Table 3. Logistical and statistical implications of three scenarios of blood culture bottle (BCB) validation.
| Optimal scenario | Intermediate scenario | Minimal scenario | |
|---|---|---|---|
| Number of strains tested | 20 | 20 | 10 |
| Number of lots tested | 3 | 2 | 2 |
| Total number of bottles needed per bottle type | 360 | 240 | 120 |
| Total number of bottles needed (all bottle types combined) | 720 | 480 | 240 |
| Total volume of blood needed | 4320 ml | 2880 ml | 1440 ml |
| Number per bottle type adult | 180 | 120 | 60 |
| Number of bottle type pediatric | 180 | 120 | 60 |
| Detectable difference in yield * | 5% | 7% | 10% |
| Detectable relative yield * | 95% | 93% | 90% |
| 95% confidence interval ** | 86–93% | 85–93% | 83–95% |
| Number of extra bottles needed per extra strain tested per bottle type | 18 | 12 | 12 |
| Volume of extra blood needed per extra strain tested | 216 | 144 | 144 |
Calculation of detectable differences in yield is based on the normal approximation of the binomial distribution (https://www.stat.ubc.ca/~rollin/stats/ssize/b2.html); confidence intervals are calculated based on the binomial distribution (http://vassarstats.net/prop1.html).
* Compared to reference system, assuming 80% power, 95% confidence and 97% yield of the reference system
** Assuming an observed yield of BCB under evaluation of 90%; confidence interval becomes narrower when observed yield of BCB under evaluation is higher. Confidence intervals for proportions are not symmetrical due to binomial distribution; the uncertainty for these proportions is larger on the lower side of the interval than on the higher side.