Table 8.

Patient welfare with observed misweighting vs. in simulations with no misweighting

	False positive rate of 4%
	Actual testing decisions	No misweighting, simulation without cross validation	No misweighting, simulation with cross validation
	(1)	(2)	(3)
Description of results:
Percent of patients tested	3.8%	4.3%	4.3%
Number of patients tested	71314	81410	79734
Test yield among tested patients	7.0%	9.2%	8.6%
Number of positive tests detected	5019	7526	6872
Welfare analysis:
Total financial costs of testing ($ millions)	35.6	45.2	43.4
Total medical cost of testing ($ millions)	8.5	12.4	11.7
Total medical benefits of testing ($ millions)	57.5	106.8	96.7
Net benefits of testing ($ millions)	13.5	49.1	41.6
Total (financial + medical) costs per test ($)	618.9	707.8	690.8
Total benefits per test ($)	806.9	1311.3	1213.1
Net benefits per test ($)	188.1	603.5	522.2

Notes: We compare testing behavior and social welfare under the observed physician weighting of patient risk factors (in column 1) to simulated behavior assuming that physicians target testing to patients with the highest expected probability of a positive test based on observable demographics and comorbidities (in column 2). The simulated results in Panel B allow τ_d to follow the estimated posterior distribution (i.e. without correcting for overtesting).