Fig. 1

Odds ratio and 95% CI for subgroup analysis. The odds ratio (OR) and 95% CI are represented as black dots and horizontal bars respectively. The subgroup analysis was performed 10 times, while each time exclude one centre from the 10 centres in this study. We observed that the ORs from the 10 experiments were all greater than 1, indicating the 10 centres had consistent harmful outcome of using HFOV in terms of 28-day mortality. The 95% confidence interval of the ORs also support our finding that HFOV was harmful. The p values in 9 out of 10 experiments were less than 0.05. By comparing the ORs and 95% CI from the subgroups, we found there was significant association of HFOV treatment with 28-day mortality in PARDS