Intravenous albumin for initial resuscitation and mortality in septic shock patients: propensity score analyses using a nationwide inpatient database

ABSTRACT

BACKGROUND

A recent large randomized controlled study suggested that albumin administration may reduce mortality in patients with septic shock. However, it remains unclear whether albumin should be used for initial resuscitation only without subsequent supplementation. The present study aimed to assess whether intravenous albumin infusion for initial resuscitation without subsequent supplementation is associated with reduced mortality in patients with septic shock using a national inpatient administrative claims database.

METHODS

In this retrospective cohort study using the Japanese Diagnosis Procedure Combination inpatient database from July 2010 to March 2016, we identified patients with septic shock who received ≥3000 mL of total fluids on day 1 after the start of noradrenaline. We defined patients who received ≥75 g of albumin within 2 days after the start of noradrenaline as the albumin group and other patients as the control group. We performed one-to-one propensity-score matching analyses to analyze the associations of albumin administration with 28-day and 90-day mortality.

RESULTS

We identified 85,563 eligible patients during the 69-month study period. Of these, 6888 patients (8.1%) were allocated to the albumin group. After propensity-score matching, no significant differences were seen between the two groups for 28-day mortality (relative risk 0.95, 95% confidence interval 0.88–1.02) and 90-day mortality (relative risk 0.97, 95% confidence interval 0.92–1.03).

CONCLUSIONS

Intravenous albumin for initial resuscitation without subsequent supplementation was not significantly associated with reduced short-term mortality in patients with septic shock. Intravenous albumin for initial resuscitation only may therefore be ineffective.

INTRODUCTION

For decades, albumin has been administered to patients with sepsis and septic shock during initial resuscitation to achieve adequate blood pressure and intravascular volume. A recent large randomized controlled study suggested that albumin administration may reduce mortality in patients with septic shock [1, 2].

Hypoalbuminemia is an independent risk factor for increased morbidity and mortality in critically ill patients [3] and patients with sepsis and septic shock [4, 5]. The SAFE trial, in which 4% albumin was compared with 0.9% saline for fluid resuscitation until day 28, indicated that albumin administration was safe and had a potential benefit in reducing 28-day mortality in patients with severe sepsis [1].

The ALBIOS trial was the first large randomized controlled trial on patients with severe sepsis or septic shock to examine the beneficial effect of intravenous albumin injection targeting a serum albumin level of ≥3.0 g/dL until day 28 [2]. While the study found no advantages of intravenous albumin for 28-day and 90-day survival, a post-hoc subgroup analysis in septic shock patients showed a significant absolute reduction of 6.3% in 90-day mortality in the albumin group compared with the control group [2]. Because of the low quality of evidence, International and Japanese Guidelines for Management of Sepsis and Septic Shock, both published in 2016, weakly recommend the use of albumin in addition to crystalloids for initial resuscitation and subsequent intravascular volume replacement in patients with sepsis and septic shock who require substantial amounts of crystalloids [6, 7]. However, these clinical guidelines differ from the ALBIOS trial, which only focused on the initial resuscitation phase and did not describe supplemental albumin use with a particular target for the serum albumin level or albumin treatment period within 28 days after resuscitation [6, 7]. Thus, patients treated in accordance with the guidelines may not have the same conditions as those in the ALBIOS trial. In fact, after patient resuscitation, physicians in Japan usually target a serum albumin level of ≥2.0 g/dL following the Japanese albumin guideline and the national health insurance policy [8]. It remains unclear whether albumin should be used for initial resuscitation in patients with septic shock.

We hypothesized that albumin administration for initial resuscitation does not have a beneficial effect even in patients with septic shock requiring substantial amounts of fluids. The present study aimed to assess whether the albumin administration for initial resuscitation can reduce mortality in patients with septic shock using a national inpatient database in Japan.

METHODS

The Institutional Review Board of the University of Tokyo approved the study. The requirement for informed consent was waived because of the anonymous nature of the data.

DATA SOURCE

Relevant data were obtained from the Japanese Diagnosis Procedure Combination inpatient database, which includes discharge abstracts and administrative claims data from more than 1200 acute-care hospitals and covers approximately 90% of all tertiary-care emergency hospitals in Japan. The database includes information on patient age, sex, smoking history, body height, body weight, diagnoses, comorbidities on admission, complications after admission, procedures, prescriptions, and costs. Diagnoses, comorbidities, and complications are recorded by International Classification of Diseases, Tenth Revision (ICD-10) codes. A previous study on records of diagnoses and procedures in the Japanese Diagnosis Procedure Combination inpatient database established the validity of this database [9].

