Characteristics of the study cohort
A total of 380 septic patients with positive blood cultures in the ED between 2017 and 2019 were enrolled in this study. 22 patients had a viral or fungal infection, 4 patients refused to participate, and 6 patients were transferred to other hospitals. Finally, 358 participants completed the 28 follow-up days (Fig. 1).
In our study, 358 septic patients with positive blood cultures were divided into ICU (n = 104) and non-ICU (n = 254) admission groups. When comparing the characteristics of these patients, we found that there were significant differences in age (P = 0.040), chronic obstructive pulmonary disease (COPD) (P = 0.013), hypertension (P = 0.042), mean arterial pressure (MAP) (P = 0.000), heart rate (HR) (P = 0.000), white blood cells (WBC) (P = 0.000), lactate (Lac) (P = 0.000), PCT (P = 0.000), C-reactive protein (CRP) (P = 0.000), albumin (ALB) (P = 0.000), SOFA (P = 0.000), MEDS (P = 0.000), APACHE II (P = 0.000), and PIRO (P = 0.000) between ICU and non-ICU admission groups (Table 2, Fig. 1).
Table 2
Characteristics of the study cohort
Characteristics
|
ICU
(n = 104)
|
Non-ICU
(n = 254)
|
P
|
MODS
(n = 75)
|
Non-MODS
(n = 283)
|
P
|
Survivor
(n = 292)
|
Non-Survivors
(n = 66)
|
P
|
Age, years
|
78 (69,84)
|
70 (59,78)
|
0.040
|
82(78,85)
|
66(55,75)
|
0.000
|
66(59,73)
|
84(80,86)
|
0.000
|
Male, %
COPD, %
Hypertension, %
CVD, %
CHF, %
CRD, %
DM, %
MAP, mmHg
Temperature, ℃
Heart rate, beats/min
Respiratory rate, bpm
WBC, °¡109/L
Lac, mmol/L
PCT, ng/mL
CRP
ALB
NLR
SOFA
MEDS
APACHE II
PIRO
|
59(56.7)
33(31.7)
54(51.9)
23(22.1)
50(48.1)
15(14.4)
42(40.4)
77(72,81)
38.51 ± 0.46
102(90,115)
24(20,28)
14.4
(12,18.8)
3.7 ± 1.4
8.6
(2.5, 12)
111 ± 39
29 ± 6
14(8,24)
6 (6,8)
11(9,14)
16.3 ± 4.8
16 (12,20)
|
147(57.9)
49(19.3)
102(40.2)
43(16.9)
99(39.0)
25(9.8)
84(33.1)
85(77,93)
38.46 ± 0.52
98(80,112)
24(23,30)
12.3
(10.8,14.9)
2.4 ± 1.4
5
(2.2, 10.1)
73 ± 23
33 ± 7
12(8,20)
4 (3,7)
8 (7,10)
11.3 ± 2.6
10 (8,12)
|
0.843
0.013
0.042
0.251
0.113
0.212
0.188
0.000
0.359
0.000
0.555
0.000
0.000
0.000
0.000
0.000
0.080
0.000
0.000
0.000
0.000
|
44(58.7)
21(28.0)
38(50.7)
17(22.7)
33(44.0)
9(12.0)
28(37.3)
79 (75,83)
38.57 ± 0.41
116(104, 134)
30(29,33)
15
(14.5,20.3)
3.3 ± 1.5
8.6
(2.8, 12.5)
115 ± 38
25 ± 4
19(10,25)
6 (5,9)
13 (10,16)
15.6 ± 4.2
16(12,20)
|
163(57.6)
65(23.0)
125(44.3)
49(17.3)
123(43.5)
31(11.0)
91(32.2)
84(76,92)
38.43 ± 0.55
90(78,101)
23(20,24)
11.2
(10.3,12.6)
2.5 ± 1.6
3.5
(1.8, 10)
70 ± 21
34 ± 5
11(8,19)
4 (4,7)
10 (8,10)
12.4 ± 2.5
10(9,12)
|
0.868
0.364
0.315
0.288
0.934
0.798
0.397
0.000
0.075
0.000
0.000
0.000
0.040
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
|
168(57.8)
55(19.0)
136(46.7)
52(17.8)
109(37.5)
28(9.8)
86(29.5)
87(79,90)
