Investigating equity and fairness concerns in the data, prediction model, and implementation of the personalized renal function monitoring tool for individuals with heart failure (RENAL-HF) project: Expanding user guidelines and information

Heart failure affects approximately one million individuals across the UK. Kidney problems are a major concern for these patients, and they need to be monitored closely in primary care. However, there is limited guidance on how to monitor heart failure patients.

A personalised renal function monitoring tool in people living with heart failure (RENAL-HF) is being developed to facilitate patient monitoring and aid clinical decision making.

A clinical prediction model - an algorithm using patient information to predict future kidney function - is the foundation of the tool and will be used to determine the advice and information given to the clinician.

The model is developed using real-world primary care data, but there is growing concern and awareness that such models can underperform in groups of individuals defined by protected characteristics, such as minority ethnic groups. This can lead to worse clinical decision making and hence outcomes in these groups: inequity or unfairness.

This unfairness can be a result of biases or other issues in the data, the model, or the implementation of the model.

Hence, fairness and unintended consequences, such as delayed diagnosis of worsening renal function or hospitalisation, of the RENAL-HF decision support tool needs to be investigated.

This research project is dedicated to exploring the equity and fairness implications associated with the RENAL-HF decision support tool.

The study involves a comprehensive analysis of patient subgroups, examining the RENAL-HF model's performance across diverse scenarios, such as different measurement and healthcare interaction patterns. Through thorough investigation and simulated assessments, the project aims to quantify any biases present in the tool.

Collaboration with the existing RENAL-HF team, including patients, clinicians, and researchers, is integral to this project.

The methodology includes a rapid review and stakeholder meetings to identify scenarios and subgroups susceptible to fairness concerns.

Subsequently, the RENAL-HF data, model, and tool implementation will be evaluated in these groups, utilizing simulations to uncover potential unintended consequences.

The project will conclude by effectively communicating any identified fairness caveats associated with RENAL-HF to users, patients, and researchers developing similar tools.

This communication includes providing clear guidance on tool usage, navigating potential biases, and incorporating insights from RENAL-HF patient representatives into the development of patient materials.

The ultimate goal is to enhance the tool's fairness, usability, and positive impact on patient outcomes while addressing concerns related to algorithmic fairness and unintended consequences.