STUDY COHORT

We included all patients with a diagnosis of septic shock between July 2010 and March 2016. Septic shock patients were defined as patients who were admitted with sepsis and required noradrenaline during hospitalization. Previous studies estimated the prevalence of sepsis based on ICD-9 Clinical Modification and ICD-10 Australian Modification codes using administrative databases [10–12]. In the present study, sepsis was defined as conditions with any bacterial, fungal, tuberculous, anaerobic, or microbiologically undefined infection recorded using ICD-10 codes (Supplementary Table 1) based on the previous studies [10–12]. We evaluated the Angus organ failure score, which identifies severity of illness by patient organ failures based on ICD-10 codes and Japanese procedure codes and has a maximum score of six, with higher scores indicating more organ failures [10–12]. The Angus organ failure score within the first 2 days of noradrenaline was calculated using the ICD-10 codes and Japanese procedure codes shown in Supplementary Table 2.

We excluded patients who: i) were aged <20 years; ii) had congestive heart failure with New York Heart Association class 3 or 4; iii) had pathological conditions for which albumin administration was clinically indicated (hepatic cirrhosis, intestinal malabsorption syndrome, nephrotic syndrome, burns); iv) died within 2 days after the start of noradrenaline; v) were discharged alive within 2 days after the start of noradrenaline; or vi) lacked body weight data. To identify patients requiring substantial amounts of fluids for initial resuscitation, we included patients with ≥3000 mL of total fluids administered on the first day of noradrenaline administration.

We defined patients who received ≥75 g of albumin in total within 2 days after the start of noradrenaline as the albumin group and other patients as the control group.

STUDY VARIABLES

The following study variables were examined: (i) patient characteristics including age, sex, smoking history (nonsmoker, current/past smoker, missing data), body mass index at admission, Barthel index at admission [13], Japan Coma Scale status at admission [14], emergency admission, updated Charlson comorbidity index score [15], treatment year, hospital case volume quartile [16], site of infection (Supplementary Table 1), Angus organ failure score within the first 2 days of noradrenaline calculated from the codes shown in Supplementary Table 2 [10–12], ambulance use, and intensive care unit admission within the first 2 days of noradrenaline; (ii) treatment variables within the first 2 days of noradrenaline including mechanical ventilation, renal replacement therapy (continuous, intermittent), polymyxin B hemoperfusion, cardiac output measurement, arterial pressure monitoring, central venous pressure monitoring, pulmonary artery pressure monitoring, cardiac pacing, cardiac massage, defibrillation, extracorporeal membrane oxygenation, therapeutic hypothermia, cardiac surgery, noradrenaline dose, dopamine, dobutamine, adrenaline, vasopressin, low-dose corticosteroids, landiolol, sivelestat sodium hydrate, sodium bicarbonate, immunoglobulin, recombinant human soluble thrombomodulin, antithrombin, synthetic protease inhibitors, enteral nutrition, probiotics, proton pump inhibitors, anti-ulcer medication, crystalloids, total fluids administered, red blood cells, fresh-frozen plasma, and platelets.

Body mass index was categorized as <18.5, 18.5–22.9, 23.0–24.9, 25.0–29.9, ≥30.0, or missing [17]. The Barthel index, a standardized scale for evaluating patient self-care ability, reflects performance of activities of daily living after discharge from hospital, with a total score ranging from 0 (complete dependence) to 100 (complete independence) [13]. Barthel index score was categorized as 0–70, 75–95, 100, or missing [18]. Japan Coma Scale status was categorized as alert consciousness, dizziness, somnolence, or coma. Japan Coma Scale status was shown to be well correlated with the Glasgow Coma Scale score [14]. Charlson comorbidity index was calculated from the recorded diagnoses for each patient and categorized as 0, 1, 2, 3, or ≥4 [15]. Total fluids administered were calculated as the sum of albumin, crystalloids, other blood products (red blood cells, fresh-frozen plasma, platelets), parenteral nutrition, and other fluids.

OUTCOME VARIABLES

The primary outcomes were 28-day and 90-day mortality after initiation of noradrenaline. Secondary outcomes were interventions during hospitalization (mechanical ventilation, renal replacement therapy), length of hospital stay, length of intensive care unit stay, length of mechanical ventilation, total cost, and fluid volumes within the first 7 days of noradrenaline (albumin, crystalloids, total fluids administered, percentages of albumin in total fluids administered).

No information allowing identification of individual patients, hospitals, or physicians was obtained.

STATISTICAL ANALYSIS

Propensity score analyses were performed to account for differences in baseline characteristics between the albumin and control groups [19, 20]. Logistic regression models were used to predict each patient’s probability of receiving albumin ≥75 g within the first 2 days of noradrenaline, incorporating all baseline variables shown in Supplementary Table 3. The C-statistic was calculated to evaluate the goodness of fit. One-to-one nearest-neighbor matching without replacement was then performed based on the estimated propensity scores of each patient with a caliper width set at 20% of the standard deviation [19, 20]. Absolute standardized differences of <10% were considered negligible imbalances in the baseline characteristics between the albumin and control groups after propensity-score matching [21].

