To compare the efficiency of detecting Haemophilus influenzae type b bacteremia by the BACTEC radiometric system (RS; Johnston Laboratories, Inc., Towson, Md.) and a conventional Trypticase soy broth (BBL Microbiology Systems, Cockeysville, Md.) blood culture system (TSB), we developed an in vivo model of bacteremia in rats. After intravenous injection of 50 to 200 CFU into adult rats, there was a linear logarithmic increase in CFU per milliliter of rat blood during the first 10 h (r = 0.98), allowing accurate prediction of the level of bacteremia with time. Culture bottles were inoculated with 0.5 ml of blood obtained by cardiac puncture and processed as clinical samples in the microbiology laboratory with our RS and conventional protocols. We found the following. (i) The first detection of bacteremia by RS was similar to that by TSB if a Gram stain of the TSB was done on day 1 and was superior if that smear was omitted (P less than 0.01). (ii) The detection times in both systems were comparable at different magnitudes of bacteremia (10(1) to 10(4) CFU/ml). (iii) Supplementation of inoculated bottles with 2 ml of sterile rat blood interfered with Gram stain detection in TSB but resulted in increased 14CO2 production in RS. (iv) No difference in detection time was found between RS and TSB for four different clinical isolates. These studies show that, in a biologically relevant model, the detection of positive blood cultures for H. influenzae type b by RS was comparable to or better than detection by TSB when blood was processed analogously to clinical specimens.