Specific parameters for obtaining blood cultures (volume of blood per culture bottle, number of samples obtained) vary widely. While detection of bacteremia has greatly improved with modern blood culture techniques, problems still exist with false-positive (contamination) and false-negative (inadequate sample) results. Lamy and colleagues noted that empiric studies to determine the optimal blood culture parameters would be laborious and difficult because of the low overall incidence of bacteremia, the common occurrence of low-grade levels of bacteremia, and the lack of a “gold standard” definition of true bacteremia. They presented a mathematical model that takes into account the known historical data on various blood culture parameters and attempts to define an optimal strategy.

The parameter that most influenced the model's predicted sensitivity of blood culture to detect bacteremia was the known variability of bacterial concentration in the blood of bacteremic patients. The authors pooled data from several studies done over the past century (1916, 1947, 1998); the studies they selected had the largest numbers of subjects and, in their opinion, used the most reliable quantitative blood culture techniques.

Other parameters for which the authors combined historical data included the variability in blood bacterial concentration within different samples from the same patient, the number of samples that were drawn, the risk of contamination, and the volume of blood inoculated into each culture bottle. In each case, the authors used statistical methods to best estimate a range of values for each parameter based on data from different studies.

After deciding on the estimated ranges for each parameter, Lamy and colleagues performed numeric simulations to determine the predicted sensitivity for a given number of blood samples and a given volume of cultured blood. Sensitivity of blood culture was maximized and contamination risk minimized by inoculating six bottles (a total blood volume of 35 to 42 mL) at a single sampling time.

Based on their mathematical model, the authors concluded that when blood cultures are ordered, a six-bottle, single-sampling method is the best way to maximize sensitivity and minimize false-positive results.