A contingency plan needs a distribution, not one forecast

Kim Lowe’s NASA article, published on 16 September 2026, explains how the Integrated Medical Model supports decisions before a crewed mission launches. The tool draws on decades of flight records and clinical research, then uses Monte Carlo simulation to model thousands of potential mission paths. Inputs include crew size, mission duration, planned spacewalks and the spacecraft environment.

The model evaluates more than 100 medical conditions and produces several types of output: projected medical events, an index of crew functional capacity, quality-adjusted time lost, probabilities of evacuation or loss of life, and expected use of medicines, consumables and tools. The resource output matters because a spacecraft has strict mass and volume limits; planners cannot carry every possible response in unlimited quantities.

A supporting NASA overview makes the assumptions visible. The model’s standard baseline reflects International Space Station conditions, including regular resupply and the possibility of a rapid return. It also assumes correct diagnosis and effective treatment. If a required resource is exhausted during a simulation, the condition is left untreated. NASA cautions that results should be weighted according to their credibility for the decision being made.

Dunes Bureau interpretation: a remote mine, offshore facility, field-service network or critical spare-parts operation can use the same decision structure without treating the medical model as an industrial benchmark. Define a range of demand, delay and failure paths; simulate how stock and capacity are consumed; measure service loss and recovery options; then choose the resource bundle that meets a stated resilience threshold. The useful output is a distribution of consequences, not one precise forecast.

This is an operating principle, not evidence that Monte Carlo modelling will reduce cost or improve reliability in a particular company. The quality of the decision still depends on the event rates, dependencies, treatment or repair assumptions and scenario coverage supplied to the model. A buyer should therefore review those inputs, test sensitivity to uncertain values and record which decisions the model is—and is not—credible enough to support.