Overall, 14237 records were identified. After exclusion of the duplicates and irrelevant studies by screening of the study titles and abstracts, 530 studies were assessed for eligibility. All eligible studies (107) were used to evaluate the risk factors of surgical site infection, and 81 eligible cohorts were used to estimate the global incidence of surgical site infection in abdominal surgery. The study selection process is presented in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram (Fig. 1).
107 studies that were included in this review were from different regions of the world. There were 22 studies from the USA(31–52), 12 from Spain(53–64), eight from China (65–72), seven from Japan (73–79), seven from Brazil (80–85), four from Korea (86–89), four from Canada (90–93), three from the UK (94–96), three from Germany (97–99), five international studies (100–104), two from each of Sierra Leone (105, 106), Israel (16, 107), Tanzania (108, 109), Norway (110, 111), Poland (112, 113) and Ethiopia (114, 115), and one from each of Thailand (116), Croatia (117), Vietnam (118), Taiwan (119), Mexico (120), Egypt (121), South Africa (122), Nepal (123), Switzerland (124), Netherlands (125), Italy (126), Ireland (127), India (128), France (129), Ghana (130), Saudi Arabia (15), Myanmar (131), Kosovo (132), Belgium (133) and Turkey (134) (Fig. 2). The full list of the included studies and related quality assessment is presented in Supplementary Table 1, Additional file 2.
Importance of risk factors
The importance of risk factors evaluated based on criteria is presented in Table 2. Patient-related factors are shown in Table 3 and operation-related factors in Table 4. Among patient-related factors, the National Nosocomial Infection Surveillance (NNIS) System Risk Index was a strong risk factor in development of surgical site infection. Educational status of patients, functional status, malnutrition, and history of Neurological/Psychiatric Disease were categorized as probable strong risk factors with insufficient evidence. Albumin or pre-albumin level, blood glucose level, male gender, remote infection, abnormal BMI, and ASA class were moderate risk factors. Other variables were less important or unimportant (Table 3).
Table III. Consistency and relative importance of patient related variables in studies
|
|
Variables
|
-2
|
-1
|
0
|
1
|
2
|
Total (N=107)
|
Relative consistency
|
Relative importance
|
NNIS risk index
|
|
|
2
|
13
|
5
|
20
|
0.90
|
Strong risk factor
|
infection
|
|
|
5
|
2
|
6
|
13
|
0.62
|
Moderate risk factor needs more evidence
|
Albumin/prealbumin
|
|
|
6
|
3
|
7
|
16
|
0.63
|
Moderate risk factor
|
blood glucose level
|
|
|
8
|
4
|
7
|
19
|
0.58
|
Moderate risk factor
|
ASA class
|
|
|
26
|
17
|
15
|
58
|
0.55
|
Moderate risk factor
|
Abnormal BMI
|
|
|
33
|
7
|
25
|
65
|
0.49
|
Moderate risk factor
|
Sex(male)
|
4
|
|
27
|
10
|
17
|
58
|
0.40
|
Moderate risk factor
|
Hypertension
|
|
|
14
|
2
|
5
|
21
|
0.33
|
Less important
|
Cardiovascular disease
|
|
|
11
|
4
|
2
|
17
|
0.35
|
Less important
|
Malignancy
|
|
|
10
|
4
|
2
|
16
|
0.38
|
Less important
|
Previous surgery
|
|
|
11
|
4
|
4
|
19
|
0.42
|
Less important
|
Respiratory disease
|
|
|
9
|
2
|
4
|
15
|
0.40
|
Less important
|
Immunosuppression
|
|
|
14
|
5
|
5
|
24
|
0.42
|
Less important
|
Blood loss
|
|
|
15
|
8
|
2
|
24
|
0.38
|
Less important
|
Low hemoglobin
|
|
|
13
|
5
|
3
|
21
|
0.38
|
Less important
|
Education
|
|
|
1
|
1
|
2
|
4
|
0.75
|
Probable strong risk factor with insufficient evidence
|
Functional status
|
|
|
1
|
1
|
2
|
4
|
0.75
|
Probable strong risk factor with insufficient evidence
|
Malnutrition
|
|
|
1
|
|
3
|
4
|
0.75
|
Insufficient evidence
|
Neurological/Psychiatric Disease
|
|
1
|
1
|
2
|
4
|
0.75
|
Probable strong risk factor with insufficient evidence
|
Renal disease
|
|
|
10
|
1
|
2
|
13
|
0.23
|
Less important needs more evidence to confirmed
|
Radiotherapy
|
|
|
9
|
2
|
4
|
15
|
0.40
|
Less important needs
|
Hematocrit
|
|
|
3
|
|
|
3
|
0.00
|
Probable unimportant factor (Insufficient evidence)
|
Smoking
|
|
|
28
|
6
|
3
|
37
|
0.24
|
Unimportant
|
Diabetes
|
|
|
30
|
15
|
6
|
51
|
0.41
|
Unimportant
|
Age
|
3
|
2
|
56
|
11
|
9
|
81
|
0.19
|
Unimportant
|
Chemotherapy
|
|
|
8
|
2
|
2
|
12
|
0.33
|
Less important needs more evidence to confirmed
|
Comorbidities (yes VS no)
|
|
|
6
|
2
|
2
|
10
|
0.40
|
Less important needs more evidence to confirmed
|
Among operation-related factors, the length of operation and higher wound class were very strong risk factors. Surgeon’s low experience/grade was categorized as a strong risk factor, but it needs more evidence to be confirmed. Hair removal with razor and non-use of prophylaxis (oral) were categorized as probable strong risk factors with Insufficient evidence to have a conclusion. Bowel preparation, use or non-use of prophylaxis, pre-operative hospital stays, and stoma use were less important factors (Table 4).
