costs and outcomes/decision analysis Flashcards
costs beyond drugs
cost to administer (nurse, pump)
testing/labs
drug-related complications
treatment failure
treatment effects beyond efficacy
increase survival
work/social functionality, QoL
patient satisfaction
efficacy vs effectiveness
efficacy: expected outcome when a drug is used under ideal conditions (phase 1-3, clinical trial)
effectiveness: expected outcome when a drug is used in a naturalistic setting (actual practice, phase IV)
problems with RCTs
unrepresentative population, short duration of study, protocol-induced elements, intermediate endpoints, inappropriate comparators, artificial environment
confounding factors in the naturalistic setting
differences in disease or patient severity, comorbidities, prescribing preferences & biases
what is decision analysis
a visual road map to help us measure costs and consequences (under conditions of uncertainty). by combining the probabilities that events will occur with the value of each possible outcome, DA prescribes which option to select to maximize the outcome of the decision
advantages of decision analysis
-can use multiple sources of evidence
-time horizon can be extended (beyond a clinical trial)
-can assess the uncertainty surrounding outcomes
why do we have to do a decision analysis?
addresses limitations of cost analysis conducted as part of RCTs such as:
-RCT won’t compare all available options
-RCT won’t look at all potential outcomes
-RCT won’t last long enough to model long term cost effectiveness
-RCTs ignore efficacy/outcomes from other trials
essentially, DAs are conducted alongside RCTs to estimate cost effectiveness.
8 steps in decision analysis
1 identify the problem & decision options
2 identify perspective
3 identify timeframe
4 structure decision & consequences of each decision over time
5 assess probability that each consequence will occur
6 determine value of each outcome
7 select decision option with the best outcome/value
8 conduct a sensitivity analysis
perspective in a decision analysis
the POV from which the analysis is conducted (society, payer, provider, patient)
impacts what costs/consequences are measured & how they’re valued
DA from patient perspective
not common; costs measured would be those only the patient incurs (co-pays, etc)
DA from provider perspective
usually considers inpatient costs (direct medical) or other costs to the provider (drugs, labs, bed, OR time, inpatient procedures, personnel time)
DA from payer perspective
includes insurance companies, government, employers. would consider all inpatient & outpatient charges covered by the payer program. Costs from provider, outpatient rx’s, outpatient MD visits, home health care, etc
DA from societal perspective
broadest perspective
recommended
considers costs borne by the patient, provider, payer, and also considers indirect & intangible costs (loss of income, productivity, pain/suffering due to health)
a ______ is a graphic representation of how all the possible choices relate to the possible outcomes
decision tree
a ______ is defined as a possible choice among all options
decision option
the first point of choice in the decision tree is represented by a ______, usually drawn as a square box
decision node
an event whose outcomes are not under control of the decision maker is denoted by a _______, symbolized by a circle
chance node
for each decision alternative, a __________ is assigned that represents the chance/likelihood of an event/outcome occurring
numerical probability
for probabilities in a decision tree: what do you do if a range of reasonable values for the probability exists
conduct a sensitivity analysis on the lowest and highest values of the range
probability estimates for decision trees can come from
published literature
internal data sources
expert opinion, surveys
what is the last step to a decision tree
assign values to all outcomes (can be costs, utilities, life years, QALYs). determine and add all the costs that would be incurred along each tree path. the total cost of a tree path is placed at the end of the tree
are there assumptions when making a decision tree?
yes when developing your tree, there are frequently items whether either no info is available, conflicting info is available, or a range of values/outcomes exist. so yes you make assumptions about your model
the costs you include in your decision tree will depend on the ________
PERSPECTIVE
what do you do after you enter in the probabilities and path “pay-offs” for your decision tree
average out and fold back (roll back analysis)
calculate each tree path from right to left
important things to know for decision trees
time flows left to right
don’t have to be symmetric
perspective should be clearly described
list assumptions
represent all important outcomes
why is there uncertainty in PE analysis and what do we do for that
there are assumptions (some may be inaccurate) and the results may not be generalizable–> for all of these reasons, it is essential to perform a sensitivity analysis
sensitivity analysis
method to handle model uncertainty; also known as a “what if” analysis
-assesses whether the use of alternative estimates would change the results
limitations of DA
-may oversimplify medical problems to a point where they do not reflect reality (don’t include all potential outcomes)
-inadequate availability of data
-may not reflect true concerns to the patient
-inappropriate assumptions
-traditional statistical analyses cannot be applied to the outcomes
what is a markov model used for
-more complex events occurring over time or for a greater number of event repetitions
-useful when the risk of an outcome changes over time (risk of GI bleed w/ NSAID increases w/ age)
-used for transitions between various stages of health where movement can go back and forth between health states (remission & recurrence of cancer)
discounting
used in PE analyses to express costs & benefits that occur in the future to “present values”
usually only discount for costs/benefits that occur >1 year from present
can range from 3-10%