WORKS
Judge it by what runs.
Claims Data × Prescription Gaps
Calculates the days between consecutive prescriptions for each patient and shows the distribution by drug, disease and period, including the share of gaps over a threshold.
Claims Data × Co-prescription
Pick a disease and a target drug to see which drugs were prescribed on the same day, scored by Support, Confidence and Lift. Teams can explore it themselves without writing SQL.
Economic Evaluations × Auditable Models
A reviewer approves the assumptions a paper leaves unstated, and an R model is generated from the locked specification alone. It then shows which assumption explains any gap from the published result.
Cost-Effectiveness Model × Sensitivity Analysis
Runs a two-arm cost-effectiveness model in the browser. Switch survival distributions and review one-way, probabilistic and scenario analyses as tornado charts and acceptability curves.
Past Projects × Knowledge Search
Search past analysis plans and programs by chat to find definitions and analysis logic. Each answer shows the source passage and original file, easing the load of internal questions.
A tool that turns an analysis listing, with one column per line of therapy, into a Sankey diagram of how patients move between treatments; click a node or band to narrow the cohort. Prescription data never leaves the browser, and the AI is given only column names, regimen names and counts. The public version lets you try the AI suggestions on fictitious rheumatoid arthritis data.
A tool that checks page breaks and figure colours in TLFs (tables, listings and figures) that run to hundreds of pages, without the PDF ever leaving the browser. It flags pages with unusual blank space at the bottom or right, and colour pairs that are hard to tell apart in black and white or under different types of colour vision. A dummy TLF with fictitious data is ready to try.
A dashboard you operate by talking to it, built on Japan's DPC open data (FY2022–FY2024, 6,448 hospitals). Claude only chooses what to display; every number comes straight from pre-aggregated data. It is a public-data demo of a tool we are building in support work for a pharmaceutical company.
A workflow that rebuilds the TrialMind paper's procedure with Claude Code: subagents handle PubMed search, abstract screening and full-text data extraction, and the results are checked with the paper's own metrics. The output can be explored in a sample report.
An app for exploring 80.7 million rows of US prescription data in plain language, without writing SQL. Ask a question and Claude writes and runs the SQL on BigQuery, then returns maps, charts and an interpretation.
Calculates the days between consecutive prescriptions for each patient and shows the distribution by drug, disease and period, including the share of gaps over a threshold.
Pick a disease and a target drug to see which drugs were prescribed on the same day, scored by Support, Confidence and Lift. Teams can explore it themselves without writing SQL.
A reviewer approves the assumptions a paper leaves unstated, and an R model is generated from the locked specification alone. It then shows which assumption explains any gap from the published result.
Runs a two-arm cost-effectiveness model in the browser. Switch survival distributions and review one-way, probabilistic and scenario analyses as tornado charts and acceptability curves.
Search past analysis plans and programs by chat to find definitions and analysis logic. Each answer shows the source passage and original file, easing the load of internal questions.