Wisnivesky and colleagues1 reported their prospective validation of a prediction rule to determine if inpatients should be isolated for suspected pulmonary tuberculosis (TB). Because missing a single case of active TB can result in great cost and danger to the health care system, health care workers, other patients, and anyone who comes into contact with the patient,2- 4 any prediction rule that may increase the false-negative rate even slightly could be dangerous to implement. Therefore, a validation study for a TB decision rule must be especially rigorous.
Further information about the patients not included in the study, the study personnel, and the reasons and circumstances behind the initial isolation decisions would be helpful. First of all, more information on the socioeconomic status, comorbid conditions (especially immunological status), length of time in the United States, and other TB risk factors of the 237 patients not included in the study would help determine how similar they were to the 516 patients enrolled in the study. In turn, this would help ascertain the likelihood of a false-negative “hiding” among the patients not included in the study. Second, information on the experience, training levels, and specialties of the study personnel is important because the effectiveness of any prediction rule that includes potentially subjective and interpretive variables, such as history and physical examination findings, may vary when used by different people. Finally, knowing the reasons why the patients were initially isolated could allow us to determine how different the initial decision making was from the prediction rule. It is unclear whether the isolation decision makers had the same circumstances and information as the study personnel. For example, the initial isolation decision makers may have been under more time pressure or not have had access to the same resources if the admission occurred during the night or on weekends. Patient history information or test results may not be available on admission or may change during the hospital stay, as the health care team has more time to collect information, the patient’s mental status or recall changes, and family members or friends become available. All of this information will help physicians to better understand the validity, generalizability, and applicability of the prediction rule and potentially avoid the significant costs and dangers of failing to isolate patients with pulmonary TB.
Correspondence: Dr Lee, Division of General Internal Medicine, University of Pennsylvania School of Medicine, 1125 Blockley Hall, 423 Guardian Dr, Philadelphia, PA 19104 (email@example.com).
Lee BY, Chen EH. Additional Information Can Enhance Validation of Tuberculosis Isolation Prediction Model. Arch Intern Med. 2005;165(15):1794. doi:10.1001/archinte.165.15.1794-a