The Health Service Journal reports on some eyebrow-raising expectation management by NHS England’s chief data and analytics officer, who admitted recently that the controversial £330 million Federated Data Platform (FDP) is ‘not pretty, it’s not that spectacular’.

But that wasn’t the most interesting thing she said that day. Ms Tang added: “It’s really hard convincing Treasury that 30 per cent of your business case should be around deployment.’

It is heartening that she tried. No matter how spectacular (or otherwise) a data platform may be, if the data feeding into it is poor quality, then the data coming out will be poor quality too. And when advanced analytics are deployed, or data is compared from different sources, then the problem is even worse.

The usual approach to feeding data into the FDP, or indeed any other system, starts with a specification of the data tables required. Then information professionals in each NHS trust write queries to deliver those data tables. There is nothing wrong with this approach in theory. However in practice it leads to different trusts delivering different things, because of differences in their underlying data systems and how they are configured and used.

In our experience, much, much more care is needed to achieve fully consistent data. We typically put source data through four transformations before really getting started, followed by a further nineteen transformations (with configurable localisations and other business rules injected along the way), to achieve proper standardisation.

If that sounds like overkill, it isn’t. The software exists and is ready to deploy. The Unified Data Layer (UDL) that it creates can be reliably used for analytics, patient pathway management, data warehousing, or indeed to feed the FDP. It enables continuity even when source systems are replaced or upgraded, and allows like-for-like data comparisons between trusts – even between sectors – across an ICB or more widely.

Now that NHS England are asking all trusts to confirm they will adopt the FDP, it may as well have good enough data feeding into it to be useful. So why not drop us a line? Whether you use the tools provided with the FDP, or something else, the first step is to have consistent and standardised data.