Subgroup analyses were performed in patients with: i) 3,000–10,000 mL of total fluids administered on day 1 after the start of noradrenaline; ii) 10,000–15,000 mL of total fluids administered on day 1 after the start of noradrenaline; iii) 15,000–20,000 mL of total fluids administered on day 1 after the start of noradrenaline; and iv) ≥20,000 mL of total fluids administered on day 1 after the start of noradrenaline. Propensity score analyses were re-performed in each subgroup.

Continuous variables were expressed as median (interquartile range). Categorical variables were presented as number (percentage). Length of hospital stay and total cost were evaluated among survivors. Length of intensive care unit was examined for patients with intensive care unit admission and survivors. Length of mechanical ventilation was investigated for patients with mechanical ventilation and survivors. Binary outcomes were compared by the chi-square test, and continuous outcomes were compared by analysis of variance or the Wilcoxon rank-sum test. Relative risk (RR) and 95% confidence interval (CI) were calculated for 28-day mortality, 90-day mortality, mechanical ventilation, and renal replacement therapy. Comparisons of fluid volumes over time were performed by two-factor analysis of variance for repeated measurements. Value of P < 0.05 were considered statistically significant. All analyses were performed with STATA/MP 14.2 software (StataCorp, College Station, TX, USA).

RESULTS

We identified 85,563 eligible patients during the 69-month study period (Fig. 1). Of these, 6888 patients (8.1%) were allocated to the albumin group. One-to-one propensity-score matching created 6760 matched pairs. Supplementary Table 3 shows the patient characteristics before and after propensity-score matching. After propensity-score matching, the patient characteristics were well balanced between the two groups.

Fig. 1 Flowchart of patient selection.

NYHA class, New York Heart Association functional classification

Table 1 shows the outcomes after propensity-score matching. No significant differences were seen between the two groups for 28-day mortality (RR 0.95, 95% CI 0.88–1.02) and 90-day mortality (RR 0.97, 95% CI 0.92–1.03). Proportion of mechanical ventilation was significantly lower in the albumin group compared with the control group (RR 0.98, 95% CI 0.97–1.00). Length of hospital stay and length of mechanical ventilation were significantly longer in the albumin group compared with the control group (median, 48 vs. 43 days and 6 vs. 5 days, respectively). Total cost was significantly higher in the albumin group compared with the control group (median: 5148 × 10³ vs. 4847 × 10³ yen).

Table 1 Outcomes after propensity-score matching

	Albumin group (n = 6,760)	Control group (n = 6,760)	Relative risk (95% CI)	P-value
Primary outcomes
28-day mortality	1,203 (18)	1,272 (19)	0.95 (0.88–1.02)	0.13
90-day mortality	1,842 (27)	1,896 (28)	0.97 (0.92–1.03)	0.3
Secondary outcomes
Mechanical ventilation	5,960 (88)	6,058 (90)	0.98 (0.97–1.00)	0.007
Renal replacement therapy	2,772 (41)	2,754 (41)	1.01 (0.97–1.05)	0.75
Length of hospital stay*	48 (30–77)	43 (26–70)		<0.001
Length of ICU stay**	7 (3–14)	7 (3–14)		0.29
Length of mechanical ventilation***	6 (2–15)	5 (2–13)		<0.001
Total cost × 103 (yen)*	5,148 (3,462–7,740)	4,847 (3,131–7,235)		<0.001

CI: confidence interval. ICU: intensive care unit

Continuous variables are expressed as median (interquartile range)

Categorical variables are presented as number (percentage)

*Length of hospital stay and total cost are presented for survivors (n = 4,692 in albumin group; n = 4,634 in control group)

**Length of ICU stay is presented for patients with ICU admission and survivors (n = 3,336 in albumin group; n = 3,350 in control group)

***Length of mechanical ventilation is presented for patients with mechanical ventilation and survivors (n = 3,949 in albumin group; n = 3,997 in control group)

Supplementary Table 4 shows the detailed information for fluid therapy during the first 7 days after the start of noradrenaline after propensity-score matching. There were no differences in total crystalloids and total fluids administered within 7 days between the two groups.

Table 2 shows the outcomes after propensity-score matching in the subgroup analyses. After propensity-score matching, the patient characteristics were well balanced between the two groups in each subgroup analysis. All subgroup analyses showed similar results to those observed in the main analyses.

