J Neurol Neurosurg Psychiatry 83:1225-1230 doi:10.1136/jnnp-2012-302194
  • Cerebrovascular disease
  • Research paper

Decision-making in the diagnosis and treatment of stroke-associated pneumonia

  1. Andreas Meisel1,2,3
  1. 1NeuroCure Clinical Research Center (NCRC), Charité—Universitaetsmedizin Berlin, Berlin, Germany
  2. 2Center for Stroke Research Berlin (CSB), Charité—Universitaetsmedizin Berlin, Berlin, Germany
  3. 3Department of Neurology, Charité—Universitaetsmedizin Berlin, Berlin, Germany
  1. Correspondence to Dr Hendrik Harms, NeuroCure Clinical Reseach Center (NCRC), Center for Stroke Research Berlin (CSB), Department of Neurology, Charité—Universitaetsmedizin Berlin, Charitéplatz 1, 10117 Berlin, Germany; hendrik.harms{at}
  1. Contributors HH: conception and design, analysis and interpretation, drafting and revising the article, final approval, corresponding author. SH: conception, revising the article. UM: analysis and interpretation of data. SO: drafting and revising the article. PH: drafting and revising the article, interpretation of data. AM: conception and design, analysis and interpretation, final approval.

  • Received 5 January 2012
  • Revised 26 July 2012
  • Accepted 27 July 2012
  • Published Online First 28 September 2012


Background Stroke-associated pneumonia (SAP) is associated with impaired outcome in acute stroke patients. Current European and American guidelines for acute stroke care are lacking standardised recommendations for the management of SAP. We investigated current diagnostic and treatment practice for SAP in German stroke units (SU).

Methods We developed a standardised questionnaire including characteristics of SU, questions related to antibiotic treatment approaches of SAP and five case vignettes describing relevant clinical scenarios based on Centers for Disease Control and Prevention (CDC) criteria for ‘clinically defined pneumonia’. All certified German SU were invited to take part in the survey.

Results The survey took place from April to August 2010. Of all 162 German SU contacted, 83 (51%) responded. Classification and regression trees analysis suggested that SAP was diagnosed on the basis of clinical criteria such as fever and stroke severity. Chest x-ray showed only limited influence on the diagnosis of SAP. C-reactive protein was frequently requested as additional diagnostic information (38–76%). Group 3 cephalosporines and (acyl-) aminopenicillins/β-lactamase inhibitors are the most frequently used antibiotics (46–60%) in empiric mono (58%) and combination (42%) therapy. A minority of SU (5%) use prophylactic antibiotic treatment. Standardised procedures are available in 61% of SU.

Conclusion Clinical criteria were the main determinants for SAP diagnosis. In contrast, chest x-ray—the central diagnostic item in CDC criteria—was of minor importance. Our survey demonstrates heterogeneous diagnostic and therapeutic strategies in German SU. Future studies need to establish and to evaluate standardised criteria for SAP care.


Stroke-associated pneumonia (SAP) remains an important medical complication after stroke.1 ,2 It prolongs hospitalisation3 and is associated with mortality and morbidity after stroke.4 ,5 Current guidelines for acute stroke treatment in Europe and the USA lack standardised recommendations for its diagnosis and treatment.6 ,7 The US Centers for Disease Control and Prevention (CDC) have provided criteria for a ‘clinically defined pneumonia’, which requires a combination of clinical, laboratory and radiological findings (see supplementary material, available online only). Based on these criteria treatment guidelines for hospital-acquired pneumonia (HAP) are provided by several medical societies.8 The aim of our survey was to investigate current diagnostic and treatment practice for post-stroke pneumonia in stroke units (SU).


Sample and settings

Between April and August 2010, a survey was performed among all certified SU in Germany (n=162). Addresses were provided by the German Stroke Society in March 2010 (see supplementary material, available online only). The survey was anonymous, and therefore did not allow for the identification of the participating hospitals. Consequently, evaluation of the accuracy of the answered questionnaire by site visits was impossible. However, all returning copies of the questionnaire were checked for misinterpretations, comments, or unclear answers before they were accepted for analysis.


