Epidemiology: The SIR model

Simulation of differential equations with turtle graphics using JSXGraph.

SIR model without vital dynamics

The SIR model measures the number of susceptible, infected, and recovered individuals in a host population. Given a fixed population, let [math]S(t)[/math] be the fraction that is susceptible to an infectious, but not deadly, disease at time t; let [math]I(t)[/math] be the fraction that is infected at time [math]t[/math]; and let [math]R(t)[/math] be the fraction that has recovered. Let [math]\beta[/math] be the rate at which an infected person infects a susceptible person. Let [math]\gamma[/math] be the rate at which infected people recover from the disease.

A single epidemic outbreak is usually far more rapid than the vital dynamics of a population, thus, if the aim is to study the immediate consequences of a single epidemic, one may neglect birth-death processes. In this case the SIR system can be expressed by the following set of differential equations:

[math] \frac{dS}{dt} = - \beta I S [/math]

[math] \frac{dR}{dt} = \gamma I [/math]

[math] \frac{dI}{dt} = -(\frac{dS}{dt}+\frac{dR}{dt}) [/math]

Example Hong Kong flu

• initially 7.9 million people,
• 10 infected,
• 0 recovered.
• estimated average period of infection: 3 days, so [math]\gamma = 1/3[/math]
• infection rate: one new person every other day, so [math]\beta = 1/2[/math]

Thus S(0) = 1, I(0) = 1.27E-6, R(0) = 0, see [1].

The lines in the JSXGraph-simulation below have the following meaning:

* Blue: Rate of susceptible population
* Red: Rate of infected population
* Green: Rate of recovered population (which means: immune, isolated or dead)

The underlying JavaScript code

var brd = JXG.JSXGraph.initBoard('box', {axis: true, boundingbox: [-5, 1.2, 100, -1.2]});

var S = brd.create('turtle',[],{strokeColor:'blue',strokeWidth:3});
var I = brd.create('turtle',[],{strokeColor:'red',strokeWidth:3});
var R = brd.create('turtle',[],{strokeColor:'green',strokeWidth:3});
            
var s = brd.create('slider', [[0,-0.3], [30,-0.3],[0,1.27E-6,1]], {name:'s'});
brd.create('text', [40,-0.3, "initially infected population rate (on load: I(0)=1.27E-6)"]);
var beta = brd.create('slider', [[0,-0.4], [30,-0.4],[0,0.5,1]], {name:'β'});
brd.create('text', [40,-0.4, "β: infection rate"]);
var gamma = brd.create('slider', [[0,-0.5], [30,-0.5],[0,0.3,1]], {name:'γ'});
brd.create('text', [40,-0.5, "γ: recovery rate = 1/(days of infection)"]);

var t = 0; // global

brd.create('text', [40,-0.2, 
        function() {return "Day "+t+": infected="+(7900000*I.Y()).toFixed(1)+" recovered="+(7900000*R.Y()).toFixed(1);}]);
            
S.hideTurtle();
I.hideTurtle();
R.hideTurtle();

function clearturtle() {
  S.cs();
  I.cs();
  R.cs();

  S.hideTurtle();
  I.hideTurtle();
  R.hideTurtle();
}
            
function run() {
  S.setPos(0,1.0-s.Value());
  R.setPos(0,0);
  I.setPos(0,s.Value());
                
  delta = 1; // global
  t = 0;  // global
  loop();
}
             
function turtleMove(turtle,dx,dy) {
  turtle.moveTo([dx+turtle.X(),dy+turtle.Y()]);
}
             
function loop() {
  var dS = -beta.Value()*S.Y()*I.Y();
  var dR = gamma.Value()*I.Y();
  var dI = -(dS+dR);
  turtleMove(S,delta,dS);
  turtleMove(R,delta,dR);
  turtleMove(I,delta,dI);
                
  t += delta;
  if (t<100.0) {
    active = setTimeout(loop,10);
  }
}

function stop() {
  if (active) clearTimeout(active);
  active = null;
}
function goOn() {
   if (t>0) {
     if (active==null) {
       active = setTimeout(loop,10);
     }
   } else {
     run();
   }

}

See also

• Epidemiology: The SEIR model
• Population growth models
• Lotka-Volterra equations

References

• http://en.wikipedia.org/wiki/Compartmental_models_in_epidemiology
• http://mathworld.wolfram.com/SIRModel.html
• http://www.cs.princeton.edu/introcs/94diffeq