Skip to content

Dosage Window Protocol

Operating protocol — instantiates Effective-Input Delivery Assurance

Sets a standing acceptable range for a managed input — a floor for effect and a ceiling for harm — with codified rules for correcting back into it.

Dosage Window Protocol is the pre-specified operating envelope for an input: it must stay between a floor (below which it fails to have the intended effect) and a ceiling (above which it causes off-target harm), and the protocol writes down the if-then rules for moving back into the band when the input drifts. Its defining move is that it codifies rather than discovers — it does not work out where the right level is (that is the dose-response curve) or tune the level to one subject (that is calibration); it fixes the acceptable band and the correction rules so that different operators, shifts, and sites all act the same way. The output is a standing document anyone can follow, not a bespoke judgment.

Example

A municipal water utility runs a Dosage Window Protocol for chlorination. The window is a free-chlorine residual of roughly 0.2–2.0 mg/L measured across the distribution network: enough at the far taps to keep a disinfectant residual (floor), not so much that it drives disinfection-byproduct formation past the regulatory limit or triggers taste complaints (ceiling). The adjustment rules are explicit: if the residual at the network extremity drops below the floor, step the plant dose up by a fixed increment; if byproduct precursors run high, cap the dose and add a booster station rather than pushing more chlorine in at the source. The effect is that a night-shift operator and a day-shift operator dose identically, over-and under-dosing become visible against a written band, and "add chlorine to taste" becomes a bounded, auditable operation.

How it works

The protocol is built from three fixed decisions: a floor taken from the minimum level that still produces the effect, a ceiling taken from the harm boundary, and a set of adjustment rules — triggers, increment sizes, and hard limits — for correcting excursions. What distinguishes it from its siblings is that it is the standing specification itself: the band and the rules are settled in advance and applied uniformly, so operating the input requires following the envelope, not re-deriving it each time.

Tuning parameters

  • Window width — narrow bands give tight control but more adjustment churn and more out-of-range excursions; wide bands are stable but let the input drift toward its edges.
  • Floor and ceiling placement — how much margin to leave below the harm boundary and above the point of ineffectiveness. Conservative ceilings trade some efficacy for safety headroom.
  • Adjustment increment and deadband — the step size for corrections and how far out of range you tolerate before acting. Big steps correct fast but overshoot.
  • Monitoring cadence — how often the controlled variable is checked against the window. Rarer checks let excursions run longer before they are caught.
  • Override policy — who may operate outside the window, under what documented conditions, and with what sign-off.

When it helps, and when it misleads

Its strength is consistency and auditability: it strips out operator-to-operator variance and makes both over- and under-dosing legible against an explicit band, which is exactly what a one-off judgment cannot provide at scale.

Its weakness is that a single standing band can be wrong for the tails. A population-level window ignores the unusually sensitive subject or segment for whom the shared floor or ceiling does not apply, and the band is often set from convenience or precedent rather than fresh evidence. The classic misuse is to treat the window as physics — "anything inside is safe" — rather than as a bounded bet with margins, or to widen it simply to reduce nuisance alarms. The discipline that keeps it honest is to derive the bounds from the dose-response evidence and the real harm boundary, revisit them as data accumulates, and keep an explicit exception path for outliers rather than forcing everyone through the average case. This is the therapeutic window[^twindow] generalized from a drug to any managed input.

How it implements the components

Dosage Window Protocol fills the standing-specification slots — the band and the rules, held in the abstract:

  • target_range — its core artifact is the explicit acceptable band: the floor-to-ceiling range the input must stay within.
  • adjustment_rule — it codifies the triggers, increments, and limits for correcting the input back into range when it drifts.

It does not derive the band from response evidence (managed_input_or_exposure / sensitivity_profile, handled by Exposure Dose Curve), tune it to one subject (response_monitoring / side_effect_signal, handled by Medication Dose Calibration), or drive the input toward its floor (minimum_effective_input, handled by Minimum Effective Dose Review); it holds the standing envelope those inform.