Fifty studies (1,501 prognostic factors and outcome associations) were included in the review. The mean quality score was 4.6 (standard deviation=0.9, range 2 to 6). One third of studies provided data enabling the calculation of odds ratios, and meta-analysis was possible for 8% of the factors.
A large number of prognostic variables were reported as predictors of poor recovery, including psychosocial variables, pain, physical impairment, clinician factors and therapeutic response. The most common factors with significant odds ratios were in the pain and psychosocial domains, but these were also the most frequently investigated and provided the required outcome data. There was a statistically significant inverse relationship between study quality and size of significant odds ratios; significantly higher odds ratios were derived from lower-quality studies (p=0.03).