Risk Management - Probability vs. Frequency

Hello Peter,

I think it is better to start a new thread to discuss the interesting topic which shown as the caption.

Using the frequency as the basis of the risk is quite new for me. Please allow me simply summarize your point of view: if the events are relatively high frequency (>1 events/device/year), it is correct to use the frequency to estimate the risk. That said, use of the probability in that case is wrong.

It seems to make sense for external events (e.g. misuse). Could you explain it in detail? Please give some examples, if possible.

Thanks in advance.

Rather than frequency or probability we tend to use likelihood and define it in terms of common cause events and special cause events, each modified as to likely to occur for most users or for a smaller group of users. Are twice as many events twice as important? I don't think that this type of question can be easily answered, so numerically based methods of assessing occurrence seem inappropriate to me (although they are obviously used almost everywhere).

Accurately estimating probability (or frequency) can be difficult so in most cases this is purely a "theoretical" discussion.

ISO 14971 specifically states that risk is based on the probability of harm.

The correct parameter is frequency (average events per period), to account for events that occur at a frequency > 1. Since probability cannot exceed 1, under ISO 14971, all events with a frequency exceeding one are treated as having the same "risk". So for example a minor burn that occurs 100 times a year has the same "risk" of a minor burn that occurs 1-2 times a year.

Obviously, in this extreme example no-one would consider the risk the same. But I have seen cases where risks that are different by factors of 10 are considered treated the same.

Also, this can also cause errors if there is a sequence of events, and overall probability of harm is based on product of each probability. Again if one of the intermediate events occurs a frequency greater than 1, the result may underestimate the risk.

The potential for "error" increases as the unit for time increases. Calculations based on per use or per day are usually OK, but calculations on a per year or per lifetime might have problems.

Obviously, risk should be risk no matter what units are used. Units are just for convenience. The ratio of risk for event A vs event B should be constant no matter what units are used. If frequency is used, the ratio never changes. But if probability is used, it can change ...

My thoughts a bit messy now. I am wondering if the overall probability will be overestimated when using frequency.

I would think the frequency (adverse events) is highly relevant to the number of use of device during a defined period. For example, device A is used 1000 times per year and 1 adverse events occurred (1 in 1000). If a similar device B is used 100 times per year but without adverse events, can we say the risk of device B is lower than the device A? Frequency based method may tell us that the risk of device A is higher. I think it is not true. There is not enough evidence to judge which device has higher risk.