HHS Public Access

Author manuscript
J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
Published in final edited form as:
J Stroke Cerebrovasc Dis. 2015 March ; 24(3): 680–686. doi:10.1016/j.jstrokecerebrovasdis.2014.11.014.

A Simple Prediction Score for Developing a Hospital-Acquired

Infection after Acute Ischemic Stroke

Adam J. Friedant, BS1,*, Brittany M. Gouse, BS1,*, Amelia K. Boehme, PhD, MSPH2,3, James
E. Siegler, MD4, Karen C. Albright, DO, MPH5,6,7, Dominique J. Monlezun, MPH1, Alexander
J. George, BS1, T. Mark Beasley, PhD8, and Sheryl Martin-Schild, MD, PhD1
1Stroke Program, Department of Neurology, Tulane University Hospital, New Orleans, LA 70112

2Gertrude H. Sergievsky Center, Department of Neurology, Columbia University, New York, NY

3Department of Neurology, School of Medicine, University of Alabama at Birmingham, 35294

4Stroke Program, Department of Neurology, Hospital of the University of Pennsylvania,

Philadelphia, PA 19103

5Department of Epidemiology, School of Public Health, University of Alabama at Birmingham,

35294

6Health Services and Outcomes Research Center for Outcome and Effectiveness Research and
Education (COERE)

7Center of Excellence in Comparative Effectiveness Research for Eliminating Disparities

( CERED) Minority Health & Health Disparities Research Center (MHRC )

8Section on Statistical Genetics, Department of Biostatistics, School of Public Health, University of

Alabama at Birmingham, Birmingham, AL 35249
 Friedant et al. Page 2

Abstract

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 3

Background—Hospital-acquired infections are a major cause of morbidity and mortality in acute

ischemic stroke patients. While prior scoring systems have been developed to predict pneumonia
in ischemic stroke patients, these scores were not designed to predict other infections. We sought
to develop a simple scoring system for any hospital-acquired infection.

Methods—Patients admitted to our stroke center (07/08-06/12) were retrospectively assessed.

Patients were excluded if they had an in-hospital stroke, unknown time from symptom onset, or
delay from symptom onset to hospital arrival >48 hours. Infections were diagnosed via clinical,

© 2014 by National Stroke Association. All rights reserved.

Corresponding Author: Sheryl Martin-Schild, MD, PhD, Stroke Program at Tulane University Hospital, Department of Neurology,
1440 Canal Street, TB-52, Suite 1000, New Orleans, LA 70112-2715, smartin2@tulane.edu, (504) 988-0972 (phone), (504) 988-6263
(fax).
*
These 2 authors contributed equally to the production of this manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our
customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of
the resulting proof before it is published in its final citable form. Please note that during the production process errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
laboratory, and imaging modalities using standard definitions. A scoring system was created to predict
infections based on baseline patient characteristics.

Results—Of 568 patients, 84 (14.8%) developed an infection during their stays. Patients who
developed infection were older (73 vs. 64, p<0.0001), more frequently diabetic (43.9% vs. 29.1%,

Conclusion—In our sample, clinical, laboratory, and imaging information available at admission
identified patients at risk for infections during their acute hospitalizations. If validated in other
populations, this score could assist providers in predicting infections after ischemic stroke.

Keywords
infection; acute ischemic stroke; outcome; risk factors; modeling

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 4

Introduction

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 5

Hospital-acquired infections (HAI) are a major cause of morbidity and mortality among
patients admitted with an ischemic stroke.[1] The National Healthcare Safety Network
defines a HAI, or nosocomial infection, as a “localized or systemic condition that is
preventable and results from an adverse reaction to the presence of an infectious agent with
no evidence that the infection was present or incubating at the time of admission to the acute
care setting”.[2] Post-ischemic stroke infections are particularly problematic because they
increase the risk of death and disability after discharge through fever, immobilization of the
patient, and end-organ damage resulting from shock.[3][4] Furthermore, infections are known
to complicate ischemic stroke recovery by increasing hospital costs and by prolonging
hospitalization.[5][6][7]

