Dear Editor,
I was interested to read Maughan and colleagues’ education piece on pulmonary embolism, particularly the factors that predict a delayed diagnosis and the impact of early anticoagulation.
I would like to draw attention to the use of positive likelihood ratios when describing elements from the patient history and examination in this article.
In a defined population, the positive likelihood ratio describes the proportion of patients with a positive finding (such as “patient has dyspnoea”) that are subsequently diagnosed with pulmonary embolism, divided by those with the finding who have an alternative diagnosis (1).
However, likelihood ratios consider all potential diagnoses, and cannot inform a comparative diagnostic exercise between the relatively few competing differentials that remain after history taking and examination. Furthermore, positive predictive values rarely allow us to exclude a diagnosis with confidence; a necessary part of the diagnostic process.
For example, even if the positive likelihood ratio is very high, a finding such as “syncope” may not help the diagnostician distinguish between differentials, such as pulmonary embolism and myocardial infarction.
Additionally, likelihood ratios apply only to a defined population. The parameters of this population are not defined in the article, though perhaps we may infer from author affiliations that these ratios apply to emergency department attendees. Such information should be clarified for the benefit of the reader.
1. Llewelyn H, Ang AH, Lewis K, Abdullah A. The Oxford Handbook of Clinical Diagnosis. 2nd ed. Oxford University Press, 2009:754-60
Positive predictive values do not inform comparative diagnostic exercises