Table S1. Frequently Identified Organisms from Cultures

Table S1. Frequently identified organisms from cultures

Organisms / SCT (%) / LCT (%)
Escherichia coli / 48 (55.8) / 65 (36.7)
Klebsiella pneumoniae / 22 (25.6) / 50 (28.2)
Klebsiella oxytoca / 6 (7.0) / 15 (8.5)
Enterobacter cloacae / 3 (3.5) / 10 (5.6)
Enterococcus faecalis / 2 (2.3) / 13 (7.3)
Aeromonas hydrophila / 2 (2.3) / 10 (5.6)
Enterococcus faecium / 4 (4.7) / 4 (2.3)
Bacteroides sp. / 2 (2.3) / 2 (1.1)
Clostridium perfringens / 3 (3.5) / 3 (1.7)
Enterococcus casseliflavus / 1 (1.2) / 9 (5.1)
Streptococcus sp. / 2 (2.3) / 6 (3.4)

SCT = short-course therapy, LCT = long-course therapy.

Table S2. Frequently used initial antimicrobials

Organisms / SCT (%) / LCT (%)
Cefoperazone/sulbactam / 42 (48.8) / 86 (48.6)
Ampicillin/sulbactam / 20 (23.3) / 24 (13.6)
Piperacillin/tazobactam / 14 (16.3) / 36 (20.3)
Meropenem / 6 (7.0) / 14 (7.9)
Levofloxacin / 0 (0) / 6 (3.4)
Cefmetazole / 3 (9.3) / 2 (1.1)
Cefepime / 0 (0) / 3 (1.7)
Ceftriaxone / 0 (0) / 2 (1.1)
Others / 1 (1.2) / 1 (0.6)

SCT = short-course therapy, LCT = long-course therapy

Table S3. Logistic regression analysis for primary outcome (30-day mortality).

Logistic regression Number of observation = 232

Area under ROC curve = 0.80

Odds ratio (95% confidence interval) / P value
Gram positives / 3.21 (0.77-13.4) / 0.11
qSOFA (median, range) / 2.03 (0.98-4.18) / 0.06
Empirical antimicrobials covering causative organisms / 0.18 (0.04-0.85) / 0.03
Polymicrobial infection / 0.44 (0.07-2.65) / 0.37
Time to drainage / 1.01 (0.99-1.02) / 0.47
SCT vs LCT / 1.07 (0.25-4.52) / 0.93

LR=likelihood ratio, qSOFA= Sepsis-related Organ Failure Assessment, SCT=short course therapy, LCT=long course therapy

Variables were selected both from univariate analysis and clinical inference, while removing variables which are likely to be co-related. For example, we did not include Tokyo Guideline Grade deliberately despite of its clinical significance on univariate analysis (P=0.02). Tokyo Guideline included neurological, cardiovascular, and respiratory dysfunction, and it overlaps with qSOFA.

Table S4. Logistic regression analysis for secondary outcome (composite outcome).

Logistic regression Number of observation = 234

Area under ROC curve = 0.71

Odds ratio (95% confidence interval) / P value
Gram positives / 1.43 (0.51-3.99) / 0.50
qSOFA (median, range) / 0.996 (0.61-1.62) / 0.99
Empirical antimicrobials covering causative organisms / 0.49 (0.16-1.51) / 0.21
Polymicrobial infection / 0.71 (0.23-2.18) / 0.55
Time to drainage / 1.01 (1.00-1.02) / 0.02
SCT vs LCT / 1.08 (0.48-2.45) / 0.85

LR=likelihood ratio, qSOFA= Sepsis-related Organ Failure Assessment, SCT=short course therapy, LCT=long course therapy

Variables were selected both from univariate analysis and clinical inference, while removing variables which are likely to be co-related. For example, we did not include Tokyo Guideline Grade deliberately despite of its clinical significance on univariate analysis (P=0.02). Tokyo Guideline included neurological, cardiovascular, and respiratory dysfunction, and it overlaps with qSOFA.

Table S5. Logistic regression analysis for propensity score analysis

Logistic regression Number of observation = 232

Area under ROC curve = 0.81

Odds ratio (95% confidence interval) / P value
WBC / 0.94 (0.83-1.07) / 0.35
Liver mass / 4.59 (0.96-21.9) / 0.06
ID specialist consultation / 0.58 (0.13-2.63) / 0.48
Gram positives / 3.67 (0.83-16.3) / 0.09
qSOFA (median, range) / 2.35 (1.08-5.13) / 0.03
Empirical antimicrobials covering causative organisms / 0.19 (0.04-0.90) / 0.04
Polymicrobial infection / 0.58 (0.09-3.63) / 0.56
Time to drainage / 1.01 (0.09-3.63) / 0.53
SCT vs LCT / 0.97 (0.21-4.41) / 0.97

WBC=white blood cell, ID=infectious diseases, qSOFA= Sepsis-related Organ Failure Assessment, SCT=short course therapy, LCT=long course therapy

Variables were selected both from univariate analysis and clinical inference, while removing variables which are likely to be co-related.