Clinical Decision Support and Malpractice Risk

Clinical decision support (CDS) refers to electronic technology used to enhance clinical decision making. For example, computerized physician order entry with integrated CDS in principle offers an electronic layer of review for ordering prescriptions.

An important feature of CDS is automated warnings issued whenever potential drug interactions or other contraindications arise. In practice, however, CDS systems often have been overinclusive in the warnings they generate, to a point at which physician “alert fatigue” may in large part undermine the utility the systems offer.^{1 - 2} The current generation of CDS systems includes alert parameters for thousands of drug interaction types. Meanwhile, a recent review of empirical studies on computerized physician order entry with integrated CDS observed that physicians override automated warnings a substantial fraction of the time—according to one study,³ in as many as 19 out of 20 instances. One paradoxical result of overly abundant warnings may be to exacerbate malpractice risk for physicians who either ignore or turn off CDS alerts, even as CDS systems create an audit trail to show that those physicians have done so.³ Another paradoxical result may be impeded adoption of CDS technologies⁴ because of physician and institutional concerns about malpractice risk. These sorts of results could prevent the technologies from achieving their potential benefits in making patients safer and in reducing the risks of medication error.⁵

The Office of the National Coordinator for Health Information Technology (ONC) is addressing this problem through its Advancing Clinical Decision Support project. The ONC effort includes a series of tasks designed to promote CDS development and adoption and to overcome barriers to CDS use. One element of the project involved a review of some liability challenges that have impeded CDS; in particular, the development of an optimized clinically meaningful drug-drug interaction list. The use of this list could potentially ameliorate the problem of overinclusive CDS warnings but may create the risk of another set of liability concerns on the part of vendors. Consequently, policy makers may need to untie a liability knot that threatens to strangle CDS. Better understanding of the liability knot, and of strategies available to loosen it, is important for both the policy and the health care communities.

For physicians, the main effects of CDS on liability depend on whether CDS systems are well designed and well implemented. A good CDS system, by definition, will not overwhelm the user with extraneous alerts to a point at which fatigue sets in and CDS warnings are ignored. By extension, a good system should enable physicians to detect and prevent some medication adverse events that would otherwise go undetected: a result that inherently reduces liability for physicians.⁶ In other words, physicians actually have a malpractice incentive, rather than a disincentive, to adopt good CDS systems.

However, many CDS systems available today are not good, in the sense that they cause alert fatigue.^{2 ,7} This problem could perhaps be addressed by retuning those systems to reduce the frequency of automated warnings. The idea is to adopt a nonproprietary, optimized drug-drug interaction (DDI) list, which would focus CDS warnings on a much smaller set of interactions most strongly associated with harm. Such a list might guide CDS systems to eliminate some categories of low-risk warnings while restricting others only to circumstances in which the risk is most clinically meaningful (eg, on first administration of a potentially harmful combination of drugs). Although the ONC has now performed some of the initial work to develop a consensus-based DDI list, the adoption of such a list into CDS also carries some liability risk. In particular, technology vendors may be concerned that CDS systems that deliberately neglect to provide warnings for known categories of adverse drug events, no matter how obscure, might create a basis for new products liability claims against vendors. Similar liability concerns might arise for physicians and hospitals, in their decisions about whether to tweak local CDS systems to restrict warnings solely to those specified by an optimized DDI list.

From the standpoint of policy makers, the basic challenge is to ensure that liability concerns do not derail the clinical value of new CDS technology. On balance, CDS is supposed to make patients safer and reduce the occurrence of preventable medication injuries. To the extent that physician and vendor liability risks result in suboptimal CDS technology or in the inability to adopt a tailored DDI list, the ultimate effect of tort law will be to make patients less, rather than more, safe.

Applying CDS to make patients safer involves solving a complex risk-management problem. Automated warnings to physicians can prevent some, but not all, adverse events. The question then becomes how to prioritize clinical events so that only a manageable number of automated warnings are given and how to protect physicians and vendors from new liability, given CDS systems that deliberately omit warnings for relatively low-risk drug interactions, despite the likelihood that some patients will nonetheless be harmed by them.

The first step to ameliorate the problem is to develop an expert-consensus DDI list. Because automated warnings to physicians fundamentally involve a limited opportunity to identify potential mistakes, someone must decide which events to prioritize. Those decisions about priority ought to correspond to the standard of care in medicine, based on the best available knowledge and state of the art in practice. “Expert consensus” is a formal procedure for defining what that standard is. With regard to DDI, expert consensus will relieve technology vendors and physicians of some of the onus of deciding what the appropriate risk trade-offs in automated CDS warning systems ought to be.

The second step involves endorsement of a consensus-based DDI list by relevant professional societies and certification or endorsement of related CDS technologies by regulators like the ONC, the Centers for Medicare & Medicaid Services, and the Joint Commission. These endorsements would provide professional and government imprimatur to a consensus DDI list, to the risk-management priorities the list embeds, and to the correspondence between the list and the formal standard of care in medicine. Professional and governmental endorsement would also constitute a collective affirmation that CDS risk management is being dealt with primarily through formal standard-setting efforts, rather than by tort liability after the fact.

Perhaps the most controversial approach would involve enacting new federal legislation to create a liability “safe harbor” for physicians who use CDS. Legal safe harbors can take many different forms. In this case, the idea would be to establish that if a physician has managed medications using CDS and a consensus-based DDI list, the choice not to receive a more inclusive set of automated warnings cannot be used as a basis for malpractice liability or as evidence of negligence. With regard to technology vendors, a safe harbor could further stipulate that the fact that a CDS software package incorporates a consensus-based DDI list (or allows users to modify automated warnings to comport with a DDI list) cannot be used as a basis for products liability claims against the vendor.

Whether this kind of legislative carve-out from tort liability would be prudent or practical remains to be seen. What does seem clear, however, is that CDS represents a situation in which malpractice and products liability can too easily lead to a perverse equilibrium in which the law has a detrimental effect on technology and in which patients, physicians, institutions, and the government are all made worse off as a result.