Overall Incidence
Pooled incidence of surgical site infection in cohort studies was 10.6 (95% CI 9.02–12.55) per 100 patients. Heterogeneity was substantial (I2 = 99%, t2 = 0.68), and there was no significant difference between prospective and retrospective cohort studies (X2 = 0.01, df = 1, P = 0.92), among WHO regions (X2 = 7.88, df = 3, P = 0.05) and income group (X2 = 3.89, df = 3, P = 0.27) of countries (Table 6).
Table IV. Consistency and relative importance of operation-related variables in studies
|
Variables
|
-2
|
-1
|
0
|
1
|
2
|
Total (N=101)
|
Relative consistency
|
Relative importance
|
Length of operation
|
|
|
17
|
25
|
35
|
77
|
0.78
|
Very strong risk factor
|
Higher wound class
|
|
|
9
|
12
|
15
|
36
|
0.75
|
Very strong risk factor
|
Low surgeon experience/grade
|
|
|
2
|
4
|
3
|
9
|
0.78
|
Strong risk factor needs more evidence
|
Hair removal with razor
|
|
|
|
|
2
|
2
|
1.00
|
Probable risk factor with Insufficient evidence
|
Prophylaxis(oral)
|
2
|
|
|
|
|
2
|
-1.00
|
Probable protective factor with Insufficient evidence
|
Hair removal
|
|
|
5
|
|
|
5
|
0.00
|
Probable unimportant factor (Insufficient evidence)
|
Opens VS Minimally invasive
|
|
|
9
|
4
|
20
|
33
|
0.73
|
Moderate risk factor
|
Emergency/elective
|
|
|
21
|
10
|
11
|
42
|
0.50
|
Moderate risk factor
|
Blood transfusion
|
|
|
8
|
8
|
7
|
23
|
0.65
|
Moderate risk factor
|
Diagnosis
|
|
|
11
|
8
|
6
|
25
|
0.56
|
Moderate risk factor
|
Drains
|
|
|
9
|
7
|
5
|
21
|
0.57
|
Moderate risk factor
|
Bowel preparation
|
|
|
9
|
3
|
3
|
15
|
0.40
|
less important
|
Prophylaxis(pre-op)
|
|
2
|
14
|
3
|
5
|
24
|
0.25
|
Less important
|
Pre-operative hospital stays
|
|
|
14
|
3
|
6
|
23
|
0.39
|
Less important
|
Stoma
|
|
|
10
|
1
|
6
|
17
|
0.41
|
Less important
|
Type of surgery procedure
|
|
|
6
|
6
|
4
|
16
|
0.63
|
Moderate risk factor
|
Prophylaxis(type)
|
|
|
5
|
5
|
3
|
13
|
0.62
|
Moderate risk factor needs more evidence
|
Prophylaxis(time)
|
|
|
6
|
5
|
2
|
13
|
0.54
|
Moderate risk factor needs more evidence
|
Additional procedure
|
|
|
5
|
2
|
3
|
10
|
0.50
|
Moderate risk factor needs more evidence
|
Anesthesia
|
|
|
6
|
3
|
2
|
11
|
0.45
|
Moderate risk factor needs more evidence
|
Prophylaxis(dose)
|
1
|
|
3
|
|
3
|
7
|
0.29
|
Less important needs more evidence
|
Table VI. Summary statistics of meta- analysis of the incidence of surgical site infections after abdominal operations
|
factor
|
subgroup
|
Study(n)
|
Incidence per 100 surgical procedures (95% CI)
|
I2%
|
t2
|
Test for subgroup differences
|
Global
|
|
81
|
10.66(9.02–12.55)
|
99
|
0.69
|
|
Design
|
Retrospective
|
25
|
