a b s t r a c t

Purpose: The predictive value of Serum albumin in adult Aspiration pneumonia patients remains unknown. Methods: Using data collected during a 3-year retrospective cohort of hospitalized adult patients with aspiration pneumonia, we evaluated the predictive value of serum Albumin level at ED presentation for in-hospital mortal- ity.

Results: 248 Patients were enrolled; of these, 51 cases died (20.6%). The mean Serum albumin level was 3.4 +-

0.7 g/dL and serum albumin levels were significantly lower in the non-survivor group than in the survivor group (3.0 +- 0.6 g/dL vs. 3.5 +- 0.6 g/dL). In the multivariable logistic regression model, albumin was associated with in-hospital mortality significantly (adjusted odds ratio 0.30, 95% confidential interval (CI) 0.16-0.57). The area under the receiver operating characteristics (AUROC) for in-hospital survival was 0.72 (95% CI 0.64-0.80). The Youden index was 3.2 g/dL and corresponding sensitivity, specificity, positive predictive value, negative pre- dictive value, positive and negative likelihood ratio were 68.6%, 66.5%, 34.7%, 89.1%, 2.05 and 0.47, respectively. High sensitivity (98.0%) was shown at albumin level of 4.0 g/dL and high specificity (94.9%) was shown at level of 2.5 g/dL.

Conclusion: Initial serum albumin levels were independently associated with in-hospital mortality among adult patients hospitalized with aspiration pneumonia and demonstrated fair Discriminative performance in the pre- diction of in-hospital mortality.

(C) 2017

Introduction

Background

The incidence of aspiration pneumonia is increasing because of the increased life span and increased number of patients with debilitating conditions such as stroke, dementia and Parkinsonism [1]. A previous study estimated that aspiration pneumonia accounted for 15% of com- munity-based pneumonia cases [2].

Various factors have been shown to be associated with mortality among patients with aspiration pneumonia. Most of the identified fac- tors have been patient characteristics such as older age, living in a nurs- ing home, a high degree of dependence, and hemiplegia [3] [4]. Among widely used laboratory markers, lymphocyte counts, serum creatinine

* Corresponding author at: Department of Emergency Medicine, Research Institute of Clinical Medicine of Chonbuk National University and Biomedical Research Institute of Chonbuk National University Hospital, 567 Baekje-daero, Deokjin-gu, Jeonju-si, Jeollabuk-do 54907, Republic of Korea.

E-mail address: akynei@naver.com (S. Jo).

levels, serum LDH levels, and serum albumin levels have been identified in association with mortality [3] [4]. Serum albumin levels may also serve as a prognostic factor for community-acquired pneumonia , and the diagnostic performance of this parameter in CAP is well-known [5]. However, the diagnostic performance of serum albu- min levels has not yet been reported among patients with aspiration pneumonia.

In the present study, we determined the association between in- hospital mortality and serum albumin levels at the presentation to the emergency department (ED) among patients with aspiration pneumo- nia. Additionally, the diagnostic performance of serum albumin levels was investigated.

Methods

Study design and setting

We conducted a 3-year retrospective cohort study using data from adult patients hospitalized with aspiration pneumonia from 2013.1.1 to 2015.12.31. This study was approved by the Institutional Review

http://dx.doi.org/10.1016/j.ajem.2017.06.039

0735-6757/(C) 2017

Board (IRB) of the study hospital, and the IRB waived the requirement for informed consent for all subjects included in this study. The study hospital is a 1200-bed urban academic, tertiary-care, university hospital.

Selection of participants

All adult patients aged 18 years or older who consecutively admitted to the study hospital via the emergency department (ED) with a diagno- sis of aspiration pneumonia were enrolled in the present study. To iden- tify study subjects, we reviewed the charts of patients with an admission diagnosis of aspiration pneumonia (International statistical Classification of Diseases (ICD), 10th Revision, codes J690 (pneumonitis due to food and vomit), J690-00 (aspiration pneumonia due to food re- gurgitated), J690-01 (Aspiration pneumonia due to gastric secretions), J690-02 (aspiration pneumonia due to milk), J690-03 (aspiration pneu- monia due to vomit), J690-04 (aspiration pneumonia NOS), P239 (neo- natal aspiration syndrome, unspecified), and P249-00 (neonatal aspiration pneumonia NOS). The following exclusion criteria were employed: 1) admission via an out-patient clinic, 2) pediatric patients, and 3) patients without a pneumonia severity index score.