While patient-related factors such as stroke severity and age are indicators of outcome
following acute ischemic stroke (AIS), HAIs may also play a long-term role.[8][9][10] Prior
research has shown that infections present on admission (POA) are not associated with poor
outcomes after ischemic stroke, whereas HAIs are a major contributor to poor functional
outcomes.[11] Urinary tract infection (UTI), pneumonia (PNA), and bacteremia are among
the most common types of HAI, presenting a barrier to long term recovery in this
population.[6][12] Together, these three infection types comprise nearly half of all HAIs.13]

Early identification of risk factors for infection during hospitalization for AIS is important,
considering effective management may prevent the development of a HAI and subsequently
improve long-term outcomes, although this remains controversial.[14][15][16] Prior research
has shown that infections occur more commonly in the acute phase following ischemic
stroke.[17] Several investigators have identified risk factors for hospital-acquired PNA after
stroke and prediction models have been generated,[18][19] but to date, no model has been
generated to assess the risk of any nosocomial infection. The purpose of this study was to
develop a simple risk prediction model useful in a broad spectrum of infection subtypes in
patients hospitalized for AIS.

Methods
Study Population
A retrospective analysis of previously collected data of all patients with AIS who presented
to our single academic institution between July 2008 and June 2012 was performed using
previously described methods.[20] The registry includes demographic variables, baseline
clinical, laboratory, medication, and imaging variables, as well as inpatient clinical,
laboratory, medication, and imaging variables on all patients admitted with a stroke. Patients
were excluded if they were last seen normal >48 hours prior to admission, had an unknown
time of last seen normal, were transferred to our center from an outside hospital, experienced
an in-hospital stroke, or had an infection present on admission (defined as an infection
diagnosed within the first 24 hours of admission).

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 6

Definition of Outcomes

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 7

All-cause HAIs were defined as any type of bacterial, fungal, or viral infection. Infection
types not included in the detailed scoring mechanisms (e.g. cellulitis, pseudomembranous
colitis, meningitis, ventriculitis) were diagnosed clinically or via laboratory/imaging
findings. Subsequent analyses were performed to assess predictors of UTI, PNA, and
bacteremia. UTIs were defined as >100,000 colony forming units per millimeter of urine in
a patient with signs and symptoms. Urinalysis, which is ordered for all patients on admission
as part of a standardized order set, distinguishes HAI from bacteriuria present on admission.
PNA was defined as an infiltrate on chest radiography with appropriate clinical correlates,
but is considered present on admission (and therefore excluded) if identified on baseline
chest radiography, which is also part of a standardized admission order set.[21] Bacteremia
was defined as >100,000 colony forming units in at least two venous blood samples
(excluding contaminants). We considered coagulase-negative staphylococci, diphtheroids,
Micrococcus spp., Bacillus spp., and Viridans group streptococci as contaminants if these
bacteria did not grow out of all available blood culture vials from a given date and time (e.g.
if only one out of two blood culture vials speciated the organism).

Statistical Analysis
We compared admission variables of interest between patients who contracted a HAI and
those who did not contract a HAI. Pearson Chi-Square (or Fisher's exact test where
appropriate) was used to compare proportions. The Wilcoxon Rank Sum test was used to
compare medians of continuous data. A prediction score for HAIs was created by dividing
the patient sample into a random sample of 55% of the dataset (build group). The remaining
45% constituted the test group. Once the score was tested in the test group, the score was
tested in the entire dataset. Logistic regression models were used to assess the association
between admission variables and the outcome of interest, HAI. Every variable collected at
the time of admission as part of the registry was tested in a univariable logistic regression

score variables and were evaluated at different values and dichotomizations by calculating
the sensitivity and specificity of each binary exposure. Further testing on the categorized
variable through crude logistic regression models to identify cutoff points was conducted.
Each continuous variable was evaluated using receiver operator characteristics (ROC)
curves. Spearman's correlation and ROC curves were used to evaluate the final score. The
points assigned to the variables in the score were determined using the beta coefficients from
the final adjusted logistic regression model for predicting all-cause infections. This process
was repeated to create a prediction score for UTIs, PNA and bacteremia. Logistic regression
was then used to assess what prediction score cut off was most predictive of each outcome
of interest. As this was an exploratory analysis, no adjustments were made for multiple
comparisons.[22] An alpha of 0.05 was used as the level of significance.