Table 2 Outcomes of subgroup analyses after propensity-score matching in patients with 3,000–10,000, 10,000–15,000, 15,000–20,000, and ≥20,000 mL of total administered fluids on day 1 after starting noradrenaline

	Albumin group	Control group	Relative risk (95% CI)	P-value
Patients with 3,000–10,000 mL of total administered fluids on day 1 after starting noradrenaline
28-day mortality	430/1,576 (27)	425/1,576 (27)	1.01 (0.90–1.13)	0.84
90-day mortality	604/1,576 (38)	571/1,576 (36)	1.06 (0.97–1.16)	0.22
Patients with 10,000–15,000 mL of total administered fluids on day 1 after starting noradrenaline
28-day mortality	208/1,204 (17)	236/1,204 (20)	0.88 (0.74–1.04)	0.14
90-day mortality	307/1,204 (26)	338/1,204 (28)	0.91 (0.80–1.04)	0.15
Patients with 15,000–20,000 mL of total administered fluids on day 1 after starting noradrenaline
28-day mortality	205/1,320 (16)	201/1,320 (15)	1.02 (0.85–1.22)	0.83
90-day mortality	314/1,320 (24)	321/1,320 (24)	0.98 (0.85–1.12)	0.75
Patients with ≥20,000 mL of total administered fluids on day 1 after starting noradrenaline
28-day mortality	356/2,667 (13)	392/2,667 (15)	0.91 (0.80–1.04)	0.16
90-day mortality	605/2,667 (23)	637/2,667 (24)	0.95 (0.86–1.05)	0.30

CI: confidence interval

Data are shown as number/total number (percentage)

Deaths within 2 days after starting noradrenaline were excluded to prevent immortal time bias

DISCUSSION

Our retrospective nationwide study demonstrated no significant association between in-hospital mortality and intravenous albumin for initial resuscitation in septic shock patients requiring substantial amounts of fluids. Subgroup analyses showed similar findings in all stratified groups of total fluids administered on day 1 after the start of noradrenaline.

To date, no randomized controlled trial specifically designed to assess the effect of intravenous albumin for initial resuscitation in septic shock patients has been conducted. The present study investigated the associations between intravenous albumin for initial resuscitation and clinical outcomes with and without stratification for volume of initial fluids. After applying careful methodology involving propensity score analyses to address the potential impact of confounders, the findings of our study contradicted the previous subgroup analyses on septic shock patients in the ALBIOS trial and recommendations from the International and Japanese Guidelines for Management of Sepsis and Septic shock [2, 6, 7]. While albumin resuscitation has been considered more efficient in severe septic patients [22], our study did not show advantages of albumin resuscitation for reducing 28-day and 90-day mortality in patients with substantial amounts of administered fluids.

Intravenous albumin is used for initial resuscitation only in real-world clinical practice in Japan. The amounts of albumin on days 1 and 2 of resuscitation were similar among our study, the SAFE trial, and the ALBIOS trial (mean: 62.5, 60, and 47 on day 1 and 25, 40, and 24 on day 2, respectively) [1, 2]. However, in our study, the percentages of albumin in the total fluids administered after the first 2 days of resuscitation were almost zero in both groups (Supplementary Table 4). It is conceivable that, in accordance with the International and Japanese sepsis guidelines, physicians used intravenous albumin during the initial resuscitation phase only and did not continue albumin infusion [6, 7].

In the ALBIOS trial, albumin was used for supplementation following initial resuscitation to maintain a serum albumin level of ≥3.0 g/dL until day 28 [2]. Our results contradict the results of the subgroup analyses in septic shock patients in the ALBIOS trial. The findings may suggest a survival advantage for an albumin supplementation strategy, i.e. initial resuscitation and subsequent supplemental administration of albumin for at least 28 days.

Our study suggested that albumin resuscitation may be associated with longer duration of mechanical ventilation, longer hospital stay, and higher total cost. Taken together with the failure in mortality reduction, the present results may be against the use of albumin for initial resuscitation only, a recommendation based on the recent guidelines for sepsis and actual clinical practice in sepsis in Japan [6, 7].

This study has some limitations. First, serum albumin levels were not measured, and it thus remains unclear whether these levels were corrected in patients treated with albumin. Second, decisions to start albumin were not assigned randomly. This could cause confounding by indication. We attempted to control for measured confounding factors by performing propensity-score matching analyses. Third, clinicians were likely to use albumin in patients who required very large volumes of fluids for initial resuscitation, and thus our cohort after propensity-score matching may have become biased. To deal with this problem, we performed subgroup analyses stratified by total fluids administered on day 1, and the results for mortality were similar to those in the main analyses. Fourth, we used total fluids administered within the first 2 days of noradrenaline as one of the propensity score estimators, but intravenous albumin may have reduced fluid volumes during resuscitation, which leads to underestimation in the effect of albumin solutions on patient outcomes. However, the effect of albumin on the total administered fluids were small in the ALBIOS trial; total fluids administered on day 2 were significantly lower in the albumin group compared with the control group (median: 3800 mL vs. 4000 mL), while no difference was seen in total fluids administered on day 1 (median: 4300 mL vs. 4250 mL) [2].

