QCovid: risk calculator
Designing accessibility into a clinical risk calculator
In short
QCovid pairs a clinical risk model, built by a University of Oxford research team for NHS Digital's shielding programme, with a public-facing website. The model was designed to give both doctors and patients a shared, nuanced understanding of individual risk — which meant the interface had to work for a genuinely mixed audience, not just clinical staff. My part was that interface: designing qcovid.org to GDS patterns and WCAG AA standards, and testing it with NHS workers. I didn't build the model itself — the two need to be kept separate, and I'm careful to do that whenever this project comes up.
Designing for a mixed, non-specialist audience
A tool meant to be read by both doctors and patients couldn't assume clinical literacy on one side or design sophistication on the other. Following GDS patterns and WCAG AA gave the interface a baseline that worked regardless of who was reading it — plain language, predictable structure, and accessibility as a starting point rather than an add-on.
Testing with the people who would use it
Usability testing was carried out with NHS workers. One finding shaped the interface directly: the Start test button needed to stay available without users scrolling back to the top of the page or hunting for the action, so I kept it close to hand throughout the flow. Testing and review of the site also led to a mandatory warning being added to the interface, so people understood what the score could and couldn't tell them before they read a result into it.
Result
The QCovid model was used by NHS Digital as part of the shielding programme, helping identify around 1.7 million additional people as high risk and prioritising them for vaccination (February 2021). That result belongs to the clinical model and the research team behind it — my contribution was the design and usability testing of the interface people used to reach it.
Closing principle
In a regulated, high-stakes context, the interface's job isn't to make a result feel more certain than the model behind it. It's to make sure the warnings, the limitations and the next step are as easy to find as the number itself.
QCovid pairs a clinical risk model, built by a University of Oxford research team for NHS Digital's shielding programme, with a public-facing website. The model was designed to give both doctors and patients a shared, nuanced understanding of individual risk — which meant the interface had to work for a genuinely mixed audience, not just clinical staff. My part was that interface: designing qcovid.org to GDS patterns and WCAG AA standards, and testing it with NHS workers. I didn't build the model itself — the two need to be kept separate, and I'm careful to do that whenever this project comes up.
Designing for a mixed, non-specialist audience
A tool meant to be read by both doctors and patients couldn't assume clinical literacy on one side or design sophistication on the other. Following GDS patterns and WCAG AA gave the interface a baseline that worked regardless of who was reading it — plain language, predictable structure, and accessibility as a starting point rather than an add-on.
Testing with the people who would use it
Usability testing was carried out with NHS workers. One finding shaped the interface directly: the Start test button needed to stay available without users scrolling back to the top of the page or hunting for the action, so I kept it close to hand throughout the flow. Testing and review of the site also led to a mandatory warning being added to the interface, so people understood what the score could and couldn't tell them before they read a result into it.
Result
The QCovid model was used by NHS Digital as part of the shielding programme, helping identify around 1.7 million additional people as high risk and prioritising them for vaccination (February 2021). That result belongs to the clinical model and the research team behind it — my contribution was the design and usability testing of the interface people used to reach it.
Closing principle
In a regulated, high-stakes context, the interface's job isn't to make a result feel more certain than the model behind it. It's to make sure the warnings, the limitations and the next step are as easy to find as the number itself.