Results
Baseline Characteristics
Of the 568 patients included in this study, 84 (14.8%) were found to have a HAI. Of these
patients, 56 (66.7%) developed a UTI, 28 (33.3%) developed PNA, and 20 (23.8%)
developed bacteremia. These infection groups were not mutually exclusive as 20 patients
(23.8%) in our cohort with HAI experienced more than one HAI during admission. In the
multivariable models, an age of greater than or equal to 70 years old on admission was a

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 8

significant independent predictor of HAI (OR 2.49, 95%CI 1.55-4.00, p=0.0002). History of
diabetes was also a significant independent predictor of HAI (OR 1.91, 95%CI 1.18-3.09,
p=0.0084). We categorized baseline NIHSS into three categories (NIHSS 0-7, 8-14, >14) as
reported in a prior prognostic study,[23] which was also found to be significantly higher in
patients with HAIs than patients without HAIs (OR=2.10, 95% CI 1.60-2.77, p<0.0001).

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 9

HAI Prediction Score

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 10

In the HAI model, history of diabetes met the <0.2 univariable p-value cut off. Glucose on
admission was not included in the final prediction model because of colinearity with history
of diabetes. History of diabetes was selected over admission glucose due to better sensitivity
and specificity in predicting HAIs. The cutoff age was determined by testing different
cutpoints in the final multivariable model to assess which threshold was more predictive in
the total model. NIHSS on admission categories were determined using previously
established categories.[23] The final prediction score for HAIs is shown in Table 2. The HAI
Prediction Score ranged from 0 to 7 and produced an area under the curve (AUC) of 0.7311.
Using the entire cohort, 76.8% of patients with a HAI Prediction Score greater than or equal
to 4 developed a HAI. Figure 2 illustrates the distribution of the HAI Prediction Score
among patients without HAI compared to patients who developed a HAI. Additionally, the
odds of a patient with a HAI Prediction Score greater than or equal to 4 of developing a HAI
was nearly 6 times the odds of patients with HAI Prediction Scores of 0-3 of developing a
HAI ( OR 5.67 95% CI 3.28-9.81, p<0.0001).

UTI Prediction Score

In the UTI model, history of diabetes met the <0.2 univariable p-value cut off. The
previously described method for creating a HAI prediction score was applied to patients
with UTI.[23] The final prediction score for UTI is shown in Table 2. The UTI Prediction
Score ranged from 0 to 7 and produced an AUC of 0.6867 (Figure 1). Figure 3 illustrates the
distribution of the UTI Prediction Score among patients without nosocomial UTI compared
to patients who developed a nosocomial UTI. The odds of a patient with a UTI Prediction
Score greater than or equal to 4 of developing a nosocomial UTI was nearly 5 times the odds
of patients with UTI Prediction Scores of 0-3 of developing a nosocomial UTI (OR 4.84 95
% CI 2.53-9.28, p<0.0001).

PNA Prediction Score

In the PNA model, history of diabetes met the <0.2 univariable p-value cut off. The
previously described method for creating HAI and UTI prediction scores was applied to
patients with PNA.[23] The final prediction score for PNA is shown in Table 2. The PNA
Prediction Score ranged from 0 to 7 and produced an AUC of 0.7862 (Figure 1). Figure 4
illustrates the distribution of the PNA Prediction Score among patients without nosocomial
PNA compared to patients who developed a nosocomial PNA. The odds of a patient with a
PNA Prediction Score greater than or equal to 5 of developing nosocomial PNA was 7 times
the odds of patients with PNA Prediction Scores of 0-4 of developing nosocomial PNA (OR
7.26 95 % CI 3.12-16.9, p<0.0001).

