Evaluating Cost-effectiveness using decision trees: The HPV case

· One of the most widely used decision-analytic models for evaluating health interventions

· Represents the possible outcomes for an individual following an intervention using a series of pathways

10 of 82

"Tips From Former Smokers" (TIPS)

· Launched in 2012 by the Centers for Disease Control (CDC)

· The first federally funded national mass media antismoking campaign

· Resulted in an increase of 12% in population quit attempts

· From 31.1% of current smokers to 34.8%

· An increase of 1.6 million quits and 100,000 sustained quits

20 of 82

The CDC evaluated cost per QALY saved and cost per life year saved from the CDC budget perspective.

· $47,932,809 campaign budget

· $6.7 million for creative development

· $38 million for advertising purchases

· $3.1 million for evaluation

· $1.0961 for each of 43.7 million adult smokers

23 of 82

· Primary effect size: 12% increase in quits; measured 6.1% sustained quit rate at 6 months

· Modeled prevented premature deaths and discounted (3%) life years and QALYs gained by quitting within age and sex groups, using outside data

· Estimated average of 1.79 QALYs per sustained quit

24 of 82

27 of 82

· Decision node (square node at start of tree): Represents a decision between two or more competing intervention alternatives

· Chance nodes (circular nodes following the decision node): Connect a series of different pathways that represent the range of possible outcomes to the root of the tree

34 of 82

· Branch probabilities: Probabilities assigned to each branch coming out of a chance node, based on their likelihood of occurring

· Conditional probabilities: Probabilities emanating from each subsequent new chance node (from left to right), conditioned on having experienced the events earlier in that particular pathway

42 of 82

Once the entire tree is built, there are a set number of pathways through which one can progress through the tree to the final outcome.

· Pathway probabilities: Probability of completing a specific pathway, determined by multiplying initial branch probability by all of the subsequent conditional probabilities

· Pathway costs: Cost associated with each pathway, calculated by summing the cost of all of the events occurring along the pathway

· Pathway effects: Measure of health effect associated with each pathway, usually a measure of the QALYs for that specific pathway

54 of 82

58 of 82

To estimate the expected cost and effect of each decision:

· Multiply probability of each pathway by the cost and effect for that pathway.

· Sum values for each decision.

59 of 82

74 of 82

· Incremental QALYs: 20.038 − 20.034 = 0.004

· Scaled to population: 0.004 × 43,729,007 = 174,916

· Incremental cost: $1.0961 − $0 = $1.0961

· Scaled to population: $1.0961 × 43,729,007 = $47,931,000

· ICER: $1.0961/0.004 QALYs = $274/QALY

· Scaled to population: $47,931,000/174,916 QALYs = $274/QALY

81 of 82

· Events are assumed to happen instantaneously.

· No easy mechanism for capturing time

· E.g., TIPS's lifetime discounted life years and QALYs calculated outside of decision tree model

· With chronic diseases (outcomes over long period of time), there are so many potential future health outcomes that it becomes difficult to characterize all future probabilities in tree form.

· Number of branches multiplying too quickly

· Exceedingly difficult to program and analyze the tree

82 of 82