Vol. 15 •Issue 12 • Page 25
Data Quality in the EMPI
Sealing the cracks in the patient identification foundation is critical to ensuring safe, successful clinical initiatives in health care facilities.
For health care providers to deliver high-quality care and be successful both clinically and financially, they must have high-quality data—especially the data that identifies patients and links them with their complete medical records. And yet, according to first-time evaluations of 112 master person index (MPI) files performed over a 4-year span, the quality of data in these files is alarming. Nearly 99 percent of the MPI files had error rates higher than the “gold standard” needed to identify patients accurately, efficiently and consistently. Sealing these cracks in the patient identification foundation is critical to ensuring safe, successful clinical initiatives.
Between January 2000 and December 2003, Initiate Systems, a provider of software for creating a complete view of data about people, households and organizations, analyzed more than 300 MPI files at facilities across the country. Just over one-third of those files (112) were first-time evaluations of single-provider MPIs. These files represent the typical MPI that has not benefited from the advanced accuracy or improved data integrity that comes from enterprise master person index (EMPI) software and services.
The MPI error rate is the percentage of records in an MPI that do not present the complete information for the individual; that is, the records that have some amount of duplication. Another way to think of MPI error rate is as the proportion of records that would need to be examined manually to eliminate duplication.
The overall mean error rate of the files analyzed was 8 percent, which is four times worse than the maximum level of 2 percent that health care providers generally deem acceptable for efficient, effective patient identification. Only two files achieved the 2 percent or less “acceptable” level and less than one-third had an error rate fewer than 5 percent. In larger MPI files, those with 1 million or more records and supporting multiple facilities, over half had duplication rates of 15 percent or greater.
In addition to analyzing single-facility MPIs, a number of EMPI files were analyzed during the past several years. An EMPI combines data from across multiple facilities, sources or systems.
When data errors from individual facilities are rolled up into a larger conglomeration, the errors are magnified. The EMPI files had error rates ranging from 7.3 percent to 39.1 percent. The best EMPI was still more than three-and-a half times worse than the standard for accurate, effective patient identification.
These error rates point to the fact that patient safety is at risk. It is generally accepted that the potential for an adverse impact on patient care is substantial when more than 1 in 10 medical records has only part of the patient’s information within the hospital’s system. Presenting only partial or fragmented data to a clinician means that key patient history, such as allergies, previous surgeries, adverse reaction, medication, etc., are therefore not available to facilitate effective decision-making and higher patient safety.
These findings indicate a substantial number of files exceed this benchmark. There is also a clear upward trend in error rates as file size increases. The 49 files with fewer than 500,000 records had a mean error rate of 7.2 percent, while the group with more than 1 million records had a mean error rate of 9.4 percent.
What Causes These Errors?
Errors can be caused by three categories of factors: underlying discrepancy in patient identification attributes, operation deficiencies in the registration process or limitations of search technology in existing registration systems.
Underlying discrepancy in patient identification attributes happens due to data fragmentation, which is the result of natural variation in person-identifying information such as misspellings of names, transpositions of letters or numbers, use of aliases or nicknames, and variation in other identifying attributes, such as address and phone numbers, which change over time. It is clear that registrars face a significant challenge in ensuring that data errors do not occur and that duplicate or split records are not created. Facilities should place an emphasis on developing procedures to protect against input errors on the part of registrars.
Errors can also be caused due to limitations of search technology. Most of the admission, discharge, transfer (ADT) systems still use some form of exact-match logic in the registration process. If a registrar enters a name into a search, but commits an error, such as transposing letters in a name, the ADT system will not find the correct record. The converse is also true; if the registrar keys in the information correctly, but there are errors in the corresponding record that is already in the database, the ADT system will not be able to find that record either. Typically in these scenarios, registrars give up on identifying a patient’s record to start the process to assign a new medical record number—thereby creating a duplicate record.
The Damaged Foundation
As organizations move toward electronic health records (EHR) and the clinical data repository (CDR), errors in patient records will become increasingly problematic. At present, often if a record is not found, the HIM staff conducts a manual search. In the future, as the EHR becomes the rule rather than the exception and information from multiple facilities and systems are linked, clinicians will increasingly use the EMPI to find and retrieve patient records.
In this electronic process, clinicians will expect the EMPI to have the most accurate and up-to-the-moment data and will no longer necessitate a manual safety net. The safety of the patient and the effectiveness of the entire care delivery system will depend on having a single, trusted system of record for each patient’s data.
What’s the Solution?
The ideal solution would provide an accurate, on-demand, trusted view of every patient at every point of care, in every system, in every facility throughout a delivery network. Only when hospitals and other health care delivery points have accurate data when and where it’s needed can old errors be corrected and new errors prevented from occurring.
This solution is reached by the implementation of an EMPI solution that has a higher level of accuracy in searching and matching patient records. Leading EMPI software solutions have advance “probabilistic algorithms,” which can find and link patient records even if they have the types of errors listed above. EMPI solutions accurately integrate data across disparate sources and deliver synchronized information throughout an organization, helping to ensure increased patient safety and improved quality of care.
Michele O’Connor is the senior manager for the Healthcare Services Group with Initiate Systems, Chicago. She also serves as the current president for the New Jersey Health Information Management Association.