I thank Drs. Wu and Shah for their thoughtful letter. Just to be clear, I was not criticizing bar-code systems in general. Bar coding is the perfect solution for many identification problems, for example, the identification of groceries at the checkout counter; we all appreciate the resultant faster checkout times. My criticism was limited to bar-coding systems used to verify medication dispensing at the bedside. Both published reports and recent conversations with disappointed CEOs suggest that these systems have not delivered the nursing time savings and error elimination they promised. Furthermore, they appear to change nursing priorities—an unintended consequence. I am sure that with time and improved technology these systems will meet their mark, but the available evidence contradicts the frothy hype for rushing to implement these systems today.

There is a general lesson here. The health care industry expects information technology, such as bar-coded medication dispensing and physician order entry, to solve many or most of its problems. These are plausible expectations, and many will eventually be realized. However, today, the hype and the hope have run far ahead of the evidence and experience. We need to critically examine the benefits and harms of these systems as implemented in operating environments to determine what works and what doesn't and to identify design flaws and misassumptions as Patterson and colleagues (1) did for bar coding and others have done for computerized physician order entry. Only with such knowledge will these products improve and reach their full promise.

The expectations about Six Sigma's potential for health care may also be a bit inflated. Even in manufacturing, the real achievement may only be four and a half, rather than Six, Sigma, and some question the applicability of these principles to fields other than manufacturing (2). Given that health outcomes are not influenced by many-fold increases in care process investment (3) and the fact that care costs have reached crisis levels, one might wonder whether this is the time to eliminate costly but marginal processes rather than investing even more to perfect them.

I agree that methods for capturing data derived from patients have value and am a fan of the work by Hripcsak and colleagues (4). Natural language processing should become a major asset to medical information management.