Problems in assessing rates of infection with methicillin resistant Staphylococcus aureus

Published as supplied by the author

Table 1 Chance of Type I and Type II errors for three strategies for assessing whether a target has been met for an median trust with 32 expected cases of MRSA per year, both with and without adjustment of the criterion for over-dispersion illustrated with a 20% reduction (the target each year) and 50% reduction (the overall target over 3 years)

All results assume data exhibits 33% over-dispersion.

The high Type II errors for criterion 1, which requires a statistically significant reduction in rates, show that few trusts will pass this threshold on an annual basis, and a substantial proportion would not be expected to show such a reduction over 3 years, even if the risk has truly been reduced by 50%. Criterion 2, simply demanding the observed reduction exceeds the target, means that 50% of trusts will fail the criterion even if the risk has truly been reduced according to the target. Criterion 3, in which one only penalises a trust if there is significant evidence that the target has not been met, will accept almost all results on an annual basis, and even over 3 years most trusts that do not reduce their risks will still pass this criterion if over-dispersion is (appropriately) taken into account.