One-way Sensitivity Analysis (part 1)

A sensitivity analysis is used to look at the effect of different values for probabilities and utilities on the model.  For example, look at the decision tree for managing sore throat:

The model assumes that the probability of strep is 0.2, or 20%.  That is a reasonable average for the family practice setting, but there is quite a range among different patient groups.  A clinical decision rule validated by McIsaac and colleagues showed that one can use the patient’s age and 4 key symptoms to get a very accurate picture of the probability of strep:

1.  Add up the points for your patient

Symptom or sign	Points
History of fever or measured temperature > 38 C	1
Absence of cough	1
Tender anterior cervical adenopathy	1
Tonsillar swelling or exudates	1
Age < 15 years	1
Age >= 45 years	-1
Total:

 2.      Find their risk of strep below

Points	% with strep
-1 or 0	1% (2/179)
1	10% (13/134)
2	17% (18/109)
3	35% (28/81)
4 or 5	51% (39/77)

Note:  Baseline risk of strep 17% in this population

If the probability of strep is over 50%, does it still make sense to do a rapid antigen test?  What if the probability is below 2%?  That raises the possible strategy of doing nothing for some patients.  Let’s add that strategy to our model and roll back the tree: