Abstract
In a recent paper, Phillips and associates stated that “based on professional expertise, prior experience and intuition, clinicians were able to successfully predict engagement in aggressive behaviour during patient admission to a child psychiatric inpatient units (sic)” (Phillips et al., 2011). This is both misleading and concerning.
The Oxford Australian Dictionary defines “to predict” as to “foretell” or to “prophesy” (Australian Concise Oxford Dictionary, 2009). Hence the authors implied that clinicians are able to make truthful statements about the future behaviour of child inpatients. The paper is misleading because the data they reported does not support their conclusions. A positive predictive value (PPV) of 49.3% can be calculated using 69 children classified as at high or very high risk and the 34 true positives who did go on to be aggressive. PPV in this context is the portion of correct prophesies of any aggression (Large et al., 2011). This PPV is a little worse than flipping a coin, and means that on balance, a statement about the presence of future aggression will be more likely to be false than true. The PPV for more severe aggression can also be calculated to be 25.3%. Of the total of 127 children, 79 were found to be at high or very high risk of more severe aggression but only 20 engaged in more severe aggression. The authors correctly concede that risk assessment was not useful in distinguishing children who would engage in severe aggression from those who do not engage in severe aggression, but they failed to recognize a similar limitation when predicting any aggressive conduct.
The conclusions of the paper are concerning because clinical decisions might be made on the basis of a risk assessment for future aggression (Ryan et al., 2010). When children are deemed to be at high risk it is possible that they might be given more medication or be further contained in a more restrictive setting. Half of these interventions would be inappropriate.
What the authors should have reported is that although they were able to define populations of children at a statistically higher risk of future aggression, these categories were of very limited value because of the high proportion of false positives.