OBJECTIVES: A range of linked electronic health records (LEHR) datasets are becoming available in the NHS in England.  Understanding the benefits such large LEHR datasets can offer when performing cost-effectiveness analysis and overcoming the challenges inherent in utilising such datasets using as a case study: the CALIBER (Cardiovascular Disease Research using Linked Bespoke Studies and Electronic Records) dataset to assess the cost-effectiveness of treatment of patients with chronic stable angina. METHODS: The CALIBER dataset links primary care data from CPRD with secondary care data from HES, mortality data from ONS and disease-specific data from MINAP.  This dataset is used to provide a generalisable baseline for cost-effectiveness models and to provide effectiveness estimates for treatments observed in the dataset.  It is also used to explore decisions that, due to ethical concerns, could not be researched through RCTs - for example, optimal durations and combinations of treatments used in current practice. The statistical methods of matching, differences in differences and instrumental variables are used to overcome the selection bias problems associated with inferring treatment effects using observational data. The models are also combined with RCT results for novel treatments to assess their cost-effectiveness in a real world as opposed to trial setting. RESULTS: There were significant challenges involved in working with LEHR ranging from correctly linking the data to plausibly imputing missing data and correcting for selection bias in estimates. Having addressed these challenges we were left with a rich dataset, from which to estimate costs, risks of health events and treatment effects in a generalisable real world setting. CONCLUSIONS: While the selection biases associated with the use of LEHR within the NHS makes them challenging to work with, the large sample sizes, generalisability of the results and relatively low cost of the research makes them a hugely valuable resource for economic evaluation.