Like Dr. Rastegar, I too am a primary care physician and agree that this role is essential in ensuring continuity and quality of care. However, even if a primary care physician had been the only provider involved in the case I discussed in my article, the test result may not have been seen, especially since many primary care physicians are dissatisfied with their systems for tracking test results (1). If the patient's own primary care physician had cared for him in the hospital, perhaps that physician would have remembered the previous outpatient evaluation. However, having systems to ensure that outside records are available is still essential, since it is difficult for any primary care physician to remember the details of every patient seen. Furthermore, we have to acknowledge that many primary care physicians now use the hospitalist model because of time constraints, competing demands, and evidence for improved quality of care (2). Therefore, we need strategies to improve the hospitalist model and the resulting handoffs in care (3).

In response to Dr. Kessler, I agree that tuberculosis (as well as lung cancer) should have been part of the differential diagnosis when the patient first presented. This is why the chest computed tomography scan was ordered: to further clarify the diagnosis. The misdiagnosis occurred because the ordering provider never saw the result of the test, and the patient subsequently presented to the hospital with no pulmonary symptoms. Errors in diagnosis are one of the most common types of errors in the ambulatory setting (4), and work clearly needs to be done to better understand how these errors can be prevented. However, I take exception to Dr. Kessler's argument that we can blame the patient's outcome on sloppiness and that increased vigilance is the answer. Patient safety and human factors literature clearly state that even the best-trained individuals will make potentially serious errors, and vigilance is a very weak error-prevention strategy (5). Physicians practice in a health care system where test results are not easily tracked, patients are sometimes poor historians, multiple handoffs exist, and information gaps are the norm. Most human errors are induced by these kinds of system failures. Therefore, we need to redesign systems to ensure that physicians' clinical decision making and workflow make it easier to achieve the highest quality of care, and that errors, which are guaranteed to happen, are caught and mitigated.