University of Southampton

ILD100

Data about patients with chronic kidney disease (CKD), end stage kidney disease (ESKD) or an acute kidney injury (AKI) cared for at the UK’s 70 adult and 13 paediatric renal centres, are received by the UK Renal Registry (UKRR) via electronic feeds. The majority of centres currently submit these data quarterly, but in time all centres will submit data via a daily feed – at this stage only a few centres submit on a daily basis. Data validation is important to eliminate errors from data and provide an accurate and complete dataset. The transition towards a daily flow of data from renal centres into the UKRR database requires the UKRR’s current manual data validation processes to be revised and automated. We want to investigate how the current processes can be enhanced and/or automated by some of the newer AI and machine learning techniques. Establishing automated validation for the daily data feed will result in improved checking and correction of patient data and a resulting improvement in patient data quality and completeness that can be used for audit, research and care quality improvement for people with kidney disease. The project will initially look at some of the key blood test result data for sets of patients. Blood test data received via daily feeds for approximately 10,000 adults and children with CKD will be included in the project. We will aim to build some test models which implement machine learning and possibly AI methods that can identify patterns and issues in this data. We will also aim to determine whether different validation models are required for data with different treatment modalities or other groupings in the data. We can then establish the process through which errors can be auto corrected and determine which will still need manual intervention.

25/05/2021