10.52(7.78–14.07)
|
100
|
0.69
|
X2 = 0.01, df = 1(P = 0.92)
|
Prospective
|
56
|
10.72(8.75–13.06)
|
99
|
0.70
|
|
Who region
|
European Region
|
27
|
10.71(8.12-14.00)
|
100
|
0.61
|
X2 = 7.88, df = 3(P = 0.05)
|
African Region
|
7
|
16.01(12.19–20.75)
|
89
|
0.13
|
|
Region of the Americas
|
26
|
12.13(8.98–16.19)
|
99
|
0.74
|
|
Western Pacific Region
|
15
|
8.77(6.04–12.56)
|
99
|
0.61
|
|
Income level
|
High income
|
54
|
9.94(8.11–12.13)
|
100
|
067
|
X2 = 3.89, df = 3(P = 0.27)
|
Upper-middle-income
|
16
|
11.91(7.34–18.73)
|
99
|
1.16
|
|
Lower-middle-income
|
5
|
10.47(7.35–14.70)
|
97
|
0.18
|
|
low-income
|
3
|
16.82(10.21–26.45)
|
92
|
0.23
|
|
Surgical procedure
|
Caesarean and gynaecological
|
15
|
8.71(6.18–12.15)
|
99
|
0.51
|
X2 = 79.81, df = 7(P < 0.01)
|
Bowel surgery (small bowel, colon and rectum)
|
22
|
13.65(10.62–17.37)
|
99
|
0.44
|
|
Mixed abdominal
|
17
|
12.18(9.41–15.61)
|
99
|
0.34
|
|
Appendectomy
|
8
|
7.57(3.90-14.17)
|
99
|
1.01
|
|
Gastric surgery
|
4
|
4.66(3.19–6.77)
|
74
|
0.11
|
|
Pancreatic surgery
|
3
|
16.45(5.41–40.39)
|
99
|
1.17
|
|
Liver transplantation
|
5
|
27.44(20.76–35.31)
|
91
|
0.14
|
|
Cholecystectomy
|
3
|
2.50(1.18–5.25)
|
94
|
0.42
|
|
Operation time
|
Operation time > t
|
23
|
14.08(10.0-18.64)
|
96
|
0.63
|
-
|
Operation time < t
|
23
|
7.24(5.04–10.30)
|
98
|
0.86
|
-
|
Wound class
|
Clean or clean-contaminated
|
24
|
7.83(6.00-10.16)
|
98
|
0.48
|
-
|
Contaminated or dirty
|
24
|
20.69(15.63–26.85)
|
96
|
0.62
|
-
|
ASA class
|
ASA < 3
|
28
|
8.70(6.75–11.14)
|
98
|
0.52
|
-
|
ASA ≥ 3
|
28
|
14.84(11.88–18.38)
|
93
|
0.41
|
-
|
Sources of heterogeneity
There were differences in the pooled incidence of surgical site infections based on the type of surgical procedure, from a lower range of 2.5 (95% CI 1.18–5.25) for cholecystectomy to a higher range of 27.4 (95% CI 20.76–35.31) for liver transplantation (Additional file 3). The meta-regression results showed that the type of surgery procedures accounted for 31.17% of the heterogeneity Additional file 4. Pooled incidence of surgical site infection was up to 14.1 (95% CI 10.0–18.64) during longer operation (surgical time ≥ T) compared to 7.2 (95% CI 5.04–10.30) at normal operation time (surgical time < T); it was 20.7 which is higher in dirty/contaminated wound class (95% CI 15.63–26.85) compared to 7.8 in clean/clean-contaminated wound class (95% CI 6.00-10.16). The pooled incidences were 14.8 (95% CI 11.88–18.38) for ASA Class ≥ 3 and 8.7 (95% CI 6.75–11.14) for ASA Class < 3 (Table 6).
All the data that was used in the analysis is presented in Additional File 5.