Through searching the Cochrane Library, EMBASE, PubMed and Web of Science, 743 potentially relevant articles were retrieved. After the removal of duplicates, the titles and abstracts of the remaining 327 studies were screened for eligibility. The full texts of 71 potentially eligible studies were then strictly assessed. Finally, 24 studies[23-46]were included in the present meta-analysis (Fig. 1)
Characteristics of the included studies
All 24 studies included in the meta-analysis were published in English. The time of publication was from 1998 to 2020. Of these, 10 were retrospective studies and 14 were prospective studies. The number of patients with VAP was 2699, and the number of patients without VAP was 9511. According to the NOS, two articles received eight scores, seven articles received seven scores, thirteen articles received six scores, and two articles received five scores. The main characteristics of the included studies are shown in Table 1a,b.
Incidence rates of VAP
The pooled incidence of VAP was 20/1000 ventilator-days (95% CI 17–22%), as shown in Fig. 2a. Stratified analysis based on variation in the study design was applied because of its potential effect on the incidence of VAP (Fig. 2b). The forest plot illustrated that the incidence of VAP in retrospective studies (18%) was relatively lower than their counterparts in prospective studies (21%); We performed a subgroup analysis by region to explore its impact on VAP incidence (Fig. 2c), and the results showed that the incidence of VAP in Europe (25%) was relatively higher than North America (16%), Asia (20%) and other region(12%); We also performed a subgroup analysis by definition of VAP (Fig. 2d), and the forest plot illustrated that the incidence of VAP were 16% in CDC criteria and 25% in non-CDC criteria respectively; Additionally, considering that the 13 studies had different types of study design, we conducted a subgroup analysis based on the sample size(<100, 100-200, 200-300, 300-1000, >1000), and we found that as the sample size increased, the incidence of VAP decreased (31%, 27%, 22%, 17%, 15%)(Fig. 2e); In addition, we performed a subgroup analysis according to different settings (medical and surgical ICUs, mixed ICUs, general ICUs)(Fig. 2f). The forest plot indicated that the incidence of VAP in medical and surgical ICUs (26%) was relatively higher than their counterparts in mixed ICUs and general ICUs (23% and 15%).
Risk factors for VAP
Considering the various risk factors included in these studies, we chose 19 risk factors for the meta-analysis. These factors included tracheotomy, H2 blocker use, nasogastric tube placement, impaired consciousness, prior antibiotic treatment, steroids use, enteral feeding, parental nutrition, blood transfusion, CVC placement, previous surgery, reintubation ,diabetes, COPD, trauma, acute respiratory distress syndrome (ARDS), sedation, sucralfate, respiratory failure (Table 2).
There were 10 potential risk factors (reported in at least two studies) summarized in this meta-analysis. We found a significant relationship between VAP and the following risk factors: trauma (OR 1.51; 95% CI 1.21-1.88), steroids use (OR 1.51; 95% CI 1.25-1.82), enteral feeding (OR 3.40; 95% CI 2.56-4.52), nasogastric tube placement (OR 2.96; 95% CI 1.75-5.00), tracheostomy (OR 3.65; 95% CI 1.93-6.90), reintubation (OR 6.08; 95% CI 4.62-8.00), CVC placement (OR 3.83; 95% CI 2.55-5.77), blood transfusion (OR 1.75; 95% CI 1.16-2.64), H2 blockers use (OR 2.28; 95% CI 1.27-4.07) and COPD (OR 1.24; 95% CI 1.07-1.44) were risk factors of VAP as noted in Table 3 (P < 0.05). No significant differences were detected for impaired consciousness, prior antibiotic treatment, previous surgery, ARDS, sedation, sucralfate and respiratory failure (P > 0.05). Therefore, these factors were determined not to be risk factors for VAP. The forest plots of the risk factors were described in Fig. 3a,b,c,d,e,f,g,h,i,j. Among these risk factors, we found significant heterogeneity in tracheotomy ( I2 = 84% ) and blood transfusion( I2 = 54 %), and we conducted a subgroup analysis based on sample size(<300, 300-600, >600; ≤ 100, >100). As a result, I2 in each subgroup drops below 50%, which may indicate that the sample size was the cause of heterogeneity ( Fig. 3k,l.)
After data integration of 15 studies, pooled results showed that VAP was associated with mortality (OR 1.38; 95% CI 1.04–1.83), as shown in Fig. 4a. However, considering that I2 =77%, a subgroup analysis was performed based on the sample size (<100, 100-200, 200-500, 500-1000, 1000-2000,> 2000), we found that heterogeneity was reduced (57%, 59%, 0%, 0%,18%), but in subgroups with sample sizes of 100-500 and> 2000, there was no correlation between VAP and mortality(Fig. 4b); In addition, our pooled results also showed that duration of ICU stay and MV were increased significantly in patients infected with VAP (OR 11.09; 95% CI 7.50-14.68; OR 12.96; 95% CI 10.42-15.51), as shown in Fig. 4c,d. Considering the existence of obvious heterogeneity (I2=85%, I2=79%), we performed subgroup analyses based on CDC criteria (CDC criteria, non-CDC criteria) and sample size (<1000, >1000) respectively. Interestingly, the heterogeneity was reduced separately in each subgroup, as shown in Fig. 4e, f. However, we did not find this association in the two variables (age and APACHE II scores).