���ϳԹ�

Artificial intelligence experts and healthcare professionals in ���ϳԹ�have come together to help prevent a common and painful complication in advanced kidney failure treatment.

led by the and (PHUT) has developed an AI model to predict which patients are most at risk of their blood pressure dropping during dialysis; a condition known as intradialytic hypotension (IDH).

3 million people have Chronic Kidney Disease in the UK and 31,000 of these are on haemodialysis (), where their blood is circulated through a machine to clean it of toxins.

One of the most common complications for patients undergoing this treatment at home or in centres is IDH, which occurs when their blood pressure drops suddenly. It is associated with increased mortality and hospitalisations, and until now there has been no reliable way to predict if it will occur.

Pre-dialysis and real-time data were collected from 10 treatment centres over two decades (2000-2020), involving 3,944 patients. The team used data comprising a total of 73,323 sessions with 36,662 IDH events.

Using this information, they identified 33 variables to determine the most at-risk individuals. These were all observations that are routinely collected during clinical care, such as weight, temperature, age, blood pressure, medication and treatment details.

Machine-learning algorithms were used to build a predictor that could be useful in preventing IDH events from occurring. Of the five different algorithms tested, the Random Forest model had the highest overall predictive accuracy (75.5%), while the Bidirectional Long Short-Term Memory model achieved the highest sensitivity (78.5%). The analysis also revealed that systolic and diastolic blood pressures are key predictor variables. 

Project lead, Dr Shamsul Masum from the University’s School of Electrical and Mechanical Engineering, said: “This research highlights the value of using machine learning in healthcare, particularly in complex situations like haemodialysis. Predicting hypotension not only helps clinicians intervene early but also opens the door to personalised care.

"As we continue to develop and refine these models, the goal is to create a practical decision-support system that could enhance dialysis management, patient safety and quality of care.”

The study also tested the algorithm using only pre-dialysis data as inputs, to model the scenario at the start of a dialysis session. It found the prediction performance decreased but nevertheless remained clinically useful.

The paper, published in the , says future work will involve building a decision-support system for clinicians and conducting a clinical trial. It was co-authored by Consultant Nephrologists at PHUT, Dr Nicholas Sangala and Dr Robert Lewis.

Dr Nicholas Sangala, Consultant Nephrologist, said: “This model offers great promise that could pave the way to a future where AI/ML can be used to personalise treatments for individuals on dialysis and significantly reduce the risk of IDH and other complications.”

Robert Lewis, Consultant Nephrologist, added: “Clinical prediction of IHD is difficult and unreliable. This study indicates that AI and machine learning may be used as a tool to help deliver safer care to patients.”

The idea for the model stemmed from a previous study led by the University and the Trust. Two years ago, the team announced the development of an algorithm which can estimate how long a patient might spend in hospital if they’re diagnosed with bowel cancer. 

Using artificial intelligence and data analytics, they were able to predict the length of the patient’s hospital stay, whether they would be readmitted after surgery, and their likelihood of death over a one or three-month period.

Emeritus Professor , who co-authored both studies during his time at the ���ϳԹ�, said: “Although generative AI is grabbing the headlines, these studies show that AI to support decision-making remains just as important, and that machine learning can be effective using existing moderately sized datasets.”

Both studies were part of Portsmouth’s Future and Emerging Technologies research theme; one of five thematic areas that are written into the University's Strategy 2025 that support collaboration to extend knowledge and impact in interdisciplinary research, innovation and education.

The paper, Prediction of hypotension during haemodialysis through data analytics and machine learning, is available to .