Screening Programme Analytics: PPV by Prevalence, Yield, Cost per Case & Biases

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Turn one year of district reports from the fictional Northvale bowel screening programme into a 2×2 table and calculate sensitivity and PPV, with specificity and NPV as worked examples.
• Use natural frequencies per 10,000 people to calculate the PPV of the same stool test in five age bands and a symptom clinic, then chart how the PPV rises with prevalence.
• Compare six cut-offs of the stool test: Youden's J picks the statistically best one, and the expected positives per 1,000 people show which cut-offs the colonoscopy service can actually handle.
• Combine two tests in series (only people positive on both are referred), calculate the combined sensitivity and specificity, and compare cancers found, false referrals and PPV with the one-test programme.
• Turn six years of Northvale's activity data into the yield measures a programme board reads: cancers per 1,000 screened, the number needed to screen to find one cancer, and colonoscopies per cancer.
• Calculate uptake by deprivation quintile, size the equity gap as the extra people who would be screened at the best group's uptake, calculate the programme's coverage and chart the social gradient.
• Cost Northvale's 2026 screening round, calculate the cost per cancer detected, subtract the treatment savings from earlier diagnosis and divide by the QALYs gained.
• Show with numbers why longer survival after screen detection does not prove benefit (lead time), why screening finds the slow tumours (length time), and how many false alarms a person collects over 13 rounds.