Introduction: learning health systems

Friedman and colleagues1 outline a vision of the learning health system founded on the sharing of data and achieved through alignment of information technology, advanced analytics and clinical expertise. The Institute of Medicine2 recognises the potential of the learning health system to generate new information automatically during the delivery of healthcare, offering continual opportunities to improve healthcare processes for the benefit of public health. The learning health system is promoted as a mechanism to accelerate the adoption of effective treatments into clinical practice, shortening the extended delay3 from publication of research findings to implementation. Furthermore, it harbours the ambition to deliver personalised medicine to each service user, rather than the systematic provision of identical care to groups of patients who share the same characteristics. Worldwide escalating costs in healthcare provision due to demographic changes, compounded by ongoing use of ineffective tests and treatments, make it critically important to harness the efficiency gains of a learning health system. Of its many potential characteristics, one distinct attribute of the learning health system is to enable efficient investigation of whether strategies promoting implementation of best practice (such as educational initiatives or care bundles) actually work. A systematic review identifying a low frequency of reports on the evaluation and impact of learning health systems4 prompted us here to reconsider the central requirements for evaluation of improvement interventions within the learning health system.

Here, we reflect on two recent implementation studies, both using efficiencies of the learning health system (minimising data acquisition and relying heavily on data acquisition from existing medical records), to illustrate the key issues arising when incorporating this type of research into the learning health system. In this paper, we argue that the presence of infrastructure which facilitates data sharing, combined with appropriate research design, analysis and reporting, …