38.43 ± 0.51
90(80,100)
23(20,25)
10.6
(10.2,12.1)
1.8 ± 0.5
2.4
(1.4, 7)
67 ± 22
34 ± 6
9(7,11)
4(4,7)
10(8,10)
12.5 ± 2.6
10 (9,12)
|
38(58.1)
15(23.3)
29(44.2)
12(18.6)
34(51.2)
7(11.6)
20(30.2)
82(71,91)
38.61 ± 0.46
119(107,126)
30(29,34)
16.8
(14.9,20.3)
3.7 ± 1.6
8.5
(3.6, 12.5)
114 ± 34
25 ± 4
19(9,25)
6(5,9)
12(10,16)
15.5 ± 4.3
15(12,20)
|
0.500
0.317
0.760
0.894
0.084
0.715
0.924
0.000
0.080
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
|
Abbreviations: ICU, Intensive care unit; MODS: multiple organ dysfunction syndrome; COPD, chronic obstructive pulmonary disease; CVD, cerebrovascular disease; CHF, congestive heart failure; CRD, chronic renal disease; DM, diabetes mellitus; MAP, mean arterial pressure; WBC, white blood cell; PCT, procalcitonin; Lac, lactate; CRP :C-reactive protein; ALB, albumin; NLR, neutrophils and lymphocytes ratio; SOFA, Sequential Organ Failure Assessment; APACHE II, Acute Physiology and Chronic Health Evaluation II; MEDS, Mortality in emergency department sepsis; PIRO, Predisposition, Infection, Response, and Organ dysfunction |
In addition, we also divided the participants into two other categories, MODS (n = 75) and non-MODS (n = 283) according to the development of MODS within 3 days. There were obvious differences in age (P = 0.000), MAP (P = 0.000), HR (P = 0.000), respiratory rate (RR) (P = 0.000), WBC (P = 0.000), Lac (P = 0.040), PCT (P = 0.000), CRP (P = 0.000), ALB (P = 0.000), (NLR) (P = 0.000), SOFA (P = 0.000), MEDS (P = 0.000), APACHE II (P = 0.000), and PIRO (P = 0.000) between these two groups (Table 2, Fig. 1).
After 28 days follow-up, statistical differences were evident in age (P = 0.000), MAP (P = 0.000), HR (P = 0.000), RR (P = 0.000), WBC (P = 0.000), Lac (P = 0.040), PCT (P = 0.000), CRP (P = 0.000), ALB (P = 0.000), (NLR) (P = 0.000), SOFA (P = 0.000), MEDS (P = 0.000), APACHE II (P = 0.000), and PIRO (P = 0.000) between survivor (n = 292) and non-survivor (n = 66) groups (Table 2, Fig. 1).
The median of the SOFA, MEDS, and PIRO scores, and mean of the APACHE II score in ICU-admission, MODS development, and 28-day mortality groups are shown in Table 2, Fig. 2. The average SOFA, MEDS, APACHE II, and PIRO scores differed significantly between patients who did, and did not, meet the outcome criteria (P = 0.000).
Spearman’s correlations between 28-day mortality and SOFA, MEDS, APACHE II and PIRO scores.
To investigate the correlations between 28-day mortality and SOFA, MEDS, APACHE II and PIRO scores, Spearman’s correlation analysis was performed. Significant negative liner correlations of 28-day mortality with SOFA, MEDS, APACHE II and PIRO scores were presented in Fig. 3A-D (r values were − 0.62, -0.75, -0.64, -0.49 respectively, and all P < 0.001).
