As healthcare entities move towards the deployment of artificial intelligence (AI) models to streamline patient engagement, being acquitted of good customer relationship management (CRM) system data is of rising importance, an expert has said.
In April 2025, the UK National Health Service (NHS) released guidance on the importance of using high-quality data in training AI models. The guidance highlighted that poor quality data can lead to unintended biases and other inaccuracies that may stymy the effectiveness of AI models used for varying functions in healthcare systems.
Go deeper with GlobalData
Discover B2B Marketing That Performs
Combine business intelligence and editorial excellence to reach engaged professionals across 36 leading media platforms.
CRM data includes information on patient demographics, individuals’ medical history, insurance information, billing details, and more. The consolidation of this data is intended to improve patient engagement, personalise care, and track individuals’ medical history.
A 2022 report indicated that revenue streams from 44% of healthcare respondents were directly affected bypoor quality CRM data. In addition, 69% of respondents said their organisations did not pursue or complete potentially valuable initiatives due to low-quality CRM data.
According to Rachel Mak-McCully, senior data scientist at digital twin developer Unlearn.AI, the cleaning and harmonisation of CRM datasets is not an area that is given enough consideration – despite its significant influence on AI models being deployed in healthcare.
“In thinking about clean data, there’s a few issues, including how accurate and comprehensive the data is,” Mak-McCully told Medical Device Network.
US Tariffs are shifting - will you react or anticipate?
Don’t let policy changes catch you off guard. Stay proactive with real-time data and expert analysis.
By GlobalData“In the US, we have a very fractured medical system, so the information that you get about someone may be quite fractured.”
Fractured data can relate to data that is missing because it hasn’t been collected at the same time or stored together, Mak-McCully said, or in the sense that it’s been measured in different ways.
“Imagine the data cleaning process as a spreadsheet. You might have a lot of different columns that are named differently, but they’re the same thing. You want to make sure that all of those are harmonised into the same column,” Mak-McCully said.
“And you want to make sure that the units that you’re measuring across a range of different data sources are the same. The ultimate goal is to have a clean, tabular dataset at the end that incorporates all relevant sources of data, so that a single, clean source of data can be used to train your AI model.”
Mak-McCully concludes that while there is a lot of talk about the performance of AI models in healthcare and other areas, such as the new benchmarks that have been made, very few people talk about the underlying data.
“For us to clearly communicate what it is that we’re doing, the implications of that and for people to understand how data is used, particularly in healthcare, is a really important conversation.”
