Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: A retrospective study based on MIMIC-Ⅲ and Ⅳ databases

Fei Yin; Zhenguo Qiao; Xiaofei Wu; Qiang Shi; Rongfei Jin; Yuzhou Xu

doi:10.1371/journal.pone.0300012

. 2024 Mar 7;19(3):e0300012. doi: 10.1371/journal.pone.0300012

Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: A retrospective study based on MIMIC-Ⅲ and Ⅳ databases

Fei Yin ¹, Zhenguo Qiao ², Xiaofei Wu ¹, Qiang Shi ¹, Rongfei Jin ¹, Yuzhou Xu ^3,^*

Editor: Tanja Grubić Kezele⁴

PMCID: PMC10919588 PMID: 38452113

Abstract

Background

To investigate the correlation between albumin-corrected anion gap(ACAG) within the first 24 hours of admission and in-hospital mortality in trauma patients in intensive care unit(ICU).

Materials and methods

We utilized the MIMIC-Ⅲ and MIMIC-Ⅳ databases to examine trauma patients admitted to the ICU. The relationship between ACAG and in-hospital mortality in trauma patients was analyzed using Receiver Operating Characteristic(ROC) curve, Kaplan-Meier (K-M) survival curve, and Cox regression model. Propensity score matching (PSM) and subgroup analysis were conducted to enhance stability and reliability of the findings. Mortality at 30-day and 90-day served as secondary outcomes.

Results

The study enrolled a total of 1038 patients. The AUC for ACAG (0.701, 95%CI: 0.652–0.749) was notably higher than that for anion gap and albumin. The Log-rank test revealed that the optimal cut-off point of ACAG for predicting in-hospital mortality was determined to be 20.375mmol/L. The multivariate Cox regression analysis demonstrated an independent association between high ACAG level and a higher risk of in-hospital mortality (HR = 3.128, 95% CI: 1.615–6.059). After PSM analysis, a matched cohort consisting of 291 subjects was obtained. We found no signifcant interaction in most stratas. Finally, The in-hospital, 30-day, and 90-day survival rates in the high ACAG group exhibited a statistically decrease compared to those in the low ACAG group both pre- and post-matching.

Conclusion

The elevated level of ACAG was found to be independently associated with increased in-hospital mortality among trauma patients in the ICU.

Introduction

Trauma ranks as the fourth leading cause of death worldwide, accounting for approximately 10% of all mortalities [1]. More than 2.8 million individuals were annually hospitalized in the United States due to trauma [2]. Hemorrhage caused by trauma often leads to shock, which was a primary contributor to early mortality among trauma patients. During the initial stages of trauma, it could be challenging to detect occult shock based on general clinical manifestations and physiological parameters such as heart rate, blood pressure, respiratory rate, and urine volume. Additionally, individuals with hypertension, atherosclerosis or prolonged use of certain cardiovascular medications may exhibit delayed responses to shock [3, 4]. Tissue hypoxia and hypoperfusion under shock conditions result in severe metabolic acidosis. The combination of acidosis along with hypotension and coagulopathy was referred to as the "trauma triad of death", which significantly predicts adverse outcomes within 24 hours following trauma [5].

The Anion gap (AG) reflected the disparity between unmeasured cations and anions concentration in serum and serves as one of the most commonly utilized biomarkers for diagnosing acid-base imbalances and identifying causes of metabolic acidosis. A study conducted by Ahmed et al. demonstrated that AG was an independent prognostic factor for severe trauma patients with an adjusted hazard ratio (HR) of 2.460 [6]. However, literature had noted that during the first hour after hemorrhagic shock onset, there was a greater increase in anion gap compared to serum lactate, this discrepancy may be attributed to uncorrected serum albumin’s influence on AG [7, 8]. Anion Gap Corrected for Albumin (ACAG) represents AG values adjusted according to ALB [9], potentially offering improved evaluation capabilities regarding metabolic acidosis and prognosis among trauma patients. The scarcity of previous studies on this topic necessitated the present study, which aimed to ascertain whether ACAG can provide a more accurate prediction of trauma outcomes.

Method

Database

This study was a retrospective analysis, utilizing data from MIMIC-Ⅲ Clinical Database CareVue subset and MIMIC-Ⅳ v2.2 databases. The patient population consisted exclusively of individuals admitted to intensive care units at the Beth Israel Deaconess Medical Center (BIDMC) in Boston, Massachusetts. The MIMIC-Ⅲ Clinical Database CareVue subset was derived from the lager MIMIC-Ⅲ Clinical Database v1.4, encompassing patients admitted between 2001 and 2008 [10]. MIMIC-Ⅳ v2.2 included 299,712 patients from 2008 to 2019 [11]. These two databases were mutually exclusive without any overlap or interference. The study involved an analysis of a de-identified publicly available database, which had received prior approval from the Institutional Review Board (IRB) at MIT and Beth Israel Medical Center. No additional ethics approval was necessary. The CITI Program course on Human Research and Data or Specimens Only Research had been successfully completed by us in order to obtain permission for accessing the databases (Record ID: 41,696,976). All individual patient information within these databases remained anonymous, with the exemption of ethical review and informed consent.

Patients

In this study, we utilized Structured Query Language (SQL) to extract data from the MIMIC-Ⅲ CareVue subset and MIMIC-Ⅳ databases using Navicat Premium (version 16.1.11). Patients meeting the criteria of having an initial diagnosis corresponding to trauma diagnosis codes in either ICD-9 (ranging from 800 to 959) or ICD-10 (ranging from S00-S99, T00-T14, and T20-T32) were included [12, 13]. In cases where patients had multiple admissions to the ICU, only the first admission was considered. The screening criteria consisted of: (1) excluding patients aged <18 years or >89 years; (2) excluding patients with an ICU stay duration of less than 24 hours; (3) excluding patients with missing important data such as AG, ALB, ACAG within 24 hours of ICU admission. Recorded variables included age, sex, race, Sequential Organ Failure Assessment (SOFA) score, Acute Physiology Score Ⅲ (APS Ⅲ), Glasgow Coma Scale (GCS) score, Simplified Acute Physiology Score Ⅱ (SAPS Ⅱ), Oxford Acute Severity of Illness Score (OASIS), comorbidities, and other clinical data. Additionally recorded were hematocrit, hemoglobin, platelet counts, white blood cell counts(Wbc), albumin(ALB), anion gap(AG), albumin corrected anion gap(ACAG), bicarbonate, urea nitrogen(BUN), creatinine, chloride, sodium, potassium, glucose, international normalized ratio(INR), prothrombin time(PT), partial thromboplastin time(PTT), mechanical ventilation status, length of hospital stays,and length of ICU stays. If a variable was recorded multiple times within the initial 24-hour period, the mean value was utilized.The AG values were calculated using the formula: AG(mmol/L) = sodium + potassium—chloride–bicarbonate [14],while ACAG values were calculated using the formula: ACAG(tendency) = [4.4—albumin(g/dl)] * 2.5 + AG [15]. The primary outcome measure was in-hospital mortality,while the secondary outcome measures included mortality at 30-day and 90-day.

Statistical analysis

Statistical analysis was performed by IBM SPSS software(Version 25.0), RStudio (Version 2022.07.0). A P-value < 0.05(two sided) was considered statistically signifificant. Variables with normal distributions were presented as the means ± SD and compared using a student T-test. The non-normally distributed variables were represented as medians and interquartile ranges (IQRs) and compared with the Mann-Whitney U-test. The counting variables were expressed as percentages and compared using the Chi-square test. According to the in-hospital survival outcome, the patients were classified into two distinct groups: the group that survived and the group that experienced mortality. We employed Random Forest method with multiple imputation to handle missing data [16]. Variables with a missing ratio exceeding 20% were excluded, and extreme values were mitigated using a 1% tail reduction approach [17]. Although the variable of lactate was missing by more than 20%, we included it due to its conventional role as a prognostic indicator for critical illness severity. Sensitivity analysis was performed using the complete data set. Receiver operating characteristic (ROC) curves for albumin (ALB), AG, and ACAG were plotted, and the area under the ROC curve was compared. The surv-cutpoint function [18] was employed to determine the optimal cut-off value of ACAG, which was subsequently utilized for stratifying patients into high and low ACAG level groups. Propensity score matching (PSM) analysis was conducted to minimize bias between these two patient groups. The patients were matched in a 2:1 ratio using nearest neighbor algorithm with a caliper width of 0.3, and standardized mean differences (SMDs) were subsequently calculated post-matching to assess balance between the groups. Cox proportional hazards models and subgroup analysis were utilized to examine the association between ACAG levels and in-hospital mortality among trauma patients. The log-rank test and survival curve analysis were performed to compare in-hospital, 30-day and 90-day survival rates between high ACAG level group and low ACAG level group before and after PSM.

Results

Baseline characteristics

A total of 1038 eligible patients were ultimately included in our study, obtained from MIMIC-Ⅲ and MIMIC-Ⅳ databases (Fig 1). The patients were stratified into two cohorts based on their hospital survival outcome, comprising 900 patients in the survival cohort and 138 patients in the mortality cohort. The mortality group exhibited a significantly prolonged duration of ICU stay compared to the survival group, whereas individuals who survived had a comparatively shorter overall length of hospital stay. The mortality group exhibited higher incidences of liver disease, congestive heart failure, cancer, and diabetes in comparison to the survival group. Additionally, SOFA score, SAPSⅡ score, APSⅢ score, OASIS score, age, respiratory rate, anion gap, ACAG, sodium, BUN, creatinine, lactate, INR, PT PTT, glucose, mechanical ventilation rate, mechanical ventilation duration were all lower among survivors than non-survivors. Conversely MBP, hematocrit, hemoglobin, platelet, albumin, bicarbonate were lower in non-survivors as opposed to survivors. Furthermore racial characteristics exhibited statistical differences between both groups (Table 1).

Table 1. Characteristics of the study population between in-hospital survival and in-hospital mortality group.