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 11

Bacteremia Prediction Score

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 12

In the HAI model, history of diabetes and history of atrial fibrillation met the <0.2
univariable p-value cut off. The previously described methods for creating HAI, UTI, and
PNA scores was applied to patients with bacteremia.[23] The final prediction score for
bacteremia is shown in Table 1. The Bacteremia Prediction Score ranged from 0 to 8 and
produced an AUC of 0.6891 (Figure 1). Figure 5 illustrates the distribution of the
Bacteremia Prediction Score among patients without nosocomial bacteremia compared to
patients who developed nosocomial bacteremia. The odds of a patient with a Bacteremia
Prediction Score greater than or equal to 6 of developing bacteremia was 3 times the odds of
patients with Bacteremia Prediction Scores of 0-5 of developing bacteremia (OR 3.34 95%
CI 1.27-8.75, p=0.0143).

Discussion
In this current study, we derived a simple risk score for HAI during post-stroke
hospitalization using clinical and demographic variables available on admission.
Additionally, we developed individual risk scores for the three main types of HAI: UTI,
PNA, and bacteremia. These models predict the potential risk of a HAI using information
available at the time of hospital presentation, regardless of subsequent in-hospital
management. These simple scores can identify patients at risk for a HAI, and have the
potential to subsequently guide in-hospital care to mitigate HAIs and their effects on
outcomes.

In our cohort, we found that 1 in 7 patients experienced a HAI during their stroke admission.
Our results are consistent with previous studies in showing that HAIs occur in older patients
with higher NIHSS at baseline and in patients with diabetes.[18][24][25][26] Our data is also
consistent with previous reports where the prevalence of HAI, while lower in our sample
than other reports of AIS (14% vs. 16-42%),[27][28] remains higher than the general
prevalence of HAI reported in larger cohort studies of non-stroke patients (16-42% vs. 6-
9%).[27][29][30]

Predictive scores have been developed in prior research for PNA. The A2DS2 score, a
10point scoring system that included age, history of atrial fibrillation, dysphagia, sex, and
stroke severity as calculated by NIHSS, was used to determine the risk of PNA following
ischemic stroke.[31] Several other studies assessed risk factors and created scores for
nosocomial post-stroke PNA.[18][24][32][33] While these previous models are adequate for
predicting post-stroke PNA, our model incorporates multiple HAI subtypes. Further, these
studies included lab values in their algorithms, while our score requires no more than patient
history and short clinical assessments.[34][35][36] Our study demonstrates that a score using
only variables available at the time of admission is feasible. Through our model, we were
able to identify patients who were at 6-fold higher odds of developing a HAI after an
ischemic stroke. This score may eventually prove valuable in an effort to prevent HAIs;
however, this remains unknown at this time.

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 13

A prior meta-analysis investigated the effects of prophylactic antibiotic use in acute ischemic

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 14

stroke patients. Unfortunately, the clinical trials used in the meta-analysis were not designed
or powered to investigate the use of prophylactic antibiotics to reduce post stroke
infections.[16] Further research is needed in the form of prospective studies to evaluate the
efficacy of certain safety precautions or prophylactic antibiotic use in AIS patients in
preventing HAIs and improving outcome.

Our study is limited by its retrospective nature and relatively small sample size. Our sample
is derived from a large, urban area seeking services at a single academic center.
Furthermore, a weakness in the development of any prediction score is the difficulty
inherent to widespread implementation, which could be mitigated by the use of centralized
freeware sources. Due to the high prevalence of UTI, PNA, and bacteremia within HAIs, we
focused our prediction models on these three subtypes only. Confirmation of additional
infection subtypes is technically challenging (e.g., rhinosinusitis, lower extremity cellulitis)
or infrequent (e.g., meningitis/encephalitis), which thereby limits the broad range of
infections we could include in our modeling. Because of the relative rarity in these infection
subtypes, we do not believe their inclusion would significantly affect the results presented
here. Additional validation is needed to examine the generalizability of our HAI prediction
score.

Despite its limitations, our study is unique in that we were able to develop sensitive and
specific prediction scores for all-cause HAI, as well as specific types of HAIs in AIS
patients. If validated, our scoring system has the potential to identify AIS patients at higher
odds of developing a HAI. This could then be used for future prospective studies designed to
evaluate safety precautions and treatment methods to reduce HAIs and improve patient
outcome. Future work may incorporate blood-based immune and stress markers to further
improve upon our simple HAI scoring system.
Acknowledgments
Sources of Funding: Dr. Boehme is supported by NINDS NIH T32 NS007153-31. The content is solely the
responsibility of the authors and does not necessarily represent the official views of the NINDS or the NIH. This
study was conducted at Tulane Medical Center.