Measurement and data collection

A trained abstractor extracted data from the charts using a struc- tured data-collection form according to the guidelines recommended by Gilbert et al. [6]. The medical record review and data abstraction was performed by an emergency medicine resident who had 3 year of emergency medicine training. The resident had undergone a data-col- lection training that included defining eligibility criteria and identifying other variables for study inclusion. After the completion of data abstrac- tion, random chart reviews were performed to ensure data accuracy. The following data were abstracted:age; sex; nursing home residency; previous hospitalization; activities of daily living (ADL) dependence; L-tube in situ; Foley catheter in situ; percutaneous endoscopic gastrostomy (PEG) in situ; transferred from another hospital; comor- bidities (hypertension, diabetes mellitus , Chronic liver disease (CLD), chronic kidney disease (CKD), stroke, dementia, heart failure, malignancy and chronic airway disease); symptoms at presentation (cough, vomiting, and confusion); performance status, as assessed using the Eastern Cooperative Oncology Group (ECOG) scale; nursing assessment (presence of sores and risk of falls); physiological parame- ters, as measured at ED presentation (systolic blood pressure (SBP), di- astolic blood pressure (DBP), heart rate (HR), respiratory rate (RR), body temperature, mentality using Alert Verbal Pain Unresponsive (AVPU) scale, and National Early Warning Score ; initial labora- tory findings (albumin level, pH level, PaO₂ level, lactate level, Blood urea nitrogen level, sodium level, glucose level, white blood cell count, hematocrit (Hct) level, erythrocyte sediment rate (ESR), and C-reactive protein level); pleural effusion; PSI score; disposi- tion (admitted to the intensive care unit (ICU) or ward); length of ED stay (ED LOS); length of hospital stay; and survival status at hospital discharge.

Outcome measures

The primary outcome was in-hospital mortality. The secondary out- comes were 14 day mortality and 28 day mortality.

Statistical analysis

Continuous data are presented as means and standard deviations (SDs). If the data were not normally distributed, continuous data are presented as medians and interquartile ranges (IQRs). Discrete data are presented as both counts and percentages.

The Student’s t-test for independent samples was used to compare the means of normally distributed variables between the survivor group and the non-survivor group. The Mann-Whitney U test was used for the comparison of variables that were not normally distributed. For categorical data, either the chi square test or chi square test with Fisher’s exact test for 2 x 2 tables was used. The results were considered significant at a threshold of p b 0.05 (two-tailed).

An area under the Receiver operating characteristic curve

analysis was performed to determine the predictive value of serum al- bumin level. The cut-off value that maximized the sum of the sensitivity and specificity (Youden index) was identified for serum albumin level [7]. The diagnostic performance of serum albumin level was determined at various cut-off points using the Youden point, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), posi- tive likelihood ratio (LR+), and negative likelihood ratio .

We used locally weighted scatterplot smoothing (LOWESS) to graphically illustrate the association between in-hospital mortality and serum albumin levels. The LOWESS is a type of nonparametric re- gression used to summarize the association between two variables that demands relatively few assumptions regarding the strength or form of the relationship [8].

The associations between the primary outcome and each eligible variable were evaluated using univariate logistic regression analyses. Then, a multivariable logistic regression analysis was performed for fac- tors identified as associated with mortality at (p b 0.10) in the univariate analysis. The logistic regression analysis results are presented as odds ratios with 95% confidence intervals (95% CIs). Multicollinearity be- tween variables was tested.

We additionally compared the baseline characteristics of groups cat- egorized by the presence or absence of hypoalbuminemia (hypoalbu- minemia vs. non-hypoalbuminemia). The cut-off serum albumin level was defined as the value of Youden index in the hypoalbuminemia group. A logistic regression analysis was performed to examine the as- sociation between baseline characteristics and the occurrence of hypoalbuminemia.

We referred to the Standards for Reporting Diagnostic Accuracy (STARD) statement for the analysis of the results [9]. All analyses were performed using STATA 11.1 (StataCorp LP, College Station, TX) and SAS 9.1 (SAS Institute, Inc., Cary, NC).

Results

During the study period, the charts of 311 patients were reviewed. After excluding 45 patients (27 patients admitted via out-patient clinics and 18 pediatric patients), 266 patients met the eligibility criteria. Among these patients, we excluded 18 patients without a PSI score (Fig. 1). There was no missing data among 248 enrolled patients.

Table 1 shows the baseline characteristics of enrolled patients. A total of 248 patients were enrolled in the study. The mean age of partic- ipants was 73.9 +- 13.5, and 185 male patients were included (74.6%). Approximately one half of patients were ADL dependent (46.4%), and one third of patients were nursing home residents (30.2%). An L-tube was placed in 38 (15.3%) patient; a foley catheter was placed in 85 pa- tients (34.3%), and a PEG was placed in 11 patients (4.4%). Hypertension (44.0%) was the most frequent identified co-morbidity, followed by DM (27.0%) and stroke (23.0%).

Overall, 51 cases died (20.6%). Significant differences were not iden- tified between the survival group and non-survival group for the major- ity of patient. Unexpectedly, the rates of DM and malignancy were lower and higher in the non-survival group, respectively. More patients were assessed as being at high risk of falls in the non-survival group (65.5% vs. 84.3%). Among the evaluated physiologic parameters, SBP, DBP, and body temperature were significantly lower in the non-survival group; however, significant differences were not identified in any of the other assessed parameters. 23 patients were determined dead at day 14 and 35 patients at day 28.

