A total of 364 children were admitted to the PICU during the 3-month study period, most of whom were from the Pediatric General Ward and Pediatric Emergency Department (Figure 1). The main reasons for PICU admission were postoperative care and disease progression (deterioration of physical status). Thirty-two of these patients were excluded because they were neonates, and 131 others were excluded based on the predefined exclusion criteria.
We included 201 children in the final statistical analysis, 93 children with infections and 103 without infections (Table 1). Overall, the median age was 49 months, the median length of hospital stay was 15 days, and the median length of PICU stay was 2 days. The infection group was younger, had longer hospital and ICU stays, and was more likely to receive ventilator therapy (all p < 0.05), but the two groups had no significant difference in prognosis.
Table 1. Baseline characteristics of patients in the two groups at PICU admission.
|
Variable
|
Infection
Group
|
Non-infection Group
|
All
|
p value
|
n=93
|
n=108
|
N=201
|
Age, months
|
20(6, 20)
|
67.5(36.25, 96.75)
|
|
<0.001
|
Sex
|
|
|
|
|
Male
|
55 (59.1%)
|
73 (67.6%)
|
128 (63.7%)
|
0.214
|
Female
|
38(40.9 %)
|
35 (32.4%)
|
73 (36.3%)
|
Reason for admission
|
|
|
|
|
Disease progression
|
51 (54.8%)
|
32 (29.6%)
|
83 (41.3%)
|
<0.001
|
Postoperative care
|
42 (45.2%)
|
76 (71.4%)
|
118 (58.7%)
|
Source
|
|
|
|
|
Pediatric ward
|
16 (17.2 %)
|
1 (0.9%)
|
17 (8.4 %)
|
<0.001
|
Pediatric surgical ward
|
44 (47.3%)
|
87 (80.6 %)
|
131 (65.2 %)
|
Pediatric emergency
|
31 (33.3%)
|
17 (15.7%)
|
48 (23.9 %)
|
Others
|
2 (2.2 %)
|
3 (2.8 %)
|
5 (2.5 %)
|
Treatment at PICU
|
|
|
|
|
Mechanical ventilation
|
31 (33.3 %)
|
8 (7.4 %)
|
39 (19.4 %)
|
<0.001
|
Hemodialysis/-filtration
|
6 (6.5 %)
|
5 (4.6 %)
|
11 (5.5 %)
|
0.571
|
Outcome
|
|
|
|
|
Survival
|
90 (96.8%)
|
107 (99.1 %)
|
197 (98.1 %)
|
0.244
|
Death
|
3 (3.2 %)
|
1 (0.9 %)
|
4 (1.9 %)
|
ICU length of stay, days
|
4(1,9.5)
|
2(1, 3)
|
2 (1, 4)
|
<0.001
|
Hospital length of stay, days
|
20 (13, 38)
|
14 (9, 19)
|
15(11, 26)
|
<0.001
|
nCD64 index
|
0.18 (0.12, 0.27)
|
0.09 (0.06, 0.12)
|
0.11 (0.07, 0.19)
|
<0.001
|
PCT, ng/mL
|
0.19 (0.09,0.83)
|
0.08 (0.04, 0.22)
|
0.13 (0.05, 0.31)
|
<0.001
|
CRP, mg/L
|
5 (0.5, 30)
|
0.5 (0.5, 8)
|
2 (0.5, 12.5)
|
<0.001
|
WBC (×109/L)
|
12.16 (8.89, 15.4)
|
12.29 (8.25, 15.97)
|
12.16 (8.82, 15.88)
|
0.0825
|
Data are indicated as n (%) or median (IQR).
Among the 93 children in the infection group, 46 had clinical confirmation but no microbiological confirmation; pathogenic microorganisms were isolated from the body fluids of the other 47 infected children (Figure 2). Among patients with microbiological confirmation, 28 were infected by a single bacterial species (22 Gram-negative, 6 Gram-positive), 4 had viral infections, 14 had mixed infections (bacterial and fungal), and 1 had a Mycoplasma pneumoniae infection. Among all infected children, there were 34 lower respiratory tract infections, 30 digestive tract infections, 20 central nervous system infections, 4 bloodstream infections, and 5 skin soft tissue infections (Figure 3).
Comparisons of the different biomarkers in the two groups (Table 1) indicated the infected group had a significantly greater median nCD64 index (0.18 vs. 0.09, p < 0.001), median CRP level (5 vs. 0.5 mg/L, p < 0.001), and median PCT level (0.19 vs. 0.08 ng/mL, p < 0.001). However, the two groups had similar levels of WBCs. We then performed ROC analysis to compare the value of three biomarkers for the diagnosis of infection (Table 2, Figure 4). For the nCD64 index, the optimal cutoff was 0.14 and the area under the curve (AUC) was 0.811. Pair-wise analysis using Z-test indicated the AUC of CD64 was significantly greater than the AUC values for CRP (0.661, p < 0.05) and PCT (0.677, p < 0.05). The nCD64 index had a sensitivity of 68.8%, specificity of 90.7%, PPV of 0.86, NPV of 0.77, PLR of 0.34, and NLR of 0.34. Thus, the nCD64 index had greater diagnostic value than CRP and PCT.