A standardised questionnaire was developed (see supplementary material, available online only). The first part consisted of questions regarding general characteristics of the respective SU, for example, annual number of treated stroke patients, number of beds, and the existence of written standard operating procedures (SOP). The second part consisted of five case vignettes based on selected CDC criteria (see supplementary material, available online only). The case vignettes provided typical clinical situations after acute stroke including varying information on clinical parameters (age, state of consciousness, comorbidities, stroke localisation, stroke severity (according to the National Institutes of Health stroke scale; NIHSS), presence or absence of oropharyngeal dysphagia, body temperature, auscultatory findings), laboratory findings (C-reactive protein (CRP), white blood cell count, procalcitonin, arterial blood gas analysis) and radiological findings (result of chest x-ray). Survey participants were asked to decide on the diagnosis of pneumonia and initiation of antibiotic treatment for the respective patient based on this information. In addition, in each of the case vignettes a set of optional clinical and diagnostic criteria could be marked if further information was needed for making a decision on SAP diagnosis. The third part of the questionnaire consisted of different questions regarding standard therapeutic procedures including questions on groups of antibiotics routinely used for treating SAP, utilisation of antibiotic combination therapy (yes/no, combination of which antibiotic groups), type of administration of antibiosis (oral/intravenous), treatment duration (in days), monitoring of treatment success (by clinical, laboratory or imaging findings), and practice of prophylactic antibiosis (if yes: in all patients, patients with a predefined stroke severity, or patients with dysphagia). The questionnaire was developed to investigate diagnostic approaches and not to judge diagnostic accuracy in terms of CDC-based ‘correct’ answers.

The case vignettes were developed by the study group and reviewed by an independent clinician and a clinical microbiologist. A pre-test was performed with four doctors experienced in stroke treatment and not involved in the study in order to investigate acceptance and correct understanding of the questionnaire. In order to achieve optimal compliance with the survey, only one questionnaire was sent to each SU. As we aimed at achieving insights into the practice of SAP care on SU we asked in the covering letter of our survey for an experienced stroke physician working on the SU to answer the questionnaire.

Statistical analysis

Frequencies and proportions were used to characterise participating SU and results of the case vignettes. Fisher's exact test was used to compare proportions for dichotomous traits between independent groups. All tests were two-tailed and resulting p values less than 0.05 are considered statistically significant. However, as this was an exploratory study, no method for multiple hypothesis testing was used and therefore we did not control the family-wise error rate. Due to limitations in the data structure the application of statistical methods for data analysis was constrained and we used the classification and regression trees (CART) method to explore further the main factors that determine the diagnostic decision concerning SAP in this survey. The CART method is a non-parametric procedure that is not based on model assumptions known from parametric multivariable procedures. It is based on recursive partitioning of the data. We selected relevant factors from our case vignettes that might influence the probability of the decision on SAP diagnosis: age, stroke localisation, stroke severity (NIHSS), fever, leucocytosis, dysphagia and at least one of the following clinical symptoms: unconsciousness, tachypnoea, cough, rhonchus. Variables, originally categorical with more than two categories or variables of quantitative nature, were dichotomised in the following manner: age (<70/≥70 years), stroke severity (NIHSS <10, moderate stroke; NIHSS ≥10, severe stroke). The variable chest x-ray was dichotomised into ‘pathological’ and ‘no or missing findings’. All statistical analyses were performed using the SPSS V.18.0 software package.


Completion of data and characteristics of SU

The response rate was 51% (N=83); characteristics and structures of the participating SU, for example, number of beds and treated stroke patients per year are summarised in table 1. SOP for SAP management are available in 51 (61%) of SU.

Table 1

Structural characteristics of participating SU

Case vignettes

The aim of the five case vignettes was both to investigate which parameters effectively determine clinicians' diagnosis of SAP and to find out if CDC criteria are applied in clinical practice. They were not developed to control the accuracy of diagnostic decisions, because a commonly accepted ‘gold standard’ for the diagnosis of post-stroke pneumonia does not exist. The key information of each case vignette is summarised in table 2.

Table 2

Synopsis of case vignettes

In case vignette number 1, the clinical findings were highly suggestive of SAP (fever, pathological auscultatory findings, pulmonary secretion). This patient had severe stroke (NIHSS 12), dysphagia was apparent but information on chest x-ray was not provided. Despite this, 99% of all participants diagnosed SAP and would start antibiotic treatment (98%). Further diagnostic information was requested by 73% of participants, mainly chest x-ray (70%) and CRP level (63%).

Information provided in case vignette number 2 incompletely fulfilled the CDC criteria for ‘clinically defined pneumonia’ (patient confused, cough, pathological chest x-ray, no fever, normal white blood cell count). In this case, only a minority of participants diagnosed pneumonia (38%) or would start antibiotic treatment (29%). Additional information was requested by 84% of participants, above all CRP level (76%) and dysphagia diagnostics (46%).

Only 55% of participants rated for pneumonia in case number 3 (patient with fever, leucocytosis and dysphagia). In contrast, 72% would start antibiotic treatment even though auscultatory and chest x-ray findings were normal. Additional information was requested by 72% of participants, who asked predominantly for CRP level (67%).

In case vignette number 4, a patient with leucocytosis, no fever, tachypnoea, but no auscultatory findings was described. A minority of participants (23%) opted for the diagnosis of SAP and treatment with antibiotics (17%). The vast majority (90%) of participants requested more information, mainly chest x-ray (88%), CRP level (72%) and information on dysphagia (54%).

In case vignette number 5, the information provided (pathological clinical findings, chest-x-ray and laboratory parameters) fulfilled the CDC criteria of a ‘clinically defined pneumonia’. All clinicians (100%) rated this case as SAP and would start antibiotic treatment. Nonetheless, 44% of participants requested further diagnostic information to ensure the diagnosis of SAP.

The results of microbiological testing were requested by 34–48% of the participants, apparently irrespective of the case information given in the vignettes.

Empirical antibiotic treatment of SAP

The majority of SU (98%) administer antibiotics for SAP treatment intravenously. Group 3 cephalosporins (60%), aminopenicillins/lactamase inhibitors (LI) (51%) and acylaminopenicillins/LI (46%) are first-choice antibiotic agents if administered as monotherapy (58%, figure 1A); 42% of SU choose antibiotic combination therapy as initial treatment (figure 1B) using predominantly group 3 cephalosporins combined either with macrolids (24%) or metronidazole (20%). Antibiotic treatment duration is predefined by SOP in 37 SU (45%) with a mean of 8 days (SD 2). A minority of SU (5%) use prophylactic antibiotic treatment depending on stroke severity.

Figure 1

Empirical antibiotic treatment of stroke-associated pneumonia in German stroke units (SU). Proportions of participating SU (%) using monotherapy (A) or combination therapy (B) are displayed. Monotherapy (A; N=46, 58%) and combination therapy (B; N=34, 42%). Ceph, cephalosporins; FC, fluoroquinolones; LI, lactamase inhibitor.

Influence of SOP on diagnosis and treatment of SAP

Internal SOP showed only little influence on diagnostic and treatment decisions in the case vignettes (table 3). In the cases in which the broad majority rated for pneumonia (case 1 and case 5), the availability of SOP was associated with increased requests for additional diagnostic information. Existence of SOP was furthermore significantly associated with fixed treatment durations (p=0.01) and with an increased use of chest x-ray as treatment control (p=0.05). No association between the existence of SOP and type of administered antibiotic drugs was found (data not shown). The existence of SOP had no impact on the use of prophylactic anti-infective treatment.

Table 3

Impact of SOP on diagnosis and treatment of SAP

Identification of decision-making case information

In CART analysis fever, stroke severity and chest x-ray were identified as important decision-making parameters based on the information provided in the case vignettes (figure 2). The decision for the diagnosis of SAP was mainly influenced by the factor ‘fever’, which was associated with the diagnosis of SAP in 84%. If ‘no fever’ was given, 70% would not have diagnosed SAP. In the case of ‘fever’, ‘stroke severity’ was considered to refine the diagnostic approach: the combination of ‘fever’ and ‘severe stroke’ yielded a diagnosis of SAP in 99%. In the case of ‘fever’ and ‘moderate stroke’, participants additionally considered findings of chest x-ray. ‘Not available/normal x-ray’ led to an inconsistent diagnostic decision (54%), whereas ‘pathological findings in chest x-ray’ determined the diagnosis of SAP in 100%. In cases with ‘no fever’, chest x-ray was also taken into account in making the decision. However, even in cases with ‘pathological findings in chest x-ray’ only 37% of participants would have diagnosed SAP.

Figure 2

Decision-making in the diagnosis of stroke-associated pneumonia. Classification and regression trees analysis based on the five case vignettes. The following factors were included: age (<70/≥70 years), stroke localisation, stroke severity (National Institutes of Health stroke scale (NIHSS) <10, moderate stroke; NIHSS ≥10, severe stroke), fever, leucocytosis, dysphagia and at least one of the following clinical symptoms: unconsciousness, tachypnoea, cough, rhonchus. The variable chest x-ray was dichotomised in ‘pathological’ and ‘no or missing findings’.


