Our previous analyses of decision strategies in children 3–24 months with acute-onset fever ⩾,39°C and no evident bacterial focus of infection indicated that the risks of routine blood cultures (the unnecessary hospitalization and treatment of children who clear their bacteremia spontaneously) outweigh its benefits (the prevention of a few cases with major infectious sequelae). Because those analyses were based on parents' values for beneficial and adverse clinical outcomes, we wished to examine whether those values differed in physicians and, if so, whether the differences were sufficient to change the results of the decision analysis. Using a pre-tested linear analog utility (value) scale, we evaluated eight potential clinical outcomes in potentially bacteremic children by surveying 121 parents of healthy 324-month-old children attending a private pediatric group practice and 57 attending physicians of a tertiary-care children's hospital emergency room. Utilities were based on a 0–1 normalization, where 0 is the utility of the worst outcome (meningitis or other major bacterial infection, plus venipuncture), and 1 the utility of the best outcome (complete recovery without venipuncture or hospitalization), and were analyzed using a recently developed statistical model of utility. The majority of parents and physicians combined the imputed components of the outcomes (disease, pain of venipuncture, and stress of hospitalization) in a nonlinear fashion. Parents assigned substantially lower utility (i.e. greater disutility) to venipuncture, minor infection, and hospitalization than' did physicians, and these utilities were even lower in parents with other children at home. There were no consistent associations between parents' utilities and their age or years of schooling, nor between physicians' utilities and their sex or years of experience. Even based on the physicians' utilities, expected utilities [E(u)s] favored the “no blood culture”, strategy [E(u) = 0.994] over “universal blood culture” [E(u) = 0.898] or even “selective blood culture” (blood cultures only in children deemed to be at highest risk of bacteremia) [E(u) = 0.982]. Moreover, the result was insensitive to feasible ranges of probability estimates. Unless greater benefits can be demonstrated for the test, our findings suggest that physicians should either obtain fewer blood cultures or develop less aggressive protocols for treating children whose cultures are positive.