Prescribing Data: Using the Dictionary of medicines and devices

Recently, we’ve been experimenting with integrating the Dictionary of medicines and devices (dm+d) into our prescribing data. dm+d is the standard dictionary for the medicines and devices used across the NHS, and it contains codes and descriptions for these medicines.

There are several benefits to using dm+d; the most useful side-effect is to allow us to show user-friendly names for drugs. The canonical names for drugs in the NHS prescribing data are sometimes very hard to read. They are taken from the NHS’ own version of the British National Formulary, which uses heavily truncated names, full of abbreviations, so they can fit within an arbitrary 15-character limit.

By mapping the BNF codes supplied in prescribing data to the dm+d, we can call these presentations by their full name. When we do this, a product like Hydrocort Sod Succ_Inj 100mg Vl + Dil becomes the much more readable Hydrocortisone sodium succinate 100mg powder and solvent for solution for injection vials.

In addition to getting user-friendly names, the process of mapping prescribing to the dm+d gives us further metadata about each presentation. For example, we can identify the list price, the Drug Tariff section, or the different pack sizes available, or the route (for example, the route grocerysolid allows us to restrict analyses to things like biscuits and bread rolls).

One of these fields identifies if a product is valid to be prescribed in primary care, so we took a quick look at if any of these are being actively prescribed despite this being impossible. The answer is yes: there is a handful every month. For example, the most common such product is Generic Earex ear drops, which was prescribed four times in March 2017, despite not having existed as a product in the real world for some time.

In the cases we were able to check easily, these were all paper prescriptions (rather than ones submitted electronically), and therefore probably represent a coding error. The NHSBSA, who are responsible for assembling this data, aim for a 99% accuracy rate, and it would be natural to expect more inaccuracy in the scanned paper prescriptions.

We’ll be launching features that use the readable product names in the next couple of months. If you’d like to do analyses using any of the more granular classification data (like route, bioequivalence, etc) then get in touch.