Opportunities to mine EHRs for malpractice risk management and patient safety

The malpractice system is an important, albeit imperfect, mechanism to compensate patients for healthcare harms. In theory, the system incentivizes clinicians and health systems to provide safe and effective medical care. Yet US malpractice claims are still frequent, averaging around 50,000 annually, with payouts totaling approximately $4 billion per year. Moreover, costs associated with defending claims are large: approximately $85,000 per case. Underlying these figures is our limited understanding of malpractice risk. While sophisticated prediction models now support clinical decision-making, malpractice risk assessment largely relies on coarse categories such as specialty (e.g. obstetrics) or simple trends (e.g. a clinician who has been the subject of multiple lawsuits or patient complaints). As a result, those working to reduce malpractice risk and improve patient safety cannot identify individual clinicians or individual patients at highest risk at a given point in time, which impedes optimal targeting of resources. Further, many resources are devoted to addressing risk after an incident occurs. For example, in communication and resolution programs, health systems and insurers encourage risk management teams and clinicians to reach out to patients and families after unanticipated outcomes, to seek a resolution that may include an apology and an offer of compensation. This approach is still reactive and addresses harm events on a case-by-case basis. Where might we find a more scalable, precise, and proactive approach? Perhaps the answer is hiding in plain sight. A detailed medical record review is the cornerstone of every malpractice case. Yet, even with near-universal electronic health records (EHRs), there has been little effort to mine records in real time—and before an untoward event—for predictors of safety events or for lawsuit risk. We believe that EHRs—and specifically the data on clinician behaviors that they contain—offer untapped potential to advance malpractice risk mitigation and patient safety. Of course, some domains of malpractice risk—namely, medication errors—have been targeted for improvement by EHRs. Yet other, equally risky domains remain unaddressed. For example, failures to act on abnormal results and concerning symptoms remain major contributors to malpractice. There are many opportunities for EHR data to identify and mitigate risks by detecting when clinicians do not open or address abnormal lab or radiology results. Clinicians with significant delays can be identified and the behavior addressed, even before one of these delays leads to harm. Taken further, automated EHR-based rules could detect when a result has not been viewed, or when it has been viewed but expected subsequent actions, such as subspecialty referrals or additional testing, have not occurred. Kaiser’s SureNet program offers one example of this approach that has been implemented at scale. This is a centralized effort to identify specific high-risk unaddressed results and intervene before patients experience harm. Examples include identifying abnormal prostatespecific-antigen tests that may be indicative of cancer as well as improving diagnosis of chronic kidney disease by identifying when an elevated blood creatinine level is not followed by a second test within 90 days. Even though SureNet has been implemented at scale, such efforts are the exception, not the norm.