Uncertainty & Sensitivity Flashcards
Bilcke on uncertainty
Scope of analysis is often limited
- Methodological choices (time horizon/discount rate)
- Structure uncertainty (what parameters to use?)
- Parameter uncertainty (what value of parameter?)
Stadhouder on uncertainty
Uncertainty can bias thresholds
e.g. Mortality rates; internal shifting
Variability
Variation/randomness within homogenous sample of patients. Central tendency can be obtained after large number of iterations
- Standard deviations
- Microstimulation
Heterogeneity
Differences between patients (Bayesian Model)
Subgroup analysis
Dealing with uncertainty
- Model discrepancy (judgements can be quantified about the difference between model output and reality)
- Sensitivity analysis (parameter estimates across range and its impact on model results)
Bias: Only use of complete cases when having missing values.
Sensitivity analysis (3)
- Oneway
- Multiway
- Threshold
Highly dependent on the researcher. Therefore can be used alongside the PSA.
One-way analysis
Each parameter is varied independently.
It underestimated the overall uncertainty and correlation
Multi-way analysis
More paramters at once (low-high).
Hard to find a suitable mix
Threshold analysis
The critical value (above-below) depicts the conclusion of the study
Probabilistic Sensitivy Analysis (PSA)
Can measure joint uncertainty across all parameters at the same time
BUT! Tend to use homogenous parameters (costs, LYs, QALYs)
- Primary data
- Secundary data
- Expert opinion
Petrou on uncertainty
Dealing with:
- Skewed data: Bootstrapping
- Missing data: Multivariate regression or censoring
Set of sample stimulations, predicting value or missing full-time
Assume that WTP is known and look out for heterogeinity
Biases (5)
- Selection bias
- Information bias
- Attrition bias
- Reporting/publication bias
- Confounding bias
Selection bias
Uneven dispersion of sample in population
Information/measurement bias
a distortion in the measure of association caused by a lack of accurate measurements of key study variables
Attrition bias
a systematic error caused by unequal loss of participants from a randomized controlled trial (RCT)
Reporting/publication bias
a distortion of presented information from research due to the selective disclosure or withholding of information by parties involved with regards to the topic selected for study
Confounding bias
Other factors influence outcome, by indication.
Can be reduced via comparison simular prognosis, restriction of the population or individual matching between the (non)exposed
