- Block 22 — Guided laboratories
- Objective
- build, evaluate and operate a two-stage medical screening classifier on a
cohort whose prevalence is 1.5 percent, whose missing values are informative rather than
accidental, and whose enrollment sites do not supply data of equal quality; derive the
operating threshold from the clinical consequence of a miss rather than from any statistical
optimum; confront the measured fact that a model with an AUC of 0.9327 produces a positive
predictive value of 0.0417 at the required sensitivity; measure the subgroup performance gap
and evaluate three honest responses to it; compute the end-to-end characteristics of the
screen-then-confirm cascade including the confirmatory test's own imperfection; and produce
the clinical summary in the vocabulary a clinical committee uses — likelihood ratios,
pre-test and post-test probabilities, and volumes per ten thousand people screened.
- Estimated duration
- 180 minutes
- Prerequisites
- chapters 018, 019, 026, 028, 034, 036, 050, 052, 054, 055, 056, 058,
062, 063, 064, 075, 076, 079, 080 and 082, and lab 1 (
085.0-lab01-churn-classification.md) - Associated quizzes
086.1-quiz-screening-versus-diagnosis.md to
086.9-quiz-clinical-reporting.md