Variables	Overall	in-hospital survival	in-hospital mortality	t/z/χ²	P
N	1038	900	138
Age (year)	58.36 [39.19, 74.59]	56.23 [38.11, 72.91]	68.40 [52.66, 80.63]	-5.293	<0.001
Female, n(%)	335 (32.3)	280 (31.1)	55 (39.9)	4.185	0.051
Weight (kg)	76.50 [65.00, 89.10]	76.70 [65.00, 90.00]	74.75 [63.85, 85.00]	-0.996	0.319
Race, n(%)				12.097	0.017
Others	267 (25.7)	222 (24.7)	45 (32.6)
White	656 (63.2)	570 (63.3)	86 (62.3)
Black	39 (3.8)	37 (4.1)	2 (1.4)
Hispanic	55 (5.3)	54 (6.0)	1 (0.7)
Asian	21 (2.0)	17 (1.9)	4 (2.9)
Vital signs
Heart rate (bpm)	86.07 [75.66, 100.00]	86.07 [76.03, 99.41]	86.42 [74.11, 102.80]	-0.269	0.788
MBP (mmHg)	81.08 [74.22, 88.31]	81.43 [74.41, 88.44]	79.40 [72.60, 86.45]	-2.118	0.034
Respiratory rate (bpm)	18.25 [16.37, 20.70]	18.11 [16.32, 20.41]	19.75 [17.73, 22.30]	-4.492	<0.001
Temperature (°C)	37.06 [36.73, 37.47]	37.07 [36.74, 37.46]	36.99 [36.52, 37.62]	-0.952	0.341
SpO2 (%)	98.14 [96.79, 99.30]	98.08 [96.80, 99.27]	98.61 [96.78, 99.47]	-1.869	0.062
Scoring systems
GCS	15.00 [13.00, 15.00]	15.00 [13.00, 15.00]	15.00 [12.00, 15.00]	-0.693	0.488
SOFA	4.00 [2.00, 6.00]	3.00 [2.00, 5.00]	6.00 [4.00, 9.00]	-8.152	<0.001
SAPSⅡ	31.00 [23.00, 41.00]	29.00 [22.00, 38.00]	45.00 [34.00, 52.75]	-10.523	<0.001
APSⅢ	38.00 [28.00, 52.00]	36.00 [28.00, 48.00]	54.50 [41.00, 73.00]	-9.010	<0.001
OASIS	32.00 [27.00, 38.00]	31.00 [26.00, 37.00]	39.00 [34.00, 43.00]	-9.825	<0.001
Laboratory parameters
Hematocrit (%)	33.24 [29.47, 37.37]	33.74 [29.72, 37.79]	30.65 [27.95, 34.64]	-4.915	<0.001
Hemoglobin (g/dl)	11.50 [10.07, 12.89]	11.62 [10.20, 13.00]	10.54 [9.42, 11.70]	-5.609	<0.001
Platelet (10^9/L)	192.67 [144.57, 241.50]	196.00 [151.00, 244.27]	165.50 [112.75, 218.42]	-4.744	<0.001
Wbc (10^9/L)	11.27 [8.40, 14.44]	11.26 [8.42, 14.25]	11.32 [8.27, 16.72]	-1.328	0.184
Albumin (g/dl)	3.40 [2.90, 3.80]	3.50 [3.00, 3.85]	3.20 [2.63, 3.60]	-5.031	<0.001
Anion gap (mmol/L)	14.10 [12.21, 16.00]	14.00 [12.00, 16.00]	16.00 [13.53, 18.00]	-5.990	<0.001
ACAG (mmol/L)	16.50 [14.75, 18.75]	16.38 [14.66, 18.42]	18.80 [16.25, 21.75]	-7.600	<0.001
Bicarbonate (mmol/L)	23.00 [20.84, 25.46]	23.08 [21.00, 25.50]	21.00 [19.00, 23.79]	-5.805	<0.001
Bun (mg/dL)	14.00 [10.00, 20.15]	13.50 [10.00, 19.33]	18.00 [11.68, 29.16]	-5.138	<0.001
Creatinine (mg/dL)	0.83 [0.70, 1.10]	0.82 [0.68, 1.05]	1.00 [0.70, 1.40]	-4.009	<0.001
Sodium (mmol/L)	139.50[137.00, 141.67]	139.38[137.00, 141.50]	140.67[138.36, 143.40]	-4.487	<0.001
Potassium (mmol/L)	4.02 [3.74, 4.37]	4.01 [3.75, 4.38]	4.02 [3.73, 4.34]	-0.237	0.813
Lactate(mmol/L)	2.30 [1.55, 3.30]	2.26 [1.50, 3.16]	2.78 [1.87, 4.67]	-4.729	<0.001
INR	1.20 [1.10, 1.34]	1.17 [1.10, 1.30]	1.31 [1.13, 1.60]	-5.954	<0.001
PT	13.30 [12.20, 14.70]	13.20 [12.12, 14.40]	14.53 [12.93, 17.03]	-6.638	<0.001
PTT	27.64 [25.20, 31.30]	27.38 [24.94, 30.52]	30.24 [26.64, 36.15]	-5.926	<0.001
Glucose (mg/dL)	133.00 [112.89, 159.25]	129.68 [111.19, 154.56]	157.07 [133.05, 181.37]	-6.937	<0.001
Comorbidities
Liver disease (%)	90 (8.7)	68 (7.6)	22 (15.9)	10.627	0.002
Paraplegia (%)	43 (4.1)	35 (3.9)	8 (5.8)	1.097	0.413
Chronic pulmory disease (%)	132 (12.7)	118 (13.1)	14 (10.1)	0.948	0.403
Congestive heart failure (%)	119 (11.5)	95 (10.6)	24 (17.4)	5.508	0.028
Peripheral vascular disease (%)	32 (3.1)	27 (3.0)	5 (3.6)	0.156	0.897
Renal disease (%)	69 (6.6)	54 (6.0)	15 (10.9)	4.572	0.051
Cancer (%)	382 (36.8)	312 (34.7)	70 (50.7)	13.266	<0.001
Diabetes (%)	430 (41.4)	357 (39.7)	73 (52.9)	8.634	0.004
Treatment
Ventilation (%)	649 (62.5)	525 (58.3)	124 (89.9)	50.739	<0.001
Ventilation duration (hours)	18.00 [0.00, 92.00]	12.47 [0.00, 76.93]	84.92 [38.55, 163.38]	-8.625	<0.001
Length of hospital stays (days)	8.73 [4.86, 16.69]	9.35 [5.37, 17.40]	5.93 [3.35, 11.31]	-5.393	<0.001
Length of icu stays (days)	3.24 [1.78, 7.54]	2.94 [1.74, 7.01]	5.51 [2.63, 8.46]	-4.352	<0.001

Open in a new tab

MAP: mean arterial pressure. Wbc: white blood cell. ACAG: albumin corrected anion gap. INR: international normalized ratio. PT: prothrombin time. PTT: partial thromboplastin time. GCS: Glasgow Coma Scale. SOFA: Sequential Organ Failure Assessment. SAPS Ⅱ: Simplified Acute Physiology Scores Ⅱ. APS Ⅲ: Acute Physiology Score Ⅲ. OASIS: Oxford Acute Severity of Illness Score.

The number of complete samples before and after the imputation of lactate value was 634 and 1038, with a median of 2.40(1.60,3.50)mmol/L and 2.30(1.55,3.30)mmol/L, respectively. Non-parametric testing revealed no significant difference between the two groups (Z = -1.461, P = 0.144). The scatter plotted in S1 Fig illustrated the distribution of lactate values before and after imputation. The imputed data exhibited the same distribution as the observed data, indicating that the missingness was completely at random (MCAR). Sensitivity analyses conducted using the complete dataset, excluding cases with missing lactate values, demonstrated that the unadjusted ACAG and the results adjusted for model 1 and model 2 (S1 Table in S1 File) did not significantly differ from those obtained using imputed lactate data (Table 4). This confirmed the robustness of the imputation method and indicated that the missing components did not have an impact on the final results.

Table 4. Cox proportional hazard analysis of ACAG of in-hospital mortality in patients with trauma.

Variable	Crude		Model 1		Model 2		Model3
Variable	HR(95% CI)	P	HR(95% CI)	P	HR(95% CI)	P	HR(95% CI)	P
ACAG<20.357mmol/L	1(ref)		1(ref)		1(ref)		1(ref)
ACAG≥20.357mmol/L	4.451(3.157–6.276)	<0.001	4.166(2.889–6.005)	<0.001	3.128(1.615–6.059)	0.001	1.981(1.222–3.213)	0.006
Continuous	1.212(1.163–1.262)	<0.001	1.177(1.126–1.231)	<0.001	1.111(1.027–1.203)	0.009	1.121(1.040–1.209)	0.003

Open in a new tab

Crude: No covariates were adjusted before PSM. Model1: adjusted for age, race, sex, liver disease, congestive heart failure, renal disease, cancer, diabetes before PSM. Model2: adjusted for age, race, sex, MBP, respiratory rate, SpO2, SOFA, SAPSⅡ, APSⅢ, OASIS, hematocrit, hemoglobin, platelets, albumin, anion gap, bicarbonate, bun, creatinine, sodium, lactate, INR, PT, PTT, glucose, liver disease, congestive heart failure, renal disease, cancer, diabetes, ventilation, ventilation duration before PSM. Model3: Univariate analysis after PSM. ACAG:albumin corrected anion gap. PSM: propensity score matching.

ROC curve analysis

The predictive efficacy of SOFA, ACAG, AG, and ALB in assessing in-hospital mortality among trauma patients was compared using ROC curve analysis. The AUC (95%CI) values for SOFA, ACAG, AG, and ALB were 0.713(0.666–0.761), 0.701 (0.652–0.749), 0.658 (0.607–0.709), and 0.633 (0.583–0.682), respectively (Fig 2). ACAG demonstrated superior predictive ability over AG for in-hospital mortality (Z = -3.420, P < 0.001) as well as over ALB(Z = -2.381, P = 0.017). Furthermore, it did not demonstrate a statistically significant difference when compared to sofa (Z = 0.425, P = 0.671).

Fig 2 — SOFA: Sequential Organ Failure Assessment. ACAG: albumin corrected anion gap; ALB: albumin; AG: anion gap.

Determination of optimal cut-off value for survival analysis

The optimal cut-off point of ACAG for predicting in-hospital mortality in trauma patients was determined to be 20.375mmol/L using the surv_cutpoint function from the R package survminer in the R programming language. The low level ACAG group was defined as having ACAG< 20.375mmol/L, while the high level ACAG group was defined as having ACAG≥20.375mmol/L. These two groups of patients exhibited the most significant disparity (Fig 3).

Fig 3 — ACAG: albumin corrected anion gap.

Post-PSM characteristics

The covariates (excluding AG, ACAG, and outcome variables) at baseline were included in the propensity score matching (PSM) analysis. We employed the nearest neighbor algorithm with a caliper width of 0.3 and maintained a 2:1 ratio between the control group and treatment group. Following matching, the cohorts exhibited excellent balance, with more comparable ACAG observed between the two groups (Fig 4). Table 2 and Fig 5 present the characteristics and standardized mean differences (SMDs) of patients in both high and low ACAG groups before and after PSM. The SMDs of the matched variables were all below 0.1 after PSM.

Fig 4 — Jitter plot of distribution of propensity scores (A). Histogram of distribution of propensity scores (B).

Table 2. Characteristics of the study population were compared between low and high ACAG groups before and after PSM.

