With the exception of skin cancers, breast cancer is the most common cancer affecting women. While progress has been made in the detection of primary breast tumours, there are few genomic tests that are able to accurately predict outcome. Current genomic tests such as as Mammaprint and Oncotype DX are not widely available and are only suitable for early stage tumours, with additional restrictions applying depending on the test used. We sought to derive a new predictor of breast cancer outcome from TCGA RNA-Seq data that can provide an accurate indication of prognosis, even in later stage tumours. Alternative polyadenylation (APA) is the process whereby the poly (A) tail is added to the 3’ untranslated region (3' UTR) of a messenger RNA (mRNA) at one of multiple possible sites, changing 3' UTR length and potentially the regulatory elements that bind to it. APA has been suggested to be predictive of tumour outcome and can be inferred from RNA-Seq data. We used elastic net linear modelling to select coefficients that best predict relapse free survival from clinical, APA and gene expression data. The best model was generated using a combination of all 3 data types, with common clinical indicators playing only a small role. When validated using 10 fold cross validation, patients with a score higher than the median generated by our model were at least 3 times less likely to die of cancer than those with a score below the median (p << 0.01). Our ultimate aim is to derive an accurate genomic test for breast cancer outcome that can be applied to all breast tumours and is less reliant on clinical data. This test could potentially be implemented using the in house M-PAT approach, for substantially less than the cost of a full RNA-Seq experiment.