Thank you, very interesting discussion ... I guess we also should take into account another parameter which could be the "duration" of the trend (sorry for my English). It allows to stay vigilant while not taking decisions too quickly?
I agree with your point. I think SPC can help. If a company is collecting this data, an XmR chart is an easy way to look for rate changes beyond what is normally expected. There are also methods to look at very rare events (1 or 2 per year) and charting them. I don't mean to say only use statistics, but I like using SPC and my gut instinct. Those interested can see Understanding Variation by Donald Wheeler.
My statistics instructor was a pupil of Don Wheeler, and my instructor was dynamite. Maybe that is the wrong comparison because my instructor worked at Oak Ridge (on bombs). But anyway we must be very careful in our use of statistics and have a great understanding on the proper use and Wheeler is a great source..
ISO TR 24971:2020 5.5.2 indicates "When sufficient data are available to estimate the probability of occurrence of harm with adequate confidence, a quantitative methods should be used. Otherwise, a qualitative method based on expert judgement is preferable to a quantitative estimate with high uncertainty." FDA has issued Warning Letters where there are problems with quantitative estimates.
Remember, that Medical Device Risk Management recognizes uncertainty and the technical committee has indicated this in use of terms such as "expert judgement" and "estimates". Until you have statistically valid data you have only estimates. This usually does not occur until you have actually use data on the device. Some try to claim that data from early design is probability, but it is only estimates until you have actual use data on the device in question.
Thanks Ed, I understand and agree with what you are saying (and I knew someone might bring this up). I've seen groups get tied up in knots using only qualitative/expert judgement looking at something like Naveen's example. That is why I have found it helpful along along with gut instinct (judgement).
Thank you, very interesting discussion ... I guess we also should take into account another parameter which could be the "duration" of the trend (sorry for my English). It allows to stay vigilant while not taking decisions too quickly?
That is an excellent point Stephanie! Thank you for sharing.
I agree with your point. I think SPC can help. If a company is collecting this data, an XmR chart is an easy way to look for rate changes beyond what is normally expected. There are also methods to look at very rare events (1 or 2 per year) and charting them. I don't mean to say only use statistics, but I like using SPC and my gut instinct. Those interested can see Understanding Variation by Donald Wheeler.
My statistics instructor was a pupil of Don Wheeler, and my instructor was dynamite. Maybe that is the wrong comparison because my instructor worked at Oak Ridge (on bombs). But anyway we must be very careful in our use of statistics and have a great understanding on the proper use and Wheeler is a great source..
ISO TR 24971:2020 5.5.2 indicates "When sufficient data are available to estimate the probability of occurrence of harm with adequate confidence, a quantitative methods should be used. Otherwise, a qualitative method based on expert judgement is preferable to a quantitative estimate with high uncertainty." FDA has issued Warning Letters where there are problems with quantitative estimates.
Remember, that Medical Device Risk Management recognizes uncertainty and the technical committee has indicated this in use of terms such as "expert judgement" and "estimates". Until you have statistically valid data you have only estimates. This usually does not occur until you have actually use data on the device. Some try to claim that data from early design is probability, but it is only estimates until you have actual use data on the device in question.
Thanks Ed, I understand and agree with what you are saying (and I knew someone might bring this up). I've seen groups get tied up in knots using only qualitative/expert judgement looking at something like Naveen's example. That is why I have found it helpful along along with gut instinct (judgement).