At what risk scoredo you draw the line between screen and don't screen?
ATOF is a decision-analytic framework that integrates clinical net benefit, overdiagnosis harms, and regional workforce capacity into a single answer to that question - for every region of Australia, under every plausible workforce scenario.
The three-component story, in one frame
One master threshold drives all three panels. Drag below - or press â–¶ Play to walk the cascade from 1% to 7% and watch the curves deflate as the map turns red.
Net benefit
Where each model beats both 'screen everyone' and 'screen no one'. Useful range = green band.
Overdiagnosis penalty
Same curves, deflated by the 54% OD rate (Glasziou/Lindsay). The useful range shrinks.
Capacity by SA4
Workforce-feasibility at this threshold under Combined (Derm + GP). Red = over capacity.
Workforce capacity vs screening demand
Each SA4 coloured by capacity utilisation at the selected threshold. Drag the slider above and watch Australia respond.
Net benefit at every threshold
Decision curve analysis (Vickers & Elkin 2006) plots the population-level net benefit of using a risk model at each threshold, against screen everyone and screen no one. Curves are modelled from published AUCs using a binormal approximation.
Where can screening start tomorrow?
For every SA4, the minimum feasible thresholdis the lowest risk cut-point at which the region's annual screening flow stays within available workforce capacity. Regions below the current national slider can run screening today; regions above it cannot, regardless of willingness.
Minimum feasible threshold by SA4 - strip plot
Hardest 5 regions to screen
| Region | State | Remoteness | Min threshold | Source |
|---|---|---|---|---|
| Moreton Bay - North | QLD | MM1 | 10.7% | Lindsay 2026 (real) |
| Brisbane - East | QLD | MM1 | 10.2% | Lindsay 2026 (real) |
| Wide Bay | QLD | MM3 | 9.8% | Lindsay 2026 (real) |
| Darling Downs - Maranoa | QLD | MM4 | 7.9% | Lindsay 2026 (real) |
| Toowoomba | QLD | MM3 | 7.3% | Lindsay 2026 (real) |
Where the workforce is vs. where the disease is
Four lenses on the same 88 SA4s. The same region appears in every panel - hover any panel to inspect it across all four at once. The maldistribution between workforce (cluster in MM1) and incidence (rural sun-exposed) is the equity problem ATOF's adaptive threshold is designed to reconcile.
Who bears the burden of a uniform threshold?
A single national threshold is mathematically simple but pretends that workforce is evenly distributed. ATOF makes the trade-off explicit: every region's min feasible threshold, stratified by SEIFA disadvantage or ARIA+ remoteness.
| SEIFA quintile | SA4s | Population | Median min | Mean min | Pop in over-capacity SA4s |
|---|---|---|---|---|---|
| Q1 (most disadvantaged) | 10 | 1.38M | 2.8% | 3.3% | 847k(61%) |
| Q2 | 28 | 6.12M | 3.1% | 3.5% | 2967k(49%) |
| Q3 | 26 | 8.85M | 2.7% | 3.5% | 2581k(29%) |
| Q4 | 10 | 3.91M | 2.6% | 4.2% | 1306k(33%) |
| Q5 (most advantaged) | 14 | 4.22M | 1.9% | 2.2% | 199k(5%) |