In this episode, I’ll give real world patient care examples of how to remember and apply the concepts of sensitivity, specificity and predictive value as they relate to medical tests.
I struggled with remembering these concepts for a long time. Then I learned about the mnemonics SpPINS and SnOUTS which really helped me. Here is what they mean:
SpPINS is for specificity. If a test is highly specific, a positive value rules in the syndrome.
SnOUTS is for sensitivity. If a test is highly sensitive, a negative value rules out the syndrome.
Two examples to help remember sensitivity and specificity are troponin for detecting acute myocardial infarction (MI) and d-dimer for detecting pulmonary embolism (PE).
Troponin
In a patient with chest pain or ECG changes, an elevated troponin is highly specific for ruling in an MI. If the patient has crushing chest pain or ECG changes and a negative troponin, this is not enough to rule out an MI. This is because a single troponin has a low sensitivity and cannot be used to rule out an MI in the presence of symptoms. It may take 2-4 hours from the onset of the MI for the troponin to become elevated, and so the patient is always kept under observation and serial troponin levels are checked several hours apart.
D-dimer
An elevated d-dimer cannot be used alone to diagnose PE. The d-dimer is a measure of fibrin degradation products, and an elevated level only indicates an abnormal amount of fibrin degradation is occurring “somewhere”. This could be from pregnancy, stroke, MI, PE, heart failure, surgery, sepsis, malignancy and several other causes of elevated d-dimer levels. However in a patient with a low likelihood of having a PE, a normal d-dimer level has high sensitivity and can be used to rule out PE.
Predictive values and the influence of disease prevalence
Determining the usefulness of a test does not stop with specificity and sensitivity. Predictive values must also be considered. The positive predictive value of a test describes the likelihood of the patient actually having the disease given a positive result. A negative predictive value of a test describes the likelihood of the patient not having the disease given a negative result.
The prevalence of the disease in the population tested has a tremendous influence on the predictive value of any given test. Heparin induced thrombocytopenia (HIT) 4T scores and nasal methicillin-resistant staph aureus (MRSA) screens are two examples to help remember this. I’ll explain.
HIT 4T scores
The 4T scoring system for HIT involves checking the degree of thrombocytopenia, timing of the platelet drop, findings of new thrombosis, and presence of other causes of thrombocytopenia. Depending on the score the patient may be judged to be at a low, medium or high risk of having HIT. Further testing is of no value in a patient with low likelihood of HIT based on a low 4T score, but is of high value if the patient has a high 4T score. The purpose of applying the 4T scoring system before ordering further testing is to determine whether the patient is representative of a population with a low or high prevalence for HIT. The next test – platelet factor 4 antibody – has a much higher positive predictive value if used in patients with a high risk of HIT. But the test has a poor positive predictive value if used in a population with a low risk of HIT.
Nasal MRSA screens
When is it OK to use a negative nasal MRSA screen to discontinue MRSA coverage in a patient with pneumonia? It depends on the negative predictive value of this test which in turn depends on the prevalence of MRSA pneumonia in the population the test is run. Patients without gram positive cocci in their sputum cultures have a low risk of having MRSA pneumonia. But patients with gram positive cocci in their sputum culture have a higher risk of it being MRSA. Therefore a negative nasal MRSA screen can only be used to support stopping MRSA antibiotic coverage if the sputum culture is also negative. The negative predictive value of the nasal screen is 99.2% in this situation.
How do you use sensitivity, specificity and predictive values in your practice? Do you have a different way to explain it? I’d love to hear from you!
If you like this post, check out my book – A Pharmacist’s Guide to Inpatient Medical Emergencies: How to respond to code blue, rapid response calls, and other medical emergencies.
This statement is problematic: “In a patient with chest pain or ECG changes, an elevated troponin is highly specific for ruling in an MI.”
Elevated troponin is not specific, and is in fact getting less specific all the time as the assays get more and more sensitive and the threshold for a positive test is lowered. Sure, if those two other tests are positive (chest pain and EKG) AND they have a positive troponin, there’s a fairly good chance they’re having a heart attack, but that is a function of the pretest probability in patients with chest pain and EKG changes, not the specificity of troponin.
Thank you for the comment! I’m not sure I understand what you mean. I would consider the specificity of all tests a function of the pretest probability. In this study when troponin I greater than 2.0 ng/mL was used to define a positive test, the sensitivity was 85% and specificity was 91% when compared to CK-MB and ECG.