Problem: The problem we address is the general lack of effective and efficient integration of health-related data, methods and expertise, across defined populations, for public benefit. This is a problem for most health systems throughout the world. Population: We focused on the Northwest of England, which has a rich diversity of social structures, health outcomes and research-active healthcare organisations. In particular, we studied the Salford health economy (population 216,000) because it has relatively advanced healthcare information, with an integrated electronic health record (EHR) across primary care, secondary care and the public sector commissioner/payer. Objectives and Methods: The requirements for epidemiological integration and processing of data were identified through an iterative process, over two years, involving pilot studies, using crude or enhanced EHR data. Salford clinicians were engaged in each study, an Informatician (Baker) was embedded in Salford, and statistical, epidemiological and software engineering input was provided by the University of Manchester. The criteria that we used for identifying, and engineering, requirements were: governance; patient/citizen privacy; ethics; data quality; skill-gaps; clinical information utility; public health information utility; and scalability. Findings: A key finding was that the EHR data alone were insufficient for scientific purposes, analyses were enriched by extracting tacit knowledge, through playback of emerging findings to local clinicians, encoding this as explicit metadata, and using the metadata in data-cleansing or statistical-modelling. We found that large-scale record linkage across the health economy was acceptable, provided it took place within the healthcare setting, specifically, as close to primary care physicians as possible. In the UK, ‘Primary Care Trusts’ bring primary care providers and commissioners/purchasers together. Academics worked effectively on the data, in the Primary Care Trust setting, under honorary contracts covering information governance. Data cleansing and organisation was more computationally expensive than anticipated, typically involving re-coding, filtering and re-organising up to 1,000,000