10.6 Measurement quality: Sensitivity, specificity, and predictive value

Some measures used in research are used to determine whether a particular characteristic or condition exists within a person. These can be assessment tools to screen for depression, cognitive impairment, or other diagnosable conditions. We can evaluate these types of diagnostic instruments by examining their sensitivity, specificity, or predictive values. Table 10.5 shows how each of these are calculated.

Sensitivity tells you the percentage of all people in the sample with the condition who were identified by the instrument as having it. For example, what percentage of people who are actually depressed will get a positive result on their depression screener? Answering this question tells us the depression screener’s sensitivity. If the screener’s sensitivity is 85%, then the screener will detect 85% of all the people with depression. If an instrument has very high sensitivity, then it is likely it will be able to detect the condition if the condition exists. This means there will be relatively few false negatives. So if you get a negative result from an instrument with high sensitivity, you can be fairly confident the person really doesn’t have the condition. This leads to the rule of thumb called SNOUT, or “Sensitivity – Negative ruled OUT”.

Specificity, on the other hand, refers to the percentage of people who don’t have the condition that a measure correctly identifies as not having it. In our example, this would be the percentage of people who are not depressed who get a negative result on the depression screener. If the depression screener has a specificity of 90%, then it identifies 90% of all the people without depression as not having depression. When an instrument has very high specificity, there are few false positives which means you can have a high degree of confidence that a positive result really is positive. This leads to a rule of thumb called SPIN. SPIN comes from “Specificity – Positive ruled IN.”

Ideally, you want measures with both high sensitivity and high specificity in your research, but this is not always possible. Sometimes you will have to make a decision about which to prioritize based on your research question and goals.

Predictive value refers to the probability that a result from an instrument is accurate. Positive predictive value is the percentage of positive results that actually are positive. If there is high positive predictive value, then you can be fairly confident that a positive result is correct. Similarly, negative predictive value is the percentage of negative results that are actually negative. If an instrument has high negative predictive value then you can be fairly confident that a negative result really means you don’t have the condition.