The health care industry is plagued with a problem that harms patient safety and exacts an annual toll of $6 billion: patient record matching. Patient record matching refers to the issue of correctly identifying a patient within the same facility or across different health care organizations. Besides the monetary issue, patient matching challenges can also cause severe harm to patients. The issue is so acute that it impacts 1 in 5 patient records within the same health care system, and up to 50% of patient records are not matched in transfers.
Most of the issues associated with patient record matching stem from a lack of standards and interoperability between health information systems, which creates significant challenges when hospitals and health systems attempt to integrate vast amounts of clinical data from an array of disparate systems. Over the course of a patient’s medical history this issue exacerbates even further since the patient data can potentially get accumulated across eight different data sets: clinical data, financial data, claims data, pharmaceutical data, patient-generated data using devices and/or apps, employer enrollment and workers compensation data, clinical trials data and DNA/genome data.
A survey revealed that 33% of all denied claims result from inaccurate patient identification or information, which cost the average hospital $1.5 million in 2017 and the U.S. health care system more than $6 billion annually. Further, duplicate records account for 18% of all hospital patient records, according to the survey.
So, how do we fix it? Here are the best solutions to the patient-matching problem.
National Patient Identifier
The idea of a national patient identifier seems like an attractive solution to the problem. Other countries (registration required) such as England and Scotland have experimented with national patient identifiers. Based on the past experience of such initiatives, we can say without a doubt that we need a federated model for maintaining patient identifiers. One of the options could be to use biometrics, which has shown good results in the work done by the Bill and Melinda Gates Foundation in Africa. The challenge to its immediate adoption in the U.S. lies in the fact that it would require significant changes to workflow and significant investment in terms of costs, training and infrastructure by health care facilities and software vendors.
A ‘Smart’ Enterprise Master Patient Index (EMPI)
Another approach to the problem of patient record matching is utilizing enterprise master patient index (EMPI) software to keep track of patient identity. An EMPI is a data registry that health care organizations use to maintain consistent and accurate data on all patients, with each patient assigned a unique identifier that is used to refer to his or her records across the organization, thus eliminating the problem of duplicate records.
Hospitals that used an EMPI tool correctly identified patients in 93% of registrations and 85% of records shared with non-networked providers. In contrast, hospitals without an EMPI posted a match rate of just 24% when organizations shared records.
A key attribute of an EMPI tool is its ability to consolidate and standardize data from a wide variety of sources — claims, electronic health records, practice management systems, revenue cycle management solutions — into an accurate and comprehensive view that enables providers to follow a single patient’s journey across the entire continuum of care. This is where machine learning and the concept of “Smart EMPI” comes into play.
Faster, more accurate record matching through machine learning
Machine learning is the study of computer algorithms that automatically improve with experience. Using machine learning, a computer program can discover rules from data and can refine those rules as more data becomes available. In most cases, computers can discern these rules far better and faster than humans.
Most good EMPI solutions today employ machine learning approaches that use a combination of probabilistic record linkage algorithms coupled with methods using regional weightages and reinforcement learning. A good EMPI solution should have three main capabilities:
1. Identify all true positives and all true negatives.
2. Minimize the number of errors and assign scores to groups.
3. Minimize false positives and false negatives.
After execution of the EMPI software, the result is a longitudinal patient record that can be transparently shared among the patient's care team, optimizing care coordination. The goal of a good EMPI solution is larger than just identifying true matches. The key is to avoid false positives. It is the wrongly linked patient records that are at the root of costly medical errors.
Other Possible Solutions
Besides the national unique identifier, the biometrics solution and the EMPI-based solution, there are other approaches that can also help solve this problem, but the adoption of those solutions would be over a large period of time, since those involve standards and technology advancements that need to happen. Broadly speaking, those changes fall into these categories:
• Standardization enforced by CMS. Adoption of standards for data elements in systems that input patient data through workflow-based solutions as well as adoption of standards for data elements for the HL7/FHIR interfaces and getting all the patient management (PM)/EMR/EHR vendors to comply with the standards.
• PM/EMR/EHR vendors utilizing integration with NCOA, U.S.PS, credit bureaus, etc., for ensuring patient address information is up to date and factually correct.
• Adoption of distributed ledger technology (aka blockchain) to finally integrate patients into the health care ecosystem and allow them complete control over their data.
Different solutions out of the above options have worked in different parts of the world due to laws, policies, need and level of technology available. A robust EMPI solution, a biometrics-based solution or a national unique identifier have proven to be the best options so far.
Written by Rahul Sharma, CEO of HSBlox. Senior technology leader with a successful track record of managing and launching innovative product solutions.
Article originally published in Forbes Technology Council