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 15

References

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 16

1. Elmsley HC, Hopkins SJ. Acute ischaemic stroke and infection: Recent and emerging concepts.
Lancet Neurol. 2008; 7(4):341–353. [PubMed: 18339349]
2. Horan TC, Andrus M, Dudeck MA. CDC/NHSN surveillance definition of health care-associated
infection and criteria for specific types of infections in the acute care setting. Am J Infect Control.
2008; 36(5):309–332. [PubMed: 18538699]
3. Diederichs C, Berger K, Bartels DB. The measurement of multiple chronic diseases--a systematic
review on existing multimorbidity indices. J Gerontol A Biol Sci Med Sci. 2011; 66:301–311.
[PubMed: 21112963]
4. Chamorro A, Urra X, Planas AM. Infection after acute ischemic stroke: A manifestation of
braininduced immunodepression. Stroke. 2007; 38:1097–1103. [PubMed: 17255542]
5. Ingeman A, Andersen G, Hundborg HH, et al. In-hospital medical complications, length of stay,
and mortality among stroke unit patients. Stroke. 2011; 42(11):3214–3218. [PubMed: 21868737]
6. Spratt N, Wang Y, Levi C, et al. A prospective study of predictors of prolonged hospital stay and
disability after stroke. J Clin Neurosci. 2003; 10(6):665–669. [PubMed: 14592613]
7. Raut M, Schein J, Mody S, et al. Estimating the economic impact of a half-day reduction in length
of hospital stay among patients with community-acquired pneumonia in the US. Curr Med Res
Opin. 2009; 25(9):2151–2157. [PubMed: 19601711]
8. Koennecke HC, Belz W, Berfelde D, et al. Factors influencing in-hospital mortality and morbidity
in patients treated on a stroke unit. Neurology. 2011; 77(10):965–972. [PubMed: 21865573]
9. Weimar C, Mieck T, Buchthal J, et al. Neurologic worsening during the acute phase of ischemic
stroke. Arch Neurol. 2005; 62(3):393–397. [PubMed: 15767504]
10. Siegler JE, Boehme AK, Kumar AD, et al. Identification of modifiable and nonmodifiable risk
factors for neurologic deterioration after acute ischemic stroke. J Stroke Cerebrovasc Dis. 2013;
22(7):e207–213. [PubMed: 23246190]
11. Boehme AK, Kumar AD, Dorsey AM, et al. Infections present on admission compared with
hospital-acquired infections in acute ischemic stroke patients. J Stroke Cerebrovasc Dis. 2013;
22(8):e582–589. [PubMed: 23954599]
12. Weimar C, Roth MP, Zillessen G, et al. Complications following acute ischemic stroke. Eur
Neurol. 2002; 48:133–140. [PubMed: 12373029]
13. Magill SS, Edwards JR, Bamberg W, et al. Multistate Point-Prevalence Survey of Health Care–
Associated Infections. N Engl J Med. 2014; 370:1198–1208. [PubMed: 24670166]
14. Lampl Y, Boaz M, Gilad R, et al. Minocycline treatment in acute stroke: An open-label,
evaluatorblinded study. Neurology. 2007; 69(14):1404–1410. [PubMed: 17909152]
15. Westendorp WF, Vermeij JD, Vermeij F, et al. Antibiotic therapy for preventing infections in
patients with acute stroke. Cochrane Database Syst Rev. 2012; 1:CD008530. [PubMed: 22258987]
16. Meisel C, Prass K, Braun J, et al. Preventive antibacterial treatment improves the general medical
and neurological outcome in a mouse model of stroke. Stroke. 2004; 35(1):2–6. [PubMed:
14684767]
17. Langhorne P, Stott DJ, Robertson L, et al. Medical complications after stroke: a multicenter study.
Stroke. 2000; 31(6):1223–1229. [PubMed: 10835436]
18. Ji R, Shen H, Pan Y, et al. Novel risk score to predict pneumonia after acute ischemic stroke.
Stroke. 2013; 44(5):1303–1309. [PubMed: 23482598]
19. Kwon HM, Jeong SW, Lee SH, et al. The pneumonia score: A simple grading scale for prediction
of pneumonia after acute stroke. Am J Infect Control. 2006; 34(2):64–68. [PubMed: 16490608]
20. Siegler JE, Boehme AK, Dorsey AM, et al. A comprehensive stroke center patient registry:
Advantages, limitations, and lessons learned. Med Stud Res J. 2013; 1:21–29.
21. Jones EM, Albright KC, Fossati-Bellani M, et al. Emergency department shift change is associated
with pneumonia in patients with acute ischemic stroke. Stroke. 2011; 42:3226–3230. [PubMed:
21903956]
22. Bender R, Lange S. Adjusting for multiple testing--when and how? J Clin Epidemiol. 2001; 54(4):
343–349. [PubMed: 11297884]
23. Saver JL, Yafeh B. Confirmation of tPA treatment effect by baseline severity-adjusted end point
reanalysis of the NINDS-tPA stroke trials. Stroke. 2007; 38(2):414–416. [PubMed: 17234987]
24. Sellars C, Bowie L, Bagg J, et al. Risk factors for chest infection in acute stroke: A prospective
cohort study. Stroke. 2007; 38(8):2284–2291. [PubMed: 17569875]

