Quarantine threshold in the SIR model (source code)

= Quarantine threshold in the SIR model
{title2=$I_*+S_*-\frac ar\log S_*=I_0+S_0-\frac ar\log S_0$}

Consider the closed <SIR model> in count variables, with $\dot S=-rSI$, $\dot I=rSI-aI$, transmission coefficient $r>0$ and <recovery rate> $a>0$. An ideal intervention stops all further transmission at the first upward crossing of $I=I_*$. This is a mathematical contact-removal rule; its instantaneous perfect effect is an assumption of the model.

Before intervention,
$$
\frac{dI}{dS}=-1+\frac{a}{rS},
$$
so $I+S-(a/r)\log S$ is a <first integral of an ordinary differential equation>. At the trigger, the remaining <susceptible individual> count $S_*$ is a <root of an equation> satisfying
$$
I_*+S_*-\frac ar\log S_*
=I_0+S_0-\frac ar\log S_0.
$$
This is a <transcendental equation>, with $S_*>0$, rather than a <polynomial equation>.

For an initially growing outbreak with $I_0>0$ and $S_0>a/r$, the unmodified <SIR model> reaches its infectious-count peak at $S=a/r$, with
$$
I_{\rm peak}=I_0+S_0-\frac ar-\frac ar\log\left(\frac{rS_0}{a}\right).
$$
A trigger with $I_0<I_*\le I_{\rm peak}$ is therefore reachable on this growth branch. The function $S-(a/r)\log S$ is strictly increasing for $S>a/r$, so the trigger equation has exactly one solution in $a/r\le S_*<S_0$. A threshold above $I_{\rm peak}$ is never reached; choosing another mathematical root would not describe the first upward crossing.

Once the assumed intervention stops transmission, recovery can continue but $S$ no longer changes. The number of new infections among those initially <susceptible individuals> is thus $S_0-S_*$. The initial $I_0$ is excluded from that count. An intervention with residual transmission, or a model with unobserved exposed individuals, requires a different final-size calculation.