This survey provides an overview on current diagnostic and treatment practice of SAP in German SU. The main findings of our survey are: (1) the diagnosis of SAP is mainly based on clinical criteria such as fever and stroke severity; (2) chest x-ray—the central diagnostic item in the CDC criteria of ‘clinically defined pneumonia’—is assigned only minor importance; (3) the frequent request for CRP as a diagnostic criterion suggests the demand for biomarkers as a diagnostic tool in SAP; and (4) group 3 cephalosporins play a prevailing role in the treatment of SAP.

Diagnosis of SAP in Germany

SAP can present with a wide range of symptoms, a situation that hampers diagnostic accuracy. Our data suggest that fever and stroke severity are the most important factors for clinicians in diagnosing SAP. This seems to be logical as pneumonia is the most common cause of fever in the acute phase of stroke.9 ,10 In addition, stroke severity is a well-established risk factor for SAP.11 ,12 Current guidelines routinely request pathological findings in a chest x-ray to confirm SAP diagnosis.8 Therefore, chest x-ray is frequently ordered in those case vignettes in which radiographic information is missing. However, our analysis further indicates that chest x-ray has only a minor influence on diagnostic decisions. Our results might reflect the unreliability of chest x-ray for diagnosing pneumonia, as shown in bedridden patients and in the early course of pneumonia.13 ,14 Recommendations of the stroke guidelines for an early treatment of infections conflict with guidelines for pneumonia diagnosis that mandate pathological chest x-ray findings.8 Blood-based biomarkers were commonly used for SAP diagnosis in German SU. Arterial blood gas analyses provide immediate laboratory information about oxygenation and ventilation functions, and are thus common and well-established diagnostic tools in critical care medicine.15 It seems, however, to be of minor importance for the participants in this survey, as CRP levels were more frequently requested as an additional diagnostic parameter. CRP is an established marker in the diagnostic work-up of bacterial infections. However, depending on the cut-off levels and study population investigated, the sensitivity (10–98%) and specificity (44–99%) of CRP for the diagnosis of acute infections differs substantially.16 The diagnostic value of CRP for SAP monitoring is questionable, as cerebral ischaemia causes a systemic inflammatory response tending towards elevated CRP levels early after stroke onset, independent of any infection.17 ,18 In addition, effective antibiotic prevention of infections diminishes but does not normalise serum CRP levels in stroke patients.19 Procalcitonin, proposed as an early marker of severe bacterial infections,20 was requested less frequently as additional information for diagnosing SAP. Recent data suggest that procalcitonin is a predictive marker for ventilator-associated pneumonia.21 In addition, procalcitonin has been shown to be effective in guiding the treatment of community-acquired pneumonia and sepsis.22 ,23 We cannot exclude the possibility that using simple clinical criteria for diagnosis might lead to an overestimation of SAP in Germany, and thus antibiotics are often used unnecessarily in stroke patients. However, in comparison to stroke populations in other countries,24 ,25 the frequency of SAP in Germany does not differ significantly over recent years (7% vs 6–8%5 ,26). Therefore, either SAP is also overestimated in other countries, or the CDC criteria for pneumonia are inappropriate to diagnose SAP.

Treatment practice of SAP in Germany

According to current guidelines,27 the choice of empirical anti-infectives in HAP should be influenced by the time point of infection after hospitalisation and the risk of multidrug-resistant (MDR) pathogens. Therefore, monotherapy with narrow-spectrum antibiotics (eg, group 3 cephalosporin, aminopenicillins/LI, fluoroquinolones) is recommended in early-onset HAP occurring within the first 4 days after hospitalisation, whereas broad-spectrum combination therapies (eg, acylaminopenicillines/LI plus aminoglycosides or ciprofloxaxin) are proposed to treat late-onset HAP and patients with a higher risk of MDR pathogens.27 Although SAP can be considered as early-onset HAP,19 ,28 ,29 MDR pathogens may occur19 and a broad anti-infective approach may thus be appropriate. Our survey identified group 3 cephalosporins, aminopenicillins/LI, and acylaminopenicillins/LI as first-choice antibiotics in mono and combination therapy in German SU. Combination therapy to cover MDR pathogens (eg, acylaminopenicillines/LI or group 3 cephalosporines plus fluoroquinolones) is used only in a minority of German SU. The heterogeneous choice of initial antibiotics may indicate different treatment philosophies, as acylaminopenicillins/LI will also cover MDR pathogens. Our data demonstrate that most of the SU used several antibiotics to treat SAP in their population of stroke patients. These findings might reflect the common practice of using several antibiotics for the same indication in order to minimise the risk of induction of bacterial resistance. Overall, fluoroquinolones, although recommended for treatment in early-onset HAP,27 are second-line antibiotics in German SU. This finding may reflect the results from a trial investigating antibiotic prophylaxis of SAP using levofloxacin.28

