There are many systems for "flagging" people at risk. The intent is to identify people who are predicted to have poor outcomes so that additional resources can be channelled to them, improving the probability of an injured worker's return to life. The intent is good, but, with a few notable exceptions, these flagging systems carry with them dangers that may undo the intended good.

Flagging systems are based on statistical observations about things happening in the injured person's life and their recovery outcomes. When the two things happen together more frequently than pure chance would predict, we say that there is a "correlation" or "association" between those things. There are studies that reliably show correlation between the various "flags" and delayed recovery.

"Correlation" is not the same thing as causation. First, a correlation published in an academic study says only that the two things studied happen together more often than is explained by coincidence or chance. Causation, as we use the word, is much more certain.

Secondly, merely happening at the same time doesn't necessarily mean the two things are connected in any way we can understand. A classic study showed that the number of priests and the number of prostitutes in an American city correlated very strongly. Jokes aside, few would seriously suggest that one caused the other.

This means that an injured worker who has "flags" is only statistically more likely to have a poor outcome. It is not properly understood as a prediction about this particular individual.

The reliability of flags when applied to the specific case of an injured worker isn't the only concern. There is danger in the way they are sometimes used. In our love of new studies and new tools for injury management we may have forgotten "self- fulfilling prophesy".

For those who are unfamiliar with the concept, there were a series of experiments about 60 years ago that demonstrated a phenomenon that is basic to human behaviour. When someone is told to expect a person to do something or have a certain outcome, the probability of the predicted occurrence increases. This is as true even when the person who is told is merely someone who interacts with the person about whom the prediction is made.

The classic study showed that when a teacher was told that some of his or her students were exceptional or challenged, the performance of the students matched the prediction at the end of the study, even though the students started out the with the same performance. This is because the prediction caused subtle differences in how the informed person (the teacher) interacted with the person about whom the prediction was made (the students). Little things like expressing sympathy for failure vs. pushing the student to succeed the next time, had cumulative impact. This is now psychology text book material and not actively researched.

With injury management flagging systems, the claims manager may take the role of the teacher and the injured person the role of student. The flags may create an expectation about what this particular worker's likely outcome will be. The expectations that the claims manager has about the workers recovery may subtly influence his or her treatment of the worker. Perhaps different concerns are expressed by the claims manager and internalised by the listening worker. Perhaps different decisions about the allocation of resources are made.

Is this result inevitable? Certainly not. But like any powerful tool, it makes sense to take care in the training of those who use it and the purposes for which it is used. Many can be helped by timely and sensitive provision of resources. But like any other powerful tool, flagging systems can do great harm if misunderstood or misused.