The unsung heroes of data science – defining standards and best practice 

11 Aug 2023

Professor Michelle Williams recently collaborated on a new British Standards Institution framework focused on artificial intelligence (AI). Here we talk about why such frameworks and guidelines are vital to good quality research in hot topic areas like AI.

Research is essential for improvements to health, through developing better treatments, planning care, and responding to emerging threats. Most people’s knowledge of health research is built from what they read in the media. However, high hitting news stories are only the tip of the iceberg in terms of what is required to ensure these breakthroughs can be translated into improvements to healthcare. 

Sharing Best Practice 

Before new findings from research are accepted, they must be assessed by the scientific community to ensure they are of high quality, accurate and reproducible – meaning that you get the same results if the research question is repeated.  

For these findings to be translated into clinical care they must also be assessed to ensure they are safe and will lead to improvements compared with current treatments.  

To enable these assessments, information describing the research must be available to the scientific community. This information must be of sufficient detail to allow other researchers to interpret and repeat the research, such as instructions on how to perform all steps involved. 

The role of the BHF Data Science Centre 

The BHF Data Science Centre enables data-led research to improve heart and circulatory health. We work with a wide range of partners including patients, the public, clinicians, researchers, and healthcare organisations to support research. Part of this includes ensuring the research we support is of high quality and is trustworthy, and that its methods and results are freely available.  

The researchers we work with follow policies, which define how the research should be shared. These policies specify that all parts of the research, including computer code and the instructions used to process, interpret and analyse the data, are freely available and accessible to everyone.  

We are developing and publishing guidelines to share with other scientists to improve quality, trustworthiness, and data availability, and vitally, speed up research. Following best practices can also build trust with other researchers, research funders and the general public. So how are we doing this? 

Team science 

Science is a team effort and we work with other groups to develop guidelines and standards for research and its translation into clinical practice, including, for example, Professor Michelle Williams’ work with the British Standards Institution on a recent publication, entitled Validation framework for the use of AI within healthcare, which has major policy and practice implications.

Barely a day goes by now that we don’t see coverage of the potential positive or negative impacts that AI could have on our health. It is such a hot topic and it will inevitably affect us all. This new framework lays the groundwork for thinking about how AI could and should be implemented into our healthcare systems.  

It will help AI developers and suppliers, healthcare providers, healthcare staff and other stakeholders to develop and use AI safely and ethically in healthcare. 

Working with other groups and organisations to develop guidelines and standards is a valuable part of the work of the BHF Data Science Centre as it helps with our aim of getting research results into clinical practice for the benefit of patients.  

What next? 

Guidelines, best practices and policies are constantly changing, and need to be flexible and easily updated to respond to ever-changing scientific developments.  

We are developing standards that we believe will be of great benefit and impact to scientists across a number of areas, including how health data can be used in clinical trials, and how to structure different types of data.  

We also want to make sure that shared computer code – for example the computer code to interpret the information in someone’s medical record and determine if they have a particular disease – is easy to re-use by other scientists. 

This foundational work is often overlooked, but is an essential part of getting research right, and ultimately, translating it into benefits for patients. Best practice and standards truly are the unsung heroes of quality research. 

  • AI
  • Artificial Intelligence
  • News