Screening, KPI and QTL dashboards
Trials often miss their recruitment targets. Feasibility estimates tend to be optimistic, and the plan is usually a single finish date. This dashboard shows how recruitment can be planned and monitored against published evidence instead.
It is set inside epiClinical, a fictional contract research organisation (CRO) running six synthetic studies. The company, studies, sponsors, sites and patients are invented. The evidence behind each number is real, and each number shows where it comes from and how far it can be trusted.
What it covers
A study design questionnaire covers the risk factors the research identifies, such as eligibility, visit burden and competing trials. It shows how each answer changes the recruitment rate in each country.
A planner simulates thousands of versions of each study. Each site opens at a different time and recruits at its own uncertain rate, so the result is a range of finish dates. Each range comes with a stated confidence. You can set the sites yourself, or set a deadline and let the planner find the fewest sites that will meet it.
Each study has a page showing which sites are ahead of, on or behind their expected rate. It also shows where each site loses patients between first approach and randomisation. Sites that are behind get suggestions graded by the strength of their evidence, and each suggestion waits for a person to accept or dismiss it. And the page works out how many patients a month the study must screen to finish on time.
A page of key performance indicators (KPIs) and quality tolerance limits (QTLs) tracks the measures the literature most strongly links to success. They are how quickly sites open, how long a study takes to enrol its first patient, enrolment against plan and whether the data the result depends on are complete. It also shows screening benchmarks from referral to randomisation, and which corrective actions have evidence behind them.
Access
The dashboard is open to view. If you'd like to talk about patient recruitment, KPIs or QTLs, get in touch.
How it is built
It runs inside this site, on the same Next.js codebase. The statistical model is written and tested in TypeScript, with more than 300 automated tests, including checks that it reproduces published results. The design uses EconoType, a design kit I built in the style of The Economist.