infectious tuberculosis face a significant obstacle: the common symptoms of active tuberculosis are not highly specific to the disease. Particularly in areas where the disease has been rare for a considerable period, physicians are more likely to think of other, more common conditions—for example, bronchitis or lung cancer—when they see someone with symptoms such as a persistent cough. Clinicians likewise may not initially link radiologic signs of active tuberculosis to the disease. Moreover, some individuals with active tuberculosis, especially those with HIV infection, may show radiologic signs that are not typical of the disease, and a few may experience no symptoms. In addition, as tuberculosis has become less common and clinicians and laboratory personnel see fewer instances of the disease, it becomes more difficult for them to maintain proficiency in obtaining, processing, and accurately interpreting specimens.
Two recent reviews of episodes of tuberculosis transmission in health care facilities found that nearly all instances involved source cases with undiagnosed and untreated disease (Dooley and Tapper, 1997; Garrett et al., 1999). Many also involved lapses in infection control processes. In several instances, the patient identified as the source case had atypical radiologic signs and negative smears. Thus, the failures were not just the result of inattention to obvious signs and symptoms.
Unfortunately, because the symptoms of active tuberculosis are nonspecific, early identification protocols are likely to identify a sizeable percentage of people who do not actually have the disease. Isolation of these people results in an expensive, unnecessary use of isolation rooms. For example, one study in a high-prevalence hospital found that for every eight patients isolated, only one had confirmed tuberculosis (Bock et al., 1996). Later, as tuberculosis case rates dropped, this ratio increased to 10 to 1 (Blumberg, 1999). In low-prevalence areas, the ratio may be as high as 100 to 1 (Scott et al., 1994).
Researchers and clinicians have attempted to develop more precise diagnostic criteria and decision-support tools to allow quicker and more accurate identification of people with tuberculosis (Scott et al., 1994; Bock et al., 1996; Knirsch et al., 1998). This would promote earlier isolation and treatment. It would also help conserve resources by reducing the number of people incorrectly identified and isolated (false positives). Typically, however, such adjustments in prediction rules or decision criteria will result in more missed cases of active, potentially infectious disease (false negatives). For example, in one study, the use of prediction rules that were developed to reduce overisolation of nontuberculosis patients would have reduced the number of such patients with false-positive results from 253 to 95, but it would have missed 8 of 42 patients with (those with true disease) false negative results (Bock et al., 1996).
In addition to increasing the costs of care, overisolation may be emotionally stressful for patients. It typically reduces an individual’s contact