Table 4 

Metaanalysis 

Outcome 
Followup (months) 
Treatment effect 
Heterogeneity 
Publication bias 
Number of studies 



Effect measure 
Pooled estimate (95% CI) 
P value for treatment effect 
P value for Cochran's Q test 
I squared statistic^{a} 
Number of studies trimmed^{b} 
Adjusted pooled estimate^{c} 



Functional status 
1 
Hedges' g 
0.11 (0.22 to 0.45) 
0.50 
0.009 
79% 
0 
0.11 (0.22 to 0.45) 

3 
0.07 (0.11 to 0.26) 
0.44 
0.006 
73% 
0 
0.07 (0.11 to 0.26) 

6 
0.09 (0.14 to 0.33) 
0.41 
0.034 
63% 
1 
0.15 (0.07 to 0.37) 

12 
0.01 (0.15 to 0.16) 
0.94 
0.15 
44% 
0 
0.01 (0.15 to 0.16) 

Length of stay 
Weighted mean difference 
0.35 (1.24 to 0.55) 
0.45 
0.75 
0% 
1 
0.34 (1.24 to 0.56) 

Mortality 
1 
Relative risk 
0.66 (0.40 to 1.09) 
0.10 
0.46 
0% 
NA 
NA 

3/4 
0.72 (0.44 to 1.17) 
0.19 
0.12 
47% 
0 
0.72 (0.44 to 1.17) 

6 
0.66 (0.52 to 0.85) 
0.001 
0.37 
0% 
2 
0.73 (0.52 to 1.02) 

8 
0.51 (0.31 to 0.85) 
0.009 
0.39 
0% 
NA 
NA 

12 
0.98 (0.86 to 1.11) 
0.75 
0.52 
0% 
2 
1.01 (0.87 to 1.16) 

Readmission 
1 
Relative risk 
0.65 (0.25 to 1.67) 
0.37 
0.59 
0% 
NA 
NA 

3 
0.93 (0.82 to 1.04) 
0.18 
0.74 
0% 
2 
0.94 (0.84 to 1.05) 

6 
0.88 (0.40 to 1.96) 
0.76 
0.001 
86% 
0 
0.88 (0.40 to 1.96) 

12 
0.99 (0.83 to 1.19) 
0.94 
0.34 
12% 
2 
1.09 (0.87 to 1.36) 



CI: confidence interval; NA: not applicable. ^{a}I squared statistic: values less than 25% indicate low, 25% to 50% indicate moderate, and greater than 50% indicate high heterogeneity. ^{b}There must be a least three studies published to run publication bias procedure. ^{c}Adjusted pooled estimate according to Duval and Tweedie's trimandfill analysis. 

Deschodt et al. BMC Medicine 2013 11:48 doi:10.1186/174170151148 