Share |
Michael Greenberg and Susan Ridgely, two lawyers from Rand Health, publish in this week’s JAMA, Clinical Decision Support and Malpractice Risk.[1] The plaintiff attorneys have it both ways. If the CDS suggests too many potential drug interactions for a new prescription and the physician ignores the lessor risks, he or she exposes himself or herself to a potential lawsuit. If on the other hand the software vender limits the number of risks on whatever basis the vendor too assumes greater risk. If the clinician withholds the medicine based on minimal risk of drug interaction, and the patient suffers, who knows, this too may be a potential tort.
The article goes on to suggest that an expert consensus further endorsed by the Office of the National Coordinator (ONC), Medicare and Medicaid, may provide a safe harbor for CDS.
My interest in CDS involves diagnosis rather than treatment and there may be a risk to the differential diagnosis as well. I would think that listing all of the possibilities for diagnosing patient problems would demonstrate the consideration of the items on the list. Furthermore, considering multiple possibilities reduces the likelihood of being wrong. Indeed, if the initial diagnosis does prove wrong, the list serves as evidence of having at least considered the right answer and rejecting that option for whatever stated reason.
Here too the issue arises of how long to make the list. With every conceivable possibility included, one runs the probability of exasperating clinicians into ignoring the entire list. Here again malpractice risks result from either too long a list or too short a one. A statistical appraisal of the list, however, might improve the odds. An expert consensus and bureaucratic endorsement may be problematic too in keeping pace with the rapid and accelerating changes in medical knowledge and understanding.
Electronic health records hold a promise of future excellence once the systems evolve. In the meantime, expect a difficult transition. As long as computers remember and do statistics, while clinicians think and integrate information, we should be all right. Computers should be good at remembering those lists that we memorized in medical school. (Let us not make them longer)