Binary logistic regression analysis of the prognostic outcome of septic patients with positive blood cultures in the ED
Binary logistic regression was used to analyze the independent predictors in our study cohort. The results showed that WBC (β = 0.176, odds ratio (OR) (95%CI) = 1.192 (1.031–1.379), P = 0.018), PCT (β = 0.241, OR (95%CI) = 1.273 (1.037–1.563), P = 0.021), Lac (β = 1.261, OR (95%CI) = 3.530 (1.867–6.676), P = 0.000), ALB (β=-0.142, OR (95%CI) = 0.867 (0.792–0.949), P = 0.002), NLR (β = 0.077, OR (95%CI) = 1.080 (1.011–1.155), P = 0.023), MEDS score (β = 0.895, OR (95%CI) = 2.447 (1.607–3.728), P = 0.000), APACHE II score (β = 0.623, OR (95%CI) = 1.864 (1.405–2.474), P = 0.000) and PIRO score (β = 0.328, OR (95%CI) = 1.388 (1.118–1.723), P = 0.003) were the independent predictors of ICU admission, but the SOFA score (β = 0.220, OR (95%CI) = 2.701 (1.548–4.713), P = 0.120) was not (Table 3).
Table 3
Binary Logistic regression analysis of clinical outcome for septic patients with positive blood culture.
Variables
|
Predictor
|
β
|
SE
|
Wald
|
P
|
OR(95%CI)
|
ICU-admission
|
WBC
|
0.176
|
1.453
|
5.632
|
0.018
|
1.192(1.031–1.379)
|
|
PCT
|
0.241
|
0.105
|
5.315
|
0.021
|
1.273(1.037–1.563)
|
|
Lac
|
1.261
|
0.325
|
15.055
|
0.000
|
3.530(1.867–6.676)
|
|
ALB
|
-0.142
|
0.046
|
9.501
|
0.002
|
0.867(0.792–0.949)
|
|
NLR
|
0.077
|
0.034
|
5.205
|
0.023
|
1.080(1.011–1.155)
|
|
SOFA
|
0.220
|
0.142
|
2.414
|
0.120
|
2.701(1.548–4.713)
|
|
MEDS
|
0.895
|
0.215
|
17.369
|
0.000
|
2.447(1.607–3.728)
|
|
APACHE II
|
0.623
|
0.144
|
18.644
|
0.000
|
1.864(1.405–2.474)
|
|
PIRO
|
0.328
|
0.110
|
8.840
|
0.003
|
1.388(1.118–1.723)
|
|
Constant
|
-13.820
|
2.939
|
22.112
|
0.000
|
|
MODS
|
WBC
|
0.362
|
0.129
|
7.861
|
0.005
|
1.436(1.115–1.849)
|
|
PCT
|
0.136
|
0.094
|
2.093
|
0.048
|
1.145(0.953–1.377)
|
|
Lac
|
0.177
|
0.299
|
0.351
|
0.554
|
1.194(0.664–2.145)
|
|
ALB
|
-0.665
|
0.192
|
11.967
|
0.001
|
0.514(0.353–0.750)
|
|
NLR
|
0.270
|
0.093
|
8.452
|
0.004
|
1.310(1.092–1.572)
|
|
SOFA
|
0.001
|
0.212
|
0.000
|
0.995
|
1.001(0.660–1.518)
|
|
MEDS
|
0.383
|
0.159
|
5.808
|
0.016
|
1.466(1.074–2.001)
|
|
APACHE II
|
0.372
|
0.162
|
5.286
|
0.022
|
1.450(1.056–1.990)
|
|
PIRO
|
0.196
|
0.084
|
5.461
|
0.019
|
1.216(1.032–1.434)
|
|
Constant
|
-18.945
|
4.468
|
19.978
|
0.000
|
|
28-day-mortality
|
WBC
|
0.996
|
0.354
|
7.919
|
0.005
|
2.707(1.353–5.417)
|
|
PCT
|
0.306
|
0.098
|
9.764
|
0.002
|
1.358(1.121–1.644)
|
|
Lac
|
1.822
|
0.701
|
6.753
|
0.009
|
6.184(1.565–24.436)
|
|
ALB
|
-0.226
|
0.116
|
3.807
|
0.051
|
0.798(0.636–1.001)
|
|
NLR
|
0.128
|
0.153
|
0.701
|
0.402
|
1.136(0.843–1.533)
|
|
SOFA
|
0.092
|
0.157
|
0.340
|
0.560
|
1.096(0.805–1.492)
|
|
MEDS
|
0.440
|
0.147
|
8.922
|
0.003
|
1.553(1.163–2.074)
|
|
APACHE II
|
0.484
|
0.163
|
8.832
|
0.003
|
1.623(1.179–2.234)
|
|
PIRO
|
0.322
|
0.123
|
6.884
|
0.009
|
1.380(1.085–1.754)
|
|
Constant
|
-16.492
|
3.630
|
20.646
|
0.000
|
|
