At The DataLab we like to work in the open and share our insights for the whole community, so we can fix a problem for us, then share the solution, and help fix it for everyone. We think this is efficient for us and the whole system. In this blog our magnificent coder Peter Inglesby sets out some analysis he has done of changes to the NHS prescribing data we use. It took us a few hours to investigate these problems for our own purposes, then document our findings and solutions internally; it took an extra 20 minutes to share in this blog post and now we give it to you so you don’t have to replicate our effort and can maybe find some insights we have missed!
OpenPrescribing.net uses an NHS dataset called the Detailed Prescribing Information (DPI) dataset, which drives all the dashboards, tools, and analyses we produce. The NHS Business Services Authority (BSA) publishes the dataset monthly and it contains the number of items, quantity, and cost of each presentation prescribed by each practice.
In 2020 the BSA are changing the underlying database which produces this dataset. This is called the One Drug Database (ODD) project, and this will change some of the contents of the dataset. We wanted to understand the scale of the changes, so we’ve put together this notebook and summarised our key findings in this blog post. We believe that the DPI dataset is used by at least 500 other organisations or individuals, so we hope this analysis is useful to other users of the data.
What are the changes?
There are two changes that OpenPrescribing are particularly interested in.
Firsty, there will be a rationalisation of BNF codes. In particular, the ODD documentation states:
“BNF codes will change for products (drugs and appliances) where the pack size is currently shown as part of the naming convention.”
An example of where this has caused confusion in the past can be seen with the VMP Budesonide 100micrograms/dose nasal spray. BNF codes are supposed to correspond to products (ie VMPs and AMPs), and a product pack is supposed to have the same BNF code as its product. Once the new DPI data has been ingested, we expect a product pack to always have the same BNF code as its product.
We have also found that:
- 1501 presentations are affected by changed BNF codes.
- However, presentations “where the pack size is currently shown as part of the naming convention” are just a small proportion of the affected presentations. For example, presentations of more than one flavour are now combined to a single BNF code, and a number of branded appliances (Chapter 20) now make an additional appearance in generic form in Chapter 13. More details of these kinds of changes are in the notebook.
- Presentations affected by changed BNF codes account for 1.3% of items and 1.4% of net cost in November 2019.
- 1% of affected presentations account for 55% of items and 29% of net cost, and 10% of affected presentations account for 89% of items and 82% of net cost.
- The 10 presentations most affected by the changes (by items) are shown here, and the 10 presentations most affected by the changes (by net cost) are shown here.
Secondly, a long-standing oddity whereby the quantity field for special containers refers to the number of packs dispensed (rather than eg the number of ml dispensed) is being resolved.
We have found that:
- Presentations affected by this account for 1.5% of items and 3.2% of net cost in November 2019.
- Just 1% of presentations affected account for 44% of items and 31% of net cost, and 10% of presentations account for 87% of items and 84% of net cost.
- The 10 presentations most affected by changes, by items and by cost, are shown here and here, respectively.
We think that all of the changes being made are sensible and will make the DPI dataset more useful. Switching to a new database is always challenging and we wish the BSA’s technical team all the best!
We expect that the new changes will take place in March 2020 and, while we are well prepared, we do expect that it might take us a little longer to import the data and we might have to amend some measures that are live on the website. Thanks for your patience with this, we will keep you updated via our blog, twitter and newsletter but if you have any specific questions please get in touch via firstname.lastname@example.org