Table 1
Specifications for Readers and Authors, and Comparison of Bayesian and Evidence-based Medicine Implementations
| Specification | Bayesian Communication* | Evidence-Based Medicine |
| 1. Express prior knowledge | Assess prior beliefs; sensitivity analysis for uncertainty in prior (F) | — |
| 2. View effect size and variability | Mean of posterior beliefs; contaminated models for surprise (F) | Point estimate (F); confidence interval (H) |
| 3. Express thresholds | Minimally clinically important difference (F if based on utilities) | Number needed to treat (H) |
| 4. View inferences | Tail probability, credible set, Bayes factor, equivalence (F) | Post-hoc adjustments (H) |
| 5. Receive explanations | Dynamic algorithms based on influence diagrams (F) | Static textbook explanations (H) |
| 6. Evaluate study and statistical quality | Likelihood debiasing (F) | Quality inventories (H) |
| 7. Synthesize multiple studies | Confidence profile method, Bayesian meta-analysis (H) | Meta-analysis; Cochrane trial banks (F) |
| 8. View beliefs of the community | Archived priors (F) | Postpublication peer review (H) |
| 9. Protect authors' investment | Likelihood function (F) | Sufficient statistics (F) |
| 10. Provide enough information | Information defined by decision problem (F) | Sufficient statistics, Outcomes research (F) |
| 11. Make authoring easy | Applet libraries | Current program of education and tool-provision |
-
↵* F indicates formal solution; H, heuristic.
