Why tackling the elective backlog with AI starts before the operating theatre

Tackling the elective backlog is one of the biggest challenges of our generation. It is, in fact, one of the biggest challenges in the history of the NHS. But this issue is not just a number; it's a growing list of real people waiting for essential surgery, affecting their quality of life and, in some cases, their prognosis. The solution to this problem is not simple, but one of the most promising approaches lies in leveraging artificial intelligence (AI) to optimise the planning and delivery of elective procedures. With all the hype surrounding AI, how can it actually be utilised to achieve throughput gains and why is engaging with AI technology our best hope for addressing the elective backlog??

Understanding the Elective Backlog Challenge?

This backlog has reached critical levels. Though it was exacerbated by the COVID-19 pandemic, the sizeable waiting list was by no means a brand-new challenge. The British Medical Association (BMA) estimated that between Jan 2009, when waiting lists were at their lowest, and Jan 2020, the number of people waiting for consultant-led elective care grew by 95%: a compound annual growth rate of 6.3%. If this growth rate were to have continued to 2024, we would have expected a waiting list of about 5.8 million (still ~1.3 million more than Jan 2020). Instead, it’s soared ~3 million higher to 7.6 million. Even if the demand brought on by COVID delays were to be mitigated, there are still both systemic challenges and “vicious cycle” effects that need addressing.?

Systemic challenges are things like our aging population, health inequalities, and the huge variation in outcomes across different hospitals.?

“Vicious cycle” effects are brought on because waiting times are long. Staff burnout that comes from dealing with more (and unhappier) patients causes mistakes, poor outcomes for patients, and staff sickness rates to go up (a vicious cycle that decreases capacity and further exacerbates the problem). Patient acuity increases as smaller ailments become bigger problems through delayed treatment. Denser cataracts and more debilitated hips require longer in theatres to deal with, meaning higher costs, and increased use of the private sector often comes at a premium.?

But how can we think of addressing systemic problems, when these vicious cycle effects are spiralling out of control and prevent us making progress? The bottleneck lies in the bottom of the funnel – if we can drive more throughput in theatres we can alleviate these effects, allowing us to reach our true capacity, and ensure every additional pound spent on improving the NHS is impacting the maximum number of patients.?

But how does AI solve this problem??

AI in Pre-Operative Planning?

The key to tackling the elective backlog effectively begins weeks before patients are wheeled into the operating theatre. Consultants have an incredible wealth of experience and expertise, but with a large enough dataset, even the best consultant cannot out-predict the best AI model when it comes to estimating procedure times. Taking millions of historical procedures, AI models can transform how surgeries are scheduled by predicting procedure times based on billions of variables. In a matter of seconds, AI can incorporate the free text describing the procedure, the theatre its being performed in, and both the surgeon and anaesthetists’ unique performance and behaviour. This prediction improvement leads to two things: better in-session utilisation from the avoidance of early and late finishes, but an often-overlooked benefit is the increase in confidence in forecasting.?

By having more reliable forecasts, patients can be booked further in advance and expectations better managed externally. Patients become happier that their procedures are not getting cancelled or moved and staff go to work knowing what to expect.?

But better procedure times are only one element of the booking process.?

Optimising Scheduling with AI?

With more accurate predictions of procedure times, AI models can then tackle the complex task of scheduling. Bookers typically painstakingly record procedures into Theatre Management Systems manually, going through each patient on the list one-by-one. When a procedure doesn’t fit into a session, they move onto the next session, often leaving excess capacity, or even building in a buffer (based on well-intentioned, but detrimental, ingrained ways-of-working).?

Smart models can do a much quicker and more efficient job. By factoring in the specific requirements of each procedure, including handover time, anaesthetic time, and urgency, schedules can be optimised for throughput. By having at hand every possible procedure that could fit into a slot, the AI can immediately “stack” these to achieve a statistical optimum – i.e., a target session-fill that minimises the risk of early or late finishes. If there are any complications (like limitations in procedures per day due to equipment constraints) this can be built into the model and adapted to seamlessly.?

Armed with AI, bookers can save significant time and energy, whilst achieving better outcomes.?

Given this all sounds straightforward, why isn’t it already happening, and what are the challenges in implementing this type of AI model??

Engaging with AI: Our Best Hope?

The first challenge comes from engagement. Changing behaviours, especially those that are ingrained over years, is difficult. Regardless of the efficacy of any software solution, it is the positive experience of those experiencing the change that is imperative in realising value.?

Recognising the need for change and being open to adopting novel solutions is the first step. Embedding technology in an empathetic way is essential when such a seemingly black-box technology comes along. Users need training, patience and an open environment to ask what they might believe to be silly questions (but often are shared by everyone).?

If it is purely engagement, then why haven’t the most positive, switched-on organisations already done this??

The Technical Challenge?

AI models are amazing at learning from data. However, you can only learn from data if you have data! The types of models that predict things like procedure times based on lots of different variables (known as supervised machine learning models) often need very large amounts of data to produce accurate results. Due to challenges in data recording practices (OPCS can be imperfect) as well as the nature of how procedure details are captured (free text!) it can be hard to get the data needed to accurately describe the procedure being performed. Without this data alone, it is incredibly difficult to make an accurate prediction on how long a procedure might take.?

Conclusion?

The elective backlog is a complex challenge that demands innovative solutions. AI stands at the forefront of these solutions, offering a way to transform the pre-operative phase and maximise operating theatre throughput. By accurately predicting procedure times and optimising scheduling, AI can help address the backlog more effectively, improving patient outcomes and reducing the strain on healthcare systems.?

Engaging with AI technology is not just an option; it's a necessity for the future of the NHS. The path forward involves embracing AI's potential, investing in its development and uptake, and overcoming the challenges it presents. As we look toward a future where healthcare systems are more resilient, efficient, and patient-centred, AI represents the biggest hope we have for tackling the elective backlog and beyond.?

?

Real World Health is a data solutions provider and has been a partner to the NHS for more than 10 years. Our Opfeed solution has been developed with theatre experts, is trained on more than a million procedures from proprietary datasets and has been shown to deliver double-digit increases in theatre productivity. To find out more about how our solutions can help deliver your productivity targets, please contact [email protected] or [email protected].?

?