11 January 2023 by Mat Oram
A university professor once told me “don’t keep polishing the cannonball, but do get the calibre right”, an expression I fundamentally agree with. After all, when it comes to data, there is a distinction to be made between haplessly trying to improve poor-quality data and finding insight with the information you have available. The reality is that you can find insights in all data – there is always something to learn and understand from it.
Depending on where you are on your digital maturity journey, this could be as simple as learning where you have gaps in information, or as advanced as using data to predict future patient demand. Naturally, most trusts sit somewhere in between, meaning the opportunity to utilise data to create operational efficiencies is enormous across the NHS.
So, what do we mean by data quality? There’s a lot to read about ‘data quality’ online, with various models used to describe different aspects of the term. But for most trusts, when trying to establish the quality of their procurement data, we recommend starting with the basics: can you get order and invoice data out of your finance system? If you can’t, then we’ve both got real problems, and you are no doubt in breach of the fiduciary duty to analyse your data and prove that you are getting value for money for the taxpayer on what you’re spending if it’s locked in your systems.
But assuming you can send us data, we can begin to look at a few metrics; the first things we do is look at (1) completeness, (2) orderliness and (3) accuracy.
By completeness, we check to see whether or not all fields have been provided. If fields have not been provided – for example, you are unable to provide us with a requisitioner name – we make sure you are aware of the impact this will have on your analysis, or work with you to understand how you might be able to obtain it.
Orderliness is about format and structure. If a field is not in the format we would typically expect it to be – for example, a date field – then we can transform it into a format that can be read. Normally, we can transform the data, but the challenge is that every trust we work with has its own nuances, which we need to understand to create the logic to correct it.
If we multiply your unit price by your quantity, do we get the same total? It is surprising how many times this isn’t the case. So, what’s the truth? Is it the total, the unit price, or the quantity?
Once we have established that a file is ready to process, we then look at (4) consistency and timeliness and (5) auditability.
In short, we want to know how quickly you can provide consistent data. Of course, if you’re an NEP customer or work under the umbrella of NHS Wales, it is all automated and we will receive all of your files on the first day of each month. But if you are not part of a shared service, you need to send the files to us, which can take up to three weeks if it relies on an analyst in a procurement team. And if a person goes on holiday or leaves their role, it is often a struggle getting the data. We therefore advise all of our customers to automate as much as possible by working with shared service providers and 3rd party solutions.
Finally, we look at auditability. Are you sending us new POs and any lines that have been updated in the past month? If not, we see the reality and analytics slowly diverge as the change in POs are not being fed through to us. On occasion, we complete mass updates for trusts that are struggling, overwriting twelve months worth of data with a new, fresh single export.
Once the data is ready to upload, we are then in a position to pass it through our Personal Identifiable Information (PII) process. Here, we look into product descriptions and supplier names for a host of personal information – including names, bank details and personal addresses.
This enables us to report on everything we see, and feed the PO numbers and line items back to the trust. PIIs are redacted with “xxxxxx”, meaning that you can see the pricing intelligence, without the potential PII. In most cases, this is a surgeon’s name – or, at least, something that sounds like a name. For example, “Welsh cakes” in Wales gets redacted to “xxxxxx cakes” (better safe than sorry!). Beyond this, we also spot when the same cost centres or users are inputting certain data into the system – Orthotics is a major culprit due to the nature of the product (customised to an individual).
Want to read more? Check out part two of this feature to find out what happens to the raw data once it is within our systems.
