Jump to comment:
- Page navigation anchor for Why sensitivity and specificity were ‘associated with’ prevalence in a Big Data studyWhy sensitivity and specificity were ‘associated with’ prevalence in a Big Data study
Even with the help of the classic tutorial series [1] and textbook [2], the topic of diagnostic probabilities remains challenging: the terminology can be unfriendly, it is difficult to combine information, and to do the math in one’s head.[3]
The “maxim” that characteristics of a test remain constant across settings[4] was reinforced in the early 2000s[5]. However, that update did emphasize that spectrum can affect sensitivity, e.g., the sensitivity of a home pregnancy test is a function of (improves with) how many days post-conception it is used, but specificity is expected to be invariant over settings. It stressed that post-test probabilities are affected by prevalence, and the separate concepts of spectrum/prevalence.
Are the relationships between prevalence and (i) sensitivity (ii) specificity in [4] at odds with the teachings of these experts? We don’t think so, and we offer explanations for the patterns, and why they should have been anticipated.
• The 6909 data points track different diagnostic tests, for different medical conditions, across 6909 settings.
They refer to different conditions. Even within say the internal medicine category, there is no logical link between a datapoint in a study involving say pulmonary embolism with a datapoint in a study involving say H pylori infection.
So why then, across diverse tests and settings, the reported patterns?
• In a low prevalence setting a very high specificity is required to...
Show MoreCompeting Interests: None declared.References
- 1. Department of Clinical Epidemiology and Biostatistics, McMaster University. Clinical Epidemiology rounds: Interpretation of diagnostic data. 1. How to do it with pictures. CMAJ 1983;129:429-432.
- 2. Sackett DL. Clinical Epidemiology: a basic science for clinical medicine. 1st ed. Boston Little, Brown; 1985.
- 3. Berwick DM, Fineberg HV, Weinstein MC. When Doctors Meet Numbers. The American Journal of Medicine 1981; 71: 991-998.
- 4. Murad MH, Lin L, Chu H, et al. The association of sensitivity and specificity with disease prevalence: analysis of 6909 studies of diagnostic test accuracy. CMAJ 2023; 195:E925-31.
- 5. Montori VM, Wyer P, Newman TP, et al. Tips for learners of evidence-based medicine: 5. The effect of spectrum of disease on the performance of diagnostic tests. CMAJ 2005;173:385-390.