CONCLUSION

Our study suggests that albumin administration for initial resuscitation was not associated with reduced 28-day and 90-day mortality in patients with septic shock. Therefore, albumin administered for initial resuscitation only may be ineffective. The effect of supplemental albumin use following resuscitation remains to be explored.

ACKNOWLEDGMENTS

This work was supported by grants from the Ministry of Health, Labour and Welfare of Japan (H29-Policy-Designated-009 and H29-ICT-General-004); Ministry of Education, Culture, Sports, Science and Technology of Japan (17H04141); and Japan Agency for Medical Research and Development.

CONFLICT OF INTERESTS

The authors declare that they have no conflict of interests.

Supplementary Table 1 ICD-10 codes for sepsis

ICD-10 codes
Lung
A15 A16 A31 A37 B44 J01 J02 J03 J04 J05 J06 J13 J14 J15 J16 J17 J18 J20 J21 J22 J44.0 J44.1 J47 J85 J86
Abdomen
A00 A02 A03 A04 A05 A08 A09 K35 K36 K37 K57 K61 K63.0 K63.1 K65 K75.0 K75.1 K80.0 K80.1 K80.3 K80.4 K80.8 K81.0 K83.0 K91.8
Urinary tract
N10 N11 N12 N15.1 N15.9 N16.0 N30 N34 N39.0 N41 N74 N75 T83.5 T83.6
Central nervous system
A17 A39 G00 G01 G02 G03 G04 G05 G06 G07 G08 G09
Skin and soft tissues
A33 A34 A35 A46 A48 B35 B36 L03 L04 L08 L72.6 L88
Cardiovascular system
I30 I33 I80 T82.6 T82.7
Others
A18 A19 A20 A21 A22 A23 A24 A25 A27 A28 A30 A32 A36 A38 A40 A41 A42 A43 A44 A49 A50 A51 A52 A53 A54 A65 A66 A67 A69 B37 B38 B39 B40 B41 B42 B43 B45 B46 B47 B48 B49 N70 N71 N72 N73 N76 N77 M00 M86 T81.4 T84.5 T84.6 T84.7 T85.7 T81.4 T88.0

ICD-10: International Classification of Diseases, Tenth Revision

Supplementary Table 2 ICD-10 codes, Japanese procedure codes, and claims used to calculate the Angus organ failure score

Organ failure	ICD-10 codes	Japanese procedure codes or claims
Cardiovascular*	I95 R57	Vasopressor or inotrope (dopamine, epinephrine, norepinephrine, vasopressin)
Respiratory		Mechanical ventilation
Neurologic	F05 G93.4
Hematologic	D65 D69.5 D69.6 D69.8 D69.9
Hepatic	K72.0 K76.3
Renal	N17	Renal replacement therapy

ICD-10: International Classification of Diseases, Tenth Revision

*All patients received noradrenaline

Supplementary Table 3 Baseline characteristics before and after propensity-score matching