CI, confidence interval; OR, odds ratio; SE, standard error. WBC, white blood cell; PCT, procalcitonin; Lac, lactate; ALB, albumin; NLR, neutrophils and lymphocytes ratio; SOFA, Sequential Organ Failure Assessment; APACHE II, Acute Physiology and Chronic Health Evaluation II; MEDS, Mortality in emergency department sepsis; PIRO, Predisposition, Infection, Response, and Organ dysfunction |
In MODS development, WBC (β = 0.362, OR (95%CI) = 1.436 (1.115–1.849), P = 0.005), PCT (β = 0.136, OR (95%CI) = 1.145 (0.953–1.377), P = 0.048), ALB (β=-0.665, OR (95%CI) = 0.514 (0.353–0.750), P = 0.001), NLR (β = 0.270, OR (95%CI) = 1.310 (1.092–1.572), P = 0.004), MEDS score (β = 0.383, OR (95%CI) = 1,466 (1.074–2.001), P = 0.016), APACHE II score (β = 0.372, OR (95%CI) = 1.450 (1.056–1.990), P = 0.022), and PIRO score (β = 0.196, OR (95%CI) = 1.216 (1.032–1.434), P = 0.019) were the independent predictors of developing MODS, but the SOFA scores (β = 0.001, OR (95%CI) = 1.001 (0.660–1.518), P = 0.995) and Lac (β = 0.177, OR (95%CI) = 1.194 (0.664–2.145), P = 0.554) determined it was not a predictor (Table 3).
In addition, with 28-day mortality, the results showed that WBC (β = 0.996, OR (95%CI) = 2.707 (1.353–5.417), P = 0.005), PCT (β = 0.306, OR (95%CI) = 1.358 (1.121–1.644), P = 0.002), Lac (β = 1.822, OR (95%CI) = 6.184 (1.565–24.436), P = 0.009), MEDS score (β = 0.440, OR (95%CI) = 1.553 (1.163–2.074), P = 0.003), APACHE II score (β = 0.484, OR (95%CI) = 1.623 (1.179–2.234), P = 0.003), and PIRO score (β = 0.322, OR (95%CI) = 1.380 (1.085–1.754), P = 0.009) were the independent predictors of 28-day mortality, but SOFA (β = 0.092, OR (95%CI) = 1.096 (0.805–1.492), P = 0.560), NLR (β = 0.128, OR (95%CI) = 1.136 (0.843–1.533), P = 0.402), and ALB (β=-0.226, OR (95%CI) = 0.798 (0.636–1.001), P = 0.051) were not (Table 3).
Prediction of the prognostic outcomes of septic patients with positive blood cultures
The ROC curves of the PCT, combined with the severe scores for predicting outcomes in septic patients with positive blood cultures, for ICU-admission, MODS development, and 28-day mortality are shown in Fig. 4, and the AUCs are presented in Tables 4,5,and 6. The AUC values of the PCT, MEDS, APACHE II, and PIRO scores for ICU-admission were 0.620 (95% CI: 0.533–0.706, P = 0.009), 0.740 (95% CI: 0.663–0.817, P = 0.000), 0.780 (95% CI: 0.709–0.850, P = 0.000), 0.751(95% CI: 0.674–0.828, P = 0.000), respectively. The AUC values of a combination of PCT and severe scores were as follows: PCT + MEDS: 0.772 (95% CI: 0.698–0.846, P = 0.000); PCT + APACHE II: 0.821 (95% CI: 0.757–0.885, P = 0.000); PCT + PIRO: 0.795 (95% CI: 0.723–0.867, P = 0.000). The predictive ability of APACHE II (AUC:0.780) for ICU admission was slightly better than that of MEDS (AUC: 0.740) and PIRO (AUC:0.751). Consequently, PCT + APACHE II (AUC:0.821) is much better at predicting ICU admission than PCT + MEDS (AUC:0.772) and PCT + PIRO (AUC:0.795) (Fig. 4A and Table 4).
