Dynamics
Most datasets give you a photograph. These give you the film. Every dataset in this category follows the same patient — or the same synthetic cohort standing in for one — across time, through a state that keeps changing: a disease progressing, an organ recovering, a drug clearing the body hour by hour.
The underlying mathematics differs by dataset — continuous-time multi-state Markov models for disease progression, coupled ordinary differential equations for physiology, dense pharmacokinetic curves on a half-hour grid — but the discipline is the same everywhere: every transition, every trajectory, and every rate of change has a stated source, whether that's a published clinical constant, a calibrated model, or an openly declared assumption.
What you'll typically find here
- Disease-progression cohorts built as multi-state models — patients moving through real clinical stages over months or years, including backward transitions where a real treatment genuinely reverses damage, and no backward transitions where it structurally can't.
- Physiological trajectories — vital signs, organ function, and lab values evolving hour by hour under a mechanistic model, not resampled from a static table.
- Counterfactual timing scenarios — the same patient under different treatment-start times, so you can study not just what happened, but what would have happened.
Every dataset here ships with a passport that states plainly what's confirmed against the literature and what's a modelling choice — including the places where our own external validation checks did not fully pass, because a partial validation reported honestly is worth more than a clean one that wasn't checked.