Thresholds, sensitivity, models

Question 1

Cost effectiveness plane quadrants?

Answer 1

North West- Higher cost and lower benefits: The exisiting treatment dominates the new.
North East- Higher cost and higher benefits- where most new interventions lay- need economic evalaution to judge whether it is cost-effective, comapred to the cost-effectiveness threshold (CET) or willingness to pay threshold.
South East- dominates- more benefits but cheaper.
South west- lower costs but lower benefits- would the cost savings be better spent elsewhere?

Question 2

Instead of plotting interventions on a cost effectiveness plane how else can you eliminate dominated interventions?

Answer 2

QALY league tables:
If list those with greatest benefits to least, if any are cheaper above can exclude those that are below (i.e. lower benefits but higher costs)
Can do order by costs and eliminate by benefits too

Question 3

willingness to pay threshold price?

Answer 3

WHO recommends highly cost-effective if less than the countries GDP per capita, cost effective if between 1-3x, and not if above 3x.
BUt this fails to take into account healhcare budgets and affordability. Opportunity cost best to use.

Question 4

Why are sensitivity analysis done?

Answer 4

Because there is uncertainity, e.g effectiveness, discount rate, costing, population external validity etc which can affect the ICER. Do some variables need more data collection on? Or separate subgroup analysis on?

Question 5

Methodological uncertainity?

Answer 5

Relates to the methods used. How have the utilities been measured and valued etc, which resources have been valued? Discount rate? etc
e.g. the use of QALYs for outcome

Question 6

Parameter uncertainity?

Answer 6

INPUTS: Uncertainity for the true mean parameter values- how generalisable?
E.g. variation in the drug costs input into the model.

Question 7

Structural uncertainity?

Answer 7

Got correct comparators? Does the model include all of the relevent events? How are these modelled?

Question 8

Reducing methological uncertainity?

Answer 8

Guidance on methods to use e.g. NICE reference case and to test the impacts of changes e.g. in a sensitivity analysis

Question 9

Reducing structural uncertainity?

Answer 9

Look at literature see what comparators are availble,

Question 10

What is a sensitivity analysis?

Answer 10

Systematically examines the influence of uncertainity of variables and assumptions made. It aims to identify those that impact the ICER when varied across a plausable range of values. Can be one way, two way, multi way or probabilistic. SHows robustness of the model.

Question 11

One way sensitivity analysis?

Answer 11

varies only one variable at a time within a plausable range whilst holding others at their best estimate.

Can display in a tornado diagram. Those with the biggest impact on the ICER are at the top. Displays whether further research is needed for any variables

Or in a threshold analysis: sees how much need to vary the parameter to reach the threshold- i.e. cost effective until only effective at 70% etc.

Question 12

Multi way sensitivity analysis?

Answer 12

one way likely to underetimate uncertainity due to interactions between e..g if set all at worse case scenari will be very bad etc. Can plot these on graphs e.g. different lines for different quit rates and these each vary with another factor

Question 13

Probabilistic sensitivity analysis?

Answer 13

varying the parameters randomly across their spectum and combining different combinations say 1000 times. Can calculate the mean and get a 95% confidence interval for.

Can plot these on a cost effectiveness plane and work out the percentage of simulations that are cost-effective.

can create a cost-efffectivess acceptability curve- as increase the willingness to pay threshold (X axis) the probability of different interventions for being cost effetive (Y axis) will vary. Will give the most likely cost effective option at a given threshold.

Question 14

vampire vs frankenstein economic evalution?

Answer 14

Vampire= done alongside RCT/ study
Frankenstein= lots of trials and info bolted together in an economic model

Question 15

advantages and disadvantages of a vampire/ RCT economic evalution? (+ and -)

Answer 15

+ Randomised, high internal validity
+ Low chance of bias
+ low cost- piggyback onto another trial
-limited to the comparators used in the study
-trial length may be too short to value all of the costs
- low external validity- trial conditions e.g. in teaching hospital etc, and limited to the healthy age group who volunteered
-often inadequate sample size for rare events

Question 16

advantages and disadvantages of a model / frankenstein economic evalution? (+ and -)

Answer 16

• People are suspicious of models- lots of uncerainity and assumptions.
  +flexible can tailor to the situation and the info that is needed.
  +bolt together information that has been gathered with likely higher external validity

Question 17

Decision tree advantages and disadvantages?

Answer 17

+ helpful when simple with short time frame and few options.
+Has a memory follows a course e.g. can have different transition probabilities
- bad for recurrent diseases
- if lots of complications and different outcomes the tree gets too large
- Hard to add in timing or infectious diseases
- may need different trees for different populations

Question 18

Markov model/ state transition model advantages and disadvantages?

Answer 18

+Useful if recurrent disease or over long term
+ cycles can be varied so 1 week or lifetime
- Not useful if lots of options
-memoryless so not useful if transition probability depends upon past cycles
- not good for infectious disease transmission

Question 19

hybrid model?

Answer 19

E.g. short term decision tree then ends on Markov model

Question 20

How to transfer economic models?

Answer 20

Weltes general approach
apply knockout crieria- knockout the variable if its not relevent or if the estimated correspondance is low and can input the relevent data.

Question 21

Hazard ratio?

Answer 21

rate of an event with: without treatment. Used to estimatee the relative treatment effects.