The concepts of sensitivity and specificity help in both the collection and analysis of data. They even underlie some of the basic strategies of interviewing. Questions with high sensitivity, if answered in the affirmative, may be particularly useful for screening and for gathering evidence to support a hypothesis. For example, "Have you had any discomfort or pain in your chest?" is a highly sensitive question for diagnosing angina pectoris. For patients with this condition, there would be few false-negative responses. Thus, it is a good first screening question. However, because there are many other causes of chest discomfort, it is not all that specific. Pain that is retrosternal, pressing, and less than 10 minutes in duration—each a reasonably sensitive attribute of angina—would add importantly to your growing evidence for the diagnosis. To confirm your hypothesis, a more specific question, if answered in the affirmative, is needed, such as "Is the pain precipitated by exertion?" or "Is the pain relieved by rest?"
Data for testing hypotheses also come from the physical examination. Heart murmurs are good examples of findings with varying sensitivity and specificity. The vast majority of patients with significant valvular aortic stenosis have systolic ejection murmurs audible in the aortic area. Presence of a systolic murmur has a high sensitivity for aortic stenosis. This finding is present in most cases. The false-negative rate is low. On the other hand, many other conditions produce systolic murmurs, such as increased blood flow across a normal valve, or the sclerotic changes associated with aging, termed aortic sclerosis, so the finding of a systolic murmur is not very specific . . . there are many false positives. Using such a murmur as your only criterion for diagnosing aortic stenosis would lead to many false positives.
In contrast, a high-pitched, soft blowing decrescendo diastolic murmur best heard along the left sternal border is quite specific for aortic regurgitation. Such a murmur is almost never heard in normal people, and it is present in very few other conditions, so there are few false positives.
Combining data from the history and physical examination allows you to test your hypotheses, screen for selected conditions, build your case, and clinch a diagnosis even before obtaining further diagnostic tests. Consider the following list of evidence: cough, fever, a shaking chill, left-sided pleuritic chest pain, dullness throughout the left lower posterior lung field with crackles, bronchial breathing, and egophony. Cough and fever are good screening items for pneumonia, the next items support the hypothesis, and bronchial breathing with egophony in this distribution is very specific for lobar pneumonia. A chest x-ray would confirm the diagnosis.
Absence of selected symptoms and signs is also diagnostically useful, especially when they are usually present in a given condition (i.e., their sensitivity is high). For example, if a patient with cough and left-sided pleuritic chest pain does not have fever, bacterial pneumonia becomes much less likely (ex cept possibly in infancy and old age). Likewise, in a patient with severe dyspnea, the absence of orthopnea makes left ventricular failure less probable as an explanation for shortness of breath.
Skilled clinicians use this kind of logic even if they are unaware of its statistical underpinnings. They start to generate tentative hypotheses as soon as the patient describes the Chief Complaint, then build evidence for one or more of these hypotheses and discard others as they continue with the history and examination. In developing a Present Illness, they borrow items from other parts of the history, such as the Past Medical History, the Family History, and the Review of Systems. In a 55-year-old man with chest pain, the skilled clinician does not stop with the attributes of pain, but moves on to probe risk factors from coronary artery disease such as family history, hypertension, diabetes, lipid abnormalities, and smoking. In both the history and physical examination, the clinician searches explicitly for other possible manifestations of cardiovascular disease such as congestive heart failure or the claudication or diminished lower extremity pulses of atherosclerotic peripheral vascular disease. By generating hypotheses early and testing them sequentially, experienced clinicians improve their efficiency and enhance the relevance and value of the data they collect. They dig and collect less ore but find more gold.
This sequence of collecting data and testing hypotheses is diagrammed below.
Was this article helpful?