Patient matching algorithms tackle medical record duplication

Hospitals see a large volume of patients throughout the day. Some of the encounters are first visits, while others are follow-up visits or emergency care. The ongoing flow of patients registering in the system generates more data, including data sent by external referring sources. Despite the best efforts to ensure that the new data does not create duplicate records, hospitals still deal with errors associated with multiple records or incorrect patient matches.

Duplicate medical records are typically the result of an existing patient registering in the system as a new one. This can happen when the patient's existing registration is not located or someone fails to look them up by the appropriate data. This results in the patient being in the system twice and introduces risks for new data, such as lab results and medical images, to be associated with two different patients. A master patient index (MPI) plays a critical role in helping match patient information to ensure that duplicate identities can be merged. These systems typically consist of patient matching algorithms that can score the search results to offer confidence in the potential matches and assist in merging duplicates.

Unfortunately, even some of the best systems can have flaws. Because most MPIs rely on matching fields from the entry point to existing data, sometimes name pair swapping can lead to mismatching when in reality it is the same patient. This can be complicated by the rise in hospital mergers and acquisitions, which can result in even more record systems coming together. Patients are also being referred across multiple healthcare systems, creating more data along the way. The result of more data and multiple systems within one big health system can also mean increased risks for data duplication and medical errors due to mismatches.

Death due to medical errors and the link between a percentage of those medical errors to identity errors are alarming, but studies indicate that it is a reality that hospitals need to tackle. The importance of identifying the right patient is not just a priority for hospitals to reduce duplication of records and reduce patient identity issues. Patient matching algorithms serve a key role when it comes to supporting interoperability, and that's the exchange and sharing of data for patients across multiple systems that may not be under one health system.

Several initiatives encourage the use of artificial intelligence and enhanced patient matching algorithms that can deliver a higher level of matching than traditional products. The use of artificial intelligence capabilities, such as machine learning, neural networks and advanced mathematical and statistical models, helps an eMPI deliver a higher matching ratio.

To encourage and support innovation in the marketplace around newer and improved methods of patient matching and indexing, ONC has announced a new challenge. ONC will provide participants with a large data set that contains a small set of true match pairs. Participants must run their patient matching algorithms that will be evaluated and scored against this "answer key." Six cash prizes are available, for a total of up to $75,000. This program highlights ONC's efforts to encourage more innovation and address some of the challenges that result from patient identity issues.

Whether the best technology comes in the form of open source software or is sold through one of the top vendors in the marketplace, it's clear that no matter how advanced the technology may be, it cannot eliminate issues resulting from human errors. In addition to MPIs, hospitals should consider defining processes around patient lookups and validations for employees capturing the data. By driving changes in user behavior and increasing training, systems will have fewer duplicates.