Fig. 1. STARD flow diagram for the study.

The mean serum albumin level was 3.4 +- 0.7 g/dL, and this value was significantly lower in the non-survivor group than in the survivor group (3.0 +- 0.6 g/dL vs. 3.5 +- 0.6 g/dL). Among the other evaluated laboratory parameters, PaO2 and BUN levels were lower in non-survival than in the survival group; however, the remaining variables were not significantly different.

The rate of pleural effusion was slightly lower (24.4% vs. 11.8%, p =

0.052) and PSI scores were slightly higher in the non-survival group (112.5 +- 32.0 vs. 122.1 +- 36.3, p = 0.07) relative to the survival group. Approximately one half of non-survival group admitted to the ICU (58.8%).

The predictive value of serum albumin level in determining survival at discharge (the reverse outcome to in-hospital mortality) was assessed using the AUROC method, which yielded a value of 0.72 (95% CI 0.64-0.80) (Fig. 2). The AUROC value was 0.79 (95% CI 0.71-0.87)

for 14 day mortality and 0.80 (0.74-0.87) for 28 day mortality. The cut-off value for serum albumin level that maximized the sum of sensi- tivity and specificity (Youden index) was 3.2 g/dL. Table 2 shows the di- agnostic performance of serum albumin for the prediction of in-hospital mortality at various cut-off levels. At a serum albumin level of 3.2, the sensitivity, specificity, PPV, NPV, LR+ and LR- were 68.6%, 66.5%, 34.7%, 89.1%, 2.05, and 0.47. The sensitivity was 98% at an albumin level of 4.0 g/dL, and the specificity was 94.9% at an albumin level of

2.5 g/dL.

The results of the LOWESS curve analysis are shown in Fig. 3. A near linear relationship was demonstrated between the measured serum al- bumin level and the observed cumulative mortality rate at hospital dis- charge. No apparent flattening was identified at any point on either side

of the serum albumin curve, suggesting that no additional cases of in- hospital death were undetected.

In the univariate logistic regression analysis, DM, malignancy, high risk of fall, DBP, body temperature, albumin, PaO2, BUN, pleural effusion, PSI, and ICU admission were significantly associated with mortality (Table 3). In the multivariable logistic regression model, albumin remained significantly associated with mortality (adjusted odds ratio (AOR) 0.30, 95% confidential interval (CI) 0.16-0.57). Among laboratory parameters, serum albumin was the only variable identified as significantly associated with mortality. DM, malignancy, pleural effusion, ED LOS, and ICU admission were remained signifi- cantly associated with in-hospital mortality. The majority of tradi- tional patient related factors were not identified as significantly associated with mortality in logistic regression analysis of the pres- ent cohort. For secondary outcome, serum albumin was most signif- icantly associated with outcomes (AOR 0.29, 95% CI 0.13-0.66 for 14 day mortality and AOR 0.20, 95% CI 0.09-0.45 for 28 day mortali-

ty) (Supplementary Tables 1 and 2).

Comparisons of the baseline characteristics of the non-hypoalbu- minemia group and hypoalbuminemia group were performed. Hy- poalbuminemia was defined as a serum albumin level of 3.2 g/dL, which was Youden index value. Nursing home residency, previous hospitalization, Foley catheter in situ, chronic airway disease, and bedsores were more frequently identified in the hypoalbuminemia group. However, DM was less frequently observed in the hypoalbu- minemia group (Table 4).

Logistic regression analysis was performed to assess the associations between baseline characteristics and the occurrence of

