Most important criteria | A: Early development | B: Early development | C: Late phase III | D: Post-marketing |
---|---|---|---|---|
A typical survey can be conducted at relatively low costs | ✓ | ✓ | ||
Data can be collected during quick sessions with participants | ✓ | ✓ | ||
Low frequency of sessions required by patients | ✓ | ✓ | ✓ | |
Relatively quick delivery of preparation, data collection, and analysis | ✓ | ✓ | ✓ | ✓ |
A large number of attributes can be explored | ✓ | |||
Suitable to study preferences in a small sample size | ✓ | ✓ | ✓ | |
A low cognitive load on patients | ✓ | ✓ | ✓ | ✓ |
Does not need an education tool or preparatory instructions in order to enhance participant comprehension | ✓ | ✓ | ||
Publically acknowledged by your organisation’s guidelines as an acceptable method to study preferences | ✓ | ✓ | ||
New attributes can be added without making prior results invalid | ✓ | ✓ | ✓ | |
Can be used to collect data from more than one participant in a single session | ✓ | |||
The analysis can calculate risk attitudes, like risk tolerance, and calculate how value functions bend due to the presence of uncertainty in the participant | ✓ | ✓ | ✓ | ✓ |
Explores the reasons behind a preference in detail | ✓ | ✓ | ✓ | ✓ |
Can estimate weights for attributes | ✓ | ✓ | ✓ | ✓ |
Estimates trade-offs that patients are willing to make among attributes | ✓ | ✓ | ✓ | ✓ |
Can quantify heterogeneity in preferences | ✓ | ✓ | ✓ | ✓ |
Internal validity can be established | ✓ | ✓ | ✓ | ✓ |
External validity can be established | ✓ | ✓ | ✓ | ✓ |
Outcomes can refer to a course of health over time (as opposed to a constant health state) | ✘ | ✘ | ||
Sensitivity analysis is possible | ✘ | ✘ | ✘ | ✘ |
Can combine quantitative and qualitative methods | ✘ | ✘ | ✘ | |
Applies validation tests | ✘ | ✘ | ✘ | |
Results can be reproduced by an (independent) researcher for reproducibility | ✘ | ✘ | ✘ | ✘ |
Applies tests for consistency | ✘ | ✘ | ||
Can be conducted without the need for specialized software (beyond Excel) | ||||
Can be conducted without programming skills | ||||
Researcher does not need to supervise the data collection | ||||
Does not require hypothetical scenarios | ||||
Attributes and attribute levels can be determined as part of the method itself (internal identification) | ||||
Data saturation can be achieved relatively quickly | ||||
Does not require model estimations | ||||
Outcomes can be expressed in a particular format (e.g. probability scores, marginal rates of substitution, monetary values) | ||||
Outcomes can refer to a constant health state (as opposed to a course of health over time) | ||||
Uses respondent validation by asking participants to check their data or responses | ||||
Validates through triangulation |