In patients with a particular disease or health condition, prognostic factors are characteristics (such as age, biomarkers) that are associated with different risks of a future clinical outcome. Research is needed to identify prognostic factors, but current evidence suggests that primary research is of low quality and poorly/selectively reported, which limits subsequent systematic reviews and meta-analysis. This thesis aims to improve prognostic factor research, through the application, development and evaluation of statistical methods to quantify the effect of potential prognostic factors. Firstly, I conduct a new prognostic factor study in pregnant women. The findings suggest that the albumin/creatinine ratio (ACR) is an independent prognostic factor for neonatal and, in particular, maternal composite adverse outcomes; thus ACR may enhance individualised risk prediction and clinical decision-making. Then, a literature review is performed to flag challenges in conducting meta-analysis of prognostic factor studies in the same clinical area. Many issues are identified, especially between-study heterogeneity and potential bias in the thresholds (cut-off points) used to dichotomise continuous factors, and the set of adjustment factors. Subsequent chapters aim to tackle these issues by proposing novel multivariate meta-analysis methods to ‘borrow strength’ across correlated thresholds and/or adjustment factors. These are applied to a variety of examples, and evaluated through simulation, which show how the approach can reduce bias and improve precision of meta-analysis results, compared to traditional univariate methods. In particular, the percentage reduction in the variance is of a similar magnitude to the percentage of data missing at random.