	Unmatched group			Matched group
	Albumin (n = 6,888)	Control (n = 78,675)	ASD	Albumin (n = 6,760)	Control (n = 6,760)	ASD
Age, years	72 (63–79)	73 (64–81)	10.6	72 (63–79)	72 (63–79)	1.0
Men	63%	61%	5.4	63%	63%	0.6
Smoking history
Nonsmoker	50%	54%	8.1	50%	51%	1.3
Current/past smoker	30%	29%	1.6	30%	30%	1.2
Unknown	20%	16%	8.7	20%	20%	0.3
BMI category
<18.5	19%	21%	4.4	20%	20%	0.4
18.5–22.9	43%	41%	2.6	43%	42%	0.5
23.0–24.9	16%	15%	1.7	16%	15%	0.2
25.0–29.9	15%	15%	0	15%	16%	1.7
≥30.0	4%	4%	1.3	4%	4%	1.6
Missing	3%	4%	2.2	3%	3%	0.8
Barthel index at admission
0–70	45%	53%	15.2	46%	46%	0.9
75–95	5%	4%	4.8	5%	5%	0.3
100	35%	27%	18.6	35%	34%	0.9
Missing	15%	17%	5.4	15%	15%	0.2
JCS status at admission
Alert	76%	66%	20.3	75%	75%	0.4
Dizziness	12%	17%	12.7	13%	13%	1.1
Somnolence	5%	7%	8.3	5%	5%	1.1
Coma	7%	10%	9.8	7%	7%	0.1
Charlson comorbidity index
0	46%	45%	2.4	46%	46%	1.8
1	24%	26%	6	24%	24%	1.1
2	17%	17%	1.2	17%	18%	2.2
3	7%	6%	0.3	6%	6%	0.3
≥4	6%	5%	3.8	6%	6%	2.0
Treatment year
2010	8%	6%	9.1	8%	8%	0.7
2011	15%	12%	9.5	15%	14%	1.6
2012	16%	15%	2.9	16%	16%	0.2
2013	19%	18%	2.5	19%	19%	0.3
2014	21%	22%	2.8	21%	21%	1.4
2015	19%	24%	10.8	19%	19%	0.0
2016	2%	4%	11.1	2%	2%	1.0
Hospital case volume quartile
1–191	23%	25%	5.6	23%	24%	2.3
192–365	25%	25%	0	25%	25%	0.5
366–601	29%	25%	8.9	29%	29%	0.1
602–1,376	24%	25%	3.6	24%	23%	1.7
Site of infection
Lung	17%	24%	16	17%	16%	3.0
Abdomen	47%	35%	23.4	47%	47%	0.4
Urinary tract	2%	9%	29	2%	2%	1.5
Central nervous system	1%	1%	7	0.5%	0.4%	0.9
Skin and soft tissues	1%	2%	6.7	1%	2%	1.6
Cardiovascular system	9%	7%	6.1	9%	9%	1.7
Others	23%	22%	3.6	23%	23%	1.2
Angus organ failure score within first 2 days of noradrenaline
1	13%	35%	54.9	13%	10%	7.9
2	46%	41%	11	46%	47%	2.8
3	32%	20%	29.2	32%	33%	2.0
4	9%	5%	17.2	9%	10%	0.6
Transferred by ambulance	43%	50%	14.6	43%	44%	0.7
ICU admission within first 2 days of noradrenaline	68%	44%	49	68%	69%	3.5
Treatment within first 2 days of noradrenaline
Mechanical ventilation	83%	54%	64.3	83%	86%	8.9
CRRT	26%	15%	27.4	26%	26%	0.5
IRRT	2%	2%	3	2%	2%	0.5
Polymyxin B hemoperfusion	26%	12%	37.2	26%	27%	1.0
Cardiac output monitoring	38%	22%	35.1	37%	38%	1.4
Arterial pressure monitoring	86%	67%	46.8	86%	87%	5.4
CVP monitoring	52%	33%	38.7	52%	53%	2.0
PAP monitoring	14%	8%	21.5	14%	15%	2.7
Cardiac pacing	0.3%	0.5%	1.8	0.5%	0.4%	0.2
Cardiac massage	3%	3%	1.3	3%	3%	0.3
Defibrillation	3%	2%	4.2	3%	3%	0.1
ECMO	3%	1%	13.9	3%	3%	0.9
Noradrenaline dose, μg/kg/min	0.14 (0.06–0.29)	0.11 (0.04–0.23)	18.5	0.14 (0.06–0.29)	0.13 (0.06–0.28)	0.1
Dopamine	64%	42%	45.4	63%	65%	3.1
Dobutamine	29%	16%	31.8	29%	30%	1.9
Adrenaline	21%	10%	29.4	20%	20%	0.3
Vasopressin	14%	7%	20.9	14%	14%	1.8
Low-dose steroids	18%	15%	7	17%	17%	2.0
Landiolol	20%	10%	29.7	20%	21%	3.0
Sivelestat sodium hydrate	29%	11%	45.1	29%	29%	0.6
Sodium bicarbonate	55%	25%	65.5	55%	56%	3.1
Immunoglobulin	38%	22%	36	38%	38%	1.2
rhTM	22%	15%	17.7	22%	23%	3.4
Antithrombin	35%	15%	45.7	35%	35%	0.2
Synthetic protease inhibitors	47%	28%	39	47%	47%	1.3
Enteral nutrition	11%	13%	6.4	11%	11%	0.7
Probiotics	13%	12%	1	12%	12%	0.3
Anti-ulcer medication	92%	81%	32.9	92%	93%	5.1
Crystalloids within first 2 days, L	18.0 (11.5–25.8)	8.4 (5.0–14.8)	85.2	17.8 (11.3–25.5)	17.3 (10.0–25.5)	0.1
Total administered fluids within first 2 days, L	22.2 (15.2–30.7)	10.0 (6.5–17.5)	95.7	22.0 (15.0–30.2)	21.0 (13.0–30.5)	0
RBC within first 2 days, L	0.84 (0.28–1.68)	0 (0–0.56)	67.2	0.84 (0.28–1.68)	0.56 (0–1.68)	4.7
FFP within first 2 days, L	0.72 (0–1.68)	0 (0–0.48)	67	0.72 (0–1.68)	0.48 (0–1.44)	3.8
Platelets within first 2 days, L	0 (0–0.40)	0 (0–0)	50.1	0 (0–0.40)	0 (0–0.40)	2.7