Postoperative fever is very common in clinical practice, so we analyzed the diagnostic ability of these same biomarkers in the 131 children (87 in the non-infection group, 44 in the infection group) who had postoperative fevers (Table 3). Compared with the non-infection group, the infection group was younger; more likely to receive general surgery, mechanical ventilation, and type II surgical incision; had longer hospital and ICU stays; and had higher levels of the nCD64 index, CRP, and PCT (all p < 0.05). All of these children improved and were discharged. The children in the infection group had infections of the digestive system (n = 19), lower respiratory tract (n = 11), central nervous system (n = 9), skin and soft tissue (n = 4), and blood stream (n = 1), and Gram-positive bacteria were the main pathogens (n = 16; Figure 5). We also recorded the etiology of the infections in the children with post-operative fevers (Figure 6).
We then performed ROC analysis to compare the value of these three biomarkers for the diagnosis of infection in children who had post-operative fevers (Table 4). The results indicated the nCD64 index had a sensitivity of 56%, a specificity of 90%, PPV of 0.73, NPV of 0.81, PLR of 5.6, and NLR of 0.49. Pair-wise analysis using the Z test indicated the AUC for the nCD64 index (0.722) was significantly greater than the AUC values for CRP (0.641, p < 0.05) and PCT (0.649, p < 0.05).
Table 2. Performance of CRP, PCT, and nCD64 index for diagnosis of infection (n = 201).
Biomarker
|
Cut-off
|
Sensitivity
|
Specificity
|
PPV
|
NPV
|
PLR
|
NLR
|
Accuracy
|
AUC (95%CI)
|
CRP
|
30 mg/L
|
76.7%
|
59.7%
|
0.77
|
0.59
|
1.9
|
0.39
|
61.6%
|
0.661 (0.585-0.736)
|
PCT
|
0.5 ng/mL
|
30%
|
90.7%
|
0.72
|
0.6
|
3.32
|
0.7
|
62.1%
|
0.677 (0.603-0.752)
|
nCD64 index
|
0.14
|
68.8%
|
90.7%
|
0.86
|
0.77
|
7.4
|
0.34
|
80.5%
|
0.811 (0.748-0.873)
|
Table 3. Baseline characteristics patients who had post-surgical fever.
Data are indicated as n (%) or median (IQR)
Variable
|
Infection Group
|
Non-infection Group
|
All
|
p
|
n=44
|
n=87
|
N=131
|
Age, months
|
22.5 (6.5, 63)
|
66 (34, 97)
|
16 (13, 24)
|
<0.001
|
Sex
|
|
|
|
|
Male
|
24 (54.5%)
|
56 (64.4%)
|
80 (61.1%)
|
0.343
|
Female
|
20 (45.5%)
|
31 (35.6%)
|
51 (38.9%)
|
Source
|
|
|
|
|
General surgery
|
28 (63.7%)
|
31 (35.6%)
|
59 (45.0%)
|
0.001
|
Neurosurgery
|
15 (34.1%)
|
49 (56.3%)
|
64 (48.9%)
|
Orthopaedic surgery
|
1 (2.2%)
|
3 (3.5%)
|
4 (3.1%)
|
Urological surgery
|
0 (0%)
|
4 (4.6%)
|
4 (3.1%)
|
Treatment at PICU admission
|
|
|
|
|
Mechanical ventilation
|
Yes
|
11 (25.0%)
|
3 (3.4%)
|
14 (10.7%)
|
<0.001
|
|
No
|
33 (75.0%)
|
84 (96.6%)
|
117 (89.3%)
|
Hemodialysis/-filtration
|
Yes
|
0 (0%)
|
1 (1.1%)
|
1 (0.8%)
|
0.664
|
|
No
|
44 (100%)
|
86 (98.9%)
|
130 (98.2%)
|
Type of surgical incision*
|
I
|
17 (38.6%)
|
65 (74.7%)
|
82 (62.6%)
|
<0.001
|
|
II
|
23 (52.4%)
|
13 (15.0%)
|
36 (27.5%)
|
|
III
|
0 (0%)
|
1 (1.1%)
|
1 (1.1%)
|
ICU length of stay, days
|
3 (1, 5)
|
1(1,2)
|
2 (1,3)
|
0.001
|
Hospital length of stay, days
|
21 (14.5, 38.5)
|
14 (11,20)
|
16 (13, 24)
|
<0.001
|
CD64 index
|
0.15 (0.08, 0.23)
|
0.09 (0.06, 0.12)
|
0.11 (0.06, 0.15)
|
<0.001
|
PCT, ng/ml
|
0.17(0.06, 0.55)
|
0.08 (0.04, 0.2)
|
0.09 (0.04, 0.23)
|
0.005
|
CRP, mg/l
|
2(0.5, 33.25)
|
0.5 (0.5, 4.0)
|
0.5 (0.5, 10)
|
0.004
|
WBC (×109/L)
|
13.58 (9.59, 17.15)
|
12.87 (9.7, 17.46)
|
12.9 (9.7, 17.3)
|
0.845
|
*Some children did not receive surgical treatment, so the sum of surgeries is not equal to the total number.
Table 4. Performance of CRP, PCT, and nCD64 index for diagnosis of infection in patients who had post-surgical fever (n = 134).
Biomarker
|
Cut-off
|
Sensitivity
|
Specificity
|
PPV
|
NPV
|
PLR
|
NLR
|
Accuracy
|
AUC (95%CI)
|
CRP
|
30 mg/L
|
27%
|
93%
|
0.67
|
0.72
|
3.8
|
0.78
|
71%
|
0.641 (0.537-0.744)
|
PCT
|
0.5 ng/mL
|
25%
|
98%
|
0.85
|
0.72
|
12.5
|
0.77
|
73%
|
0.649 (0.546-0.752)
|
nCD64 index
|
0.14
|
56%
|
90%
|
0.73
|
0.81
|
5.6
|
0.49
|
79%
|
0.722 (0.621-0.823)
|