Pharmaceutical Press produces the most comprehensive and trustworthy range of independent pharmaceutical information in the world. Its main publishing focus is the design, manufacture and therapeutic use of medicines, building a body of knowledge that can be relied upon by pharmacists, health and social care professionals worldwide. Its flagship publication, the British National Formulary (BNF), is produced in multiple formats and used throughout the National Health Service.
Pharmaceutical Press is the publishing arm of the Royal Pharmaceutical Society. The Society has been publishing information for pharmacists since 1841, and today Pharmaceutical Press has more than 150 titles in its comprehensive range of print and digital resources.
A key consideration in prescribing, an interaction occurs when one drug affects the activity of a second drug when both are administered together. The interactions section of the BNF is thus a critical component of the publication. For some time, the way BNF interactions were created and managed had been constrained, both in terms of the content structure and in the production system. Editors lived with an old system that for a time was available only on one pc, and which consequently carried a large business risk. However, the internal technology team was consumed by business-as-usual tasks and other competing priorities. When a new contract with NICE (National Institute for Clinical Excellence) for the supply of BNF came into force, it included an obligation to refresh the interactions section, and so the need for a new way to produce interaction content became paramount.
Pharmaceutical Press asked 67 Bricks to build a new system for inputting, managing, and outputting interactions content. It was key to the success of the project that 67 Bricks spent time gaining a thorough understanding of the content. Only by doing that would it be possible to provide the most appropriate solution to Pharmaceutical Press’s needs.
Following a one-day workshop where 67 Bricks facilitated scoping and prioritising the main features of the system, the build was undertaken as an agile project with close collaboration between the teams.
The resulting system is finely-tuned to offer as much help as possible to the human editor when inputting the highly structured data, because any errors could be detrimental to patient care (Figure 1). There are many innovative touches that add up to an impressive series of time-saving and error reducing techniques. For example, the drug interaction input screen is highly fielded with copious use of controlled vocabularies. There are also parameterised fields that have a different set of controlled vocabularies to use within the parameters.
As the user fills in the fields, the system builds up a human readable statement about the interaction. A live preview of the human readable statement is created on the fly while the user inputs the fields and parameters (Figure 2). Not only does this provide essential feedback to the user to check the sense of the information they are constructing, but it also makes the writing completely consistent.
The system automatically interprets terms selected from the vocabularies into the right form (e.g. tense, singular/plural) needed for grammatical correctness in the human readable statement. This saves a lot of time and means that the vocabulary list is shorter and easier to use. Changing a term in one of the vocabularies causes all instances of that term throughout the database to be updated instantly.
There are currently around six different exports that mainly use the human readable statement, including the printed BNF, several online versions and the BNF app. In future the underlying data will be exposed by API to a variety of recipients for incorporation into external systems such as clinical decision support.