A national health survey must estimate the prevalence of undiagnosed diabetes among 2 million adults. A two-phase design is proposed:

Phase 1: Simple random sample of 5,000 people; all given an inexpensive screening test ($20 per person)

Phase 2: All 1,200 Phase-1 positives receive a gold-standard diagnostic test ($150 per person), plus a random 5% sample of the 3,800 Phase-1 negatives (190 people)

Total cost: $5,000(20) + [1,200 + 190](150) = 100,000 + 209,000 = 309,000$

Compared to giving ALL 5,000 people the gold-standard test ($750,000), what is the PRIMARY statistical advantage of two-phase sampling?