Can we predict which patients with community-acquired pneumonia are likely to have positive blood cultures?