Amazon had revenues of $280 billion in 2019- not bad for an online bookshop founded in 1994. According to McKinsey [1], 35% of what Amazon customers buy comes from recommendations generated by the company’s algorithms. Netflix, another company founded at a similar time, has revenues of $20 billion and 75% of what customers watch comes from their recommendations.
In just a few years, the analysis of customer data has become central to the way the world’s most successful businesses operate.
Are there lessons from this for the health sector?
I think there are.
The serious incident process across healthcare in the UK produces a massive amount of data. Over two million patient safety incidents are reported through the current National Reporting and Learning System each year. Work is well underway to introduce the new ‘Patient safety incident management system’, which will be a big step forward. Nevertheless, this process is based around reporting and cataloguing of incidents, rather than learning from the results of the incident investigation process.
Every incident that occurs is investigated by the relevant care provider. I know that the NHS puts an enormous amount of effort into carrying out this sort of investigation. A typical report is 20 to 30 pages long. They are usually written by clinical professionals who work in the same organisation where the incident occurred. This has the big advantage that they know the particular healthcare setting intimately. However, as they typically spend a number of days of their time on each report – time is taken away from caring for patients. This represents a colossal allocation of resources.
And what is the return from that colossal investment? It would be unfair to say that the return is low – individual trusts have extensive processes in place to review incident reports, and have often learned a lot from them. But there is so much more that could be gained from all this data – not least the healthcare system learning as a whole from the patterns that present themselves. It is this learning from big datasets that has boosted Amazon and Netflix into the commercial stratosphere.
The key elements for an effective system for learning from incidents were identified in research carried out for the Health & Safety Executive (HSE)[2]. It is worth reading their description of what such a system in full. They say that the system needs the following elements:
- “An incident/accident reporting system;
- A process for incident investigation that ensures that the underlying as well as immediate causes of accidents and incidents are understood, taking full account of human and organisational factors;
- A process for analysing cumulative information on accidents and incidents from both internal and external events;
- A process for ensuring that the findings of incident investigation and analysis of accident and incident data are acted upon in a timely fashion and suitable interventions put in place or modifications made to prevent a recurrence of the incident or similar incidents;
- A process for evaluating the success or otherwise of interventions and modifications;
- A process for disseminating information on accident and incident causation and suitable interventions/modifications to all relevant parties (both internal and external), as quickly as possible; and
- A system to capture the information in a format that is readily searchable and retrievable to allow ease of access, so that any lessons learned stay learned (corporate memory).”
The UK healthcare system clearly has something in place in place for the first two steps – reporting and investigation. It is hoped that the new system being implemented will deliver some of the third stage – analysing cumulative information on incidents. What is missing is the next stage, analysing cumulative information across the healthcare system. Without that information, the other requirements of an effective system – acting, evaluating, disseminating and learning are necessarily curtailed.
There are well over a thousand hospitals in the UK, with over a million people working in them and over two million incidents a year. The data pool is potentially vast. So why doesn’t healthcare make more of this data?
Well, the truth is that the serious incident process is often a manual one. Although some incident reports can be uploaded from data put into incident reporting systems, this is not the case for the outputs of incident investigations. We have worked with many trusts whose investigation process involves a series of Word documents tracked in an Excel spreadsheet. Nationally speaking, the outputs of the investigation process are disbursed across hundreds of providers. The technology being used is now decades out of date.
We believe that we can do better than this. We have designed Eva so that investigation reports can be completed in a common electronic format, with a system that guides users through the process and provides an audit trail that can be reviewed by colleagues as it progresses. This also provides the potential for (suitably anonymised) data to be shared securely via the Cloud allowing the data to be mined for the detailed information that they hold.
Most importantly, having access to a pool of data would be crucial in spotting patterns where things go wrong. The beauty of using large datasets is that they can spot trends that simply can’t be picked up by the human eye. We are all used to spotting patterns that we expect to see, but the value really comes in things that we couldn’t think of ourselves. The history of science is full of examples of unexpected patterns being recognised – often by accident – and being responsible for driving process.
And, like with Amazon, it isn’t necessary for every recommendation to be correct, as long as from time to time they generate something new.
Another benefit of such a large dataset is that it could present patterns with less common procedures. For less frequent procedures, each provider may only have one or two incidents per year relating to that procedure. This simply isn’t enough to recognise patterns. Bring such data across a large number of providers, however, and it will be definitely possible to see where there are associations.
The data could also be used in identifying areas where more work is needed. The data may simply throw up problems, rather than solutions. But this in itself is valuable as it indicates where understanding could be better and further research is needed.
A system like Eva could also be used to help with learning. One of the key requirements of an effective system identified by the HSE was having a process for evaluating the success of interventions. Within individual hospitals, however, there will often be too few incidents to evaluate the statistical significance of any interventions. The more data, therefore, the more it will be possible to see what really makes a difference.
In short, the potential benefit from a shared electronic system used across a significant proportion of the healthcare settings in the UK is incalculable. Our technology, Eva, does exactly that. Not only does the system guide the user through the process of an investigation (making investigations easier and quicker to carry out), it also records the data in a standard format. Suitably anonymised, that data can be shared with other care systems to identify common factors and learning.
Detailed examination of feedback is now not only commonplace, but essential to the businesses that have come to dominate the world economy. They have demonstrated beyond doubt that the use of data can transform our lives. The benefits of this sort of ‘learning system’ in the health field make the current uses of this technology look trivial. It can surely only be a matter of time before healthcare catches up with the way the modern world works.
[1] https://www.mckinsey.com/industries/retail/our-insights/how-retailers-can-keep-up-with-consumers#
[2] Principles for Learning Lessons from Incidents – A UK Perspective, Health & Safety Laboratory/Health & Safety Executive
Follow Eva on social media. You can find Eva on Facebook, Instagram, Twitter and LinkedIn.
