true positive observations
198 false positive observations a b cd false negative observations
792 true negative observations
207 total positive observations
793 total negative observations
10 persons with X
990 persons without X
You are now confronted with possibly upsetting 207 people (all those with positive observations) to detect 9 out of the 10 real cases. The predictive value of a positive observation is only 4%. Improving the specificity of your observation without diminishing its sensitivity would be very helpful, if it were possible. For example, if you could increase the specificity of the observation from 80% to 98% (given the same prevalence of 1% and sensitivity of 90%), the positive predictive value of the observation would improve from 4% to 31%—scarcely ideal but certainly better. Good observations or tests have a sensitivity and specificity of 90%.
Because prevalence strongly affects the predictive value of an observation, prevalence too influences the assessment process. Because coronary artery disease is much more common in middle-aged men than in young women, you should pursue angina as a cause of chest pain more actively in the former group. The effect of prevalence on predictive value explains why your odds of making a correct assessment are better when you hypothesize a common condition rather than a rare one. The combination of fever, headache, myalgias, and cough probably has the same sensitivity and specificity for influenza throughout the year, but your chance of making this diagnosis correctly by using this cluster of symptoms is much greater during a winter flu epidemic than it is during a quiet August.
Prevalence varies importantly with clinical setting as well as with season. Chronic bronchitis is probably the most common cause of hemoptysis among patients seen in a general medical clinic. In the oncology clinic of a tertiary medical center, however, lung cancer might head the list, while in a group of postoperative patients on a general surgical service, irritation from an endotracheal tube or pulmonary infarction might be most likely. In certain parts of Asia, in contrast, one should think first of a worm called a lung fluke. When you hear hoofbeats in the distance, according to the familiar saying, bet on horses, not on zebras, unless, of course, you're visiting the zoo.
Chapter 1. An Overview of the Physical Examination and History-Taking
Was this article helpful?