Variables	Before PSM			After PSM
Variables	ACAG < 20.375mmol/L	ACAG ≥ 20.375mmol/L	P	ACAG < 20.375mmol/L	ACAG < 20.375mmol/L	P
N	890	148		181	110
Age (year)	57.37 [38.16, 73.92]	62.13 [47.75, 77.26]	0.016	63.61 [41.49, 78.31]	60.22 [44.85, 75.27]	0.696
Female, n(%)	280 (31.5)	55 (37.2)	0.201	58 (32.0)	34 (30.9)	0.943
Weight (kg)	76.60 [64.43, 89.83]	75.80 [66.15, 87.62]	0.659	75.65 [63.60, 90.00]	77.85 [68.40, 88.90]	0.804
Race, n(%)			0.216			0.498
Others	224 (25.2)	43 (29.1)		47 (26.0)	32 (29.1)
White	567 (63.7)	89 (60.1)		112 (61.9)	64 (58.2)
Black	34 (3.8)	5 (3.4)		7 (3.9)	5 (4.5)
Hispanic	50 (5.6)	5 (3.4)		13 (7.2)	5 (4.5)
Asian	15 (1.7)	6 (4.1)		2 (1.1)	4 (3.6)
Vital signs
Heart rate (bpm)	85.48 [75.23, 98.35]	91.59 [79.39, 106.09]	0.002	92.56 [82.23, 103.24]	92.53 [79.66, 105.60]	0.846
MBP (mmHg)	81.57 [74.68, 88.31]	77.73 [71.71, 87.84]	0.004	80.96 [74.00, 88.27]	80.39 [72.94, 89.56]	0.62
Respiratory rate (bpm)	18.11 [16.31, 20.38]	19.77 [17.11, 22.50]	<0.001	19.30 [16.89, 21.50]	19.52 [17.04, 22.34]	0.365
Temperature (°C)	37.08 [36.74, 37.48]	37.00 [36.60, 37.33]	0.010	37.07 [36.69, 37.47]	37.02 [36.71, 37.44]	0.924
SpO2 (%)	98.20 [96.84, 99.35]	97.92 [96.51, 98.91]	0.041	98.20 [96.83, 99.36]	97.94 [96.70, 99.04]	0.293
Scoring systems
GCS	15.00 [13.00, 15.00]	15.00 [13.00, 15.00]	0.604	15.00 [14.00, 15.00]	15.00 [13.00, 15.00]	0.161
SOFA	3.00 [2.00, 5.00]	6.00 [4.00, 9.00]	<0.001	5.00 [3.00, 7.00]	5.00 [3.00, 7.00]	0.436
SAPSⅡ	30.00 [22.00, 39.00]	41.00 [30.00, 51.00]	<0.001	36.00 [28.00, 45.00]	36.00 [28.00, 46.00]	0.596
APSⅢ	37.00 [28.00, 48.00]	55.50 [37.00, 73.25]	<0.001	46.00 [33.00, 62.00]	47.00 [35.00, 62.75]	0.509
OASIS	32.00 [27.00, 37.00]	36.00 [31.00, 42.00]	<0.001	33.00 [29.00, 40.00]	35.00 [30.00, 40.00]	0.251
Laboratory parameters
Hematocrit (%)	33.51 [29.74, 37.70]	31.26 [28.01, 35.93]	<0.001	31.28 [28.50, 35.83]	31.92 [28.57, 36.86]	0.677
Hemoglobin (g/dl)	11.60 [10.21, 13.00]	10.65 [9.24, 12.09]	<0.001	10.70 [9.73, 12.27]	10.88 [9.61, 12.32]	0.667
Platelets (10^9/L)	195.17 [151.00, 242.83]	166.83 [108.60, 239.75]	<0.001	172.83 [125.50, 234.00]	172.50 [109.50, 240.50]	0.747
Wbc (10^9/L)	11.25 [8.43, 14.25]	11.57 [8.12, 15.96]	0.207	11.50 [8.50, 15.07]	11.57 [8.50, 15.28]	0.861
Albumin (g/dl)	3.50 [3.00, 3.85]	3.00 [2.50, 3.60]	<0.001	3.10 [2.60, 3.50]	3.20 [2.56, 3.70]	0.938
Bicarbonate (mmol/L)	23.50 [21.37, 25.50]	20.00 [17.50, 21.67]	<0.001	20.67 [19.25, 22.50]	20.33 [19.00, 22.00]	0.206
Bun (mg/dL)	13.50 [10.00, 19.00]	18.58 [11.33, 34.29]	<0.001	15.43 [11.33, 23.60]	15.54 [10.08, 28.92]	0.87
Creatinine (mg/dL)	0.80 [0.67, 1.02]	1.10 [0.77, 1.71]	<0.001	0.90 [0.75, 1.30]	0.94 [0.72, 1.40]	0.664
Sodium (mmol/L)	139.50 [137.00, 141.50]	140.00 [136.88, 142.75]	0.092	140.25 [137.17, 142.33]	140.00 [136.12, 142.00]	0.553
Potassium (mmol/L)	4.00 [3.73, 4.30]	4.16 [3.80, 4.62]	0.001	4.14 [3.87, 4.50]	4.05 [3.74, 4.55]	0.522
Lactate(mmol/L)	2.20 [1.50, 3.03]	3.81 [2.12, 5.41]	<0.001	3.03 [2.07, 4.20]	3.00 [1.91, 4.54]	0.983
INR	1.17 [1.10, 1.30]	1.29 [1.13, 1.55]	<0.001	1.22 [1.10, 1.40]	1.23 [1.10, 1.46]	0.494
PT	13.20 [12.20, 14.42]	14.31 [12.80, 16.65]	<0.001	13.70 [12.53, 15.25]	13.68 [12.33, 15.95]	0.793
PTT	27.40 [25.04, 30.50]	30.31 [26.61, 35.85]	<0.001	28.50 [25.57, 33.20]	29.80 [25.61, 34.52]	0.455
Glucose (mg/dL)	132.00 [112.45, 156.50]	142.85 [114.75, 183.70]	0.002	141.67 [119.67, 166.50]	139.67 [113.00, 175.19]	0.516
Comorbidities
Liver disease, n(%)	59 (6.6)	31 (20.9)	<0.001	24 (13.3)	15 (13.6)	1
Paraplegia, n(%)	33 (3.7)	10 (6.8)	0.133	13 (7.2)	9 (8.2)	0.933
Chronic pulmory disease, n(%)	118 (13.3)	14 (9.5)	0.250	14 (7.7)	9 (8.2)	1
Congestive heart failure, n(%)	92 (10.3)	27 (18.2)	0.008	29 (16.0)	21 (19.1)	0.608
Peripheral vascular disease, n(%)	29 (3.3)	3 (2.0)	0.585	3 (1.7)	3 (2.7)	0.844
Renal disease, n(%)	47 (5.3)	22 (14.9)	<0.001	17 (9.4)	14 (12.7)	0.485
Cancer, n(%)	331 (37.2)	51 (34.5)	0.585	60 (33.1)	38 (34.5)	0.907
Diabetes, n(%)	367 (41.2)	63 (42.6)	0.830	77 (42.5)	50 (45.5)	0.716
Treatment
Ventilation, n(%)	546 (61.3)	103 (69.6)	0.068	121 (66.9)	75 (68.2)	0.916
Ventilation duration (hours)	15.36 [0.00, 88.04]	35.09 [0.00, 124.94]	0.012	40.00 [0.00, 150.00]	28.38 [0.00, 124.17]	0.574
Outcomes
Length of hospital stays (days)	8.77 [4.90, 16.56]	8.18 [4.75, 16.97]	0.820	11.41 [5.91, 21.36]	8.81 [5.62, 16.58]	0.08
Length of icu stays (days)	3.10 [1.77, 7.33]	3.92 [1.91, 9.79]	0.041	4.16 [2.01, 11.20]	3.92 [2.00, 10.06]	0.513
in-hospital mortality, n(%)	85 (9.6)	53 (35.8)	<0.001	31 (17.1)	35 (31.8)	0.006
30-day mortality, n(%)	105 (11.8)	57 (38.5)	<0.001	38 (21.0)	38 (34.5)	0.016
90-day mortality, n(%)	136 (15.3)	65 (43.9)	<0.001	48 (26.5)	43 (39.1)	0.035

Open in a new tab

MAP: mean arterial pressure. Wbc: white blood cell. INR: international normalized ratio. PT: prothrombin Time. PTT: partial thromboplastin time. GCS: Glasgow Coma Scale. SOFA: Sequential Organ Failure Assessment. SAPS Ⅱ: Simplified Acute Physiology Scores Ⅱ. APS Ⅲ: Acute Physiology Score Ⅲ. OASIS: Oxford Acute Severity of Illness Score. PSM: propensity score matching.

Fig 5 — MAP: mean arterial pressure. Wbc: white blood cell. ACAG: albumin corrected anion gap. INR: international normalized ratio. PT: prothrombin time. GCS: Glasgow Coma Scale. PTT: partial thromboplastin time. SOFA: Sequential Organ Failure Assessment. SAPS Ⅱ: Simplified Acute Physiology Scores Ⅱ. APS Ⅲ: Acute Physiology Score Ⅲ. OASIS: Oxford Acute Severity of Illness Score.

Subgroup analyses

The association between ACAG and in-hospital mortality was also explored using subgroup analysis (Table 3). Overall, significant interactions were not observed in most strata.

Table 3. Subgroup analysis of the association between different levels of ACAG and in-hospital mortality.

Variable	N	Hazard ratio(95%CI)	P	P for interaction
Age				0.329
<62	143	2.93(1.24–6.92)	0.014
≥62	148	1.75(0.96–3.19)	0.067
Sex				0.386
Female	199	1.70(0.93–3.11)	0.086
Male	92	2.73(1.21–6.14)	0.016
Race				0.338
Others	115	1.51(0.74–3.09)	0.261
White	176	2.45(1.27–4.72)	0.008
SOFA				0.483
<5	133	2.53(1.02–6.30)	0.046
≥5	158	1.76(0.99–3.11)	0.054
SAPSⅡ				0.564
<36	137	2.48(0.94–6.52)	0.065
≥36	154	1.84(1.05–3.23)	0.032
APSⅢ				0.366
<46	142	1.45(0.63–3.36)	0.383
≥46	149	2.37(1.30–4.30)	0.005
OASIS				0.403
<34	139	1.34(0.51–3.52)	0.551
≥34	152	2.16(1.23–3.80)	0.008
Ventilation				0.338
No	95	1.05(0.25–4.41)	0.942
Yes	196	2.20(1.31–3.70)	0.003
Liver disease				0.483
No	252	1.83(1.08–3.11)	0.025
Yes	39	3.11(0.91–10.64)	0.071
Paraplegia				0.256
No	269	1.81(1.09–3.02)	0.022
Yes	22	4.76(0.92–24.63)	0.063
Chronic pulmory disease				0.698
No	268	1.94(1.17–3.21)	0.01
Yes	23	2.92(0.49–17.54)	0.242
Congestive heart failure				0.777
No	241	2.05(1.20–3.52)	0.009
Yes	50	1.69(0.57–5.03)	0.345
Renal disease				0.283
No	260	2.19(1.30–3.70)	0.003
Yes	31	0.96(0.26–3.57)	0.949
Cancer				0.363
No	193	1.60(0.78–3.24)	0.189
Yes	98	2.56(1.31–5.00)	0.006
Diabetes				0.446
No	164	2.41(1.17–4.96)	0.017
Yes	127	1.67(0.87–3.21)	0.123

Open in a new tab

Outcome measurement

To determine whether high ACAG level independently contribute to increased in-hospital mortality among trauma patients, we conducted univariate and multivariate Cox regression analyses (Table 4). In the crude univariate models, high ACAG level was significantly associated with an increased risk of in-hospital mortality (unadjusted HR = 4.451; 95%CI:3.157–6.276). After adjusting for age, race, sex, liver disease, congestive heart failure, renal disease, cancer, and diabetes in Model 1 analysis, high ACAG level remained significantly associated with higher in-hospital mortality (adjusted HR = 4.166; 95%CI:2.889–6.005). Furthermore, Model 2 incorporated additional adjustments for confounding laboratory parameters and treatments. Remarkably, elevated ACAG level still independently predicted a higher risk of in-hospital mortality (adjusted HR = 3.128; 95%CI:1.615–6.059). Notably, even within the PSM matched cohort analysis, the association between high ACAG level and in-hospital mortality remained significant (HR = 1.981;95%CI:1.222–3.213). After being included as a continuous variable in the COX regression analysis, ACAG remained statistically significant in predicting in-hospital mortality. Following adjustment for various confounding variables, the hazard ratio (HR) was 1.111 (95%CI: 1.027–1.203), and within the propensity score matching (PSM) cohort, the HR was 1.121 (95%CI: 1.040–1.209).

In addition, we employed the Log-rank test to construct Kaplan-Meier survival curves for assessing the prognostic value of ACAG. Within the original cohort, there was a significantly higher mortality rate in the high ACAG group compared to the low ACAG group at in-hospital, 30-day and 90-day time points (P<0.001). Furthermore, within the matched cohort, no significant differences were observed in baseline between the high and low ACAG level groups. However, high level of ACAG remained statistically associated with increased risks of in-hospital mortality(P = 0.004) as well as mortality rates at 30-day (P = 0.012) and 90-day (P = 0.021) (Fig 6).

Fig 6 — Kaplan—Meier survival curve of post-trauma patients with low level group of ACAG (blue curve. ACAG<20.375mmol/L) and high level group of ACAG (red curve. ACAG≥20.375mmol/L) at in-hospital(A, D), 30-day(B, E), 90-day(C, F) follow-up. (A-C)Reflect the results before PSM. (D-F)Reflect the results after PSM. ACAG: albumin corrected anion gap. PSM: propensity score matching.

