A 99% accurate newborn screen can still be wrong 91 times in 100

New story: with an assumed 1-in-10,000 condition, 99% sensitivity and 0.1% false positives, about 11 babies in 10,000 are flagged and one is real: roughly 9%.

Tools and projects · 1 min read · Source
fictionhearth-and-horizonhealthstatistics

One in Eleven is a new Hearth & Horizon story (text only), set in an invented parish in Cape Breton in 2048. A health authority's assistant has to tell two parents what a positive newborn screen means. The arithmetic, with assumed figures (not any real programme's):

  • Prevalence 1 in 10,000, sensitivity 99%, false-positive rate 0.1%.
  • In 10,000 babies: 0.99 true positives and about 10.0 false ones. About 11 are flagged; one is real. The chance a flagged baby has the condition is 0.99 / (0.99 + 10.0), about 9%, not 99%.
  • Counting natural frequencies ("a hall of ten thousand babies") is easier to follow than quoting percentages, and it is how the assistant explains it.
  • The practical rule in the story is the same either way: act as though it could be real (regular feeds, a clear emergency threshold) while speaking as though it probably is not, until the second-tier gene test.

The story also lets a grandmother's hurry to baptise, and the priest's answer, stand on their own. Read it on the Stories page. It is fiction and not medical advice.

More in Tools and projects

All writing · All in this topic · All tools and research