Numerically solve delay ODE

Question

I am trying to numerically solve a simple delay ODE in mma for the first time. I think this is likely a simple fix.

I first tried to solve system as ordinary differential equation system for up to 𝑡=l. And then use these solutions as history functions for the delay differential equation system. I keep getting the error: NDSolve::dvnoarg The function na appears with no arguments

ClearAll["Global`*"]
c = 200; Q = 4; P = 1.92*10^-6; ea = 2.6*10^-6; ga = 
6.24*10^-8; l = 0.5; ba = 98; pr = 0.2;
sol0 = NDSolve[{
 r'[t] == pr (c - r[t]) - P r[t]/(Q + r[t]) (na[t] + ma[t]),
 na'[t] == P r[t]/(ea (Q + r[t])) na[t] - pr na[t] - ga 
 na[t] p[t],
 ma'[t] == ga na[t] p[t] - pr ma[t],
 p'[t] == -pr p[t] - p[t] ga na[t] ,
 r[0] == 1000, ma[0] == 0, na[0] == 100, p[0] == 1000},
 {r, na, ma, p}, {t, 0, l}][[1]];

sol = NDSolve[{
 r'[t] == pr (c - r[t]) - P r[t]/(Q + r[t]) (na[t] + ma[t]),
 na'[t] == P r[t]/(ea (Q + r[t])) na[t] - pr na[t] - ga 
 na[t] p[t],
 ma'[t] == ga na[t] p[t] - pr ma[t] - 
   Exp[-pr l] ga na[t - l] p[t - l],
 p'[t] == ba Exp[-pr l] ga na[t - l] p[t - l] - pr p[t] - 
   p[t] ga na[t] ,
 r[0] == 1000, na[t /; t <= l] == sol0[[2]], ma[0] == 0, 
 p[t /; t <= l] == sol0[[4]]},
{r, na, ma, p}, {t, 0, 100}][[1]];

Any suggestions are much appreciated.

Answer 1

I'm not sure the separate NDSolve to initialize the initial conditions is even necessary. If you just use your second NDSolve with the initial conditions of the first, it gives a warning NDSolve::ihist but seems to run:

sol = NDSolve[{
  r'[t] == pr (c - r[t]) - P r[t]/(Q + r[t]) (na[t] + ma[t]), 
  na'[t] == P r[t]/(ea (Q + r[t])) na[t] - pr na[t] - ga na[t] p[t], 
  ma'[t] == ga na[t] p[t] - pr ma[t] - Exp[-pr l] ga na[t - l] p[t - l], 
  p'[t] ==  ba Exp[-pr l] ga na[t - l] p[t - l] - pr p[t] - p[t] ga na[t], 
  r[0] == 1000, na[0] == 100, ma[0] == 0, p[0] == 1000},
  {r, na, ma, p}, {t, 0, 400}, AccuracyGoal -> 128][[1]];

LogPlot[Evaluate[{na[t], ma[t], p[t]} /. sol], {t, 0, 400}, 
 PlotRange -> All]

However note that I had to really crank up AccuracyGoal due to the very low levels that na[t] reaches. Maybe other parameter values would be less violent.