or 
T = true | F = false | P = positive | N = negative
Sensitivity
Sensitivity is the proportion of all cases that have an abnormal test result. For example, if the detection of umbilical artery notches has a 90% sensitivity to predict pre-eclampsia, it would imply that 90% of women with pre-eclampsia will have an abnormal uterine artery notch.
Specificity
Specificity is the proportion of cases that do not have a clinical abnormality with a normal test result. The equivalent statement would be that 90% of women without pre-eclampsia have a normal uterine artery Doppler result.
Predictive Values
Although sensitivity and specificity are useful for looking at the entire population, they do not give information about the individual. Predictive values are more applicable for an individual (that is, they give the percentage chance of that individual being affected or unaffected). These are useful to the clinician but are also highly dependent on the prevalence of the disease concerned. If a disease is rare, even a bad test will have a high negative prediction. A positive prediction is a proportion of cases with an abnormal test result that have the clinical abnormality. So, the positive prediction of an abnormal uterine artery Doppler result (where 1 in 5 get the disease) for prediction of pre-eclampsia would imply that 20% of women with an abnormal uterine artery Doppler result will get pre-eclampsia. Negative prediction is the proportion of cases with a normal test result that do not have the clinical abnormality. For example, 99% of women with a normal uterine artery Doppler result will not get pre-eclampsia.
Likelihood Ratios
Other tests are sometimes used to look at the efficiency of a test. For example, a likelihood ratio is the effectiveness of a particular diagnostic test to confirm or exclude a particular diagnosis. The likelihood is a ratio of the event rate before and after the test, so therefore a likelihood ratio of 1 means that the patient is no more or less likely to have a condition once she has had the test. A good test will therefore have a high likelihood ratio. The advantage of using the likelihood ratio is that it can be applied to different populations who are at lower and higher risk. Likelihood ratios generally stay the same in high and low risk populations although this cannot be guaranteed. For example, a fetal fibronectin test to predict preterm birth generally has a high likelihood ratio of approximately 10–15. This is similar in a high‐ and low‐risk population, although the absolute risk (that is, the positive prediction) is very different depending on the risk status of the woman.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
