Table 1 presents that totally, 61,773 paediatric hospitalized patients were included in the 15 included studies, which covered 13 regions in 10 provinces of China. Specifically, 2,438 patients were identified as having HAIs, while 59,335 patients did not have HAIs. The majority (11/15) of the included studies were retrospective cross-sectional studies; 2 studies were retrospective case-control studies; one was retrospective cohort study; and one was retrospective and prospective study. Besides, most of the studies (9/15) were conducted in tertiary hospitals, while the rest were undertaken in secondary hospitals.
Frequency of the identified risk factors from the included studies
Table 2 shows a total of 20 risk factors on HAIs were identified from the 15 included studies. The most frequent risk factors were length of hospitalization (14/15), age (10/15), gender (5/15) and number of kinds of antibiotics use (5/15). The risk factors were not incorporated in the meta-analyses since the frequency was less than 2. In addition, although the classification of the age and length of hospitalization was different among the included studies, we adopted the most frequent classification used in the included studies for the meta-analyses. With regard to the length of time of antibiotics use, due to the inconsistent classification used in both studies (17, 18), it was also excluded in the meta-analysis.
Quality assessment of all the included studies
With regard to the 13 cross-sectional studies, the median points were 10 (medium points). It indicated the overall quality of the cross-sectional studies was fair, which needed to be improved. All of the studies clearly defined the inclusion for the pediatric hospitalized patients and set the outcomes measured in a valid and reliable way. However, some of the cross-sectional studies needed to be improved in study design, including describing the study setting in more detail, providing the specific definition and inclusion and exclusion criteria for the HAIs pediatric hospitalized patients etc. Only two case-control and one cohort studies were included in this systematic review. The overall quality of the case-control studies was good since one study got 17 points, while the other study got 18 points. Both of them also needed to provide the clear information about the outcome measurements and how to measure the exposure. The quality of the cohort study was poor (10 points). It did not clearly depict the follow-up time and how to deal with the incomplete follow-up. Also, it did not provide the clear information about the outcome measurements and how to measure the exposure. Quality assessments of the included studies are exhibited in Supplementary material (Appendix 1: Table 1, 2, and 3).
Meta-analyses of the included risk factors
Table 3 presents the pooled ORs of the risk factors estimated in the meta-analyses between the paediatric hospitalized patients with HAIs and the paediatric hospitalized patients without HAIs. 11 risk factors were analysed in the meta-analysis. Specific information is as follows. The forest plots of the included risk factors are shown in Supplementary material (Appendix 2: Figure 1-1 to 1-11).
Age (≤ 1 year vs. >1 year): 6 studies (17, 19-23) reported the age (≤ 1 year) of 429 with and 6,001 patients without HAIs. The random-effects model showed that the prevalence of HAIs was significantly higher (P=0.003) in patients who were younger than 1 year old compared with those who were over 1 year old (OR: 2.25, 95% CI: 1.32-3.85). There was a high heterogeneity among the studies (2=0.35 (P<0.01), I2=84%).
Gender (Male vs. Female): 6 studies (17-19, 22, 24, 25) reported the gender of 790 with and 15,289 patients without HAIs. The fixed-effects model showed that the prevalence of HAIs was almost equal (P=0.26) between male and female patients (OR: 1.07, 95% CI: 0.95-1.21). The heterogeneity did not exist among the studies (P=0.63, I2=0%).
Hormone:2 studies (17, 18) reported the hormone of 25 with and 31 patients without HAIs. The fixed-effects model showed that the prevalence of HAIs was significantly higher (P=0.0007) in patients using hormone than those not using hormone (OR: 3.66, 95% CI: 1.73-7.74). The heterogeneity did not exist among the studies (P=0.61, I2=0%).
Invasive procedures: 4 studies (17, 21, 26, 27) reported the invasive procedures of 191 with and 1,183 patients without HAIs. The fixed-effects model showed that the prevalence of HAIs was significantly higher (P<0.01) in patients experiencing invasive procedure than those not experiencing invasive procedure (OR: 5.62, 95% CI: 4.27-7.40). The heterogeneity did not exist among the studies (P=0.77, I2=0%).
Length of hospitalization (>7 days vs. ≤7 days): 12 studies (17, 19-25, 28-31) reported the length of hospitalization (> 7 days) of 1,617 with and 12,534 patients without HAIs. The random-effects model showed that the prevalence of HAIs was significantly higher (P<0.01) in patients whose hospitalization were more than 7 days than those whose hospitalization were less than 7 days (OR: 7.79, 95% CI: 6.38-9.50). There was a middle level of heterogeneity among the studies (2=0.06 (P=0.001), I2=64%).
Malnutrition:2 studies (17, 18) reported the malnutrition of 29 with and 54 patients without HAIs. Fixed-effects model showed that the prevalence of HAIs was significantly higher (P=0.0004) in patients with malnutrition than those without malnutrition (OR: 3.72, 95% CI: 1.80-7.70). The heterogeneity did not exist among the studies (P=0.75, I2=0%).
Number of kinds of antibiotics (>3 vs. ≤3): 6 studies (17, 18, 20, 22, 28, 31) reported the number of kinds of antibiotics (> 3) of 177 with and 1,206 patients without HAIs. The fixed-effects model showed that the prevalence of HAIs was significantly higher (P<0.01) in patients who used more than 3 kinds of antibiotics than those who used at most 3 kinds of antibiotics (OR: 3.25, 95% CI: 2.66-3.96). The heterogeneity did not exist among the studies (P=0.15, I2=39%).
Number of underlying disease (>3 vs. ≤3): 4 studies (17, 21, 31) reported the number of underlying disease (> 3) of 68 with and 79 patients without HAIs. The random-effects model showed that the prevalence of HAIs was significantly higher (P=0.0002) in patients who had more than 3 kinds of underlying disease than those who had at most 3 kinds of underlying disease (OR: 4.24, 95% CI: 1.84-9.78). The heterogeneity existed among the studies (2=0.30 (P=0.11), I2=55%).
Room size (Large vs. Small): 2 studies (19, 23) reported the room size (Big) of 77 with and 1,061 patients without HAIs. Fixed-effects model showed that the prevalence of HAIs was significantly higher (P=0.005) in patients who stayed in a large room than those who stayed in a small room (OR: 2.22, 95% CI: 1.28-3.85). The heterogeneity did not exist among the studies (P=0.72, I2=0%).
Seasons (Autumn + Winter vs. Spring + Summer): 4 studies (19, 20, 23, 32) reported the seasons (Autumn + Winter) of 235 with and 6,145 patients without HAIs. Random-effects model showed that the prevalence of HAIs was significantly higher (P=0.03) in patients who were admitted to the hospitals in autumn or winter than those who were admitted to the hospitals in spring or summer (OR: 1.56, 95% CI: 1.04-2.35). The heterogeneity existed among the studies (2=0.11 (P=0.02), I2=69%).
Use of antibiotics: 6 studies (20, 22, 24, 26, 28, 31) reported the use of antibiotics of 788 with and 20,957 patients without HAIs. Random-effects model showed that the prevalence of HAIs was higher (P=0.09) in patients using antibiotics than those not using antibiotics (OR: 3.18, 95% CI: 0.83-12.15). There was a high heterogeneity among the studies (2=1.87 (P<0.01), I2=85%).