The Guide to Health Informatics 3rd Edition

Learning health systems are the next big thing. Through the use of information technology, the hope is that we can analyse all the data captured in electronic health records to speed both the process of scientific discovery and the translation of these discoveries into routine practice^1,2. Every patient’s data, their response to treatment, and final outcome, will no longer be filed away, but feed the care of future patients³. It’s an exciting vision, and if we can achieve it, there is no doubt healthcare delivery would be transformed.

If we were to step back, we might conclude that although this is an admirable vision, for all its failings, the machinery of science is already working faster than we can handle it. The arena where organizational learning most needs to take hold is in the way we deliver health services. It is clear that we could do so much better in this arena. There is too much variation in patient care, too much waste and harm in the system.

So, if what we have today is not yet a learning health system, then by definition we must have the opposite – a forgetting health system. If that is the case, then here are two working definitions to contrast the two ideas:

1. Learning system: The past shapes the future. Today’s mistake is tomorrow’s wisdom.
2. Forgetting system: The past is the future. Today’s mistakes are forgotten quickly and are repeated tomorrow.

With this perspective it is easy to see examples everywhere of such ‘forgetting’. The history of large-scale e-health is a litany of the same case study being repeated. A large health IT project is started (usually by a government and usually as a technology innovation project rather than to fix a defined health problem). It quickly runs over time, over budget, and is treated with dismay as its users find it doesn’t do what they were promised. The end result is new problems, workarounds to circumvent the intruder technology, and in some cases, the eventual removal of the system.

The solution to this mess is of course seems to be to start a new large-scale e-health system, run by different people. Yet, like moths to a flame, these new protagonists seem to make the same set of strategic mistakes, but in new ways. Today’s large-scale health IT projects seem to be in a perpetual state of Groundhog Day, and must be a classic example of a forgetting system. This may be because there is no learning, because we convince ourselves that this time is ‘different’ and the past has nothing to teach us, or because when we look at past failures, we are unable to drawn any conclusions because of a lack of ability or imagination.

Information is lost

If a system ‘forgets’ then by implication that means that information that existed in the system has been lost, preventing its reuse to guide future events. If you think about it for one moment, the health system loses information every moment of every day, in unimaginable amounts. We only record a fraction of the events that occur. Most of these are not directly captured but rather are reconstructed from the memory of the individual making the record.

If we were to transform our forgetting system into a learning system, so that health system improvements are driven by experience, what do we need to do? For a system to ‘remember’ an event, a teachable moment, it firstly needs to be detected. You can’t remember what you never saw. Next, after detection, it needs to be recorded. Finally, this recording needs to be somehow aggregated with other events, for discovery of new knowledge to occur.

When it comes to treatment and diagnosis, we have built very precise mechanisms to detect important events or processes through a variety of diagnostic methods. The data captured are increasingly being recorded in electronic systems, and analysed. That’s the ‘big data’ movement in healthcare in a nutshell.

When it comes to health services, we still don’t know what events we should detect, what processes of service delivery should be instrumented, and there is still little thought to pooling system measurements in a way to allow across organizational analysis and learning. So, if we are to make our health services become learning ones, the task is much bigger than installing patient records.

One of the challenges to learning in health services is that, whilst it is easy to imagine that organizational memories can be stored in bits on a database, more often they will sit in the heads of those who work within them⁴. So, a good indicator that you are working in a forgetting system is that your organization has a high staff turnover – because it is very likely that when staff walk out the door they also walk with everything they know about the system that they are leaving, and no one has tried to capture that rich web of knowledge (if it could so easily be captured).

When learning also requires forgetting

Memories however also exist in other ways in an organisation. Crucially, they persist in the processes, protocols and built structures of the organisation, and in the workarounds and annotations that happen to physical spaces⁵. These structural memories are not inert or idly awaiting someone’s analysis. They sit there every moment shaping work, and altering human perceptions, actions and intent.

With time, accreted structural memories can lead to inertia, to an immovable status quo⁶. These kinds of memories therefore need to be managed. Some need to be revisited and revised, others are the jewels of the past that we should never forget and some, some are impeding the emerging purpose of an organization, and need to be actively forgotten. The need for organizational apoptosis, or active forgetting, is something I’ve written about before⁶.

From learning to adaptive systems

So, learning clearly is not nearly enough. Recording and analysing, at least as far as organisations are concerned, can only take you so far. The living walls and praxis of an organisation are already busy recording, even learning, but may not make it possible to do anything about what has been learned. For organisational inertia to be broken, and for repeated failures to be avoided, we not only need to learn, we need to actively, wisely, carefully, forget.

That will be a true learning system, because learning systems also need the capacity to change.

References

1. Etheredge LM. A rapid-learning health system. Health affairs. 2007;26(2):w107-w118.
2. Friedman CP, Wong AK, Blumenthal D. Achieving a nationwide learning health system. Science translational medicine. 2010;2(57):57cm29-57cm29.
3. Gallego B, Walter SR, Day RO, et al. Bringing cohort studies to the bedside: framework for a ‘green button’ to support clinical decision-making. Journal of Comparative Effectiveness Research. 2015(0):1-7.
4. Coiera E. When conversation is better than computation. Journal of the American Medical Informatics Association. 2000;7(3):277-286.
5. Coiera E. Communication spaces. Journal of the American Medical Informatics Association. May 1, 2014 2014;21(3):414-422.
6. Coiera E. Why system inertia makes health reform so hard. British Medical Journal. 2011;343:27-29.