SH24 is more than just a bespoke website or IT system, something which we find ourselves often reminding people about. Nonetheless we still need a suitable platform to build the digital service on.
There were a number of backend and frontend platform options available to us to host the SH24 service but we were acutely aware that by selecting one before the other may prevent us from developing the service we want to provide in the future (and prematurely lock us into a technology – even during prototyping). There was also a political undercurrent of uncertainly around platform choice, not so much within our team, but amongst some of SH24’s other stakeholders. Statements such as ‘the electronic patient record (EHR) platform needs to be rock solid to protect patient data’ and ‘we don’t want to reinvent the wheel’ were often muted.
In layman’s terms, we were presented with four options:
- Rent an mainstream existing platform (such as Microsoft HealthVault)
- Rent a propriety existing platform
- Use an open source platform (such as Indivo)
- Build our own
To help us decide we to commission an independent, impartial review to appraise us of the technology platforms available to us (principally relating to the storage and handling of data).
In November 2013 we commissioned an agency to:
- Consider the characteristics and behaviours of both backend and frontend platforms against the working capabilities and functionality we sought (which we had a growing map of)
- Challenge, validate and add to our working considerations.
On completion of their review, the external agency recommended that SH24 build a custom platform.
This recommendation sparked an interesting debate in the team, especially as the Government Digital Service (GDS) advise to work with existing software (avoiding reinventing wheels). GDS do however caveat their advice with ‘where possible’ – which we interpreted to mean; as long as your choice does not compromise experience and/or functionality. In our case we thought it would force compromises, and we also had to consider our IP, our limited dataset, the speed we’d like to work at and above all, our highly iterative approach, meaning there was no natural ‘fit’.