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 17

25. Hilker R, Poetter C, Findeisen N, et al. Nosocomial pneumonia after acute stroke: Implications for
neurological intensive care medicine. Stroke. 2003; 34(4):975–981. [PubMed: 12637700]
26. Westendorp WF, Nederkoorn PJ, Vermeij JD, et al. Post-stroke infection: A systematic review and
meta-analysis. BMC Neurol. 2011; 11:110. [PubMed: 21933425]
27. Wartenberg KE, Stoll A, Funk A, et al. Infection after acute ischemic stroke: Risk factors,
biomarkers, and outcome. Stroke Res Treat. 2011; 14:830614. [PubMed: 21789273]
28. Meisel C, Schwab JM, Prass K, et al. Central nervous system injury-induced immune deficiency
syndrome. Nat Rev Neurosci. 2005; 6(10):775–786. [PubMed: 16163382]
29. Andersen BM, Ringertz SH, Gullord TP, et al. A three-year survey of nosocomial and
communityacquired infections, antibiotic treatment and re-hospitalization in a Norwegian health
region. J Hosp Infect. 2000; 44(3):214–223. [PubMed: 10706805]
30. Emmerson AM, Enstone JE, Griffin M, et al. The second national prevalence survey of infection in
hospitals--overview of the results. J Hosp Infect. 1996; 32(3):175–190. [PubMed: 8690881]
2 2
31. Hoffmann S, Malzahn U, Harms H, et al. Development of a clinical score (A DS ) to predict
pneumonia in acute ischemic stroke. Stroke. 2012; 43(10):2617–2623. [PubMed: 22798325]
32. Harms H, Grittner U, Dröge H, et al. Predicting post-stroke pneumonia: The PANTHERIS score.
Acta Neurol Scand. 2013; 128(3):178–184. [PubMed: 23461541]
33. Zhang X, Wang F, Zhang Y, et al. Risk factors for developing pneumonia in patients with diabetes
mellitus following acute ischaemic stroke. J Int Med Res. 2012; 40(5):1860–1865. [PubMed:
23206467]
34. Chumbler NR, Williams LS, Wells CK, et al. Derivation and validation of a clinical system for
predicting pneumonia in acute stroke. Neuroepidemiology. 2010; 34(4):193–199. [PubMed:
20197702]
35. Qian J, Payabvash S, Kemmling A, et al. Variable selection and prediction using a nested, matched
case-control study: Application to hospital acquired pneumonia in stroke patients. Biometrics.