To reduce the overestimation effect of ACAG, we attempted to hierarchically split the original dataset into a training set (n = 726) and a testing set (n = 312) at a 7:3 ratio. After conducting balance testing (S2 Table in S1 File), there were no significant differences in patient characteristics between the two datasets. ACAG was included as both a continuous variable and a categorical variable in model1 and model2 of the multivariate Cox regression analysis, respectively, in the testing set. Similar and significant results were obtained as those in the training set (S3 Table in S1 File).

Discussion

The present study conducted a retrospective analysis on clinical data from the MIMIC-Ⅲ and MIMIC-Ⅳ databases to assess the potential of AG and ACAG in predicting the prognosis of trauma patients. The findings demonstrated a significant positive correlation between AG and ACAG within the initial 24-hour period of admission to ICU and the risk of in-hospital mortality among trauma patients. Hemorrhage resulting from trauma often leaded to shock and might subsequently be accompanied by dilutive coagulopathy and hypothermia, frequently associated with severe metabolic acidosis. This condition prolonged hospitalization duration and increases mortality rates [4]. The occurrence of metabolic acidosis in severe trauma patients could be attributed to the heightened production of organic acids, where unmeasured anions served as indicators for dissociated organic acids and were the primary contributors to metabolic acidosis [19]. As early as 1983, Stewart proposed identifying unmeasured ions through physicochemical acid-base analysis, elucidating that the charge difference between ions formed the foundation of acid-base physiology. By adhering to principles of electric neutrality and mass conservation laws, the missing charge in plasma was identified as a "gap" [20]. Both strong ion gap (SIG) and anion gap (AG) could typically serve as indicators for estimating ion gaps [21]. Kaplan et al. discovered that SIG and AG could differentiate between survivors and non-survivors of severe vascular injuries more effectively than lactic acid levels, standard base excess (SBE), or injury severity scores (ISS). The clinical utility could be enhanced by employing AG due to the relatively complex nature of SIG calculation [22]. Its easy accessibility had garnered scholarly attention in recent years, providing valuable insights into the diagnosis or prognosis of trauma patients. Leskovan et al., through a retrospective study, demonstrated that an AG level exceeding 16mmol/L was significantly associated with unfavorable clinical outcomes in elderly trauma patients [23]. Zhang et al. showed that patients with critical hip fracture and AG>12.5mmol/L had a 1.7-fold higher 30-day mortality rate compared to those with AG≤12.5mmol/L [24]. Trauma-related morbidity and mortality are frequently associated with hemorrhage, shock, tissue hypoperfusion resulting in metabolic acidosis and microcirculatory dysfunction, which could further lead to complications such as acute kidney injury (AKI), acute traumatic coagulopathy (ATC), adult respiratory distress syndrome (ARDS), ultimately culminating in fatality [21]. The two groups exhibited statistically significant differences in terms of renal function markers (creatinine, BUN), coagulation function markers (INR, PT, PTT), and others (P<0.001).

However, the AG could be influenced by various factors, including charged serum albumin [25]. Researches had demonstrated that for every 10 mg/L decrease in serum albumin, there was a corresponding 2.5 mmol/L decrease in AG. Hypoalbuminemia could result in a reduced measured AG, thereby concealing the presence of a high AG [26]. Adjusted corrected anion gap (ACAG) was calculated after accounting for serum albumin to mitigate this issue [27]. Hypoalbuminemia had been demonstrated to be strongly associated with unfavorable outcomes in surgical trauma patients. This correlation might arise from protein-energy malnutrition (PEM) induced by hypoalbuminemia, which could hinder wound healing, increase susceptibility to infection, exacerbate multiple organ dysfunction, and prolong hospitalization duration. Moreover, there was an elevated risk of in-hospital mortality [28]. Gonzalez et al. revealed that trauma patients with low albumin levels were more susceptible to developing traumatic endotheliopathy and subsequently experiencing protein extravasation, leading to a grim prognosis [29]. The clinical significance of ACAG lied in its ability to accurately reflect the dual pathological conditions of hypoalbuminemia and metabolic acidosis. When assessing AG levels in ICU patients, it was crucial to consider serum albumin correction as extensively as possible. In a case-control study involving 2160 individuals with acute myocardial infarction (AMI), Jian et al. found that ACAG exhibited superior predictive value compared to AG for 30-day all-cause mortality among patients in ICU. Additionally, a high ACAG level (ACAG≥21.75mmol/L) was identified as independent prognostic risk factors [30]. Moreover, numerous studies had demonstrated a significant association between ACAG and in-hospital mortality among patients suffering from cardiac arrest (CA), sepsis, and acute pancreatitis (AP) [31–33]. However, there was limited research on the correlation between ACAG and the prognosis of trauma patients. A total of 1038 trauma patients were stratified into two cohorts based on their in-hospital mortality status. Baseline data revealed a notable disparity in serum albumin within 24 hours after admission (P < 0.001), which could potentially impact the actual anion gap measurement. The ACAG was calculated using AG and serum albumin values according to the calibration formula, resulting in a higher area under the curve (AUC) for ACAG compared to AG (△AUC = 0.043, Z = -3.420, P < 0.001).

To further investigate the prognostic significance of ACAG in trauma patients, survival analysis was employed to demonstrate the correlation between elevated ACAG level mortality rates, including in-hospital mortality as well as 30-day, 90-day mortality rates. This study collected comprehensive and complete baseline data including demographic characteristics, vital signs, laboratory parameters, comorbidities, etc. Statistical techniques such as PSM (Propensity Score Matching) and COX regression were utilized to adjust the data, thereby enhancing the reliability of the findings. Given the absence of previous literature determining an optimal cut-off value for ACAG in trauma prognosis, we employed R language’s surv-cutpoint function to identify a threshold value of 20.375mmol/L for predicting in-hospital death as the primary outcome. Subsequently, patients were categorized into two groups according to this criterion. The results demonstrated that high ACAG level remained an independent risk factor for in-hospital mortality among patients with trauma, irrespective of adjustments made through COX multivariate analysis or after PSM. Furthermore, the Kaplan-Meier survival curves provided additional evidence by confirming significantly higher rates of in-hospital, 30-day and 90-day mortality among patients with high ACAG level prior to PSM compared to those with low ACAG level. Even after achieving balance between the baseline characteristics of deceased and surviving groups following PSM, a statistically significant difference persisted between patients with high and low ACAG levels regarding various mortality outcomes. Some studies had suggested the need for lactate correction when measuring anion gap [34]. In our dataset, lactate values were missing in more than 20% of cases. However, considering the significance of lactate as a biomarker, we conducted analyses after imputing missing values and performing sensitivity analyses. We incorporated this into multivariate regression before and after PSM and observed that the results remained robust.

The present study also had certain limitations. Firstly, due to its retrospective nature and limited sample size, as well as the utilization of average values for calculating ACAG, it would be beneficial to include a larger sample or prospective external data for further validation. Secondly, despite the inclusion of numerous covariates to control for confounding variables, there was a possibility that unexplored factors might have influenced the results. Lastly, although this study focused on sever trauma patients admitted to the ICU, it did not investigate the specific trauma sites within the study population. Nevertheless, notwithstanding these limitations, our study holds significant importance in comprehending the association between ACAG and trauma.

In conclusion, elevated ACAG(>20.375mmol/L) was found to be independently associated with in-hospital mortality as well as increased 30-day and 90-day mortality rates among critically injured patients. ACAG outperformed both albumin (ALB) and anion gap (AG) in predicting in-hospital mortality for trauma patients admitted to ICU. Given its low cost and ease of measurement, ACAG may prove useful for initial risk stratification of trauma patients, identification of high-risk individuals, and guiding clinical management.

Supporting information

S1 Data

(ZIP)

pone.0300012.s001.zip^{(2.7MB, zip)}

S1 Fig

(TIF)

pone.0300012.s002.tif^{(251.9KB, tif)}

S1 File

(DOCX)

pone.0300012.s003.docx^{(178KB, docx)}

Acknowledgments

The authors express their sincere gratitude to all the participants involved in this study.

Data Availability

All relevant data are within the manuscript and its Supporting Information files.

Funding Statement

The author(s) received no specific funding for this work.

