=8820; r=0.99, P<0.001), internal validation cohort (B) (n=5882; r=0.99, P<0.001), and external
validation cohort (C) (n=3037; r=0.98, P<0.001), which indicated excellent calibration.
Appendix: The CNSR investigators
Yongjun Wang, Beijing Tiantan Hospital; Qi Bi, Beijing Anzhen Hospital; Weiwei Zhang, Beijing Military District
Gengral hospital of Chinese Peoples Liberation Army; Liying Cui, Peking Union Medical College Hospital of
Peking University; Yuheng Sun, Beijing Jishuitan Hospital; Maolin He, Beijing Shijitan Hospital; Dongsheng Fan,
Peking University Third Hospital; Xunming Ji, Beijing Xuanwu Hospital; Jimei Li, Beijing Friendship Hospital
Affiliated to Capital Medical University; Fang Zhang, Beijing Guangwai Hospital; Kai Feng, Beijing Shunyi
District Hospital; Xiaojun Zhang, Beijing Tongren Hospital; Yansheng Li, Shanghai Renji Hospital; Shaoshi Wang,
Shanghai First Municipal Peoples Branch hospital; Wei Fan, Zhongshan Hospital of Fudan University; Zhenguo
Liu, Xin Hua Hospital Affiliated to Shanghai Jiao Tong University; Xiaojiang Sun, The sixth PeoplesHospital
Affiliated to Shanghai JiaoTong University; Wei Li, Shanghai Ninth Peoples Hospital Affiliated to Shanghai
JiaoTong University; Jianrong Liu, ShanghaiRuijin Hospital; Xu Chen, Shanghai 8th Peoples Hospital; Qingke
Bai, Pudong New Area Peoples Hospital; Dexiang Gu, Shanghai Yangpu Area Shidong Hospital; Xin Li, Shanghai
Yangpu Area Center Hospital; Qiang Dong, Huashan Hospital of Fudan University; Yan Cheng, Tianjin Medical
University Gengeal Hospital; Lan Yu, Tianjin Huanhu Hospital; Bin Li, Dagang Oilfield Gengeal Hospital; Tongyu
Wang, Bohai Oilfield Hospital; Kun Zhao, Baodi District Peoples Hospital of Tianjin; Chaodong Zhang, The First
Affiliated Hospital of China Medical University; Dingbo Tao, The First Afflicated Hospital of Dlian Medical
University; Lin Yin, The Second Affiliated Hospital of Dlian Medical University; Fang Qu, Dlian Second Peoples
hospital; Jingbo Zhang, Dlian Third Peoples hospital; Jianfeng Wang, Dalian Central hospital; Ying Lian, Dalian
Economic and Technological Development District Hospital; Fang Qu, Shenying Military District General hospital
of Chinese Peoples Liberation Army; Jun Fan, Shenyang Military District 202 Hospital; Ying Gao, National
Traditional Chinese Medicine (TCM)Thrombus Treatment Center of Liaoning Province; Mingdong Cheng,
Enliang hopital of Taian County; Jiang Wu, The First Clinical College of Jilin University; Huashan Sun, Jilin
Chemical Industrial Group General hopital; Jinying Li, Jilin Oilfield General Hospital; Guozhong Li, The First
Clinical College of Harbin Medical University; Yulan Zhu, The Second Clinical College of Harbin Medical
University; Zichao Yang, The Fourth Clinical College of Harbin Medical University; Fengmin Yang, Daqing
Oilfield General Hospital; Jun Zhou, Mudan Jiang Second hospital of Hailongjiang Province; Minxia Guo,
Shaanxi Provincial Peoples Hospital; Zhengyi Li, The First Afflicated Hospital of Medical College of Xian
Jiaotong University; Qilin Ma, The First Hospital of Xiamen; Renbin Huang, Chenzhou First Peoples Hospital;
Bo Xiao, Xiangya Hospital of Centre-south University; Kangning Chen, Southwest Hospital; Xinyue Qin, The
First Affiliated Hospital of Chongqing Medical University; Changlin Hu, The Second Affiliated Hospital of
Chongqing Medical University; Li Gao, Chengdu Third Municipal Peoples Hospital; Jinsheng Zeng, The First
Affiliated Hospital of Sun Yat-Sen University; Anding Xu, The First Affiliated Hospital of Jinan University; Xiong
Zhang, Guangdong Peoples Hospital; Ming Shao, The First Affiliated Hospital of Guangzhou Medical University;
Feng Qi, LiWan Hospital of GuangZhou Medical College; Weimin Xiao, Dungun Municipal Peoples Hospital;
Suping Zhang, Guangzhou Red Cross Hospital; Xiaoping Pan, Guangzhou First TMUNICIPAL Peoples Hospital;
Suyue Pan, Nan Fang Hospital; Yefeng Cai, Guangdong Provincial Hospital of Traditional Chinese Medicine; Qi
Wan, Jiang Su Peoples Hospital; Yun Xu, Drum Tower Hospital Affiliated to Nanjing Medical University Upper
First-class Hospital; KaiFu Ke, he Affiliated Hospital of Nantong University Upper First class Hospital; Yuenan
Kong,Wuxi Second Peoples Hospital Upper First-class Hospital; Qing Di, Neurology Hospital Affiliated to
Nanjing Medical University Upper First-class Hospital; Fengyang Shao, Jiangsu Province Lianyungang Hospital
of TCM Upper First-class Hospital; Yajun Jiang, Jiangsu Province Hospital of TCM Upper First-class Hospital;
Daming Wang, The First Peoples Hospital of Changzhou Upper First-class Hospital; Li Guo, The Second Hospital
of Hebei Medical University; Wencui Xue, Qinhuangdao C.