Different Meta-Analysis Methods, Different Answers: The Case of Exercise for Older Acute Medical Patients

The aim of this study was to illustrate the potential for confounding when interpreting group mean data from systematic reviews for a heterogeneous participant population. A case report comparing the results obtained from a Coch- rane review and meta-analysis compared to an individual patient data (IPD) meta-analysis was conducted. Participants were 396 older patients admitted to a general medical ward at two acute public hospitals. For the intervention group, an exercise program was provided for 20-30 minutes twice per day in addition to usual care. The control group received usual hospital care. The primary outcome measure was changed in Barthel Index scores between hospital admission and discharge. Meta-analysis of group mean data provided evidence that additional exercise for older medical patients does not significantly improve Barthel Index scores at hospital discharge (fixed effects model, 0.17 (-0.06 to 0.40) I 2 = 0%). Re- sults of IPD meta-analysis indicated that additional exercise significantly improved discharge Barthel Index scores for pa- tients who required assistance to ambulate at hospital admission (coefficient for group 5.4 (1.38 to 9.40), p=0.01) but not for those who were non ambulant or independently ambulant at hospital admission. This case report provides an example where performing meta-analysis using group mean data for heterogeneous populations can result in effective interventions being discarded.