Yes, it is, if its sufficiently random.

In fact, 385 people is enough for 350M people, if sufficiently random.

n = (ZZp(1-p))/(EE)

n = sample size

Z = confidence (studies are usually 95%)

p(1-p) = variance of binormal proportion, usually set p=.5

E = margin of error (is 1.00-.95 or .05)