5 Recap: bias and error in AMR data

Once we have conducted a null hypothesis test, we now have more information about the importance of the finding. But this isn’t the only piece of information we need to consider when interpreting the result. Sources of error and bias in AMR studies and surveillance programs (described in the module Fundamentals of Data for AMR) include:

• random error
• systematic bias (or systematic error)
  • selection bias
  • information bias (including misclassification error, recall bias and observer bias)
• confounding.

For a refresher, check the glossary for the definitions of these terms.

Hypothesis testing allows us to address random error. However, other types of error and bias may still affect the inference (conclusions) drawn from the data. Systematic bias should be accounted for when interpreting findings, as the analysis cannot correct for it. Lastly, although there is no test for confounding, it can be accounted for by using multivariable statistical modelling, which is outside the scope of this course.