Introduction
Innovate UK’s Digital Catapult provides funding to help businesses commercialise innovation, explore and test new products and seek expert advice and guidance. It was in the latter capacity that Smith Institute recently supported a grant recipient called Flush-Tech with a solution that can potentially improve the lives of millions of people globally.
Flush-Tech is an innovative company looking to improve early-stage detection of acute kidney injury (AKI), a global public health challenge. AKI is a complex clinical condition which affects more than 13 million people worldwide each year (Abebe, 2021) and occurs when your kidneys suddenly stop functioning properly. Accounting for confounding factors such as age and sex, even a mild occurrence of AKI has been found to increase the mortality rate by more than 25% (Linder, 2014). Furthermore, failure to diagnose AKI early can lead to complications in treatment of other illnesses such as cancer.
Working with Smith Institute was like talking to NASA engineers.
Flush-Tech wanted to provide a proof of concept in their solution to early-stage diagnosis of AKI. Through advising on AI techniques and different machine learning (ML) processes, Smith Institute were able to help them to improve their algorithms and data gathering process. We were also able to help them to plan how to progress to the next technology readiness level, bringing this advance in medical technology closer to reality.
Problem
The National Institute for Health and Care Excellence (NICE, 2019) have emphasised that early-stage detection of AKI could significantly reduce deaths or complications in patients. One early-stage indicator of AKI is a sudden decrease in urine output (UO) (Oliguria). Up to 100,000 AKI related deaths in the UK occur annually (Think Kidneys, 2017) and accurate UO-monitoring is key in preventing 20-30% of these deaths as estimated to be preventable (NHS, 2014).
The NHS spends over £1.02 billion annually in addressing AKI (Think Kidneys, 2017)
AKI is a major health problem, affecting 1 in 5 hospital inpatients in the UK (Davies et al., 2017)
The accurate monitoring of UO plays a key role in preventing AKI (Davies et al., 2017)
This means, conservatively speaking, a 10% avoidance of AKI cases within NHS Trusts saves the NHS approximately £102 million annually, while in tandem reducing ward pressures due to the early discharge of inpatients and avoidance of readmissions
Currently, UO monitoring in hospital is an intermittent manual process which is therefore prone to errors. Flush-Tech is aiming to tackle this problem through automated monitoring of UO. Digitising the process could further alleviate the monitoring and analysis burden on nurses who are overworked and short-staffed. It could also improve the patient experience. Flush-Tech partnered with us through the UK’s Digital Catapult to help them to improve the accuracy and speed of their algorithms.
Solution
When automating UO collection, the solution needs to be reliable and able to provide real-time data which improve on the existing human data-collection methods. Through our expertise in ML and AI, as well as our capabilities in product development and scientific method, we were able to provide technical guidance and a dynamic response through a series of weekly workshops. In these workshops we could review their existing approach together, provide feedback and iterate on recommendations we put forward.
Together, we were able to discuss recommendations for each stage of algorithm development from ML and AI best practices to performance evaluation. We were also able to identify the limiting steps. This allowed us to provide a plan for improving the accuracy, reliability, and speed of their algorithms.
We were further able to provide guidance on best practices in data collection to improve the robustness of their methods in operation and suggest future avenues for exploration to ensure that their machine learning has the best design for their needs.
Results
Supported by Innovate UK funding, Smith Institute were able to help Flush-Tech to bring this potentially-life-saving technology closer to launch. Additionally, we have enabled them, future users, and patients to have greater trust in the product to identify AKI. We did this by helping Flush-Tech to improve their accuracy and reliability, and by providing a roadmap to future development of their ML algorithms. Flush-Tech described working with Smith Institute as being like talking to NASA engineers. Working together has helped them to future-proof their data collection to ensure that their methods are robust in real-world deployment.
Smith Institute can collaborate with Innovate UK grant recipients by providing expert guidance and verification on optimisation engines, forecasting, machine learning and simulations to support ambition and the solutions of tomorrow. Discover more about our solutions here.
References
Abebe, A. a. (2021). Mortality and predictors of acute kidney injury in adults: a hospital-based prospective observational study. Scientific Reports, 11(1), 1–8.
Davies, A., Srivastava, S., Seligman, W., Motuel, L., Deogan, V., Ahmed, S. and Howells, N., 2017. Prevention of acute kidney injury through accurate fluid balance monitoring. BMJ Open Quality, 6(2), p.6.
Linder, A. a. (2014). Small acute increases in serum creatinine are associated with decreased long-term survival in the critically ill. American journal of respiratory and critical care medicine, 189(9), 1075–1081.
NICE. (2019). Acute kidney injury: prevention, detection and management. National Institute for Health and Care Excellence. Retrieved from https://www.nice.org.uk/guidance/ng148/chapter/Context
Think Kidneys, 2017. [online] Thinkkidneys.nhs.uk. Available at: <https://www.thinkkidneys.nhs.uk/aki/wp-content/uploads/sites/2/2016/02/Care-Homes-AKI-guide-FINAL-160217.pdf> [Accessed 1 January 2020].
