Risk of Stroke-Associated Pneumonia During Hospitalization: Predictive Value of Combined A2DS2 Score and Hyperglycemia

Background: Stroke-associated pneumonia (SAP) is a common complication of cerebrovascular disease. The A2DS2 score has been used to predict the risk of SAP. However, hyperglycemia is not included in this scale. The purpose of the present study was to determine whether combining the A2DS2 scoring system and hyperglycemia can improve the predictive value of the conventional A2DS2 scale for SAP. Methods: This retrospective study enrolled 2,552 patients with acute ischemic stroke. The A2DS2 scores, fasting blood glucose level and blood glucose level on admission were collected. Regression analysis was used to identify the independent risk factors of SAP. ROC curve analysis was used to evaluate the speci�city and sensitivity of the combined A2DS2 score and fasting hyperglycemia for predicting SAP. Results: Fasting hyperglycemia was an independent risk factor for SAP (OR=2.95; 95% con�dence interval: 2.11-4.12; P<0.001). The area under curve of the combined A2DS2 score and fasting hyperglycemia was signi�cantly higher than that of the A2DS2 score alone (0.814 vs. 0.793; P=0.048). Conclusion: Fasting hyperglycemia is an independent risk factor for predicting SAP. The predictive value of the modi�ed A2DS2 score (combined A2DS2 score and fasting hyperglycemia) is superior to that of A2DS2 score.


Introduction
Pneumonia is a common critical complication following acute ischemic stroke, and the risk factors include senior age, severe basic diseases, and long duration of hospitalization.Pneumonia increases the duration and expense of hospitalization, and has well been associated with high mortality.Therefore, the early identi cation and appropriate treatment of post-stroke pneumonia should be highlighted to improve the clinical prognosis.
The A2DS2 assessment can easily be completed on admission.However, it is still not widely used in clinical practice.
Several studies have indicated that hyperglycemia may also be a risk factor for SAP [5,13,14].
Hyperglycemia is not included in the A2DS2 scoring system [6], and relevant evidence has shown that diabetes is not an independent risk factor for SAP.However, these ndings do not contradict the hypothesis that hyperglycemia is a risk factor for SAP, since the blood glucose level of patients with wellcontrolled diabetes can be normal.Blood glucose level is one of the routine clinical indicators for patients with acute cerebrovascular disease.The value of the combined A2DS2 score and hyperglycemia for predicting SAP remains unclear.
The purpose of the present study was to investigate the predictive value of the combined A2DS2 score and hyperglycemia for SAP in patients with acute ischemic stroke during hospitalization.

Data Collection
Demographic data (age and gender) were collected.Previous and present medical history (history of atrial brillation, dysphagia and diabetes) were reviewed.An electrocardiogram was performed for all patients, and atrial brillation was recorded.
Admission hyperglycemia was diagnosed when the random blood glucose level on admission was ≥11.1 mmol/L, while fasting hyperglycemia was diagnosed when the blood glucose level at the next morning after admission was ≥7.1 mmol/L.SAP was diagnosed according to the Centers for Disease Control and Prevention (CDC) criteria and/or Mann's criteria [1].
The assignment rules for the independent risk factor of SAP were as follows: 1 point was assigned for 1.25 < adjusted odds ratio (OR) < 2.0; 2 points were assigned for 2.0 ≤ adjusted OR < 4.0; 3 points were assigned for 4.0 ≤ adjusted OR < 6.0 [6].
Statistical Analyses SPSS 20.0 software (IBM Corp., Armonk, NY, USA) was used for the statistical analyses.Continuous variables were expressed as the mean ± standard deviation (SD).Categorical variables were expressed as percentage.Continuous variables were compared using nonparametric tests, and categorical variables were compared using Chi-squared test.Logistic regression was used to calculate the adjusted OR value after balancing the confounding factors.Hyperglycemia was assigned according to the above assignment rules, and the a modi ed A2DS2 scoring system was designed by combining the conventional A2DS2 items and assigned hyperglycemia points.Receiver operating characteristic (ROC) curve analysis was used to calculate the area under the curve of the conventional A2DS2 scoring system and modi ed A2DS2 scoring system.The cut-off values that represented the sensitivity and speci city of each evaluation tool were analyzed.Points with the maximal Youden's index were obtained as the optimal cut-off values.Probability (p) values ≤0.05 were considered statistically signi cant.

