An arXiv preprint by Nanini et al. presents the first open reliability atlas for Pulse, an actively maintained open-source adult-physiology simulator. The authors note that maintainers had reported no formal sensitivity analysis of operational inputs. Using Morris elementary-effects sampling, they ran 1,280 simulations spanning seven inputs, four severity scenarios, and male and female bodies.
Sensitivity is highly skewed: three inputs carry a median 93% of each output's total. A four-flag classifier identifies failure modes, two of which are operator-actionable and one sex-stratified. Tidal-volume oscillation appeared in 20.6% of runs and is consistent with patient-ventilator dyssynchrony; locking the VC-AC trigger to -2 cmH2O collapsed that variability. At FiO2 = 1.0, transient hyperoxia caused 120-second runs to underestimate plateau PaO2; 300-second healthy-lung reruns gave a median 643 mmHg.
Solver divergence occurred in 20 of 1,280 runs, all in an input-extreme male regime; reducing the male blood-volume baseline to the 4,082 mL floor eliminated it. The authors release runs and flags openly so users can identify regions needing checks before training on Pulse output. Since this digest draws on a single preprint, the findings are not independently corroborated here.