Rate of Errors During Routine Biological Manipulations

Quantitative data informing biosafety practices have long been lacking. In this study, we describe the conduct of the first large-scale investigation into human reliability in the life sciences laboratory to estimate an error rate during routine biological experiments.

To generate these critical data, we conducted two sets of experiments: blinded experiments in clinical laboratories in Brazil, Jordan, and Tunisia, and volunteer experiments in training laboratories in two U.S. universities. In these experiments, GloGerm was used to indicate where spills occurred during laboratory manipulations. In the blinded clinical experiments, dummy samples were introduced into the normal workflow and workers processed them as they would a normal clinical sample. Surfaces were examined at the end of the shift for contamination. In the experiments in the United States, volunteers would repetitively pipette a solution of GloGerm into a 96-well plate and the work area was inspected after each plate.

The median volunteer is estimated to have an error rate of about 4 or 8 errors per 1,000 manipulations (for volunteers with significant laboratory experience vs. those with little laboratory experience, respectively). Estimated error rates from both experiments are comparable, suggesting that studies using volunteers who know they are working with nonhazardous materials can be used to replicate real laboratory conditions to provide critical data in biosafety. The volunteers were also asked to declare when they thought they made an error. By comparing true errors to those declared, we found that volunteers identified a maximum of 52% of their total mistakes, indicating that many mistakes go unnoticed.