Results
The average age of the enrolled patients was 61.9 ± 12.7 years.SAP was observed in 8.2% of all patients.There was no signi cant difference in gender and infarction site between the SAP group and non-SAP group (P>0.05).The incidence of SAP in patients with the following factor(s) was signi cantly higher than that in patients without the following factor(s) (P<0.05):elderly age (≥75 years old), atrial brillation, dysphagia, random hyperglycemia, fasting hyperglycemia, or random or fasting hyperglycemia.Therefore, these factors were considered as potential risk factors for SAP.These patients were classi ed into three subgroups according to stroke severity (NIHSS score), and there was a statistically signi cant difference in the incidence of pneumonia among these three subgroups (P<0.05).
The clinical characteristics of patients in the SAP group and non-SAP group, and the relevant statistical results are summarized in Table 1.
Logistic regression analysis was further used to calculate the adjusted OR value of each potential risk factor.The logistic regression analysis results revealed that elderly age (≥75 years old), male gender, atrial brillation, dysphagia, an NIHSS score of 5-15 points, NIHSS of ≥16 points, and fasting hyperglycemia were independent risk factors of SAP (P<0.05).The detailed statistical results are prsented in Table 2.The adjusted OR value of fasting hyperglycemia was 2.95, which was assigned with 2 points in the modi ed A2DS2 scoring system (Table 3).
ROC curve analysis was used to evaluate the sensitivity and speci city of each cut-off value.The statistical results are presented in Table 4, while the ROC curves are presented in Figure 1.When the cutoff value was 4 points, the sensitivity and the speci city were 80.5% and 67.9%, respectively.Hence, 4 points can be used as the optimal operating point of the modi ed A2DS2 scoring system.
The area under curve of the modi ed A2DS2 scoring system was signi cantly higher than that of the conventional A2DS2 scoring system (F test, P<0.05; Table 5 and Figure 1), indicating the modi ed A2DS2 scoring system had higher predictive e ciency.

Discussion
In the present study, the value of the modi ed A2DS2 scoring system (combining the conventional A2DS2 scoring system with hyperglycemia) for predicting SAP was investigated.It was found that fasting hyperglycemia is an independent risk factor of SAP, which may be a valuable indicator for predicting SAP.Moreover, the predictive value of the modi ed A2DS2 scoring system is better than that of the conventional A2DS2 scoring system.
In present clinical practice, evaluating the risk of SAP remains challenging.The A2DS2 scoring system, in which scoring items include age, gender, atrial brillation, dysphagia and severity of stroke, has been proven to be a simple and reliable scoring scale.However, in literature, hyperglycemia has also been reported as a potential risk factor of SAP, although the evidence remains controversial.Hoffman et al.
found that the history of diabetes was not an independent risk factor for SAP [6].It is noteworthy that the history of diabetes cannot represent an abnormal blood glucose level, and temporary hyperglycemia may indicate stress hyperglycemia, rather than diabetes.In addition, diabetic patients with poor blood glucose control would most likely suffer from cerebral infarction.Thus, the correlation between hyperglycemia and SAP, as well as the value of hyperglycemia for predicting SAP, were investigated.
According to the international diagnosis and treatment guidelines for acute stroke, blood glucose level is recommended as a routine screening index for all patients.In the present study, the random blood glucose level after the onset of stroke represented stress hyperglycemia.Fasting hyperglycemia represented diabetes with poor glycemic control or newly-onset diabetes.Random or fasting hyperglycemia represented an increased blood glucose level caused by different causes.After statistical comparisons, merely fasting hyperglycemia entered the logistic regression model.It was speculated that stress hyperglycemia may be not an independent risk factor for SAP, and occasional transient hyperglycemia cannot increase the risk of SAP.Patients with fasting hyperglycemia (diabetes with poor glycemic control or newly-onset diabetes) are more likely to have SAP, which is consistent with previous ndings [5,13,14].
The present study indicated that fasting hyperglycemia is an independent risk factor for SAP.Acute ischemic stroke can cause hyperglycemia mainly through the following mechanisms: the activation of the sympathetic and parasympathetic nervous system [15][16][17], and the immune response of the hypothalamic-pituitary-adrenal axis [16][17][18].Hyperglycemia can reduce the bactericidal ability of leukocytes, increasing the likelihood of pulmonary infection [19,20].The study conducted by Obiako et al. revealed that the proportion of hyperglycemia was greater than that of diabetes in patients with acute stroke [21], suggesting that the poor prognosis of acute stroke may be attributed to hyperglycemia induced by stress reaction, rather than diabetes.
A number of studies have shown that hyperglycemia is signi cantly correlated with the occurrence of pneumonia and the poor outcome of acute ischemic stroke, especially in patients without diabetes.Dziedzic et al. noted that the incidence of pneumonia was higher in non-diabetic patients with fasting hyperglycemia.Nevertheless, the multivariate analysis revealed that fasting hyperglycemia was not signi cantly associated with pneumonia [22].Hirata et al. reported that the mortality of pneumonia was signi cantly correlated with hyperglycemia during hospitalization, but was not correlated to the history of diabetes [23].It was speculated that hyperglycemia may be associated with the severity and poor prognosis of acute stroke in non-diabetic patients, and diabetic patients may have adapted to the longterm hyperglycemia, which can protect the brain tissue against acute blood glucose increase.
The management of hyperglycemia should be highlighted during hospitalization for reducing the risks of SAP.Blood glucose level is correlated to the functions of various intracranial systems, such as the cerebrovascular system, in ammatory system, and metabolic system [24].Appropriate blood glucose control can improve immunosuppression and decrease the incidence and severity of infection.The optimal treatment of hyperglycemia in patients with acute stroke remains to be well-elucidated.
Furthermore, the average age of patients in the present study was lower than that reported in the study conducted by Hoffman et al. (61.9 ± 12.7 vs. 71.2± 13.1) [6].This discrepancy may indicate a different age distribution between China and Germany.ROC curve analysis has been widely used for making the best diagnostic criteria, and determining the best critical value, while the area under curve can represent the e ciency of the prediction.In the present study, the area under curve of the modi ed A2DS2 scoring system was signi cantly higher than that of the conventional A2DS2 scoring system, suggesting that the modi ed system (including the item of hyperglycemia) is more effective for predicting SAP.This modi ed A2DS2 scoring system may help in the early identi cation of stroke with high SAP risks, allowing timely prophylactic treatment, such as antibiotic therapy and the prophylactic use of aspiration.
The present study has some strengths.At present, reliable tools for predicting the risk of SAP include the A2DS2 scoring system and AIS-APS scale [5,6].In the present study, the former one was utilized, since it was more simple and practical.In addition, the inclusion and exclusion criteria were strict in the present study, and all researchers were uniformly trained [1].The novelty of the present study was the combination of the A2DS2 scoring system and evaluation of hyperglycemia.The present ndings may improve the predictive value of the A2DS2 scoring system.
There were still some limitations in the present study.First, the dynamic changes of the blood glucose levels of patients were not monitored throughout hospitalization, and only the random blood glucose level on admission was assessed.Second, the single-center and retrospective design was an inherent defect of the present study.In the future research, more external veri cations are needed to arrive at a de nitive conclusion.

