Patients arrive at random and wait for a free bed. Triage serves the most urgent first. Adjust arrivals, beds and treatment time, then trigger a surge and see which patients wait.
Running check...
The science
Queueing model. Patients arrive as a Poisson process with rate λ (exponential gaps), wait in a queue and are treated by one of c identical beds with exponential treatment time of mean 1/μ (an M/M/c queue). The offered load is a = λ/μ (in "bed-equivalents") and the utilisation ρ = a/c. For ρ ≥ 1 no steady state exists: the queue grows without bound, which is why a surge so quickly overwhelms an emergency department.
Erlang C. With first-come-first-served service the probability that an arriving patient must wait is C(c,a) = [a^c/(c!(1−ρ))] / [Σ_{k<c} a^k/k! + a^c/(c!(1−ρ))] and the mean wait is Wq = C(c,a)/(cμ − λ). The page runs a hidden 150,000-patient simulation of a single-class FIFO case and compares it with this closed form. Little's law Lq = λ·Wq is checked live on the running department (the mean queue length versus the rate of leaving the queue times the mean wait).
Triage. With priority service (non-preemptive) the red (immediate) patients wait least, green (minor) wait most; the total work is the same, so priority redistributes waiting rather than removing it (Kleinrock's conservation law). Treatment time multipliers (red 2.2x, yellow 1.2x, green 0.7x) keep the average at 1x.
Simplifications. Exponential times, a single pool of beds (no separate resus bays, doctors, imaging or admission blocking), a fixed acuity mix of 10/30/60 percent, walk-outs with exponential patience, and a surge modelled as 3x arrivals for 3 hours. This is an educational queueing model, not clinical guidance and not calibrated to any real hospital.