Case Study

• Your COO has asked your team to make a recommendation on moving forward with an intervention treatment pilot.
• Your analytic team built a model that satisfies the clinicians based on variables in the model.
• You need to check out their confusion matrix and put your stamp of approval or disapproval (send them back to the drawing board). Summarize your view of the model. What is the specificity, sensitivity, accuracy and F score of the model?

Confusion Matrix	Condition Positive	Condition Negative
Test Positive	700	100
Test Negative	30	900

• Assuming their random forest model is acceptable, take the F score as an indicator of the percent of the population that can be identified correctly. If 5% of those indicated by the model are reachable and there is a $5,000 cost savings by reaching them before crisis, what is the potential savings impact of the program? Do you think it worth recommending moving forward with a pilot?