Conclusion
Fasting hyperglycemia is an independent risk factor for predicting SAP.The predictive value of the modi ed A2DS2 score (combined A2DS2 score and fasting hyperglycemia) is higher than that of the A2DS2 score.
List Of Abbreviations: analysis, paper writing and literature review.LSM provided some data and literature.JSC assists in consulting literature and revising papers.All authors read and approved the nal manuscript.There was no difference in gender and site the SAP group and non-SAP group (P>0.05).The incidence of SAP in with the following factor(s) higher than that patients without the following factor(s) (P<0.05):elderly (≥75 old), atrial fibrillation, dysphagia, random hyperglycemia, fasting hyperglycemia, or random or fasting hyperglycemia.Therefore, these factors were considered as risk factors for These patients were classified into three subgroups according to stroke severity (NIHSS score), and there was significant difference in the incidence of pneumonia among these three subgroups (P<0.05).The logistic regression analysis results revealed that elderly age (≥75 years old), male atrial fibrillation, dysphagia, an NIHSS score of points, NIHSS of ≥16 points, hyperglycemia were risk factors of SAP (P<0.05).The adjusted OR value of fasting hyperglycemia was 2.95, which was assigned with 2 points in the modified A 2 DS 2 scoring system.When the cut-off value was 4 points, the sensitivity and the specificity were 80.5% and 67.9%, respectively.

This retrospective study enrolled 2 ,
552 patients with acute ischemic stroke from the First Hospital of Shanxi Medical University between January 2012 and December 2016.Inclusion criteria: (a) magnetic resonance imaging (MRI) revealing an acute cerebral infarction; (b) the patient was admitted within seven days after onset of stroke; (c) random blood glucose level on admission and fasting blood glucose level at the next morning after admission are available.Exclusion criteria: (a) transient ischemic attack (TIA), (b) patients who were discharged or died within three days after onset, (c) patients with pre-existing pneumonia before admission, (d) the lack of more than one of the A2DS2 scoring items, or (e) mechanical ventilation.The A2DS2 scoring system comprised of the following: (1) 1 point for elderly age (≥75 years old); (2) 1 point for male gender; (3) 1 point for atrial brillation; (4) 2 points for dysphagia; (5) 3 points for an National Institute of Health stroke scale (NIHSS) score within 5-15 points; (6) 5 points for an NIHSS score >16 points.

Table 1 .
Statistical analysis of clinical characteristics between two groups

Table
Stepwise logistic regression analysis showing independent risk factors of

Table 4 .
Cut-off values and corresponding sensitivity and specificity

Table 5 .
Area under curve