Spectrum Effect in Tests for Risk Prediction, Screening, and Diagnosis
2016
This article discusses the impact of the spectrum effect on measures of test performance, and its implications for the development, evaluation, application, and implementation of such tests. The authors describe this effect as the variation between settings in performance of tests used to predict, screen for, and diagnose disease among different population subgroups. They emphasize that a test developed in a population with a higher prevalence of disease (or at higher risk) will typically have a lower sensitivity and higher specificity when applied in a population with lower disease prevalence (or at lower risk).
In addition to a discussion of the spectrum effect with examples from the literature, quantitative examples are used to illustrate variation in the sensitivity, specificity, and likelihood ratio, with different assumptions about the prevalence and population distribution of the condition. They point out the importance of considering the spectrum effect in the development, evaluation, and choice of tests.
While, in an ideal world, new tests should be developed and evaluated using data from the population(s) in which they are intended to be used, this may not be possible in a context such as a global pandemic with a new infectious agent. In this instance, an understanding of the types of biases that emerge with assumptions about data performance from a very specific population will help in using clinical reasoning when interpreting test results from the general population.
Spectrum Effect in Tests for Risk Prediction, Screening, and Diagnosis
Source:
Usher-Smith JA, Sharp SJ, Griffin SJ. The Spectrum Effect in Tests for Risk Prediction, Screening, and Diagnosis. BMJ 2016; 353. https://doi.org/10.1136/bmj.i3139