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 18

2014; 70(1):153–163. [PubMed: 24320930]

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 19

36. Daniels SK, Brailey K, Priestly DH, et al. Aspiration in patients with acute stroke. Arch Phys Med
Rehabil. 1998; 79(1):14–19. [PubMed: 9440410]

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 20

Figure 1. Receiver Operating Characteristic Curve for the HAI Predictor Score, the UTI

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 21

Prediction Score, the Pneumonia Prediction score and the bacteremia prediction score Figure 2.

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 22

Distribution of HAI prediction Score

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 23

Figure 3. Distribution of UTI Prediction Score

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 24

Figure 4. Distribution of PNA Prediction Score

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 25

Figure 5. Distribution of Bacteremia Prediction Score

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 26

Table 1
Baseline characteristics of patients with and without a hospital inquired infection (HAI) during stroke service
admission.

No HAI (N=484) HAI (N=84) p-value

Age, median y (range) 64 (24-102) 73 (19-95) <0.0001

Gender, No. female (%) 211 (45.9%) 39 (46.4%) 0.9247

Black Race, (%) 309 (67.3%) 57 (67.9%) 0.9231

Past Medical History, No. (%)

Coronary Artery Disease 77 (16.8%) 22 (26.2%) 0.1043

Diabetes 132 (29.1%) 36 (43.9%) 0.0077

Hypertension 346 (75.9%) 66 (78.6%) 0.5936

Dyslipidemia 209 (45.7%) 31 (36.9%) 0.2571

Atrial Fibrillation 50 (11.0%) 14 (16.7%) 0.1436

Congestive Heart Failure 27 (13.0%) 5 (17.2%) 0.5363

Stroke 185 (40.2%) 29 (34.5%) 0.3260

Baseline NIHSS, median (IQR) 5 (0-31) 12 (1-33) <0.0001

Simple TOAST 0.1736

Cardioembolic 116 (25.2%) 31 (36.9% )

Large vessel disease 97 (21.1%) 21 (25% )

Small vessel disease 88 (19.1%) 11 (13.1% )

Cryptogenic 114 (24.8%) 15 (17.9% )

>1 Cause 12 (2.6%) 1 (1.2% )

Other 33 (7.2%) 5 (5.9% )

Baseline Laboratory values

Serum Glucose, median mg/dl (IQR) 115 (58-569) 133 (78-481) 0.0427

Serum Leukocyte, median per mL 8 (0.2-18) 8 (2.4-26.9) 0.2024

(IQR)
Hematocrit, median % (IQR) 39.7(23-55.3) 39.3 (24.7-48.6) 0.5203

Serum Platelet count, median per mL 222 (54-751) 210 (10-750) 0.1732
(IQR)
IV tPA, No. (%) 186 (40.3%) 37 (44.1%) 0.5159

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 27

Abbreviations: IQR, interquartile range; Baseline NIHSS, National Institutes of Stroke Scale score at admission; TOAST, Trial of Org 10172 in
Acute Stroke Treatment; IV tPA, intravenous tissue plasminogen activator.
Table 2

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.
 Friedant et al. Page 28

Characteristics of the Overall HAI score, the UTI prediction score, the pneumonia
prediction score and the bacteremia prediction score

Score Variables Included ROC Characteristics

HAI Prediction Score (Range 0-7) AUC 0.731
Age ≥ 70= 1 point
History of Diabetes= 1 point
Baseline NIHSS 4-15= 3 points
Baseline NIHSS >15 = 5points
UTI Infection Score (Range 0-7) AUC 0.687
Age ≥ 70= 1 point
History of Diabetes= 1 point
Baseline NIHSS 4-15= 3 points
Baseline NIHSS >15 = 5points
Pneumonia Infection Score (Range 0-7) AUC 0.786
Age ≥ 70= 1 point
History of Diabetes= 1 point
Baseline NIHSS 4-15= 3 points
Baseline NIHSS >15 = 5points
Bacteremia Infection Score (Range 0-8) AUC 0.689
Age ≥ 70= 1 point
History of Diabetes= 1 point
Baseline NIHSS 4-15= 3 points
Baseline NIHSS >15 = 5points
History of atrial fibrillation= 1 point

J Stroke Cerebrovasc Dis. Author manuscript; available in PMC 2016 March 01.