Fast and specific detection of moderate long-term changes in occupational blood exposures

Hospital surveillance systems have been established to monitor occupational blood exposures. We compare short-term monitoring with long-term monitoring of data analysis over 11 years and 21 institutions to identify variations in the number of reported exposures.

Short-term monitoring examines the current number of exposures compared to their average over previous years. Long-term monitoring detects trends over several years by various exposure characteristics (place, staff, procedure, etc) through estimating rates of change and using the best linear unbiased predictors (BLUPs) to prevent artefactual trends due to the many categories for each characteristic. Graphical representations of estimated rates help detect change and differences in rates of change.

Annual monitoring allowed detection of significant changes in the number of reported exposures. Long-term monitoring identified moderate trends over time. The BLUP corrected the estimate of each specific annual rate of change and allowed all other rates to reduce the random variability around the mean change for more specificity. League tables showed significant increases or decreases compared to no change. League tables for two-by-two comparisons allowed reliable comparisons between estimates of the rates of change, although with spurious ranking. Funnel plots enabled quick detection of changes in trends within specified confidence intervals. Long-term trends agreed with the dominant type of annual changes over the 11 years but were not as sensitive.

The two methods have different uses. Both are helpful for assessing short-term sudden and long-term minor changes in number of exposures, possibly reflecting the success or otherwise of introducing specific safety devices or guidelines.