Summary

Of the data types discussed in this chapter, clinical call centers have perhaps the greatest potential because of the interaction between a patient, a nurse, and a decision support system (the documentation system based on clinical guidelines) relatively early in the course of illness. This combination not only leads to a high level of clinical detail elicited by highly trained nurse but opens up the opportunity for adaptive questioning driven by Bayesian priors computed from probabilistic output of regional biosurveillance systems. Moreover, there is a long-term trend towards increased use of call centers in healthcare because of potential cost savings.

Prescriptions and test orders simply have not been studied, although we expect this situation to change quickly as existing biosurveillance systems are collecting these data and are being studied.

Of the preclinical data we discussed, telephone call volume has high potential because of the earliness suggested by the described study and that telephone systems are already so highly automated. However, call volume is extraordinarily non specific (caller may be sick), so its role seems to be limited to detection of sudden outbreaks that affect a large fraction of a region. Much more study of the potential of volumetric telephone calling is needed.

With polling and self-reporting, the big question is sustain-ability and feasibility. The one existing study of polling was for a limited duration in a permissive environment, and further work is needed to assess the costs and benefits for routine polling of larger populations. Restaurant sales do not have strong evidence supporting their use, and are generally difficult to obtain. The most likely use of restaurant sales would be for site-based surveillance in a permissive environment. The potential value of presymptomatic data awaits development of technology—it hasn't been studied at all.

Permissive environments allow enhanced surveillance, and therefore could represent a promising sentinel population for a surrounding community. Further study is warranted to investigate this question.

As with all of the data discussed in Part IV of this book, each of the data types described here may have a potential role in an all-threat biosurveillance system. It is inconceivable (at least for the foreseeable future) that there will be a one-size-fits all type of data as there was in the days of notifiable disease reporting. People with the same disease behave differently (not everyone sees a clinician), and people with different diseases behave quite differently. Data type X may show the earliest signal for outbreak Y, but show no signal whatsoever for outbreak Z.

We expect that a great deal more will become known in the next few year now that research methods have been worked out by the studies described in Part IV. There is at least one biosurveillance systems that is routinely collecting nearly every one of these type of data and investigators are poised to seek data retrospectively when outbreaks occur.

A great deal more research will be needed as the computerization of the world progresses and new types of data become available for study. The speed at which research will progress is limited by funding, availability of data, and occurrence of outbreaks.

Was this article helpful?

0 0

Post a comment