One Drug Database update

We recently told you about the ‘One Drug Database’ project we’re undertaking and our plan to provide the new dataset in January 2020 (November 2019 data).

It looks like the project will be delayed by one month. We’re now aiming to go live in February 2020 (for December 2019 data).

As we’re delaying our project slightly, we’ve been able to consider other improvements to how we provide prescribing information.

Currently, the NHS releases detailed information on prescribing of medications via two datasets, both of which are sourced from NHSBSA data:

• Detailed Prescribing Information (DPI), released by NHS Business Services Authority via the ISP portal.
• Practice Level Prescribing in England (PLP), released by NHS Digital on gov.uk website.

There is inconsistency between these two datasets, in terms of the level of organisational detail available and the volume of data published in each.

From February 2020, we intend to bring together both of these sources into a single unified dataset, which will be released through the NHS BSA Open Data Portal.

The benefits of this will be to provide end users with a single, accurate and comprehensive source of prescribing information, which will increase consistency and also improve data accessibility.

To do this, we intend to make the following changes to the datasets.

The new dataset will be based on the Detailed Prescribing Information dataset.

We will add the following fields from the PLP dataset, so that existing PLP users can continue to make use of the detail in that data: 

• BNF Chemical Substance Code
• BNF Chemical Substance Description
• Practice Address and Postcode
• Total Quantity  - quantity x items
• Unidentified – unidentified practices were not included in the PLP dataset, so this flag can be used to filter out unidentified practices to provide consistency within the data for PLP users

 You can view a test file containing March 2019 data (ZIP: 144MB) for mapping purposes.

(The ZIP file contains one Excel file which is 4.7GB).

Measures

There are no differences in Items, NIC or Actual Cost in the published and new datasets at a national, CCG or practice level.

There are some differences in Items, NIC and Actual cost at BNF presentation level, because of changes in the drug classification structure brought in by the One Drug Database.

Due to the coding changes brought in by the One Drug Database, a number of BNF presentation codes that featured in the original published data for March 2019 have been removed.  The items that were associated to the removed BNF presentation codes have been mapped to existing BNF presentation codes in the One Drug Database structure.  This has caused the values for some BNF presentation codes to change in the new dataset.

Additionally, a number of items which were assigned to unspecified drugs in the original dataset have now been specified to a BNF presentation code in the new dataset.  This means it’ll not be possible to map these items between the published and new datasets.

The overall values for items, NIC and Actual Cost reported in each dataset remains equal and unchanged.

We can quantify this as follows.

There were 22,643 BNF presentation codes in the data published in March 2019.

Of these, 428 have been removed and do not feature in the new dataset.

487,619 items were associated with these BNF codes.

456,848 of these items have been mapped from removed BNF codes in the published data to existing BNF codes in the new dataset

There are 216 new BNF presentation codes in the new dataset, that did not exist in the published data.

30,771 items are associated with these new BNF presentation codes – these will relate to drugs that were unspecified in the original published data which have since been mapped to a BNF presentation code.

There are now 22,431 BNF presentation codes in the new dataset: (22,643 – 428 + 216 = 22,431).

Summary

There are three drug names with underscores listed below which are still included in the dataset but these will be corrected once the data is published.

Our plan is to decommission the PLP and DPI feeds and switch over to the new dataset on the Open Data Portal by February 2020. The historic data will be re-run back to January 2014 and will be available in six yearly files which will also be available via the open data portal.