The diagram above outlines a learning healthcare system, which can be defined as an integrated system of delivering health that learns from its own mistakes. The four stages of the learning cycle resemble a traditional learning cycle for, say, an individual learning a new skill (for example, learning to ride a bicycle) and are as follows:
- Plan: the starting point is a plan of action. This might be a treatment guideline or, for research, a trial protocol.
- Do: action is taken as per the treatment guideline or trial protocol.
- Data: data is collected. This might be routine clinical data and/or research data collected according to a trial protocol.
- Review: the results are reviewed and the learning cycle is completed by revising the plan of action.
Research is an integral part of a learning healthcare system. This contrasts with the situation in much of medicine where research and the delivery of care are separate. There are many reasons for this separation. Bad and inappropriate regulation and other related bureaucracy are one of the major barriers to the closer integration of health and research needed for a learning healthcare system.
Examples of functioning learning healthcare systems are somewhat rare. One example is the treatment of childhood cancer in the UK. The figure below shows the overall improvement in childhood cancer survival in England over a 25 year period with 10-year survival having increased from around 65% in 1990 to 80-85% in 2015:
Source: ONS website
These remarkable improvements in survival have occurred over a time period when the vast majority of children with cancer in the UK were treated as part of a randomised trial. It has been estimated that between 70-90% (depending on the cancer type) of children treated for cancer in the UK were recruited to a randomised trial. These series of trials over a number of decades evaluated the new treatment regimen against the current best available treatment option and the findings from previous trials informed the design and treatment regimens evaluated in the latter trials.
There is a lot of discussion at present about the use of “real-world evidence” in medicine and the potential it has to help create a learning healthcare system. What should not be forgotten is that real-world evidence doesn’t just mean evidence from routine observational data for, as the example above shows, the most value will result from evidence from randomised trials that are an integral component of a learning healthcare system. What we need to do is to remove barriers that make it much more difficult to do trials as part of routine clinical care.