Commission Decision of 16 December 2009 laying down guidelines for the management of the Community Rapid Information System ‘RAPEX’ established under Article 12 and of the notification procedure established under Article 11 of Directive 2001/95/EC (the General Product Safety Directive) (notified under document C(2009) 9843) (2010/15/EU) (repealed)

3.6. Probability of injury

The ‘probability of injury’ is the probability that injury scenario may indeed materialise during the expected lifetime of the product.

This probability is not easy to estimate; but when a scenario is described in distinct steps, each step can be given a certain probability, and multiplying these partial probabilities together gives the overall probability of the scenario. This stepwise approach should make it easier to estimate the overall probability. Of course, where several scenarios are developed, each scenario requires its own overall probability.

Where an injury scenario is nevertheless described in a single step, the probability of the scenario can also only be determined in a single overall step. This would only be a ‘guesstimate’, however, which could be severely criticised and thus call the entire risk assessment into question. A more transparent assignment of probabilities to a several-steps-scenario is therefore preferable, especially as the partial probabilities can be built on undisputable evidence.

These guidelines distinguish between 8 levels of probability to classify overall probability: from < 1/1 000 000 to > 50 % (see left-hand side of table 4). The following example of a hammer head that breaks when the user knocks a nail into a wall should illustrate how to assign a probability to each step, and how to classify overall probability:

Step 1:	The hammer head breaks when the user tries to knock a nail into a wall because the material of the hammer head is too weak. The weakness was determined in a test, and with the reported weakness the probability of the hammer head breaking during the otherwise expected lifetime of the hammer is put at 1/10.
Step 2:	One of the pieces of the hammer hits the user when it breaks. The probability of this happening is put at 1/10, since the area of upper body exposed to the pieces flying off is considered to be 1/10 of the half-sphere in front of the wall. Of course, if the user were standing very close to the wall, his body would take a larger share of the half-sphere, and the probability would be higher.
Step 3:	The piece hits the user on the head. The head is estimated to be about 1/3 of the upper body, and the probability is therefore 1/3.
Step 4:	The piece hits the user in the eye. The eyes are considered to be about 1/20 of the area of the head, and therefore the probability is 1/20.

Multiplying the probabilities of the above steps together gives an overall probability for the scenario of 1/10 * 1/10 * 1/3 * 1/20 = 1/6 000. This translates into > 1/10 000 (see left-hand side of table 4).

Once the overall probability has been calculated for an injury scenario, it should be checked for plausibility. This requires rather a lot of experience, thus suggesting that the assistance of persons experienced in risk assessment should be sought (see above in section ‘Let others check your risk assessment’). As experience is gained with these guidelines estimating probability should become easier, and an increasing number of examples will become available to facilitate this task.

Assigning probabilities to different injury scenarios for the same product may lead to the following:

• When the product is used by more vulnerable consumers in a scenario, the probability may have to be raised in general because more vulnerable consumers can be injured more easily. This applies in particular to children, since children do not normally have the experience to take preventive action, on the contrary (see also ‘Vulnerable consumers’ in section 3.3).

• When the risk is readily recognisable, including through warning labels, the probability may have to be lowered because the user will use the product more carefully in order to avoid injury as far as possible. This may not apply to an injury scenario with a (young) child or other vulnerable user (see table 1) who cannot read.

• When accidents have been reported that fit into the injury scenario, the probability for that scenario could increase. In cases where accidents have only rarely been reported, or are not known at all, it may be useful to ask the manufacturer of the product whether he is aware of any accident or adverse effect caused by the product.

• When a fairly large number of conditions are needed for the injury to occur, the overall probability of the scenario would normally be lower.

• When the conditions needed for the injury to occur are easily met, this may increase the probability.

• When the test results of the product fail by a large margin to come within the limit values required (by the relevant standard or legislation), the probability of the injury (scenario) occurring may be higher than if the product performed close to the limit values.

The ‘probability of injury’ in this instance is the probability that the injury scenario may actually happen. Probability does not therefore describe the general exposure of the population to the product, calculated, for example, by considering the millions of product items sold on the market and then considering that a few of them might fail. Considerations of this kind do, however, play a role when determining the appropriate risk reduction measures (see section 4).

Also, accident statistics, even if product-specific, have to be considered with care when used for to estimate probability. The circumstances of the accident may not be reported in sufficient detail, the product may have changed over time, or the manufacturer may be different, and so on. In addition, light accidents may not have been reported to those collecting the data for the statistics. None the less, accident statistics can shed light on injury scenarios and their probability.