References

1.Duque P, Mora L, Levy JH, Schöchl H. Pathophysiological Response to Trauma-Induced Coagulopathy: A Comprehensive Review. Anesth Analg. 2020;130(3):654–664. doi: 10.1213/ANE.0000000000004478 . [DOI] [PubMed] [Google Scholar]
2.Shashaty MG, Kalkan E, Bellamy SL, Reilly JP, Holena DN, Cummins K, et al. Computed tomography-defined abdominal adiposity is associated with acute kidney injury in critically ill trauma patients. Crit Care Med. 2014;42 (7):1619–28. doi: 10.1097/CCM.0000000000000306 [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Callaway DW, Wolfe R. Geriatric Trauma. Emergency Medicine Clinics of North America. 2007;25(3):837–860. doi: 10.1016/j.emc.2007.06.005 [DOI] [PubMed] [Google Scholar]
4.Shane AI, Robert W, Arthur K, Patson M, Moses G. Acid-base disorders as predictors of early outcomes in major trauma in a resource limited setting: An observational prospective study. Pan Afr Med J. 2014;17:2–8. doi: 10.11604/pamj.2014.17.2.2007 [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Messerer DAC, Halbgebauer R, Nilsson B, Pavenstädt H, Radermacher P, Huber-Lang M. Immunopathophysiology of trauma-related acute kidney injury. Nat Rev Nephrol. 2021;17(2):91–111. doi: 10.1038/s41581-020-00344-9 [DOI] [PubMed] [Google Scholar]
6.Ahmed FS, Ali L, Joseph BA, Ikram A, Ul Mustafa R, Bukhari SAC. A statistically rigorous deep neural network approach to predict mortality in trauma patients admitted to the intensive care unit. J Trauma Acute Care Surg. 2020;89(4):736–742. doi: 10.1097/TA.0000000000002888 [DOI] [PubMed] [Google Scholar]
7.Rudkin SE, Grogan TR, Treger RM. Relationship Between the Anion Gap and Serum Lactate in Hypovolemic Shock. J Intensive Care Med. 2022;37(12):1563–1568. doi: 10.1177/08850666221106413 [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Zhong L, Xie B, Ji XW, Yang XH. The association between albumin corrected anion gap and ICU mortality in acute kidney injury patients requiring continuous renal replacement therapy. Intern Emerg Med. 2022;17(8):2315–2322. doi: 10.1007/s11739-022-03093-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Silva JM Jr, Ribas Rosa de Oliveira AM, Mendes Nogueira FA, Vianna PM, Amendola CP, Carvalho Carmona MJ, et al. Metabolic Acidosis Assessment in High-Risk Surgeries: Prognostic Importance. Anesth Analg. 2016;123(5):1163–1171. doi: 10.1213/ANE.0000000000001575 [DOI] [PubMed] [Google Scholar]
10.Johnson AE, Pollard TJ, Shen L, Lehman LW, Feng M, Ghassemi M, et al. MIMIC-Ⅲ, a freely accessible critical care database. Sci Data. 2016;3:160035–160043. doi: 10.1038/sdata.2016.35 [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Johnson AEW, Bulgarelli L, Shen L, Gayles A, Shammout A, Horng S, et al. MIMIC-Ⅳ, a freely accessible electronic health record dataset. Sci Data. 2023;10(1):1–9. doi: 10.1038/s41597-022-01899-x [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Flynn-O’Brien KT, Fallat ME, Rice TB, Gall CM, Nance ML, et al. Pediatric Trauma Assessment and Management Database: Leveraging Existing Data Systems to Predict Mortality and Functional Status after Pediatric Injury. J Am Coll Surg. 2017;224(5):933–944.e5. doi: 10.1016/j.jamcollsurg.2017.01.061 [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Clark DE, Black AW, Skavdahl DH, Hallagan LD. Open-access programs for injury categorization using ICD-9 or ICD-10. Inj Epidemiol. 2018;5(1):11–18. doi: 10.1186/s40621-018-0149-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Oh MS, Carroll HJ. The anion gap. N Engl J Med. 1977;297(15): 814–817. doi: 10.1056/NEJM197710132971507 [DOI] [PubMed] [Google Scholar]
15.Figge J, Jabor A, Kazda A, Fencl V. Anion gap and hypoalbuminemia. Crit Care Med. 1998;26(11):1807–1810. doi: 10.1097/00003246-199811000-00019 [DOI] [PubMed] [Google Scholar]
16.Hayati Rezvan P, Lee KJ, Simpson JA. The rise of multiple imputation: a review of the reporting and implementation of the method in medical research. BMC Med Res Methodol. 2015;15:30–43. doi: 10.1186/s12874-015-0022-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Chen D, Fritz MS. Comparing Alternative Corrections for Bias in the Bias-Corrected Bootstrap Test of Mediation. Eval Health Prof. 2021;44(4):416–427. doi: 10.1177/01632787211024356 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Wang P, Yuan D, Zhang C, Zhu P, Jia S, Song Y, et al. High fibrinogen-to-albumin ratio with type 2 diabetes mellitus is associated with poor prognosis in patients undergoing percutaneous coronary intervention: 5-year findings from a large cohort. Cardiovasc Diabetol. 2022;21(1):46–58. doi: 10.1186/s12933-022-01477-w [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Martin M, Murray J, Berne T, Demetriades D, Belzberg H. Diagnosis of acid-base derangements and mortality prediction in the trauma intensive care unit: the physiochemical approach. J Trauma. 2005;58(2):238–243. doi: 10.1097/01.ta.0000152535.97968.4e [DOI] [PubMed] [Google Scholar]
20.Stewart PA. Modern quantitative acid-base chemistry. Can J Physiol Pharmacol. 1983;61(12):1444–1461. doi: 10.1139/y83-207 [DOI] [PubMed] [Google Scholar]
21.Zingg T, Bhattacharya B, Maerz LL. Metabolic acidosis and the role of unmeasured anions in critical illness and injury. The Journal of surgical research. 2018;224:5–17. doi: 10.1016/j.jss.2017.11.013 [DOI] [PubMed] [Google Scholar]
22.Kaplan LJ, Kellum JA. Initial pH, base deficit, lactate, anion gap, strong ion difference, and strong ion gap predict outcome from major vascular injury. Crit Care Med. 2004;32(5):1120–1124. doi: 10.1097/01.ccm.0000125517.28517.74 [DOI] [PubMed] [Google Scholar]
23.Leskovan JJ, Justiniano CF, Bach JA, Cook CH, Lindsey DE, Eiferman DS, et al. Anion Gap as a Predictor of Trauma Outcomes in the Older Trauma Population: Correlations with Injury Severity and Mortality. The American SurgeonTM. 2013;79(11):1203–1206. 10.1177/000313481307901126 [DOI] [PubMed] [Google Scholar]
24.Zhang XB, Shu WB, Li AB, Lan GH. The Anion Gap and Mortality in Critically Ill Patients with Hip Fractures. Contrast Media Mol Imaging. 2022; 2022: 1591507–1591512. doi: 10.1155/2022/1591507 [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Hatherill M, Waggie Z, Purves L, Reynolds L, Argent A. Correction of the anion gap for albumin in order to detect occult tissue anions in shock. Arch Dis Child. 2002;87(6):526–529. doi: 10.1136/adc.87.6.526 [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Wang R, Li J, Chen H, Xiao Z, Xu R, Hu Y, et al. Preoperative albumin corrected anion gap is associated with in-hospital and long-term mortality in patients undergoing coronary artery bypass grafting in a retrospective cohort study. J Thorac Dis. 2022;14(12):4894–4903. doi: 10.21037/jtd-22-1633 [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Chawla LS, Shih S, Davison D, Junker C, Seneff MG. Anion gap, anion gap corrected for albumin, base deficit and unmeasured anions in critically ill patients: implications on the assessment of metabolic acidosis and the diagnosis of hyperlactatemia. BMC Emergency Medicine. 2008;8:18–25. doi: 10.1186/1471-227X-8-18 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Kim SC, Kim DH, Kim TY, Kang C, Lee SH, Jeong JH, et al. The Revised Trauma Score plus serum albumin level improves the prediction of mortality in trauma patients. Am J Emerg Med. 2017;35(12):1882–1886. doi: 10.1016/j.ajem.2017.06.027 . [DOI] [PubMed] [Google Scholar]
29.Gonzalez Rodriguez E, Cardenas JC, Lopez E, Cotton BA, Tomasek JS, Ostrowski SR, et al. Early Identification of the Patient with Endotheliopathy of Trauma by Arrival Serum Albumin. Shock. 2018;50(1):31–37. doi: 10.1097/SHK.0000000000001036 . [DOI] [PubMed] [Google Scholar]
30.Jian L, Zhang Z, Zhou Q, Duan X, Xu H, Ge L. Association between albumin corrected anion gap and 30-day all-cause mortality of critically ill patients with acute myocardial infarction: a retrospective analysis based on the MIMIC-Ⅳ database. BMC Cardiovasc Disord. 2023;23(1):211–220. doi: 10.1186/s12872-023-03200-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Hu B, Zhong L, Yuan M, Min J, Ye L, Lu J, et al. Elevated albumin corrected anion gap is associated with poor in-hospital prognosis in patients with cardiac arrest: A retrospective study based on MIMIC-Ⅳ database. Front Cardiovasc Med. 2023;10:1099003–1099010. doi: 10.3389/fcvm.2023.1099003 [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Hu T, Zhang Z, Jiang Y. Albumin corrected anion gap for predicting in-hospital mortality among intensive care patients with sepsis: A retrospective propensity score matching analysis. Clin Chim Acta. 2021;521:272–277. doi: 10.1016/j.cca.2021.07.021 [DOI] [PubMed] [Google Scholar]
33.Li P, Shi L, Yan X, Wang L, Wan D, Zhang Z, et al. Albumin Corrected Anion Gap and the Risk of in-Hospital Mortality in Patients with Acute Pancreatitis: A Retrospective Cohort Study. J Inflamm Res. 2023;16:2415–2422. doi: 10.2147/JIR.S412860 [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Chawla LS, Jagasia D, Abell LM, Seneff MG, Egan M, Danino N, et al. Anion gap, anion gap corrected for albumin, and base deficit fail to accurately diagnose clinically significant hyperlactatemia in critically ill patients. J Intensive Care Med. 2008;23(2):122–127. doi: 10.1177/0885066607312985 [DOI] [PubMed] [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0300012.r001

Decision Letter 0

Tanja Grubić Kezele

19 Dec 2023

PONE-D-23-37440Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: a retrospective study based on MIMIC-III and IV databasesPLOS ONE

Dear Dr. Xu,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

==============================Based on the reviewers' suggestions, the paper needs major revision. The reviewers' comments can be found below.

==============================

Please submit your revised manuscript by Feb 02 2024 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Tanja Grubić Kezele, Ph.D., M.D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that your Data Availability Statement is currently as follows: All relevant data are within the manuscript and its Supporting Information files.

Please confirm at this time whether or not your submission contains all raw data required to replicate the results of your study. Authors must share the “minimal data set” for their submission. PLOS defines the minimal data set to consist of the data required to replicate all study findings reported in the article, as well as related metadata and methods (https://journals.plos.org/plosone/s/data-availability#loc-minimal-data-set-definition).

For example, authors should submit the following data:

- The values behind the means, standard deviations and other measures reported;

- The values used to build graphs;

- The points extracted from images for analysis.

Authors do not need to submit their entire data set if only a portion of the data was used in the reported study.

If your submission does not contain these data, please either upload them as Supporting Information files or deposit them to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of recommended repositories, please see https://journals.plos.org/plosone/s/recommended-repositories.

If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent. If data are owned by a third party, please indicate how others may request data access.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: No

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This article is about association between ACAG and in-hospital mortality of intensive patients with trauma using MIMIC-III and -IV databases. Though well prepared, neither the theme nor the biomarker investigated is of interest to the general audience. Here are several limitations I would like to point out so that the authors may further revise the manuscript for other journals:

1. Unlike what the authors have stated, the patients in MIMIC-III actually intersected with those in MIMIC-IV, as patients admitted to ICU from 2001-2008 (MIMIC-III) can be readmitted later in 2008-2012 (MIMIC-IV). Since identifiers of both datasets are different, the authors might not remove duplicate patients nor remove the wrong patients.

2. The association of ACAG might be overestimated since the cut-off point was chosen using statistics instead of pre-determined criteria. The authors might separate the patients into 2 groups in 8:2 ratio, respectively named as training set and validation set, so as to alleviate the possible overestimation effect.

3. Special attention should be paid to the patient group specified. Since trauma can range greatly from superficial injury to traumatic amputation. It would be impossible to figure out the appropriate spectrum to which the conclusion can be generalized. Meanwhile, the conclusion itself seemed rather dubious at the same time.

4. The manuscript lacks detailed and thorough discussion in the context of the previous literature. Rationale for this research and this biomarker should be specified. Otherwise, it seemed to be a part of huge data-mining program of limited use to clinical practice.

Reviewer #2: Thank you for the opportunity to review your research. My comments are below:

1) It is not clear when the albumin was drawn relative to the AG. For example, given that albumin is a negative acute phase reactant, if the AG was drawn on admission and the albumin was drawn close to 24 hours later, this could potentially affect the ACAG calculation. If this data on timing is not available, it should be noted in the limitations section of the discussion.

2) In Results/Baseline Characteristics, it states “…SOFA score, SAPS II score…were all higher among survivors”. I believe it should state higher among non-survivors.

3) The ROC curve analysis shows the predictive value of ACAG, AG, and albumin. It would be more meaningful to also include the Injury Severity Score or other acute severity score to determine how much ACAG adds beyond commonly available scores used to determine severity and estimate prognosis.

4) Propensity score matching is typically used to balance treatment and non-treatment groups. Its use in this manuscript is unconventional. Can you provide any references on use of PSM in this type of pathophysiology-based approach?

5) The discussion states that “…lactate values were missing by more than 20% and were subsequently excluded.” Exclusion of the lactate severely limits any conclusion that can be drawn from this study. The association between ACAG and in-hospital mortality may be confounded by or entirely due to lactate, which will increase the ACAG. Lactate should be included and can be subtracted from the ACAG and/or adjusted for in the regression analysis. Otherwise, the findings in the manuscript may entirely be due to lactate which is a well known prognostic indicator. A sensitivity analysis can be performed to ensure the missing lactate values do not contribute to uncertainty of the findings.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PLoS One. 2024 Mar 7;19(3):e0300012. doi: 10.1371/journal.pone.0300012.r002

Author response to Decision Letter 0

15 Jan 2024

Dear editor and reviewers of PLOS ONE:

Our reference: PONE-D-23-37440

Title: Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: A retrospective study based on MIMIC-Ⅲ and Ⅳ databases

By: Yuzhou Xu et al

Thank you very much for your letter and for the editors’ and reviewers’ comments concerning our manuscript entitled "Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: A retrospective study based on MIMIC-Ⅲ and Ⅳ databases" (ID: PONE-D-23-37440). These comments are of great reference value to the revision and improvement of our paper and have important guiding significance to our researches. We have studied comments carefully and have made correction. We hope that the revision is acceptable and look forward to hearing from you soon. Revised portion are marked in color in the paper. The main corrections in the paper and the responds to the reviewer’s comments are as flowing:

Reviewer 1

Since the release of MIMIC-IV, both the MIMIC-III and MIMIC-IV databases contain admissions data from 2008 to 2012, creating an overlap that complicates database analysis. To address this issue, we utilized the official MIMIC-III Clinical Database CareVue subset exclusively for our study. This subset was carefully selected to include only subject_ids not present in MIMIC-IV, ensuring that it comprises patients who were not represented in MIMIC-IV. Consequently, these two databases could be considered independent of each other without any overlapping data, thereby facilitating their combined utilization. The detailed methodology could be found on the official website: "https://physionet.org/content/mimic3-carevue/1.4/".

