Between the deep learning work I’ve done in low vision and glaucoma and my time spent as a deeptech investor, I’ve spent a great deal of time looking at diagnostic technologies of all sorts and thinking about diagnostic tests and how to make them successful.
While each technology and entrepreneurial team is different, there are some commonalities that drive whether a diagnostic test is useful around concepts like sensitivity, specificity, positive predictive value, and AUROC.
I summarized some of these views in “Why is it so Hard to Build a Diagnostic Business?” and as part of that work, I created Google Sheets and browser-based calculators for some of the figures of merit that matter. You can access them from the post but also here as well:
Google sheets
- Main link
- Tab on hypothetical HIV test from blog post
- Tab with a positive predictive value (PPV) and number needed to screen (NNS) calculator (the key figures of merit to determine clinical utility for a test)
- Tab with a calculator of cost to screen 1 true positive (the key figure of merit to determine cost-effectiveness of a test)
- Tab showing how to calculate comparative cost-effectiveness of two tests
- Tab showing the clinical utility comparison of FIT test and Cologuard from the 2014 NEJM paper
- Tab showing the clinical utility comparison of FIT test and Cologuard Plus from the 2024 NEJM paper
- Tab showing the economic comparison of FIT test and Cologuard from the 2014 NEJM paper
- Tab showing the economic comparison of FIT test and Cologuard Plus from the 2024 NEJM paper
- Tab on raw data from the 2014 NEJM paper that populates the other tabs
- Tab on raw data from the 2024 NEJM paper that populates the other tabs
- Tab on Exact Sciences financials pulled from their SEC filings
- Browser-based calculators
- Calculator for positive predictive value (PPV) and number needed to screen (NNS) calculator (the key figures of merit to determine clinical utility for a test)
- Calculator for cost to screen 1 true positive (the key figure of merit to determine cost-effectiveness of a test)
If you use these in preparing a research paper, grant proposal, or other publication, I would appreciate your acknowledging it by citing it in the references. Here is a suggested bibliography entry in APA or “author (date)” style:
Tseng, B. (2024). Calculators for diagnostic figures of merit [Computer software]. Retrieved [month day, year], from https://benjamintseng.com/portfolio/diagnostic-math/Code language: plaintext (plaintext)