Consider this hypothetical situation: A patient with new onset of a major depression is evaluated by a physician who has found no evidence of suicidal ideation. The doctor concludes that the drugs Sadoff and Newhappiness would be appropriate treatment, but she is unclear as to which is better. The doctor has just purchased a new electronic medical record (EMR) system that contains a continuously updated PDR, which she consults to review both medications. Although the doctor discovers that Sadoff has been linked to suicidal episodes, she prescribes this drug because patients are less compliant with NewHappiness (which has no known connection with suicide). When the patient does commit suicide, his family is outraged and files a wrongful death action against the doctor. During the discovery process, the family learns that the doctor knew about Sadoff ’s side effects and knowingly prescribed this drug. The family wants this information used against the doctor.
In this hypothetical, the family was able to learn about the doctor’s decisionmaking from an examination of the EMR’s metadata
. Like any computer, an EMR generates metadata during its routine operation. This metadata is functionally an audit trail of how the computer was used. Anyone who uses an EMR needs to realize that every time they open a document (including patient notes, medication lists, and radiographic images) their “digital fi ngerprint” is left behind. This digital fingerprint, which can be traced back to the physician with as much certainty as a normal fingerprint, is actually easier to detect than a normal fi ngerprint. Unfortunately, few physicians using EMRs are aware of the potential treasure trove of information available in metadata and how it might be used against them.What is metadata?
Metadata comes in two forms: system and application. System metadata are marks automatically made on electronic documents by computers. Examples include computer markings as to when a document was signed (time-signature metadata) and by whom (identity metadata). Without system metadata, a computer cannot function properly. Unlike humans, a computer cannot actually remember anything. For example, when the “back key” of a Web browser is hit, to find the last website visited, the computer must consult its system metadata. In the hypothetical presented, analysis of EMR system metadata allowed the family to discover whether the doctor consulted a PDR and what medications the doctor reviewed.
In contrast, application metadata is generated by the computer user. Microsoft Word’s “track changes” feature is probably the best example of application metadata;when the feature is “on,” Word displays some of the application metadata associated with the program (the history of changes made to that document). The potential for application metadata to “un-erase” comments made in an EMR is potentially frightening. For example, if in the heat of the moment, a physician enters into the EMR an unprofessional comment concerning another doctor, examination of the EMR’s metadata might discover the comment, even though it was deleted. Fortunately, for technical reasons that are beyond the scope of this article, such application metadata may be stored only for a limited time.Metadata uses
The ways in which metadata can be used are limited only by the imagination. Th e following examples explain how metadata might be used to objectively measure a physician’s productivity and quality of care and how it might be used as evidence.
In the business world, a classic way to obtain objective data on an individual’s productivity is to perform a time-and-motion study. To perform such a study, an observer with a stopwatch follows the subject throughout the course of a business day, timing how long it takes to perform various job-related tasks. As the need to employ an observer makes such studies labor intensive, many businesses cannot afford to routinely perform timeand- motion studies. However, in healthcare, time-and-motion studies are going to become more popular, with physicians electronically tracked throughout their workday via a combination of identity and time-signature metadata
. Such data will provide employers with an estimate of how much time a physician actually spends with patients. Indeed, some software vendors currently off er products that can automatically perform time-and-motion studies on primary care physicians and psychiatrists.