Before our analysis, there was no relevant literature documenting the association between ACAG and severe trauma, and no rationale existed for this cut-off point. Therefore, we calculated it as 20.375mmol/L based on the statistical findings of our dataset. However, it was worth noting that this result closely aligned with the ACAG cut-off value reported in prognostic studies of other diseases. For instance, Jian et al., while investigating ACAG levels and 30-day all-cause mortality in acute myocardial infarction patients, determined a calculated ACAG cut-off value of 21.75mmol/L[1]. Similarly, Hu et al. demonstrated that an elevation in ACAG (≥20mmol/L) served as an independent risk factor for in-hospital mortality among cardiac arrest patients [2].

We attempted to hierarchically split the original dataset into a training set (n=726) and a testing set (n=312) at a 7:3 ratio. After conducting balance testing (Table S2), there were no significant differences in patient characteristics between the two datasets. ACAG was included as both a continuous variable and a categorical variable in model1 and model2 of the multivariate Cox regression analysis, respectively, in the testing set. Similar and significant results were obtained as those in the training set (Table S3). To further validate whether the statistical power of ACAG was overestimated, we subsequently performed propensity score matching (PSM) analysis using Table 2, Figure 4, and Figure 5. PSM was commonly used in cohort studies to mitigate confounding bias when randomization was not feasible, serving as an alternative method for multiple regression analysis with small to medium-sized sample sizes[3-5]. Through PSM, we established a new cohort where ACAG emerged as an important predictor of in-hospital mortality after adjusting for other confounding factors' balance (Table 4). After stratifying patients into two groups based on a cut-off point of 20.375mmol/L, the distinction between survival and non-survival outcomes among severe trauma patients was evident in hospital as well as at 30 and 90 days post-admission (Figure 6).

We conducted a screening of trauma patients in the MIMIC database to identify those who required admission to an intensive care unit and based on a top-ranked diagnosis that matched the ICD-9 or ICD-10 trauma diagnostic codes. These criteria were employed to select critically ill patients in need of intensive care due to traumatic reasons. As Tsiklidis et al. The MIMIC III database was utilized to develop a risk prediction model for trauma patients, with the screening of trauma patients being conducted using ICD-9 codes[6].After reviewing the clinical data of 711 trauma patients, Leskovan et al. demonstrated a significant association between elevated AG without albumin correction and increased mortality and ISS scores in trauma patients. This association was also observed in patients with lower injury severity[7]. In their study, Hu et al. included all sepsis patients with a SOFA score ≥2 and suspected infection, and found that ACAG had a high predictive value for in-hospital mortality in sepsis patients[8]. Gundoğan reviewed the clinical records of pediatric patients admitted to PICU within 3 years and proposed that ACAG was an independent risk factor for death (OR = 1.064,95% CI:1.010-1.122)[9]. All these findings suggested that ACAG was an important prognostic indicator for critically ill ICU patients. The absence of AIS codes and ISS scores in this database poses challenges for conducting subgroup analyses on causes of trauma. This also aligns with our future research direction, which aims to further investigate the relationship between ACAG and different subgroups of trauma patients based on site and severity.

Hemorrhage resulting from trauma often leaded to shock and might subsequently be accompanied by dilutive coagulopathy and hypothermia, frequently associated with severe metabolic acidosis. This condition prolonged hospitalization duration and increases mortality rates[10]. The occurrence of metabolic acidosis in severe trauma patients could be attributed to the heightened production of organic acids, where unmeasured anions served as indicators for dissociated organic acids and were the primary contributors to metabolic acidosis[11-12]. As early as 1983, Stewart proposed identifying unmeasured ions through physicochemical acid-base analysis, elucidating that the charge difference between ions formed the foundation of acid-base physiology. By adhering to principles of electric neutrality and mass conservation laws, the missing charge in plasma was identified as a "gap"[13]. Both strong ion gap (SIG) and anion gap (AG) could typically serve as indicators for estimating ion gaps[14]. Kaplan et al. discovered that SIG and AG could differentiate between survivors and non-survivors of severe vascular injuries more effectively than lactic acid levels, standard base excess (SBE), or injury severity scores (ISS). The clinical utility could be enhanced by employing AG due to the relatively complex nature of SIG calculation[15].

References

[1] Jian L, Zhang Z, Zhou Q, Duan X, Xu H, Ge L. Association between albumin corrected anion gap and 30-day all-cause mortality of critically ill patients with acute myocardial infarction: a retrospective analysis based on the MIMIC-Ⅳ database. BMC Cardiovasc Disord. 2023;23(1):211-220. doi:10.1186/s12872-023-03200-3

[2] Hu B, Zhong L, Yuan M, Min J, Ye L, Lu J, et al. Elevated albumin corrected anion gap is associated with poor in-hospital prognosis in patients with cardiac arrest: A retrospective study based on MIMIC-Ⅳ database. Front Cardiovasc Med. 2023;10:1099003-1099010. doi:10.3389/fcvm.2023.1099003

[3] Jupiter DC. Propensity Score Matching: Retrospective Randomization?. J Foot Ankle Surg. 2017;56(2):417-420. doi:10.1053/j.jfas.2017.01.013

[4] Kane LT, Fang T, Galetta MS, et al. Propensity Score Matching: A Statistical Method. Clin Spine Surg. 2020;33(3):120-122. doi:10.1097/BSD.0000000000000932

[5] Rassen JA, Shelat AA, Myers J, Glynn RJ, Rothman KJ, Schneeweiss S. One-to-many propensity score matching in cohort studies. Pharmacoepidemiol Drug Saf. 2012;21(2):69-80. doi:10.1002/pds.3263

[6] Tsiklidis EJ, Sinno T, Diamond SL. Predicting risk for trauma patients using static and dynamic information from the MIMIC III database. PLoS One. 2022 19;17(1):e0262523. doi: 10.1371/journal.pone.0262523.

[7] Leskovan JJ, Justiniano CF, Bach JA, et al. Anion gap as a predictor of trauma outcomes in the older trauma population: correlations with injury severity and mortality. Am Surg. 2013;79(11):1203-1206.

[8] Hu T, Zhang Z, Jiang Y. Albumin corrected anion gap for predicting in-hospital mortality among intensive care patients with sepsis: A retrospective propensity score matching analysis. Clin Chim Acta. 2021;521:272-277. doi:10.1016/j.cca.2021.07.021

[9] Gündoğan Uzunay B, Köker A, Ülgen Tekerek N, Dönmez L, Dursun O. Role of Albumin-corrected Anion Gap and Lactate Clearance in Predicting Mortality in Pediatric Intensive Care Patients. Balkan Med J. 2023;40(6):430-434. doi:10.4274/balkanmedj.galenos.2023.2023-7-87

[10] Shane AI, Robert W, Arthur K, Patson M, Moses G. Acid-base disorders as predictors of early outcomes in major trauma in a resource limited setting: An observational prospective study. Pan Afr Med J. 2014;17:2. doi:10.11604/pamj.2014.17.2.2007

[11] Martin M, Murray J, Berne T, Demetriades D, Belzberg H. Diagnosis of acid-base derangements and mortality prediction in the trauma intensive care unit: the physiochemical approach [published correction appears in J Trauma. 2005 Oct;59(4):1035]. J Trauma. 2005;58(2):238-243. doi:10.1097/01.ta.0000152535.97968.4e

[12] Zingg T, Bhattacharya B, Maerz LL. Metabolic acidosis and the role of unmeasured anions in critical illness and injury. J Surg Res. 2018;224:5-17. doi:10.1016/j.jss.2017.11.013

[13] Stewart PA. Modern quantitative acid-base chemistry. Can J Physiol Pharmacol. 1983;61(12):1444-1461. doi:10.1139/y83-207

[14] Zingg T, Bhattacharya B, Maerz LL. Metabolic acidosis and the role of unmeasured anions in critical illness and injury. J Surg Res. 2018;224:5-17. doi:10.1016/j.jss.2017.11.013

[15] Kaplan LJ, Kellum JA. Initial pH, base deficit, lactate, anion gap, strong ion difference, and strong ion gap predict outcome from major vascular injury. Crit Care Med. 2004;32(5):1120-1124. doi:10.1097/01.ccm.0000125517.28517.74

Reviewer 2

1. It is not clear when the albumin was drawn relative to the AG. For example, given that albumin is a negative acute phase reactant, if the AG was drawn on admission and the albumin was drawn close to 24 hours later, this could potentially affect the ACAG calculation. If this data on timing is not available, it should be noted in the limitations section of the discussion.

Albumin values within 24 hours of admission to the ICU were not measured for every patient in MIMIC database. We observed over 3000 samples with missing albumin data, and the frequency of albumin measurement within 24 hours for samples without missing data was significantly lower compared to that of anion gap. Moreover, there was a smaller number of samples where both anion gap and albumin were measured simultaneously. These factors presented challenges in obtaining real-time ACAG from the database. Similar to other laboratory data, ACAG was calculated using the average value within 24 hours after ICU admission. This limitation of our study should be acknowledged, and it was anticipated that future research endeavors will address this issue through a real-world prospective study or validation with a larger sample size to further substantiate the relevance of real-time ACAG in determining patient outcomes following severe trauma.

2. In Results/Baseline Characteristics, it states “…SOFA score, SAPS II score…were all higher among survivors”. I believe it should state higher among non-survivors.

The errors that occurred in the writing of the results due to our negligence have been rectified in the original text, for which we sincerely apologize. SOFA score, SAPSⅡ score, APSⅢ score, OASIS score, age, respiratory rate, anion gap, ACAG, sodium, BUN, creatinine, lactate, INR, PT PTT, glucose, mechanical ventilation rate, mechanical ventilation duration were all lower among survivors than non-survivors.

3. The ROC curve analysis shows the predictive value of ACAG, AG, and albumin. It would be more meaningful to also include the Injury Severity Score or other acute severity score to determine how much ACAG adds beyond commonly available scores used to determine severity and estimate prognosis.

Thank you for your insightful feedback. We had incorporated the SOFA score into the ROC curves(Figure 2) and determined an AUC(95%CI) of 0.713(0.666-0.761) for this score. There was no statistically significant difference in AUC between ACAG and SOFA score(Z = 0.425, P = 0.671).

Figure 2 Receiver-operating characteristic curves of the SOFA, ACAG ,ALB and AG to predict in-hospital mortality among trauma patients. SOFA: Sequential Organ Failure Assessment. ACAG: albumin corrected anion gap; ALB: albumin; AG: anion gap.

4. Propensity score matching is typically used to balance treatment and non-treatment groups. Its use in this manuscript is unconventional. Can you provide any references on use of PSM in this type of pathophysiology-based approach?

In small and medium-sized sample size environments, propensity score matching (PSM) can serve as an alternative to multiple regression analysis for mitigating confounding bias in non-randomized observational cohorts. Extensive literature review reveals that PSM not only achieves balance between treatment and control groups by accounting for confounding factors but also facilitates the investigation of associations between specific factors and disease prognosis.

Tang et al. employed a multivariate logistic regression model to estimate the propensity score when investigating the association between INR and all-cause mortality in patients with cardiac arrest. They performed 1:1 matching of subjects in the low INR group and high INR group, followed by re-performing Kaplan-Meier curve analysis and COX regression analysis to examine the potential impact of INR on all-cause mortality in post-cardiac arrest patients [1]. Zhang et al., in their study on the correlation between Acidemia and 30-day/90-day mortality in AMI patients, categorized all participants into Non-acidemia group and acidemia group, implemented PSM to mitigate any imbalances between these groups, subsequently employing Kaplan-Meier survival analysis to compare patient mortality rates across both groups [2]. The method of propensity score matching (PSM) was also employed by Wu et al. in their study, wherein patients were categorized into an AF group and a non-AF group. Subsequently, the primary and secondary outcomes were compared based on the matched data obtained through PSM[3].

