Paperonium

Childhood cancer · PLoS Med · 2026

Halved cancer-diagnosis waits: chance 1 in 1,000, says the paper’s table. Same row: 1 in 6.

1 in 1001 in 1,000every time1 in 20the usual bar1 in 1,000Printed chance figure: convincingabout1 in 6Same row’s score,1.41: not convincing
1 in 1001 in 1,000every time1 in 20the usual bar1 in 1,000Printed chancefigure: convincingabout1 in 6Same row’s score,1.41: not convincing

How often chance alone gives this drop; convincing if rarer than 1 in 20

The paper says

After Northwest Ethiopian health workers were trained, children’s median wait for a cancer diagnosis (half waited less) fell 54%, from 56.5 to 25.8 days. The paper’s table says chance alone would give a drop this big 1 time in 1,000, if waits hadn’t really changed.

Its own data show

The same row prints the score behind that figure: 1.41, the drop measured against chance’s typical swing. Scores convert into chance figures by one fixed rule, like Celsius into Fahrenheit: 1.41 means about 1 time in 6; 1 in 1,000 needs 3.29. The paper’s 11 other scores obey it; only this table’s three waits don’t.

A result usually counts as convincing only if chance alone would produce it under 1 time in 20. The 1.41 and the 1 in 1,000 can’t both be right; the paper doesn’t publish each child’s wait, so readers can’t tell which is. If the 1.41 is right, the paper’s conclusion linking training to “reduced diagnostic delays” has no convincing test.

“The summary” means the paper’s abstract: the short summary printed at its top, which is the part most readers see.

Read the paper →Full technical write-up →

Multilevel onsite training and mentorship model to accelerate early childhood cancer diagnosis in Northwest Ethiopia: A quasi-experimental mixed method study · PLoS Med (2026) · doi:10.1371/journal.pmed.1005132