Deliberate deception
Genuine effort
Tests that apply a cut-off value, such as those used to diagnose particular conditions (e.g. MMSE, ACE, MOCA, BDI, Conners) or those designed to detect malingering/poor effort (TOMM, RDS, Greens) typically quote sensitivity and specificity values for various cut-off scores.
Put simply, sensitivity is the percentage of total 'true positives' in the normative sample that are below the chosen cut-off. So, using deliberate malingering as an example, it is the percentage of malingerers in the sample who would score below the chosen cut-off score. A sensitivity of 0.8 (80%) means that 80% of malingerers would score below the chosen cut-off (so 20% would score above).
Specificity, on the other hand, is the percentage of total 'true negatives' in the normative sample that are above the chosen cut-off. So, using malingering as an example, it is the percentage of people who are giving genuine effort that score above the chosen cut-off score. A specificity of 0.8 means that 80% of people giving genuine effort will score above the chosen cut-off (so 20% will score below).
Using this as an example, you can adjust the cut-offs for the chart below to see how it impacts the sensitivity and specificity of the test. You can see that a higher cut-off 'catches' a bigger proportion of malingerers (higher sensitivity = less 'false negatives') but also ends up 'catching' a bigger proportion of non-malingerers (poorer specificity = more 'false positives')
The accuracy of the tool is also strongly influenced by the base-rate in the population you are currently testing (in this case, the proportion of malingerers in the group of people who access your clinic). Try changing the population base rate and you will see how this affects the number of false positives and false negatives, even though the specificity and sensitivity remain the same.
Population base rates are extremely important in interpretation. For example, if you assume a high base rate, like when the test has been validated on forensic populations, but apply it to a group with low base-rates (e.g. amongst children, people with intellectual disabilities) you are likely to have a very high number of false positives (inaccurately labelling people as malingering).
This is where positive predictive value (in this case, the likelihood that someone who scores below the cut-off is actually a malingerer) and negative predictive value (In this case, the likelihood that someone who scores above the cut-off is not malingering) are extremely important
It is worth noting that the above graph is an illustration only- typically, tests of effort cognitive screens are relatively easy, so the scores are not normally distributed (they have 'ceiling' effects)- however the same principles apply. There are other ways to plot the sensitivity and specificity of tests, including Receiver Operating Characteristic curves.