Table 4
Statistical data of ROC curve in predicting ICU-admission in patients with positive blood culture.
Variables
|
AUC (95%CI)
|
SE
|
P
|
Cut off
|
Sensitivity
|
Specificity
|
WBC
|
0.725(0.496–0.954)
|
0.117
|
0.075
|
12.75
|
0.800
|
0.750
|
PCT
|
0.620(0.533–0.708)
|
0.045
|
0.009
|
6
|
0.294
|
0.977
|
Lac
|
0.637(0.454–0.820)
|
0.093
|
0.125
|
2.25
|
0.500
|
0.913
|
ALB
|
0.759(0.574–0.945)
|
0.095
|
0.015
|
26.35
|
0.875
|
0.636
|
NLR
|
0.615 (0.449–0.782)
|
0.085
|
0.225
|
5.03
|
0.950
|
0.353
|
MEDS
|
0.740(0.663–0.817)
|
0.039
|
0.000
|
8.5
|
0.646
|
0.759
|
APACHE II
|
0.780(0.709–0.850)
|
0.036
|
0.000
|
15.5
|
0.468
|
0.975
|
PIRO
|
0.751(0.674–0.828)
|
0.039
|
0.000
|
11.5
|
0.684
|
0.759
|
PCT + MEDS
|
0.772(0.698–0.846)
|
0.038
|
0.000
|
0.39
|
0.924
|
0.557
|
PCT + APACHE II
|
0.821(0.757–0.885)
|
0.033
|
0.000
|
0.37
|
0.899
|
0.608
|
PCT + PIRO
|
0.795(0.723–0.867)
|
0.037
|
0.000
|
0.45
|
0.911
|
0.633
|
Abbreviations: WBC, white blood cell; PCT, procalcitonin; Lac, lactate; ALB, albumin; NLR, neutrophils and lymphocytes ratio; APACHE II, Acute Physiology and Chronic Health Evaluation II; MEDS, Mortality in emergency department sepsis; PIRO, Predisposition, Infection, Response, and Organ dysfunction |
Table 5
Statistical data of ROC curve in predicting MODS in patients with positive blood culture.
Variables
|
AUC (95%CI)
|
SE
|
P
|
Cut off
|
Sensitivity
|
Specificity
|
WBC
|
0.879(0.819–0.938)
|
0.030
|
0.000
|
12.75
|
0.911
|
0.878
|
PCT
|
0.664(0.584–0.745)
|
0.041
|
0.000
|
1.01
|
0.911
|
0.426
|
ALB
|
0.789(0.722–0.856)
|
0.034
|
0.000
|
28.10
|
0.851
|
0.620
|
NLR
|
0.727(0.653–0.801)
|
0.038
|
0.000
|
7.32
|
0.873
|
0.511
|
MEDS
|
0.701(0.622–0.779)
|
0.040
|
0.000
|
12.5
|
0.43
|
0.957
|
APACHE II
|
0.761(0.691–0.832)
|
0.036
|
0.000
|
12.5
|
0.658
|
0.723
|
PIRO
|
0.811(0.746–0.876)
|
0.033
|
0.000
|
12.5
|
0.544
|
0.957
|
PCT + MEDS
|
0.758(0.685–0.831)
|
0.037
|
0.000
|
0.53
|
0.582
|
0.862
|
PCT + APACHE II
|
0.794(0.727–0.862)
|
0.034
|
0.000
|
0.47
|
0.684
|
0.787
|
PCT + PIRO
|
0.837(0.776–0.898)
|
0.031
|
0.000
|
0.48
|
0.696
|
0.883
|
Abbreviations: WBC, white blood cell; PCT, procalcitonin; Lac, lactate; ALB, albumin; NLR, neutrophils and lymphocytes ratio; APACHE II, Acute Physiology and Chronic Health Evaluation II; MEDS, Mortality in emergency department sepsis; PIRO, Predisposition, Infection, Response, and Organ dysfunction |
Table 6
Statistical data of ROC curve in predicting 28-day-mortality in patients with positive blood culture.
Variables
|
AUC (95%CI)
|
SE
|
P
|
Cut off
|
Sensitivity
|
Specificity
|
WBC
|
0.789(0.673–0.905)
|
0.059
|
0.000
|
14.05
|
0.767
|
0.896
|
PCT
|
0.782(0.683–0.882)
|
0.051
|
0.000
|
1.05
|
0.933
|
0.521
|
Lac
|
0.699(0.561–0.838)
|
0.071
|
0.003
|
2.25
|
0.567
|
0.937
|
MEDS
|
0.745(0.627–0.863)
|
0.060
|
0.000
|
12.5
|
0.400
|
0.979
|
APACHE II
|
0.805(0.707–0.904)
|
0.050
|
0.000
|
15.5
|
0.633
|
0.875
|
PIRO
|
0.831(0.742–0.920)
|
0.045
|
0.000
|
11.5
|
0.867
|
0.708
|
PCT + MEDS
|
0.812(0.705–0.919)
|
0.055
|
0.000
|
0.50
|
0.667
|
0.937
|
PCT + APACHE II
|
0.866(0.789–0.943)
|
0.039
|
0.000
|
0.23
|
0.933
|
0.687
|
PCT + PIRO
|
0.885(0.810–0.961)
|
0.039
|
0.000
|
0.39
|
0.833
|
0.896
|
Abbreviations: WBC, white blood cell; PCT, procalcitonin; Lac, lactate; APACHE II, Acute Physiology and Chronic Health Evaluation II; MEDS, Mortality in emergency department sepsis; PIRO, Predisposition, Infection, Response, and Organ dysfunction |
Furthermore, for MODS incidence, the AUC values of the PCT, MEDS, APACHE II, and PIRO scores were 0.664 (95% CI: 0.584–0.745, P = 0.000), 0.701 (95% CI: 0.622–0.779, P = 0.000), 0.761 (95% CI: 0.691–0.832, P = 0.000), 0.811 (95% CI: 0.746–0.876, P = 0.000), respectively. The AUC values of the combination of PCT and severity scores were as follows: PCT + MEDS: 0.758 (95% CI: 0.685–0.831, P = 0.000); PCT + APACHE II: 0.794 (95% CI: 0.727–0.862, P = 0.000); PCT + PIRO: 0.837 (95% CI: 0.776–0.898, P = 0.000). The ability of PIRO (AUC:0.811) to predict the onset of MODS was much better than that of MEDS (AUC: 0.701) and APACHE II (AUC: 0.761). Therefore, PCT + PIRO (AUC: 0.837) would be better used in predicting the development of MODS than PCT + MEDS (AUC: 0.758) and PCT + APACHE II (AUC: 0.794) (Fig. 4B and Table 5).
In addition, after 28-days follow-up, the AUC values of the PCT, MEDS, APACHE II, and PIRO scores for 28-day mortality were 0.782 (95% CI: 0.683–0.882, P = 0.000), 0.745 (95% CI: 0.627–0.863, P = 0.000), 0.805 (95% CI: 0.707–0.904, P = 0.000), 0.831 (95% CI: 0.742–0.920, P = 0.000), respectively. The AUC values of the combination of PCT and severity scores were as follows: PCT + MEDS: 0.812 (95% CI: 0.705–0.919, P = 0.000); PCT + APACHE II: 0.866 (95% CI: 0.789–0.943, P = 0.000); PCT + PIRO: 0.885 (95% CI: 0.810–0.961, P = 0.000). The ability of the PIRO score (AUC: 0.831) to predict 28-day mortality was better than that of MEDS (AUC: 0.745) and APACHE II (AUC: 0.805). Thus, PCT + PIRO (AUC: 0.885) are slightly better at predicting 28-day mortality than PCT + MEDS (AUC: 0.812) and PCT + APACHE II (AUC: 0.866) (Fig. 4C and Table 6). Furthermore, the survival curve was estimated by Prism 6.0. The results show that 28-day mortality was around 18.4%, which is an improvement on the 27.9% measured 5 years ago in this hospital (Fig. 5).