The majority of German SU comply with the European stroke guidelines in strongly discouraging prophylactic antibiotic treatment.6 This recommendation is based on one negative clinical trial.28 Experimental and clinical data, however, suggest that antibiotic prophylaxis in acute stroke is effective in preventing post-stroke infections.19 ,29–31 Notably, our survey revealed that in cases with an indecisive diagnosis of SAP (eg, case 3) the majority (72%) of participants favoured anti-infective treatment, suggesting that antibiotic prophylaxis is being partly implemented in current stroke treatment strategies.

Our data indicate that internal SOP on SAP management define the duration of empirical antibiotic drug treatment but have a minor impact on diagnostic and therapeutic decisions in the case vignettes. However, the implementation of SOP adapted to current antibiotic treatment guidelines and local resistance rates reduced early mortality in HAP.32 As pneumonia constitutes a leading risk factor of early mortality in acute stroke5 it would be reasonable to establish standards for SAP care and to prove their efficacy.

There are some limitations of our study that need to be addressed. First, the response rate of 51% may have introduced a selection bias. Second, the small number of case vignettes and especially the structure of case vignettes with a varying amount of information provided hampers the analysis of decision-making factors, which therefore have to be interpreted with caution. Third, a survey may only reflect parts or an optimised image of daily stroke care and therefore might not reliably reflect current treatment decisions in individual stroke patients. Fourth, we cannot exclude the possibility that the accuracy of the questionnaire was impaired by the anonymous census that we used. Fifth, we can only speculate about the reasons for the given diagnostic and treatment decisions, because the questionnaire did not ask for explanations. For example, in bedridden patients diagnostic sensitivity of chest x-ray is known to be impaired. In addition, stroke physicians are guided to treat post-stroke infections as early as possible. As positive signs of pneumonia on chest x-ray are often delayed in the course of the disease, clinicians might fear to treat pneumonia in stroke patients too late, if they rely on x-ray findings. This might also explain the high usage of blood-based biomarkers such as CRP as surrogates for infection in order to support the clinical suspicion. However, to our knowledge this is the first study to investigate common practice for the diagnosis and treatment of SAP. Our study provides the rationale for designing a study investigating the impact of different strategies of SAP care on stroke outcome.


Our survey shows a substantial heterogeneity in diagnosing and managing SAP among SU in Germany. The data suggest that mainly clinical information is used in its diagnosis. This might be reasonable given the impaired diagnostic value of chest x-ray as a specific diagnostic tool for pneumonia, especially in bedridden stroke patients.14 The survey underlines the need for further studies that evaluate and establish standardised diagnostic and therapeutic approaches for SAP. Predictive scores and biomarkers might be helpful in routine clinical practice. As current guidelines for acute stroke treatment in Europe and the USA lack evidence-based recommendations further effort has to be made to establish and standardise diagnostic and treatment approaches to SAP in daily clinical practice using different classifications levels of diagnostic confidence (eg, definite vs probable SAP).


The authors gratefully thank Drs Elke Halle, Christoph Drenckhahn, Martin Köhnlein, Witold Rogge, Alexander Stoll, and Alexander Ecke for reviewing the case vignettes, and Marret Heinold for technical support.


  • Funding This study was supported by the European Union's Seventh Framework Programme (FP7/2008-2013) under grant agreements 201024 and 202213 (European Stroke Network), the Helmholtz Gemeinschaft für Forschungseinrichtungen (SO-022NG), the German Ministry for Health and Education (01 EO 08 01), and the Deutsche Forschungsgemeinschaft (Exc 257 NeuroCure).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.


Visit the full archive of podcasts for JNNP here >>

Free sample
This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of JNNP.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.