Fig. 3

Meta-regression showing the effect of varying proportions of HIV-infected participants on inpatient mortality for sepsis and severe sepsis. 95% confidence intervals generated from 1000 bootstrap replicates. Likelihood ratio testing of nested models shows that including proportion of HIV-infected participants significantly improved model fit for sepsis (p = 0.008), but not severe sepsis (p = 0.83). Size of circle is proportional to the number of participants with sepsis or severe sepsis