Table 1

Baseline characteristics of enrolled patients

Number	All		Survival group		Non-survival group	p-Value
	248 (100%)		197 (79.4%)		51 (20.6%)
Age, yr	73.9 +- 13.5		74.2 +- 13.4		72.8 +- 14.0	0.51
b 65	42 (16.9)		36 (18.3)		6 (11.8)	0.17
65-74	56 (22.6)		43 (21.8)		13 (25.5)
75-84	106 (42.7)		79 (40.1)		27 (52.9)
>= 85	44 (17.7)		39 (19.8)		5 (9.8)
Male sex, %	185 (74.6)		150 (76.1)		35 (68.6)	0.27
Nursing home residency	75 (30.2)		58 (29.4)		17 (33.3)	0.59
Previous hospitalization	150 (60.5)		116 (58.9)		34 (66.7)	0.31
Dependence daily activity	115 (46.4)		92 (46.7)		23 (45.1)	0.84
L-tube	38 (15.3)		32 (16.2)		6 (11.8)	0.43
Foley catheter	85 (34.3)		63 (32.0)		22 (43.1)	0.14
PEG	11 (4.4)		8 (4.1)		3 (5.9)	0.70
Transferred from another hospital	131 (52.8)		103 (52.3)		28 (54.9)	0.74
Co-morbidity
Hypertension, %	109 (44.0)		83 (42.1)		26 (51.0)	0.26
Diabetes mellitus, %	67 (27.0)		62 (31.5)		5 (9.8)	b 0.01
Chronic liver disease, %	7 (2.8)		4 (3.0)		3 (5.9)	0.16
Chronic kidney disease, %	13 (5.2)		11 (5.6)		2 (3.9)	1.00
Stroke, %	57 (23.0)		45 (22.8)		12 (23.5)	0.92
Dementia, %	26 (10.5)		19 (9.6)		7 (13.7)	0.28
Heart failure, %	16 (6.5)		14 (7.1)		2 (3.9)	0.54
Malignancy, %	27 (10.9)		17 (8.6)		10 (19.6)	0.03
Chronic airway disease, %	11 (4.4)		10 (5.1)		1 (2.0)	0.47
Charlson Comorbidity Index	5.5 +- 2.5		5.5 +- 2.5		5.7 +- 2,4	0.59
Cough	50 (20.2)		42 (21.3)		8 (15.7)	0.37
Vomiting	23 (9.3)		18 (9.1)		5 (9.8)	0.88
Confusion	42 (16.9)		37 (18.8)		5 (9.8)	0.15
ECOG score						0.89
1	127 (51.2)		100 (50.8)		27 (52.9)
2	38 (15.3)		30 (15.2)		8 (15.7)
3	33 (13.3)		28 (14.2)		5 (9.8)
4	50 (20.2)		39 (19.8)		11 (21.6)
bedsores	20 (8.1)		17 (8.6)		3 (5.9)	0.77
High risk of falls	172 (69.4)		129 (65.5)		43 (84.3)	b 0.01
Physiological parameters
Systolic blood pressure, mm Hg	117.0 +- 30.0		119.5 +- 29.5		107.6 +- 30.6	0.01
Diastolic blood pressure, mm Hg	70.8 +- 17.0		72.6 +- 15.7		64.2 +- 19.9	b 0.01
Pulse rate, bpm	93.7 +- 21.7		97.5 +- 22.8		90.5 +- 16.7	0.24
Respiratory rate, bpm	19.8 +- 2.6		19.7 +- 2.6		20.0 +- 2.3	0.55
Body temperature, ?C	36.8 +- 0.8		36.8 +- 0.8		36.6 +- 0.8	0.03
SpO2	95.0 +- 5.9		95.2 +- 5.4		94.0 +- 7.6	0.18
Mentality – alert %	148 (60.0)		115 (58.4)		33 (64.7)	0.59
Mentality – verbal %	36 (14.5)		28 (14.2)		8 (15.7)
Mentality – pain	7 (2.8)		7 (3.6)		0
Mentality – unresponsive, %	57 (23.0)		47 (23.9)		10 (19.6)
NEWS	4.8 +- 3.5		4.7 +- 3.5		4.8 +- 3.6	0.95
Laboratory parameters
Albumin	3.4 +- 0.7		3.5 +- 0.6		3.0 +- 0.6	b 0.01
pH	7.4 +- 0.1		7.4 +- 0.1		7.4 +- 0.1	0.64
PaO2	90.5 +- 40.3		93.3 +- 42.1		79.9 +- 30.5	0.03
Lactate	2.4 +- 2.2		2.3 +- 2.3		2.7 +- 2.2	0.30
BUN	28.6 +- 22.9		27.1 +- 19.7		34.3 +- 32.2	0.045
Sodium	136.9 +- 7.5		136.9 +- 7.1		137.0 +- 8.9	0.99
Glucose	159.8 +- 91.6		160.4 +- 97.2		157.5 +- 66.5	0.84
WBC	12.0 +- 6.4		12.0 +- 6.3		12.1 +- 6.4	0.89
ESR	38.9 +- 29.2		37.7 +- 28.9		43.6 +- 30.1	0.20
CRP	85.9 +- 79.5		87.9 +- 79.7		78.1 +- 78.9	0.44
Hct	33.7 +- 6.2		33.6 +- 6.2		34.2 +- 6.0	0.51
Pleural effusion	54 (21.8)		48 (24.4)		6 (11.8)	0.05
PSI	114.5 +- 33.1		112.5 +- 32.0		122.1 +- 36.3	0.07
ED LOS, hr	42.6 +- 47.8		45.3 +- 46.7		32.1 +- 40.7	0.07
Hospital LOS	31.5 +- 45.5		32.6 +- 44.0		27.5 +- 51.2	0.48
ICU admission	88 (35.5)		58 (29.4)		30 (58.8)	b 0.01
Ward admission	160 (64.5)		139 (70.6)		21 (41.2)	b 0.01

Abbreviations L-tube, Levin tube; PEG, percutaneous endoscopic gastrostomy; ECOG, Eastern Cooperative Oncology Group; NEWS, national early warning score; BUN, blood urea nitrogen; WBC, white blood cell; ESR, erythrocyte sediment rate; CRP, C-reactive protein; Hct, hematocrit; PSI pneumonia severity index; ED LOS, emergency department length of stay; ICU inten- sive care unit.

hypoalbuminemia. Among the candidate variables, nursing home resi- dency, Foley catheter in situ, and DM were identified as significantly as- sociated with the occurrence of hypoalbuminemia (Table 5).

Subgroup analysis was also performed after excluding chronic liver and kidney disease, which could cause distribution disturbance of the serum albumin. Serum albumin is the only significant factor for in-

Fig. 2. AUROC of serum albumin for the prediction of in-hospital mortality among aspiration pneumonia patients.

Fig. 3. The locally weighted scatterplot smoothing curve analysis demonstrated a near- linear relationship between serum albumin level and the cumulative rate of in-hospital mortality.

hospital mortality, 14 day mortality, and 28 day mortality (Supplemen- tary Table 3).

Discussion

In the present study, serum albumin was shown to be associated with in-hospital mortality among patients with aspiration pneumonia even after adjusting for patient characteristics and laboratory parame- ters. The AUROC value was fair (0.72), and the sensitivity and specificity were relatively low (68.6% and 66.5%), even at Youden index point (3.2 g/dL). However, the use of a relatively high cut-off level for albumin (4.0 g/dL) was associated with high sensitivity (98.0%), and the use of a low cut-off level for albumin (2.5 g/dL) showed high specificity (94.9%). Nursing home residency and a Foley catheter in situ were associated with the occurrence of hypoalbuminemia.

To our knowledge, only two studies have previously demonstrated an association between mortality and serum albumin among patients with aspiration pneumonia. In 2012, Bosch et al. reported that lower al- bumin levels were associated with Six-month mortality (AOR 1.129, 95% CI 1.008-1.265) after adjusting age, prior and admission Barthel index (BI), prealbumin level, total protein level, cholesterol level, lym- phocyte count, and assistance required when eating using data from 120 patients [4]. In 2014, Ogasawara et al. reported that hypoalbumin- emia, when defined as an albumin level of 3.0 g/dL or less was associat- ed with in-hospital mortality (AOR 7.26, 95% CI 1.679-31.383) after adjusting sex, community-acquired pneumonia severity score (A- DROP age, dehydration, respiration, orientation, pressure), CRP level, BUN level, and duration of antibiotic use using data from 105 patients [10].

Results consistent with aforementioned study were found in the present study. The results of the multivariable logistic regression analy- sis revealed that the association between serum albumin level and in- hospital mortality was highly significant (AOR 0.24, 0.12-0.47, p b 0.001), as were the associations between in-hospital mortality and DM, malignancy, and high risk of falls. Discriminatory value for

predicting in-hospital mortality was fair (AUROC 0.72, 0.64-0.80) even at the Youden index (3.2 g/dL), and sensitivity and specificity were low at 68.6% and 66.5%, respectively. However, the use of a rela- tively high level of albumin (4.0 g/dL) was associated with high sensitiv- ity (98.0%), and the use of a low level of albumin (2.5 g/dL) was associated with high specificity (94.9%). These findings seem to have meaningful clinical significance. Interestingly, AUROC value (0.72) of serum albumin level for the prediction of in-hospital mortality identi- fied in this study was similar to that identified for 30-day mortality among patients with community acquired pneumonia [5].

Some theories exit regarding the reasons why lower serum albumin levels, or hypoalbuminemia, may be associated with poor outcomes. Be- cause synthesis and distribution of albumin may be directly associated with serum albumin level, factors that could affect albumin synthesis, distribution, or both need to be considered. Theoretically albumin deg- radation may cause hypoalbuminemia, however, at present the degra- dation of albumin is poorly understood.

Albumin is synthesized in the liver and derived from proalbumin. Proalbumin is derived from preproalbumin which are synthesized by polysomes (poly-ribosomes) using the amino-acids in hepatocytes. Therefore, decreased liver function or insufficient amino-acids intake may result in the hypoalbuminemia. In the present study, malnutrition was more frequently identified in the hypoalbuminemia group. A distri- bution disturbance between intravascular albumin levels and extravas- cular albumin levels may also exist. Certain conditions may decrease serum albumin levels such as pleural effusion, ascites, edema, or ne- phrosis. We confirmed that patients with debilitating conditions were more frequently identified in the hypoalbuminemia group than the non-hypoalbuminemia group in this study cohort. CLD or CKD were sel- dom identified, and no significant difference was identified in the rates of CLD or CKD between the hypoalbuminemia group and non-hypoal- buminemia group. (See Table 4.)

In addition to aforementioned effect of hypoalbuminemia in associ- ation with other pathological conditions, hypoalbuminemia may be

Table 2

Diagnostic performance of serum albumin for the prediction of in-hospital mortality among aspiration pneumonia patients

Cutoff level	% of patients	SN, % (95% CI)	SP, % (95% CI)	PPV, % (95% CI)	NPV, % (95% CI)	LR+ (95% CI)	LR- (95% CI)
<= 2.5 g/dL	9.3%	25.5 (14.3-39.6)	94.9 (90.9-97.5)	56.5 (34.5-76.8)	83.1 (77.6-87.8)	5.02 (2.34-10.80)	0.79 (0.67-0.93)
<= 3.0 g/dL	28.2%	49.0 (34.8-63.4)	77.2 (70.7-82.8)	35.7 (24.6-48.1)	85.4 (79.3-90.2)	2.15 (1.47-3.14)	0.67 (0.50-0.87)
<= 3.2 g/dL	40.7%	68.6 (54.1-80.9)	66.5 (59.4-73.0)	34.7 (25.5-44.8)	89.1 (82.9-93.6)	2.05 (1.56-2.68)	0.47 (0.31-0.72)
<= 3.5 g/dL	56.0%	80.4 (66.9-90.2)	50.3 (43.1-57.4)	29.5 (22.1-37.8)	90.8 (83.8-95.5)	1.62 (1.33-1.96)	0.39 (0.22-0.69)
<= 4.0 g/dL	81.9%	98.0 (89.6-100.0)	22.3 (16.7-28.8)	24.6 (18.9-31.2)	97.8 (88.2-99.9)	1.26 (1.16-1.37)	0.09 (0.01-0.62)

Abbreviations: SN, sensitivity; SP, specificity; PPV, positive predictive value; NPV, negative predictive value; LR+, positive likelihood ratio positive; LR-, negative likelihood ratio.

Table 3

Logistic regression analysis for in-hospital mortality among aspiration pneumonia patients

	Unadjusted odd ratio	p-Value	Adjusted odd ratio	p-Value
Diabetes mellitus	0.24 (0.09-0.62)	0.004	0.22 (0.08-0.65)	0.006
Malignancy	2.58 (1.10-6.05)	0.029	5.06 (1.57-16.34)	0.007
High risk of falls	2.83 (1.26-6.37)	0.012	2.78 (1.02-7.60)	0.046
Diastolic blood pressure	0.97 (0.95-0.99)	0.002
Body temperature	0.65 (0.44-0.97)	0.035
Albumin	0.29 (0.17-0.79)	b 0.001	0.30 (0.16-0.57)	b0.001
PaO2	0.99 (0.98-1.00)	0.036	0.99 (0.98-1.00)	0.053
BUN	1.01 (1.00-1.02)	0.060	1.01 (1.00-1.04)	0.063
Pleural effusion	0.41 (0.17-1.03)	0.058	0.27 (0.09-0.83)	0.022
PSI	1.01 (1.00-1.02)	0.067
ICU	3.42 (1.81-6.47)	b 0.001	4.26 (1.91-9.49)	b0.001

Abbreviations BUN, blood urea nitrogen; PSI pneumonia severity index; ED LOS, emergen- cy department length of stay; ICU intensive care unit.

Systolic blood pressure and diastolic blood pressure correlated with a Pearson p-value of 0.88.

directly linked to opposing treatment effects. Nearly all drugs, including antibiotics can bind with plasma protein and form protein-drug com- plexes. Because the unbound fraction of drugs exhibits a pharmacologic effect [11], lower serum protein levels would be beneficial. However, protein-drug complexes can escape via renal tubular secretion or hepat- ic metabolism [12] and may serve as a reservoir, resulting in a slow re- lease of the drug in an active unbound form. These effects may be associated with an increased biological half-life of drug; thus, contrary to the prior mechanism, higher serum protein levels would be beneficial in this case.

Table 4

Comparisons between the non-hypoalbuminemia group and hypoalbuminemia (serum albumin <= 3.2 g/dL) group

Non-hypoalbuminemia Hypoalbuminemia p-Value

Age, y	74.7 +- 12.9	72.7 +- 14.2	0.26
b65	22 (15.0)	20 (19.8)	0.44
65-74	30 (20.4)	26 (25.7)
75-84	68 (46.3)	38 (37.6)
>= 85	27 (18.4)	17 (16.8)
Male sex, %	113 (76.9)	72 (71.3)	0.32
Nursing home residency	36 (24.5)	39 (38.6)	0.02
Previous hospitalization	81 (55.1)	69 (68.3)	0.04
Dependence daily activity	62 (42.2)	53 (52.5)	0.11
L-tube	21 (14.3)	17 (16.8)	0.58
Foley catheter	41 (27.9)	44 (43.6)	0.01
PEG	5 (3.4)	6 (5.9)	0.36
Transferred from other	83 (56.5)	48 (47.5)	0.17
hospital Hypertension, %	68 (46.3)	41 (40.6)	0.38
Diabetes mellitus, %	48 (32.7)	19 (18.8)	0.02
CLD, %	6 (4.1)	1 (1.0)	0.25
CKD, %	10 (6.8)	3 (3.0)	0.25
Stroke, %	39 (26.5)	18 (17.8)	0.11
Dementia, %	12 (8.2)	14 (13.9)	0.15
HF, %	10 (6.8)	6 (5.9)	0.79
Malignancy, %	13 (8.8)	14 (13.9)	0.21
Chronic airway disease, %	3 (2.0)	8 (7.9)	0.03
Cough	29 (19.7)	21 (20.8)	0.84
Vomiting	13 (8.8)	10 (9.9)	0.78
Confusion	24 (16.3)	18 (17.8)	0.76
ECOG			0.60
1	80 (54.4)	47 (46.5)
2	22 (15.0)	16 (15.8)
3	19 (12.9)	14 (13.90)
4	26 (17.7)	24 (23.8)
Sore	8 (5.4)	12 (11.9)	0.07
High risk of falls	102 (69.4)	70 (69.3)	0.99

