Figure 4.

A meta-analysis of readmissions. (A) Exacerbation-related readmissions (subgroup analysis with observation period). A sensitivity analysis using the fixed-effect model resulted in a similar effect size [RR = 0.81, 95% CI (0.74–0.88)] compared to random-effects modeling. One study (Soriano et al.) (9) could not be included in the meta-analysis for it only collected the number of participants who have at least one exacerbation (ER visit or hospitalization) in the past 12 months, resulting in the failure of exacting the required data. Subgroup analysis. In the subgroup in which follow-up duration was 6 or 9 months, it did not reduce readmissions [P = 0.28, RR = 0.62, 95% CI (0.26, 1.46) and P = 0.10, RR = 0.45, 95% CI (0.17, 1.17), respectively]. Statistically, heterogeneity was found in both subgroups (I² = 83% and 58%). However, the subgroup of 12 months showed a greater reduction on readmissions [P = 0.03, RR = 0.87, 95% CI (0.76, 0.99), I² = 50%]. Additionally, a sensitivity analysis was performed for the primary outcome to test an overall pooled effect. The results were no different between fixed and random statistical effects (RR = 0.81; P = 0.000). (B) The rate of exacerbation-related readmissions. (C) All-cause readmissions. Low heterogeneity was found (P = 0.29, I² = 27%).