W6- Analysis of Economic Evaluations and Sensitivity Analysis Flashcards
(22 cards)
Incremental cost effectiveness ratio equation
(Total cost A – Total cost B)/
(Total QALYA – Total QALYB)
basically- Difference in cost/ difference in QALY
Average cost effectiveness ratio Equation
Total COST/Total QALY
what is dominance
• When comparing two interventions, one intervention is said to dominate the other if it is more effective and less costly
Cost-effectiveness- dominance
- Dominance can be used to identify which strategy is cost effective (for single intervention and comparator)
- Or which should be eliminated from consideration (for more than two interventions)
- Where there is no dominance:
- Thresholds
- QALY league tables
Thresholds- cost effectiveness
• Regions in the cost effectiveness plane where there is no dominance
- Compare the ICER to decision maker’s threshold value (λ) i.e. opportunity cost (value of the next best alternative)
- If the ICER is less than λ, the new intervention can be deemed ‘cost-effective’
Source of threshold values
- Threshold sources
- Opportunity cost of programmes displaced by newer, more costly technologies
- Social value of a QALY/DALY
- Implied value from a funded intervention
- For the NHS - NICE guidance - fund things £20,000-30,000/QALY
• … but no thresholds for other types of outcomes such as natural units
what are QALY league tables
- Ranking of intervention in terms of incremental cost per quality adjusted life years (lowest incremental cost per QALY at the top)
- Two separate motivations for league tables
- Placing results in a broader context
- Allocation of resources (i.e. implies that we move funding from lower placed, to higher placed interventions)
limitations of QALY league tables
- Need to be careful with in the use of league tables
- Be wary of differences between studies:
- discount rates
- methods for estimating utilities
- costing perspective
- choice of comparator
what is Uncertainty-
Every evaluation will contain some degree of uncertainty, eg what if the effectiveness of a diagnostic test was lower than that used in the analysis, different discount rate ect
• Uncertainty applies to both trial-based and model-based economic evaluations
what are 3 different types of uncertainty
• Methodological uncertainty
– How should health utilities be measured and valued?
– Which resources and how are they valued?
– Which discount rate?
• Parameter uncertainty
– Uncertainty surrounding the true mean parameter values
– How does the data we have relate to the population we are interested in?
• Structural uncertainty
– Do we have the correct comparators?
– Does the model include all relevant events?
– How are those events modelled?
Source of uncertainty
- Assumptions and methods used-
* Generalizability of the findings to other settings
importance of uncertainty
• Uncertainty leads to uncertainty in the results.
• As any assessment of cost-effectiveness is uncertain, hence any decision based on cost-effectiveness will also be uncertain
There is always a chance that we will make the wrong decision
• If results change due to different sources of uncertainty – overall results uncertain
• If results do not change as a result of uncertainty – overall results are robust
Dealing with uncertainty
- Methodological uncertainty can be dealt with by providing guidance or a ‘reference case’
- Testing the impact of different choices is also important in order to identify whether these affect results – sensitivity analysis
- Structural uncertainty requires inputs from those engaged in providing interventions, data collection and literature to ensure appropriate comparators and events
Methodological uncertainty
• To ensure a consistent approach, NICE has developed a ‘reference case’ for methods
what is Sensitivity analysis
systematically examining the influence of uncertainties in the variables and the assumptions used on the results. The aim is to identify which variables or assumptions have an impact on the ICER
what are the 2 different Sensitivity analysis
• Simple sensitivity analysis:
One-way – vary one variable at a time
Multi-way – vary 2 or more variables at a time
what is One-way sensitivity analysis
• One way – vary one variable at a time.
• Different approaches to undertaking one way sensitivity analysis:
1. across a plausible range of values whilst holding others at their ‘best estimate’ or baseline value
2. Can be shown on Tornado diagram
Type of bar chart which shows impact of changing key variables
Ones which have the biggest effect at the top
Other way can be through a- Threshold analysis
what is Multi-way sensitivity analysis
- One way sensitivity analysis likely to underestimate the uncertainty because of interactions between variables
- Need to look at changing two or more variables simultaneously.
- Can become complicated beyond two variables particularly in how to present the results
- Scenario analysis – type of multi-way analysis where assumptions are specified for a number of variables
what is Probabilistic Sensitivity Analysis
- Addresses parameter uncertainty
- Samples all uncertain parameters from their distribution and re-calculates results
- Does this a large number of times
- Calculates the mean and uncertainty
How to interpret sensitivity analysis
- How robust are the findings – do changes in parameters alter the decision?
- Which variables cause biggest impact on ICER?
- Does more data need to be collected?
- Are some variables more under the control of policy makers?
- Separate sub-group analysis may be important
Simple sensitivity analysis Advantages
Advantages
• Simple sensitivity analysis (e.g. 1-way) are easy to implement.
• May enable the decision-maker to identify the key factors of the decision problem which determine the cost-effectiveness of the technology.
• Can be used as a means of validating/verifying the underlying logic model.
Simple sensitivity analysis limitations
Disadvantages
• Extreme scenarios are unlikely to be helpful for decision-makers
• Multi-way scenario analysis can be complicated to present
• All parameters are uncertain. Varying a selection of these does not account for overall uncertainty
