Record Matching Algorithms: Close Isnt Good Enough
From the Medicare/Medicaid EHR incentive program to ACOs and HIEs, information exchange is the centerpiece of nearly every initiative designed to bend the health care quality and cost curves. The ability to access real-time clinical data wherever a patient encounter takes place drives these improvements by informing care decisions, avoiding preventable adverse events, eliminating diagnostic and therapeutic duplications and shortening lengths of stay.
Ensuring the integrity of the data upon which this transformation depends is paramount to success, both at the individual hospital level and health care system-wide. But as more systems are interfaced, significantly higher volumes of data are flowing into patient records and other clinical, administrative and financial systems. An error at any point along the way--an incorrect birth date, transposed digit in a Social Security number, missed middle initial or misspelled name --quickly snowballs as the information feeds from one system to the next.
In a single hospital, inaccurate data captured in one system can feed into as many as 50 others before it is detected (if ever). When that data is shared outside the facility’s four walls, the errors grow exponentially, impacting the integrity of all other systems through which it flows and, subsequently, any care decision for which it is accessed.
Many hospitals are taking proactive steps to eliminate duplicate and overlaid patient records from their master patient indexes. But the problem is not solved once the MPI is clean. Steps must be taken to maintain the long-term integrity of that data by preventing new duplicates from being created, as well as quickly identifying, validating and reconciling any that do sneak into the system.
Those steps must start with ensuring that record matching algorithms are capable of efficiently and accurately managing the most complex step in a resource-intense process--record identification. These algorithms aren’t often top-of-mind during the health I.T. planning and selection process, but they need to be.
Too often, facilities rely on the record matching algorithms contained within whatever clinical information system is ultimately chosen. However, these systems at best contain intermediate algorithms that rely primarily on probabilistic matching. The result is an unacceptably high number of false positives (the algorithm incorrectly identifies two records as belonging to one person) and/or false negatives (the algorithm misses identification of a true duplicate).
While these will always occur, those systems that utilize advanced algorithms focused on mathematical and/or statistical matching will deliver lower rates of each. The result is fewer duplicates entering the system, driving a much cleaner MPI and ensuring that the data that is ultimately shared is --and remains --accurate.
Whether the goal is successful participation in ACOs, HIEs and other information-sharing initiatives or strengthening strategies surrounding quality and cost improvements at the facility level, hospitals must place a priority on deploying advanced record matching algorithms. Strong algorithms ensure that MPIs remain clean and that the data used by clinicians is accurate and complete.
Beth Haenke Just (email@example.com) is CEO and president of Just Associates, a data integrity consulting firm.