In the setting of medicine Precision and Accuracy are ways to measure the quality of a given measurement (laboratory tests, imaging etc.). The concepts are somewhat similar. A test with high precision and high accuracy is a great test and a test with low accuracy and low precision is a poor test. However, accuracy and precision are not always directly proportional (AKA one can be high while the other is low). Precision and accuracy are distinct concepts and therefore care must be taken not to confuse them.

• Precisions = how consistent the given measurement is
• Accuracy = how well the given measurement represents the truth

 

Precision in this setting is a synonym of reliability, consistency or reproducibility. If the same patient receives an identical result when a given laboratory test is performed 5 different times that test is very precise. This is an example of test-retest reliability or internal reliability, concepts you should understand but terms you don’t necessarily need to memorize. Another illustration of precision is if the same patient goes to five different doctors that use 5 different laboratory services. If each of the 5 labs give the same exact result for a test performed on a single patient that is a very precise test. This example illustrates external reliability.Another way to think about precision is to correlate it to the concept of standard deviation. A precise test is one that has a low standard of deviation. This makes sense as we already covered that standard deviation is a measure of dispersion or variability that compares each measurement in a set to the mean. Click here for a link to our video on standard deviation.

Accuracy or validity is defined as how well the measurement or test represents the “truth.” What is the truth in this setting? Consider a scenario where I come up with a blood test that estimates a person’s height. Forget for a second that that is a preposterous test, and imagine that my blood test estimates a person height at 5 feet 1 inches when they have recently been measured at 6 feet 3 inches tall using traditional measuring techniques. In this case the established “gold standard” measurement (the best available measurement under reasonable conditions) of manually measuring a person’s height with a tape measure represents the truth. Now there might be slight inaccuracies with manually measuring a person’s height (they could be wearing shoes during the measurement, the person measuring their height could have made an error etc.), but for the purposes of this conversation we assume the gold standard test is the true value. Using measured height as the true value, it is easy to see that my blood test seems to be very inaccurate.

 

Another way to think about accuracy is to think of accurate tests as having low bias. Bias basically is the lack of systemic error or a lack of non-random (directional) deviation from the truth. Click here to see our video that covers bias.

 

It would seem intuitive that accuracy and precision would be directly proportional (AKA either both are high or both are low). After all good tests have high accuracy and precision, while bad tests have low accuracy and precision right? Not necessarily. There are situations where accuracy can be high and precision can be low or vice versa. Test writers like to write questions about these scenarios as it requires a more in depth understanding of the concept. Consider situation where you do the same blood test twice for a single patient. You get back 2 results for this patient, 0 and 1000. If the “true” value (gold standard value) is 500, your test has high accuracy (the average of the two tests is close to the true value) but low precision (there is huge variability among the two data points). Now consider doing a different blood test and receiving values of 5, 5.1 and 5.2. This data set is obviously precise as there is very little variability, but if the true value is 10,000 then this measurement is very inaccurate.

Specificity and sensitivity are basically ways to measure accuracy of a particular measurement. Click here to see our video that discusses specificity and sensitivity. Positive Predictive Value (PPV) can also be thought of as a measurement of precision. Click here to see our video on Positive and Negative Predictive Value.