ASD: absolute standardized difference, BMI: body mass index, CRRT: continuous renal replacement therapy, CVP: central venous pressure, ECMO: extracorporeal membrane oxygenation, FFP: fresh-frozen plasma, ICU: intensive care unit, IRRT: intermittent renal replacement therapy, JCS: Japan Coma Scale, PAP: pulmonary artery pressure, RBC: red blood cells, rhTM: recombinant human thrombomodulin

Continuous variables are expressed as median (interquartile range)

Categorical variables are presented as percentage

Supplementary Table 4 Fluid therapy during the first 7 days of noradrenaline after propensity-score matching

Variables	Albumin		Control		P-value
	No. of patients	Value	No. of patients	Value
Albumin (mg)
Day 1	6760	62.5 (50, 87.5)	6760	12.5 (0, 25)	<0.001
Day 2	6760	25 (12.5, 37.5)	6760	0 (0, 25)	<0.001
Day 3	6760	12.5 (0, 25)	6760	0 (0, 12.5)	<0.001
Day 4	6553	0 (0, 25)	6541	0 (0, 12.5)	<0.001
Day 5	6449	0 (0, 25)	6423	0 (0, 0)	<0.001
Day 6	6388	0 (0, 12.5)	6350	0 (0, 0)	<0.001
Day 7	6340	0 (0, 12.5)	6289	0 (0, 0)	<0.001
7 days total	6340	137.5 (100, 187.5)	6289	50 (12.5, 75)	<0.001
Crystalloids (mL)
Day 1	6760	14750 (8306, 21250)	6760	14000 (7000, 21500)	0.77
Day 2	6760	2000 (1000, 4750)	6760	2000 (1000, 5000)	0.06
Day 3	6760	1000 (500, 2500)	6760	1000 (500, 2500)	0.45
Day 4	6553	750 (0, 2000)	6541	500 (0, 1500)	0.45
Day 5	6449	500 (0, 1500)	6423	500 (0, 1500)	0.044
Day 6	6388	500 (0, 1200)	6350	500 (0, 1000)	0.14
Day 7	6340	500 (0, 1000)	6289	500 (0, 1000)	0.85
7-day total	6340	23250 (16000, 32750)	6289	22050 (14,500, 32500)	0.79
Total fluids administered (mL)
Day 1	6760	17355 (10540, 24680)	6760	16315 (8400, 24900)	0.88
Day 2	6760	3480 (2020, 6445)	6760	3400 (2000, 6500)	0.68
Day 3	6760	2280 (1500, 3500)	6760	2000 (1500, 3495)	0.34
Day 4	6553	2000 (1200, 3000)	6541	2000 (1000, 2780)	0.73
Day 5	6449	2000 (1000, 2680)	6423	1600 (1000, 2500)	0.003
Day 6	6388	1700 (1000, 2500)	6350	1500 (800, 2400)	0.006
Day 7	6340	1500 (1000, 2500)	6289	1500 (500, 2150)	0.11
7-day total	6340	32880 (24380, 43640)	6289	31100 (22000, 42580)	0.50
Percentages of albumin in total fluids administered (%)
Day 1	6760	7.2 (4.2, 12.5)	6760	0.5 (0, 2.9)	<0.001
Day 2	6760	8.1 (1.4, 20)	6760	0 (0, 3.7)	<0.001
Day 3	6760	2.0 (0, 12.5)	6760	0 (0, 2.9)	<0.001
Day 4	6553	0 (0, 7.2)	6541	0 (0, 2.2)	<0.001
Day 5	6449	0 (0, 5.6)	6423	0 (0, 1.0)	<0.001
Day 6	6388	0 (0, 5.0)	6350	0 (0, 0)	<0.001
Day 7	6340	0 (0, 4.5)	6289	0 (0, 0)	0.020
7-day total	6340	6.6 (4.4, 10)	6289	1.5 (0.3, 3.1)	<0.001

Values are expressed as median (interquartile range)

The 7-day total values are cumulative data for the entire study period (until discharge alive, death, or day 7, whichever came first)

Percentages of albumin in total fluids administered were calculated from total intravenous albumin (4% or 20% albumin solution; mL) divided by total fluids administered (mL)

