Review previous visualisation work

[sebbacon]

This smiley faces approach (from Tom at North Devon) is a common approach (original)

There is a similar, animated version here, which couches it in terms of Bayes' Theorem, and includes the classic "finding terrorists" example among others:

North Devon also piloted a diabetes test in a single practice (full write up), showing each patient their test results in context, and found a significant drop in HbA1c in patients with the intervention:

[sebbacon]

CRUK have a PSA infographic which doesn't address all the numbers, but is visually quite nice.

It is an illustration of the Cochrane Review that finds no reduction in prostate-cancer-related mortality, so doesn't indicate (for reasonable didactic purposes) the number of extra true cancer cases found, and doesn't attempt to illustrate harm from over-treatment (which is just a footnote).

[sebbacon]

There are some interesting (if unrelated) approaches to statistical visualisation here

[sebbacon]

An excel tool for visualising the relationships between odds rations and sensitivity / selectivity

[sebbacon]

Another tool combining screening and relative risk into data tables

[sebbacon]

From Jessica's paper:

[jamesscottbrown]

Luana Micallef had a 2012 paper comparing the effectiveness of 6 visualizations (more data):

It is cited by a couple of papers by Karin Binder et al that may be of interest:

• Effects of visualizing statistical information – an empirical study on tree diagrams and 2 × 2 tables

• Visualizing the Bayesian 2-test case: The effect of tree diagrams on medical decision making

[sebbacon]

Fab, thanks!

[sebbacon]

Tangentially related: we found this RCT recently, which compares GP understanding of post-test probability based on (a) text summary of sensitivity and specificity, (b) Fagan’s nomogram, (c) probability-modifying plot (PMP), (d) natural frequency tree. Spoiler: (c) and (d) won.

[sebbacon]

A useful nomenclature from here:

[sebbacon]

@jamesscottbrown my hypothesis is that an interactive tool with sliders to change the prior probabilities might be helpful. You've obviously been thinking about the problem too. Any thoughts yourself?

[sebbacon]

A website currently under development:

[sebbacon]

Same website has these posters: