By: Edward Bramley-Harker, Co-founder and Director at Edge Health, Tom Michaelis, Analyst at Edge Health and Christian Moroy, Co-founder and Associate Director at Edge Health
Computers, sensors, data storage and new software to analyse data are everywhere and the NHS is no exception. The NHS collects more and better data than most other health systems in the world. In fact, the English department of health mandates that data is collected and submitted for all NHS commissioned activity. Over 5 million records[1], corresponding to 98-99% of all UK inpatient hospital activity are submitted every day to central NHS databases[2], resulting in highly detailed and complete datasets.
More can be done but we are on the way
Of course, more can be done to link datasets across care settings, improve quality and frequency and make data available at scale. However, several large-scale NHS initiatives have already been working hard at harnessing these datasets to improve patient care. This includes academics, audits and registries or national programmes such as Getting it right first time (GIRFT) which aim to use NHS datasets to spot unwarranted variation in practice, improve outcomes and optimize hospital resource allocation.
Can we move from reactive to predictive on the ground?
Despite their success, programmes such as these are often retrospective: in many cases they can only improve patient care once the care has already been delivered. An example of this can be seen in the GIRFT orthopaedic national report which identified that hip replacements using an un-cemented cup made from tantalum did not improve outcomes despite increased cost[3]. This was only possible by looking at patient outcomes 5 years after the initial surgery – while in the meantime, inefficient practice continued, taking up valuable resources that could have spent improving patient care elsewhere.
This raises the obvious question: are there cases where we can put the cart before the horse and utilise NHS patient databases to predict information that can help improve decision making before it takes place? This approach has already attracted interest when it comes to complex imaging tasks; we regularly hear of how artificial intelligence (AI), trained on large medical datasets, is used to aid patient diagnoses. It is time that we considered using the same technology to support clinicians and managers in making operational decisions about patient care – enabling them to mitigate bottlenecks and reduce risks.
Predicting patient outcomes before the fact
One such example where this may be possible is predicting the likelihood of an adverse outcome from a surgery, before the surgery occurs. The potential benefits of doing so are numerous:
- Patient communication and consent: Clinicians can better communicate the risks of a procedure if they know the likely outcomes. This is important in getting complete patient consent and has been shown to improve patient satisfaction[4].
- Bed planning: Managers can better allocate beds to patients if they know a patient is likely to have a long length of stay.
- Efficient scheduling: Procedures can be better scheduled to avoid grouping patients that are high complexity.
- Clinical decision making: Making the right decision on admitting a patient as a day case or inpatient may depend on likelihood of readmission which data can help predict.
Speak to us at the NOA Annual Members’ Conference about where we can take this
With our partners Edge Health we are exploring bringing these decision-making capabilities to you. For the 2022 NOA Annual Members’ Conference we have developed a proof-of-concept tool that predicts both length of stay, and probability of an emergency readmission within 30 days for a hypothetical patient using artificial intelligence (AI). The AI calculator allows users to specify 28 different clinically relevant patient characteristics for 13 commonly performed orthopaedic procedures. We are keen to speak to you and hear of your experience on the ground and ideas of how this can improve care for patients.
[1] NHS Digital, SUS+ Essentials v13.4, 2020
[2] National Audit Office. Healthcare Across the UK: A Comparison of the NHS in England, Scotland, Wales and
Northern Ireland, 2012. https://www.nao.org.uk/wp-content/uploads/2012/06/1213192.pdf
[3] GIRFT, A national review of adult elective orthopaedic services in England, 2015
[4] Stacey et al, Decision aids for people facing health treatment or screening decisions, 2017, https://pubmed.ncbi.nlm.nih.gov/28402085/