REFERENCES

• 1.   Finfer  S,  Bellomo  R,  Boyce  N,  French  J,  Myburgh  J,  Norton  R, et al. A comparison of albumin and saline for fluid resuscitation in the intensive care unit. N Engl J Med 2004;350:2247–2256.
• 2.   Caironi  P,  Tognoni  G,  Masson  S,  Fumagalli  R,  Pesenti  A,  Romero  M, et al. Albumin replacement in patients with severe sepsis or septic shock. N Engl J Med 2014;370:1412–1421.
• 3.   Vincent  JL,  Dubois  MJ,  Navickis  RJ,  Wilkes  MM. Hypoalbuminemia in acute illness: is there a rationale for intervention? A meta-analysis of cohort studies and controlled trials. Ann Surg 2003;237:319–334.
• 4.   SAFE Study  Investigators,  Finfer  S,  McEvoy  S,  Bellomo  R,  McArthur  C,  Myburgh  J, et al. Impact of albumin compared to saline on organ function and mortality of patients with severe sepsis. Intensive Care Med 2011;37:86–96.
• 5.   Caironi  P. POINT: Should intravenous albumin be used for volume resuscitation in severe sepsis/septic shock? Yes. Chest 2016;149:1365–1367.
• 6.   Rhodes  A,  Evans  LE,  Alhazzani  W,  Levy  MM,  Antonelli  M,  Ferrer  R, et al. Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock: 2016. Intensive Care Med 2017;43:304–377.
• 7.   Nishida  O,  Ogura  H,  Egi  M,  Fujishima  S,  Hayashi  Y,  Iba  T, et al. The Japanese Clinical Practice Guidelines for Management of Sepsis and Septic Shock 2016 (J-SSCG 2016). Acute Med Surg 2018;5:3–89.
• 8.   Yasumura  S,  Makino  S,  Matsumoto  M,  Kono  T,  Kino  S,  Tanaka  A, et al. Evidence-based Guidelines for the Use of Albumin Products Japan Society of Transfusion Medicine and Cell Therapy. Jpn J Transfusion Cell Ther 2017;63:641–663.
• 9.   Yamana  H,  Moriwaki  M,  Horiguchi  H,  Kodan  M,  Fushimi  K,  Yasunaga  H. Validity of diagnoses, procedures, and laboratory data in Japanese administrative data. J Epidemiol 2017;27:476–482.
• 10.   Sasabuchi  Y,  Matsui  H,  Lefor  AK,  Fushimi  K,  Yasunaga  H. Risks and benefits of stress ulcer prophylaxis for patients with severe sepsis. Crit Care Med 2016;44:e464–469.
• 11.   Sundararajan  V,  Macisaac  CM,  Presneill  JJ,  Cade  JF,  Visvanathan  K. Epidemiology of sepsis in Victoria, Australia. Crit Care Med 2005;33:71–80.
• 12.   Angus  DC,  Linde-Zwirble  WT,  Lidicker  J,  Clermont  G,  Carcillo  J,  Pinsky  MR. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med 2001;29:1303–1310.
• 13.   Manhony  FI,  Barthel  DW. Functional evaluation: The barthel index. A simple index of independence useful in scoring improvement in the rehabilitation of the chronically ill. Md State Med J 1965;14:61–65.
• 14.   Shigematsu  K,  Nakano  H,  Watanabe  Y. The eye response test alone is sufficient to predict stroke outcome--reintroduction of Japan Coma Scale: a cohort study. BMJ Open 2013;3:e002736.
• 15.   Quan  H,  Li  B,  Couris  CM,  Fushimi  K,  Graham  P,  Hider  P, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol 2011;173:676–682.
• 16.   Walkey  AJ,  Wiener  RS. Hospital case volume and outcomes among patients hospitalized with severe sepsis. Am J Respir Crit Care Med 2014;189:548–555.
• 17.  Obesity: preventing and managing the global epidemic. Report of a WHO consultation. World Health Organ Tech Rep Ser 2000;894:i–xii, 1–253.
• 18.   Wade  DT,  Collin  C. The Barthel ADL Index: a standard measure of physical disability? Int Disabil Stud 1988;10:64–67.
• 19.   Griswold  ME,  Localio  AR,  Mulrow  C. Propensity score adjustment with multilevel data: setting your sites on decreasing selection bias. Ann Intern Med 2010;152:393–395.
• 20.   Rosenbaum  PR,  Rubin  DB. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat 1985;39:33–38.
• 21.   Austin  PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med 2009;28:3083–3107.
• 22.  Xu J-Y, Chen Q-H, Xie J-F,  Pan  C,  Liu  SQ,  Huang  LW, et al. Comparison of the effects of albumin and crystalloid on mortality in adult patients with severe sepsis and septic shock: a meta-analysis of randomized clinical trials. Crit Care 2014;18:702.