Chapter 2: decision rule for risking no more than a two-week hospital stay

To obtain the relevant information, we would simulate the possible outcomes from administering treatment on testing positive at the four gestational latencies, keeping track of the length of hospital stay in each case. Since the length of stay is related only to the time of inappropriate treatment, we can simply multiply the probability of receiving treatment incorrectly by the length of stay at each treatment time (re-using the variable definitions from the previous simulation):

FAs=[0 0 0 0]; %initialize list of simulated false alarms at each month

Hs=[0 0 0 0]; %initialize list of simulated hits at each of 4 months

for M=1:4,

for n=1:REPS,

if randi(URNall(M))>URNhm(M), %disease absent

FAs(M)=FAs(M)+(randi(URNcf(M))>CR(M));

else Hs(M)=Hs(M)+(randi(URNhm(M))>MISS(M)); end, end, end

.8*[5:10:35].*[FAs./(FAs+Hs)]

which shows that the probabilities of incorrectly administering treatment based on a positive test result are p(~D|+,i)=[0.954 0.946 0.939 0.935], and the only times that meet the criterion are at 3 and 4 months.