OpenPrescribing: normalised data and software tool to research trends in English NHS primary care prescribing 1998-2016

Abstract

Objectives: We aimed to compile and normalise England's national prescribing data for 1998-2016 to facilitate research on long-term time trends and create an open-data exploration tool for wider use.

Design: We compiled data from each individual year's national statistical publications and normalised them by mapping each drug to its current classification within the national formulary where possible. We created a freely accessible, interactive web tool to allow anyone to interact with the processed data.

Setting and participants: We downloaded all available annual prescription cost analysis datasets, which include cost and quantity for all prescription items dispensed in the community in England. Medical devices and appliances were excluded.

Primary and secondary outcome measures: We measured the extent of normalisation of data and aimed to produce a functioning accessible analysis tool.

Results: All data were imported successfully. 87.5% of drugs were matched exactly on name to the current formulary and a further 6.5% to similar drug names. All drugs in core clinical chapters were reconciled to their current location in the data schema, with only 1.26% of drugs not assigned a current chemical code. We created an openly accessible interactive tool to facilitate wider use of these data.

Conclusions: Publicly available data can be made accessible through interactive online tools to help researchers and policy-makers explore time trends in prescribing.

Keywords: national health service (nhs); prescribing; primary care.

Figure 1

BNF code normalisation process flow chart describing how drug names were matched to the current BNF. Examples of matches at each numerated stage are given in table (bottom), with code/name changes in red. The first example demonstrates a drug which was matched to the current BNF through an exact match by name but had moved from Chapter 4 to Chapter 2. For other types of matching, care was taken to avoid mistakenly matching to similar (but distinctly different) drugs across chapters/sections. BNF, British National Formulary.

Figure 2

Screen shots from Trends tool, showing items per 1000 population and inflation-corrected costs per 1000 population for selected drugs. Product names are in parentheses. (A) Prescribing trends for four selected antipsychotic chemicals following the safety alert on thioridazine. Full dashboard available at https://public.tableau.com/shared/XX7DTWSG2?:display_count=yes. (B) Prescribing trends for selected antidepressant chemicals. Full dashboard available at https://public.tableau.com/shared/72SJGGP89?:display_count=yes. (C) Prescribing trends for all chemicals within the Paragraph of Nicotine Dependence (smoking cessation medications). Full dashboard available at https://public.tableau.com/shared/6BW9J5RJB?:display_count=yes.

Figure 3

Screenshots from Trends tool, showing inflation-corrected costs per 1000 population (A) by Chapter, and (B) by Section for Chapter 2 (Cardiovascular System).