The optimal cut-off value of ACAG was initially determined and subsequently re-grouped in our study. Propensity score matching (PSM) was employed to mitigate bias between the two patient groups. Ultimately, KM curves were obtained with adjustment for confounding factors, revealing a significant association between ACAG and both primary and secondary outcomes.

5. The discussion states that “…lactate values were missing by more than 20% and were subsequently excluded.” Exclusion of the lactate severely limits any conclusion that can be drawn from this study. The association between ACAG and in-hospital mortality may be confounded by or entirely due to lactate, which will increase the ACAG. Lactate should be included and can be subtracted from the ACAG and/or adjusted for in the regression analysis. Otherwise, the findings in the manuscript may entirely be due to lactate which is a well known prognostic indicator. A sensitivity analysis can be performed to ensure the missing lactate values do not contribute to uncertainty of the findings.

The expression of gratitude is extended to you for bringing this matter to my attention. Although lactate values were missing in over 20% of cases within our dataset, it was imperative to incorporate lactate into the multivariate regression analysis in order to mitigate bias, given its pivotal role as a biomarker. The technique of multiple imputation was employed to handle missing lactate values.

The number of samples with lactate values before and after imputation was 634 and 1038, with a median of 2.40(1.60,3.50)mmol/L and 2.30(1.55,3.30)mmol/L, respectively. Non-parametric testing revealed no significant difference between the two groups (Z = -1.461, P = 0.144). The scatter plot in Figure S1 illustrated the distribution of lactate values before and after imputation. The imputed data exhibited the same distribution as the observed data, indicating that the missingness is completely at random (MCAR).

Sensitivity analyses conducted using the complete dataset, excluding cases with missing lactate values, demonstrated that the unadjusted ACAG and the results adjusted for model 1 and model 2 (Table S1) did not significantly differ from those obtained using imputed lactate data (Table 4). This confirmed the robustness of the imputation method and indicated that the missing components did not have an impact on the final results.

Figure S1 Multiple imputation scatter plot of lactate. The blue dots represent the observed data, while the red dots indicate the imputed data. The ordinate axis represents the lactate value, and the abscissa axis represents the number of imputations, with 0 representing the original data column.

Table 4 Cox proportional hazard analysis of ACAG of in-hospital mortality in patients with trauma, following imputation of lactate values.

Variable Crude Model 1 Model 2

HR(95% CI) P HR(95% CI) P HR(95% CI) P

ACAG<20.357mmol/L 1(ref) 1(ref) 1(ref)

ACAG≥20.357mmol/L 4.451(3.157-6.276) <0.001 4.166(2.889-6.005) <0.001 3.128(1.615-6.059) 0.001

Continuous 1.212(1.163-1.262) <0.001 1.177(1.126-1.231) <0.001 1.111(1.027-1.203) 0.009

Crude: No covariates were adjusted. Model1: adjusted for age, race, sex, liver disease, congestive heart failure, renal disease, cancer, diabetes. Model2: adjusted for age, race, sex, MBP, respiratory rate, SpO2, SOFA, SAPSⅡ, APSⅢ, OASIS, hematocrit, hemoglobin, platelets, albumin, anion gap, bicarbonate, bun, creatinine, sodium, lactate, INR, PT, PTT, glucose, liver disease, congestive heart failure, renal disease, cancer, diabetes, ventilation, ventilation duration. ACAG:albumin corrected anion gap.

Table S1 Cox proportional hazard analysis of ACAG in the data set excluding cases with missing lactate values

Variable Crude Model 1 Model 2

HR(95% CI) P HR(95% CI) P HR(95% CI) P

ACAG<20.357mmol/L 1(ref) 1(ref) 1(ref)

ACAG≥20.357mmol/L 4.609(3.168-6.707) <0.001 4.365(2.937-6.486) <0.001 3.259(1.536-6.915) 0.002

Continuous 1.195(1.145-1.248) <0.001 1.161(1.109-1.216) <0.001 1.146(1.049-1.252) 0.004

References

[1] Tang Y, Sun J, Yu Z, et al. Association between prothrombin time-international normalized ratio and prognosis of post-cardiac arrest patients: A retrospective cohort study. Front Public Health. 2023;11:1112623. Published 2023 Jan 20. doi:10.3389/fpubh.2023.1112623

[2] Zhang T, Guan YZ, Liu H. Association of Acidemia With Short-Term Mortality of Acute Myocardial Infarction: A Retrospective Study Base on MIMIC-III Database. Clin Appl Thromb Hemost. 2020;26:1076029620950837. doi:10.1177/1076029620950837

[3] Wu CS, Chen PH, Chang SH, et al. Atrial Fibrillation Is Not an Independent Determinant of Mortality Among Critically Ill Acute Ischemic Stroke Patients: A Propensity Score-Matched Analysis From the MIMIC-IV Database. Front Neurol. 2022;12:730244. Published 2022 Jan 17. doi:10.3389/fneur.2021.730244

Once again, thank you very much for your comments and suggestions.

Yours sincerely,

Yuzhou Xu

2024-01-15

Attachment

Submitted filename: Response to Reviewers.docx

pone.0300012.s004.docx^{(165.1KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0300012.r003

Decision Letter 1

Tanja Grubić Kezele

30 Jan 2024

PONE-D-23-37440R1Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: a retrospective study based on MIMIC-III and IV databasesPLOS ONE

Dear Dr. Xu,

==============================

Your manuscript, entitled "Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: a retrospective study based on MIMIC-III and IV databases", has been reviewed. Your efforts to revise the manuscript are appreciated. However, the peer review process continues because Reviewer 1 has a few additional comments that the author should address. Please find the reviewer's commentary below.

==============================

Please submit your revised manuscript by Mar 15 2024 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

We look forward to receiving your revised manuscript.

Kind regards,

Tanja Grubić Kezele, Ph.D., M.D.

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

Reviewer #2: (No Response)

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: (No Response)

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: No

Reviewer #2: (No Response)

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

Reviewer #2: (No Response)

**********

6. Review Comments to the Author

Reviewer #1: I am rather glad to review the new version of the manuscript which is much more acceptable for publication. One minor problem remained that the detailed ICD code for trauma or the specific spectrum for the alleged trauma population has been given. The authors are strongly encouraged to provide detailed ICD-9/10 they used to extract the patients on the Github repo or submit to the journal as supplementary materials. Without enough illustration, the conclusions of this article can hardly be put into clinical practice.

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

**********

PLoS One. 2024 Mar 7;19(3):e0300012. doi: 10.1371/journal.pone.0300012.r004

Author response to Decision Letter 1

4 Feb 2024

Dear editor and reviewers of PLOS ONE:

Our reference: PONE-D-23-37440R1

Title: Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: A retrospective study based on MIMIC-Ⅲ and Ⅳ databases

By: Yuzhou Xu et al

Thank you very much for your letter and for the editors’ and reviewers’ comments concerning our manuscript entitled "Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: A retrospective study based on MIMIC-Ⅲ and Ⅳ databases" (ID: PONE-D-23-37440R1). These comments are of great reference value to the revision and improvement of our paper and have important guiding significance to our researches. We have studied comments carefully and have made correction. We hope that the revision is acceptable and look forward to hearing from you soon. Revised portion are marked in color in the paper. The main corrections in the paper and the responds to the reviewer’s comments are as flowing:

Reviewer 1

1. I am rather glad to review the new version of the manuscript which is much more acceptable for publication. One minor problem remained that the detailed ICD code for trauma or the specific spectrum for the alleged trauma population has been given. The authors are strongly encouraged to provide detailed ICD-9/10 they used to extract the patients on the Github repo or submit to the journal as supplementary materials. Without enough illustration, the conclusions of this article can hardly be put into clinical practice.

We appreciate your valuable feedback, and as per your suggestion, we have incorporated the coding range for trauma-related ICD-9/10 in our article. Specifically, the ICD-9 trauma codes encompass the range of 800-959, while the ICD-10 trauma codes include S00-S99, T00-T14, and T20-T32[1-3].

A comprehensive inventory of trauma-related ICD-9/10 codes within the MIMIC-III/IV databases is provided in the appendix. Furthermore, we have incorporated all diagnostic codes for each patient into the original dataset. The aforementioned items have been submitted to the journal as supplementary materials..

References

[1] Flynn-O'Brien KT, Fallat ME, Rice TB, et al. Pediatric Trauma Assessment and Management Database: Leveraging Existing Data Systems to Predict Mortality and Functional Status after Pediatric Injury. J Am Coll Surg. 2017;224(5):933-944.e5. doi:10.1016/j.jamcollsurg.2017.01.061

[2] Clark DE, Black AW, Skavdahl DH, Hallagan LD. Open-access programs for injury categorization using ICD-9 or ICD-10. Inj Epidemiol. 2018;5(1):11-18. doi:10.1186/s40621-018-0149-8

[3] Wada T, Yasunaga H, Yamana H, et al. Development and validation of a new ICD-10-based trauma mortality prediction scoring system using a Japanese national inpatient database. Inj Prev. 2017;23(4):263-267. doi:10.1136/injuryprev-2016-042106

Furthermore, we have rectified certain formatting issues in the references. All accompanying figures have been adjusted by PACE.

Once again, thank you very much for your comments and suggestions.

Yours sincerely,

Yuzhou Xu

2024-02-04

Attachment

Submitted filename: Response to ReviewersR2.docx

pone.0300012.s005.docx^{(15KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0300012.r005

Decision Letter 2

Tanja Grubić Kezele

20 Feb 2024

Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: a retrospective study based on MIMIC-III and IV databases

PONE-D-23-37440R2

Dear Dr. Xu,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Tanja Grubić Kezele, Ph.D., M.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

**********

6. Review Comments to the Author

Reviewer #1: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

**********

PLoS One. doi: 10.1371/journal.pone.0300012.r006

Acceptance letter

Tanja Grubić Kezele

27 Feb 2024

PONE-D-23-37440R2

PLOS ONE

Dear Dr. Xu,

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now being handed over to our production team.

At this stage, our production department will prepare your paper for publication. This includes ensuring the following:

* All references, tables, and figures are properly cited

* All relevant supporting information is included in the manuscript submission,

* There are no issues that prevent the paper from being properly typeset

If revisions are needed, the production department will contact you directly to resolve them. If no revisions are needed, you will receive an email when the publication date has been set. At this time, we do not offer pre-publication proofs to authors during production of the accepted work. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few weeks to review your paper and let you know the next and final steps.

Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

If we can help with anything else, please email us at customercare@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Prof. dr. Tanja Grubić Kezele

Academic Editor

PLOS ONE

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Data

(ZIP)

pone.0300012.s001.zip^{(2.7MB, zip)}

S1 Fig

(TIF)

pone.0300012.s002.tif^{(251.9KB, tif)}

S1 File

(DOCX)

pone.0300012.s003.docx^{(178KB, docx)}

Attachment

Submitted filename: Response to Reviewers.docx

pone.0300012.s004.docx^{(165.1KB, docx)}

Attachment

Submitted filename: Response to ReviewersR2.docx

pone.0300012.s005.docx^{(15KB, docx)}

Data Availability Statement

All relevant data are within the manuscript and its Supporting Information files.

