Assessing disease prevalence from inaccurate test results: teaching an old dog new tricks

Med Decis Making. 1994 Oct-Dec;14(4):369-73. doi: 10.1177/0272989X9401400407.

Abstract

Estimates of disease prevalence are needed for the interpretation of test results as well as for public health decisions. Assessing prevalence may be difficult if a definitive test is unavailable, impractical, or expensive. A formula derived from Bayes' theorem can calculate the prevalence of disease in a population by incorporating test results with a knowledge of the sensitivity and specificity of a test. This paper reviews this formula and provides examples evaluating the prevalence of HIV disease, the usefulness of ventilation-perfusion scans in diagnosing pulmonary embolism, and settings where screening tests should not be applied. These examples demonstrate that precise yet inexpensive estimates of disease prevalence are possible by enhancing the usefulness of an inaccurate test.

MeSH terms

  • Bayes Theorem*
  • Blotting, Western / economics
  • Breast Neoplasms / epidemiology
  • Breast Neoplasms / prevention & control
  • Confidence Intervals
  • Costs and Cost Analysis
  • Enzyme-Linked Immunosorbent Assay / economics
  • Female
  • HIV Infections / epidemiology
  • Health Services Research / methods
  • Humans
  • Mass Screening
  • Mathematical Computing
  • Monte Carlo Method
  • Osteoporosis / epidemiology
  • Osteoporosis / prevention & control
  • Predictive Value of Tests
  • Prevalence*
  • Pulmonary Embolism / diagnosis
  • Pulmonary Embolism / epidemiology
  • Pulmonary Embolism / prevention & control
  • Risk Factors
  • Sensitivity and Specificity
  • Ventilation-Perfusion Ratio