I have previously blogged about the chronic and continuing problem of duplicate EHR records (see: Duplicate Patient Records as a Major and Costly EHR and EMPI Problem). Think of the problem in this way: it's relatively easy for a hospital registrar to create a new patient record but very risky to merge two nominally identical patient records because they may not represent the same person. This reluctance to merge records results in hidden data that may be critical in the care of a particular patient. A recent article dealt with this problem (see: Why Is It So Hard to Match Patients With Their Medical Records?) and below is an excerpt from it:
In the half-century since EHR technology tiptoed into health care, hospitals and doctors have used hundreds of vendors, each of which is constantly updating its technology, thereby creating new opportunities for inconsistencies. Today, on average, 18 percent of patient records within organizations are duplicates....And the match rates between organizations — for example, between a doctor’s office and the hospital — can be extremely low....That means physicians are often unaware of information that patients assume they know: test results, diagnoses, medications, and more....
On its face, patient-matching looks like a simple problem to fix, until you look closely. One root problem is that the health care industry has been consolidating for years, and the pace of mergers is only speeding up. Take the example of Northwell Health. Already the largest health care provider in New York State, Northwell is growing rapidly through acquisitions. Every time the system adds a new hospital, their medical records are integrated — and duplicates proliferate.... Meanwhile, Northwell’s 23 hospitals and more than 700 clinics are creating new duplicate records every day....Rather than try to figure out which of those 50 is the right one, [the registrar] will create a new case, knowing that there’s a way to reconcile it later on....In 2016, Harris Health System in Houston reported it had 2,488 records with the name “Maria Garcia;” of those, 231 shared the same birthdate, suggesting some of them refer to the same individual. ....
Pew did a deep dive into these and other potential solutions. Its finding: None of the ideas are as easy as they look at first glance. All the opportunities we examined to address matching have benefits and drawbacks — and none of the opportunities alone can solve this problem ....[One idea that has been] suggested but never implemented nationwide: data standardization. If all health care organizations collected certain pieces of demographic data in a uniform way, patient-match rates would increase significantly. Just a little bit of standardization could go a long way....[A] unique patient identifier — a single number, possibly linked to an iris scan or other biometric, that distinguishes one [person from another] might be the most logical solution to the problem. .... In the U.S., however, it is essentially against the law for the government to invest in a unique health identifier.
On the face of it, artificial intelligence could possibly provide a solution for this problem by using algorithms to scrutinize multiple duplicate records and determine which ones represent a single patient and which ones do not. However, a recent article introduced a note of caution into such process. It emphasized that bad data can produce worse models and it's a certainty that many of the duplicate records considered for merger may be replete with incorrect data (see: AI can improve EHRs and analytics, but physician input is a must). Here the relevant quote from this latter article:
...[A]utomation has its perils too, of course. For instance, when developing predictive models, "there is nothing more critical than the data. Bad data (such as from the EMR) can be amplified into worse models. Simply allowing machine learning to run without human intervention presents clear risks. Instead, clinicians should seek a partnership in which the machine predicts (at a demonstrably higher accuracy), and the human explains and decides on action.
Comments