[pone.0300012.ref001] 1.Duque P, Mora L, Levy JH, Schöchl H. Pathophysiological Response to Trauma-Induced Coagulopathy: A Comprehensive Review. Anesth Analg. 2020;130(3):654–664. doi: 10.1213/ANE.0000000000004478 . [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref002] 2.Shashaty MG, Kalkan E, Bellamy SL, Reilly JP, Holena DN, Cummins K, et al. Computed tomography-defined abdominal adiposity is associated with acute kidney injury in critically ill trauma patients. Crit Care Med. 2014;42 (7):1619–28. doi: 10.1097/CCM.0000000000000306 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref003] 3.Callaway DW, Wolfe R. Geriatric Trauma. Emergency Medicine Clinics of North America. 2007;25(3):837–860. doi: 10.1016/j.emc.2007.06.005 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref004] 4.Shane AI, Robert W, Arthur K, Patson M, Moses G. Acid-base disorders as predictors of early outcomes in major trauma in a resource limited setting: An observational prospective study. Pan Afr Med J. 2014;17:2–8. doi: 10.11604/pamj.2014.17.2.2007 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref005] 5.Messerer DAC, Halbgebauer R, Nilsson B, Pavenstädt H, Radermacher P, Huber-Lang M. Immunopathophysiology of trauma-related acute kidney injury. Nat Rev Nephrol. 2021;17(2):91–111. doi: 10.1038/s41581-020-00344-9 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref006] 6.Ahmed FS, Ali L, Joseph BA, Ikram A, Ul Mustafa R, Bukhari SAC. A statistically rigorous deep neural network approach to predict mortality in trauma patients admitted to the intensive care unit. J Trauma Acute Care Surg. 2020;89(4):736–742. doi: 10.1097/TA.0000000000002888 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref007] 7.Rudkin SE, Grogan TR, Treger RM. Relationship Between the Anion Gap and Serum Lactate in Hypovolemic Shock. J Intensive Care Med. 2022;37(12):1563–1568. doi: 10.1177/08850666221106413 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref008] 8.Zhong L, Xie B, Ji XW, Yang XH. The association between albumin corrected anion gap and ICU mortality in acute kidney injury patients requiring continuous renal replacement therapy. Intern Emerg Med. 2022;17(8):2315–2322. doi: 10.1007/s11739-022-03093-8 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref009] 9.Silva JM Jr, Ribas Rosa de Oliveira AM, Mendes Nogueira FA, Vianna PM, Amendola CP, Carvalho Carmona MJ, et al. Metabolic Acidosis Assessment in High-Risk Surgeries: Prognostic Importance. Anesth Analg. 2016;123(5):1163–1171. doi: 10.1213/ANE.0000000000001575 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref010] 10.Johnson AE, Pollard TJ, Shen L, Lehman LW, Feng M, Ghassemi M, et al. MIMIC-Ⅲ, a freely accessible critical care database. Sci Data. 2016;3:160035–160043. doi: 10.1038/sdata.2016.35 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref011] 11.Johnson AEW, Bulgarelli L, Shen L, Gayles A, Shammout A, Horng S, et al. MIMIC-Ⅳ, a freely accessible electronic health record dataset. Sci Data. 2023;10(1):1–9. doi: 10.1038/s41597-022-01899-x [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref012] 12.Flynn-O’Brien KT, Fallat ME, Rice TB, Gall CM, Nance ML, et al. Pediatric Trauma Assessment and Management Database: Leveraging Existing Data Systems to Predict Mortality and Functional Status after Pediatric Injury. J Am Coll Surg. 2017;224(5):933–944.e5. doi: 10.1016/j.jamcollsurg.2017.01.061 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref013] 13.Clark DE, Black AW, Skavdahl DH, Hallagan LD. Open-access programs for injury categorization using ICD-9 or ICD-10. Inj Epidemiol. 2018;5(1):11–18. doi: 10.1186/s40621-018-0149-8 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref014] 14.Oh MS, Carroll HJ. The anion gap. N Engl J Med. 1977;297(15): 814–817. doi: 10.1056/NEJM197710132971507 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref015] 15.Figge J, Jabor A, Kazda A, Fencl V. Anion gap and hypoalbuminemia. Crit Care Med. 1998;26(11):1807–1810. doi: 10.1097/00003246-199811000-00019 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref016] 16.Hayati Rezvan P, Lee KJ, Simpson JA. The rise of multiple imputation: a review of the reporting and implementation of the method in medical research. BMC Med Res Methodol. 2015;15:30–43. doi: 10.1186/s12874-015-0022-1 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref017] 17.Chen D, Fritz MS. Comparing Alternative Corrections for Bias in the Bias-Corrected Bootstrap Test of Mediation. Eval Health Prof. 2021;44(4):416–427. doi: 10.1177/01632787211024356 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref018] 18.Wang P, Yuan D, Zhang C, Zhu P, Jia S, Song Y, et al. High fibrinogen-to-albumin ratio with type 2 diabetes mellitus is associated with poor prognosis in patients undergoing percutaneous coronary intervention: 5-year findings from a large cohort. Cardiovasc Diabetol. 2022;21(1):46–58. doi: 10.1186/s12933-022-01477-w [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref019] 19.Martin M, Murray J, Berne T, Demetriades D, Belzberg H. Diagnosis of acid-base derangements and mortality prediction in the trauma intensive care unit: the physiochemical approach. J Trauma. 2005;58(2):238–243. doi: 10.1097/01.ta.0000152535.97968.4e [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref020] 20.Stewart PA. Modern quantitative acid-base chemistry. Can J Physiol Pharmacol. 1983;61(12):1444–1461. doi: 10.1139/y83-207 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref021] 21.Zingg T, Bhattacharya B, Maerz LL. Metabolic acidosis and the role of unmeasured anions in critical illness and injury. The Journal of surgical research. 2018;224:5–17. doi: 10.1016/j.jss.2017.11.013 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref022] 22.Kaplan LJ, Kellum JA. Initial pH, base deficit, lactate, anion gap, strong ion difference, and strong ion gap predict outcome from major vascular injury. Crit Care Med. 2004;32(5):1120–1124. doi: 10.1097/01.ccm.0000125517.28517.74 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref023] 23.Leskovan JJ, Justiniano CF, Bach JA, Cook CH, Lindsey DE, Eiferman DS, et al. Anion Gap as a Predictor of Trauma Outcomes in the Older Trauma Population: Correlations with Injury Severity and Mortality. The American SurgeonTM. 2013;79(11):1203–1206. 10.1177/000313481307901126 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref024] 24.Zhang XB, Shu WB, Li AB, Lan GH. The Anion Gap and Mortality in Critically Ill Patients with Hip Fractures. Contrast Media Mol Imaging. 2022; 2022: 1591507–1591512. doi: 10.1155/2022/1591507 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref025] 25.Hatherill M, Waggie Z, Purves L, Reynolds L, Argent A. Correction of the anion gap for albumin in order to detect occult tissue anions in shock. Arch Dis Child. 2002;87(6):526–529. doi: 10.1136/adc.87.6.526 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref026] 26.Wang R, Li J, Chen H, Xiao Z, Xu R, Hu Y, et al. Preoperative albumin corrected anion gap is associated with in-hospital and long-term mortality in patients undergoing coronary artery bypass grafting in a retrospective cohort study. J Thorac Dis. 2022;14(12):4894–4903. doi: 10.21037/jtd-22-1633 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref027] 27.Chawla LS, Shih S, Davison D, Junker C, Seneff MG. Anion gap, anion gap corrected for albumin, base deficit and unmeasured anions in critically ill patients: implications on the assessment of metabolic acidosis and the diagnosis of hyperlactatemia. BMC Emergency Medicine. 2008;8:18–25. doi: 10.1186/1471-227X-8-18 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref028] 28.Kim SC, Kim DH, Kim TY, Kang C, Lee SH, Jeong JH, et al. The Revised Trauma Score plus serum albumin level improves the prediction of mortality in trauma patients. Am J Emerg Med. 2017;35(12):1882–1886. doi: 10.1016/j.ajem.2017.06.027 . [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref029] 29.Gonzalez Rodriguez E, Cardenas JC, Lopez E, Cotton BA, Tomasek JS, Ostrowski SR, et al. Early Identification of the Patient with Endotheliopathy of Trauma by Arrival Serum Albumin. Shock. 2018;50(1):31–37. doi: 10.1097/SHK.0000000000001036 . [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref030] 30.Jian L, Zhang Z, Zhou Q, Duan X, Xu H, Ge L. Association between albumin corrected anion gap and 30-day all-cause mortality of critically ill patients with acute myocardial infarction: a retrospective analysis based on the MIMIC-Ⅳ database. BMC Cardiovasc Disord. 2023;23(1):211–220. doi: 10.1186/s12872-023-03200-3 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref031] 31.Hu B, Zhong L, Yuan M, Min J, Ye L, Lu J, et al. Elevated albumin corrected anion gap is associated with poor in-hospital prognosis in patients with cardiac arrest: A retrospective study based on MIMIC-Ⅳ database. Front Cardiovasc Med. 2023;10:1099003–1099010. doi: 10.3389/fcvm.2023.1099003 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref032] 32.Hu T, Zhang Z, Jiang Y. Albumin corrected anion gap for predicting in-hospital mortality among intensive care patients with sepsis: A retrospective propensity score matching analysis. Clin Chim Acta. 2021;521:272–277. doi: 10.1016/j.cca.2021.07.021 [DOI] [PubMed] [Google Scholar]

[pone.0300012.ref033] 33.Li P, Shi L, Yan X, Wang L, Wan D, Zhang Z, et al. Albumin Corrected Anion Gap and the Risk of in-Hospital Mortality in Patients with Acute Pancreatitis: A Retrospective Cohort Study. J Inflamm Res. 2023;16:2415–2422. doi: 10.2147/JIR.S412860 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0300012.ref034] 34.Chawla LS, Jagasia D, Abell LM, Seneff MG, Egan M, Danino N, et al. Anion gap, anion gap corrected for albumin, and base deficit fail to accurately diagnose clinically significant hyperlactatemia in critically ill patients. J Intensive Care Med. 2008;23(2):122–127. doi: 10.1177/0885066607312985 [DOI] [PubMed] [Google Scholar]

PERMALINK

Association between albumin-corrected anion gap and in-hospital mortality of intensive care patients with trauma: A retrospective study based on MIMIC-Ⅲ and Ⅳ databases

Fei Yin

Zhenguo Qiao

Xiaofei Wu

Qiang Shi

Rongfei Jin

Yuzhou Xu

Roles

Abstract

Background

Materials and methods

Results

Conclusion

Introduction

Method

Database

Patients

Statistical analysis

Results

Baseline characteristics

Fig 1. Patient selection flowchart.

Table 1. Characteristics of the study population between in-hospital survival and in-hospital mortality group.

Table 4. Cox proportional hazard analysis of ACAG of in-hospital mortality in patients with trauma.

ROC curve analysis

Fig 2. Receiver-operating characteristic curves of the SOFA, ACAG, ALB and AG to predict in-hospital mortality among trauma patients.

Determination of optimal cut-off value for survival analysis

Fig 3. The best cutoff value of ACAG was taken by the K-M curve with log-rank test.

Post-PSM characteristics

Fig 4.

Table 2. Characteristics of the study population were compared between low and high ACAG groups before and after PSM.

Fig 5. Standardized mean differences (SMDs) of variables before and after matching.

Subgroup analyses

Table 3. Subgroup analysis of the association between different levels of ACAG and in-hospital mortality.

Outcome measurement

Fig 6.

Discussion

Supporting information

Acknowledgments

Data Availability

Funding Statement

References

Decision Letter 0

Tanja Grubić Kezele

Roles

Author response to Decision Letter 0

Decision Letter 1

Tanja Grubić Kezele

Roles

Author response to Decision Letter 1

Decision Letter 2

Tanja Grubić Kezele

Roles

Acceptance letter

Tanja Grubić Kezele

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases