Fortune favors a prepared health care system.

Biosurveillance for infectious diseases is a field in evolution. The anthrax attacks of 2001 called attention to the need for early warning systems for terrorist attacks, and the 2009 H1N1 influenza pandemic underscored the need for tracking syndromic patterns. The Centers for Disease Control and Prevention established BioSense in 2004 to monitor national health data and to provide federal, state, and local agencies with biosurveillance data for detecting disease clusters (1). The need for effective biosurveillance systems arose from the increased vigilance of our “post 9/11” existence, but their utility has become even more apparent from the public health perspective as the nation prepared for an influenza pandemic. In 2012, the value of these systems is clear: They are essential for us to be able to detect incipient epidemics, detect bioterrorism, and protect the public health. As Bush and Perez remind us in their recounting of the early events of the 2001 anthrax outbreak in this issue (2), biosurveillance at the level of the individual clinician is critical. Louis Pasteur’s quip that “fortune favors the prepared mind” certainly applied to the diagnosis of the index case in that outbreak. If Perez had not considered anthrax in the initial differential diagnosis of gram-positive rods, the diagnosis could have been delayed for a substantial block of time, perhaps even precluding the recognition of a multistate bioterrorist attack. Infection caused by inhalational anthrax was rare in the United States, and even a single case captured the attention of public health authorities. In the case of less singular infections, individual clinicians seeing individual cases may not appreciate the presence of an outbreak. The increasingly widespread use of electronic medical records married to a sophisticated ability to query those records has opened the possibility of capturing a broad array of data that reflect the observations of frontline medical personnel. Also in this issue, Elkin and colleagues (3) compare the value of chief complaint surveillance used by BioSense with that of surveillance using the whole electronic medical records of 1203 influenza patients and 991 uninfected controls at the Mayo Clinic (3). The group used codes abstracted through natural language processing methods to encode data from each of the records evaluated into Systematized Nomenclature of Medicine–Clinical Terms (SNOMED CT) reference terminology to glean cogent meaning from the entropy of the written medical record. The same symptoms and signs used by BioSense to identify influenza from chief complaints were extracted from multiple sections of the electronic medical records of cases and controls. Whole-record surveillance was moderately more sensitive than chief-complaint surveillance alone at the same level of specificity. Of greater significance than that incremental increase in sensitivity was the demonstration that whole-record surveillance is feasible by using available tools. Assuming that use of the complete electronic medical record for surveillance is both feasible and practical, use of the entire record makes implicit sense. A key question that follows naturally from Elkin and coworkers’ study is how to utilize the information. Investment in increasingly sophisticated and sensitive syndromic surveillance tools will not have a significant benefit unless the public health infrastructure has a functioning mechanism for reviewing, investigating, and analyzing the results in real time. As Elkin and colleagues point out, syndromic surveillance should favor sensitivity over specificity, and protocols need to be in place for sorting through nonspecific alerts. Finally, none of these steps in data collection and analysis will bear fruit unless there are plans and means to respond to credible alerts. Uscher-Pines and colleagues (4) surveyed 30 health departments in 8 states and found a general lack of formal response protocols and a low level of preparedness for an alert requiring in-depth investigation and tangible action. In our view, the study by Elkin and colleagues has 2 significant limitations (both of which are identified by the authors). First, because the data were generated from the medical records of 1 academic tertiary referral center, the generalizability of the results to other academic centers and community hospitals—not to mention for records from freestanding clinics and other centers where walk-in health care is provided—is questionable. One might suspect that the quality of the records in these diverse clinical settings may be highly variable, perhaps resulting in an overestimate of the increased sensitivity of the “whole-record” technique. A second, and perhaps more challenging, aspect of the proposal to use the entire medical record for biosurveillance is cost. The authors do not estimate the additional cost of the approach they advocate, stating simply that “The cost of comprehensive biosurveillance monitoring was not studied” and that “We do not know what the final cost of comprehensive biosurveillance monitoring would be for the nation. A cost–benefit analysis should be undertaken before a large-scale investment can be made (to include both cost-effectiveness and the need for safety against catastrophic outcomes)” (3). In these days of dwindling health care budgets, before recommending any system for broad-scale national implementation, policymakers should have a clear assessment of the costs of implementation and, as the authors suggest, should also have a precise evaluation of whether the moderate increase in sensitivity afforded by the whole-record technique is worth the additional investment from the shrinking pot of available public health and health care infrastructure dollars. Especially in light of the current focus, both on health care reform as well as health care institutional “belt-tightening,” such decisions will need to be made quite carefully. However, these limitations do not detract from the proof-of-principle study conducted by Elkin and colleagues; they only serve as a reminder of the imporAnnals of Internal Medicine Editorial