While displaying a Pareto chart at a reliability meeting at Ciba Corning, one of the participants suggested we knock off all the little problems at the bottom of the chart. (Maybe he saw some problems that appealed to him). In any case, this is a bad idea.
So to construct this Pareto chart, take the specification of some limit (the exact limit is what people are debating) and require that 95% of glucose results are within this limit as defined by average bias times a multiple of the CV. This limit demarcates no harm from minor harm. To setup the Pareto chart, use a 1 to 5 scale with 5 being the most severe harm and 5 the highest probability of occurrence.
For the 5% that exceed this limit, assume that most that exceed the limit (4.5%) are close to the limit and cause minor harm and 0.5% cause major harm. To classify this:
Minor harm: severity = 1, probability = 3; Pareto rank = 1X3=3
Major harm: severity=5, probability = 1; Pareto rank = 5X1=5
To continue to ignore severe harm in discussions about glucose specifications is a bad idea, and severe harm does occur.
Note: One way of constructing Pareto charts is to rank all severity=5 events by decreasing probability, then list all severity=4 events and so on.