Abbreviations L-tube, Levin tube; PEG, percutaneous endoscopic gastrostomy; ECOG, East- ern Cooperative Oncology Group; NEWS, national early warning score; BUN, blood urea nitrogen; WBC, white blood cell; ESR, erythrocyte sediment rate; CRP, C-reactive protein; Hct, hematocrit; PSI pneumonia severity index; ED LOS, emergency department length of stay; ICU intensive care unit.

An association between hypoalbuminemia and Acute kidney injury has also been reported. A meta-analysis of seventeen clinical stud- ies including 3917 patients showed that lower serum albumin was an independent predictor of both AKI occurrence and death after AKI oc- currence. With each 1.0 g/dL decrease in serum albumin, the odds of AKI occurrence increased by 134%, and the pooled OR for AKI occurrence was 2.34 (95% CI 1.74-3.14). Among patients in whom AKI occurred, the odds for mortality increased by 147% with each 1.0 g/dL decrease in serum albumin and the pooled OR for mortality was 2.47 (95% CI 1.51-4.05) [13].

Even all aforementioned description, it is still not clear that hypoal- buminemia is a sign of sicker patients prior to occurrence of aspiration pneumonia, or hypoalbuminemia is a pathological process of the aspira- tion pneumonia. It seems nearly same as a chicken or egg argument. However, the results that the nursing home residency and foley catheter was associated with occurrence of the hypoalbuminemia in the present study imply that hypoalbuminemia is a sign of sicker patients before they got the aspiration pneumonia.

Actually, trials investigating the therapeutic role of albumin have shown mixed results, with findings suggesting that this application of albumin was both effective and ineffective. A multicenter, randomized, double-blind study of 6997 ICU patients (SAFE study) reported a similar rate of 28-day mortality between the 4% albumin arm and normal saline arm when investigation the use of albumin in fluid resuscitation [14]. Another multicenter, open-label trial including 1818 severe sepsis or septic shock patients (ALBIOS study) reported similar rates of 28- and 90-day mortality between the 20% albumin with crystalloid arm and the crystalloid alone arm. In that trial, the target serum albumin level was 3.0 g/dL [15]. However, a post-hoc analysis of data obtained from the severe sepsis patients included in the SAFE study showed a de- creased risk of mortality in the albumin arm in multivariable logistic re- gression model (AOR 0.71, 95% CI 0.52-0.97) [16]. In the most recent meta-analyses, albumin solution resuscitation was not found to reduce mortality among sepsis patients [17] or acute respiratory distress syn- drome (ARDS) patients [18].

In addition, it is not certain whether higher serum albumin levels are more beneficial or if a specific threshold of serum albumin level exists, because the aforementioned RCTs were not designed to answer this question. Therefore, further investigations regarding this issue are needed. The authors cautiously suggest that the strategy in which serum albumin levels are maintained at the highest level as possible may be beneficial based on the LOWESS results. No specific turning point at which no added benefit or harm existed in association with the primary outcome was identified on the LOWESS graph. This finding implies that higher serum albumin level were associated with greater chance of survival. Based on this result, the authors thought that poten- tial role of albumin administration would exist in aspiration pneumonia. Although the LOWESS result was obtained using data at initial presenta- tion and not after treatment, it could be representative of one-point in the patients’ course of illness.

Limitations

The present study had some limitations that should be noted. First, we used the initial serum albumin level measured when patients pre- sented to the ED. Strength of this approach was that the initial serum al- bumin level was unlikely to be affected by fluid resuscitation or other possible treatment during ED. However, we did not investigate the trends in serum albumin level, which could reveal more kinetic infor- mation and should be a subject of the further research. Second, we could not collect long-term mortality data. Third, this was a retrospec- tive observational single center study. Whether the results of the pres- ent study are replicable in the other regions is of question, because of potential differences in treatment quality and Hospital resources. Fourth, unknown confounding factors may have influenced the associa- tions between the predictive and outcome variables. However, we

Table 5

Logistic regression analysis for the occurrence of hypoalbuminemia (serum albumin <= 3.2 g/dL) among aspiration pneumonia patients

	Unadjusted odd ratio	p-Value	Adjusted odd ratio	p-Value
Nursing home residency	1.94 (1.12-3.36)	0.018	1.81 (1.01-3.24)	0.047
Previous hospitalization	1.76 (1.03-2.99)	0.037
Foley catheter	2.00 (1.17-3.40)	0.011	1.94 (1.08-3.48)	0.026
Diabetes mellitus	0.48 (0.26-0.88)	0.017	0.37 (0.19-0.72)	0.003
Chronic airway disease	2.58 (1.10-6.05)	0.029	3.68 (0.88-15.31)	0.074
Sore	1.34 (0.92-5.96)	0.074

attempted to collect extensive covariates data and avoid introducing potential bias. Fifth, we could not determine whether transferred pa- tients received albumin treatment before visit to the study hospital ED. In conclusion, we found that decreased serum albumin levels were associated with increased in-hospital mortality even after adjusting for covariates among patients with aspiration pneumonia. In addition, the initial serum albumin level showed fair discriminative performance

for predicting in-hospital mortality.

Conflict of interests

None.

Funding and support

None.

Author contributions

H.K and S.J designed this study. H.K supervised the overall data col- lection process, had full access to all the data in the study, and takes re- sponsibility for the integrity of the data. B.P conducted the data analysis. S.J wrote the initial draft of the article. All authors provided substantial review and feedback on the final version of the article. S.J takes responsibility for the paper as a whole.

All authors have read and approved the submitted manuscript. This manuscript has not been submitted nor published elsewhere in whole or in part.

Declaration of interest“>Declaration of interest

The authors report no conflicts of interest. The authors are responsi- ble for the content and composition of this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ajem.2017.06.039.

References

1. Akbar U, Dham B, He Y, et al. Incidence and mortality trends of aspiration pneumo- nia in Parkinson’s disease in the United States, 1979-2010. Parkinsonism Relat Disord 2015;21:1082-6.
2. Marik PE. Aspiration pneumonitis and aspiration pneumonia. N Engl J Med 2001; 344:665-71.
3. Ramos A, Asensio A, Caballos D, et al. Prognostic factors associated with community- acquired aspiration pneumonia. Med Clin (Barc) 2002;119:81-4.
4. Bosch X, Formiga F, Cuerpo S, et al. Aspiration pneumonia in old patients with de- mentia. Prognostic factors of mortality. Eur J Intern Med 2012;23:720-6.
5. Viasus D, Garcia-Vidal C, Simonetti A, et al. Prognostic value of serum albumin levels in hospitalized adults with community-acquired pneumonia. J Infect 2013;66: 415-23.
6. Gilbert EH, Lowenstein SR, Koziol-McLain J, et al. Chart reviews in emergency med- icine research: where are the methods? Ann Emerg Med 1996;27:305-8.
7. Perkins NJ, Schisterman EF. The inconsistency of “optimal” cutpoints obtained using two criteria based on the receiver operating characteristic curve. Am J Epidemiol 2006;163:670-5.
8. Bodo BA. Robust graphical methods for diagnosing trend in irregularly spaced water quality time series. Environ Monit Assess 1989;13:407-28.
9. Bossuyt PM, Reitsma JB, Bruns DE, et al. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ 2015;351:h5527.
10. Ogasawara T, Umezawa H, Naito Y, et al. Procalcitonin-guided antibiotic therapy in aspiration pneumonia and an assessment of the continuation of oral intake. Respir Investig 2014;52:107-13.
11. Bergogne-Berezin E. Clinical role of protein binding of quinolones. Clin Pharmacokinet 2002;41:741-50.
12. Lindup WE, Orme MC. Clinical pharmacology: plasma protein binding of drugs. Br Med J (Clin Res Ed) 1981;282:212-4.
13. Wiedermann CJ, Wiedermann W, Joannidis M. Hypoalbuminemia and acute kidney injury: a meta-analysis of observational clinical studies. Intensive Care Med 2010; 36:1657-65.
14. Finfer S, Bellomo R, Boyce N, et al. A comparison of albumin and saline for fluid re- suscitation in the intensive care unit. N Engl J Med 2004;350:2247-56.
15. Caironi P, Tognoni G, Masson S, et al. Albumin replacement in patients with severe sepsis or septic shock. N Engl J Med 2014;370:1412-21.
16. SAFE Study Investigators, Finfer S, McEvoy S, Bellomo R, et al. Impact of albumin compared to saline on Organ function and mortality of patients with severe sepsis. Intensive Care Med 2011;37:86-96.
17. Patel A, Laffan MA, Waheed U, et al. Randomised trials of human albumin for adults with sepsis: systematic review and meta-analysis with trial sequential analysis of all-cause mortality. BMJ 2014;349:g4561.
18. Uhlig C, Silva PL, Deckert S, et al. Albumin versus crystalloid solutions in patients with the acute respiratory distress syndrome: a systematic review and meta-analy- sis. Crit Care 2014;18:R10.