Maintaining accurate, consistent, and up-to-date patient demographics is a constant challenge. The number of application systems that store and share patient data, combined with variances in data quality, lead to duplicate, incomplete, or corrupted patient records. Even with the proliferation of electronic medical records (EMRs) and advances in system integration, the patient identification problem remains, costing organizations millions in unnecessary tests, duplicate medical record remediation, billing delays, fraud, and medical error resolution.
When the data exchange requirements of connected care initiatives, like Accountable Care Organizations (ACO) and Health Information Exchanges (HIE), are added to the mix, the challenge - and importance - of accurate patient identification increases significantly.
Fortunately, there is a pretty straightforward technical solution to the problem. An Enterprise Master Patient Index (EMPI) is designed to compare patient records from any application and determine if they relate to the same patient or not. The EMPI creates a unique enterprise identifier for each patient, and maintains a list of all the patient's other identifiers, promoting a single and comprehensive view of a patient.
Accurate patient identification will indicate that "Jonathan McDonald" in one system is the same patient as "John MacDonald" in another system. This awareness leads to better patient care, improved registration and scheduling, more efficient billing, and greater confidence in the exchange of patient information between systems.
Although application systems typically maintain their own Master Patient Index, an MPI from a single vendor system, like an EMR or HIS, is often not sufficient to handle the sophisticated cross-application logic needed to provide true enterprise identification.
When evaluating EMPI technology, there are several key attributes and capabilities to consider:
Robust Matching Algorithms - The goal of an EMPI is to automatically match and link identifiers for the same patient from different systems, but do so without incorrectly matching records that aren't for the same patient. An EMPI must employ both probabilistic and deterministic matching algorithms.
Probabilistic algorithms are used where data may not be an exact match. For example, addresses may be incomplete and formatted differently, or identifiers may have typos or transposition errors, or names may be spelled differently or nicknames may be used. Comparison algorithms are able to evaluate content and assign partial "weight" scores indicating the probability that two fields match. Weights are also assigned to the more or less important fields allowing a total score to be calculated and compared against site- specific matching thresholds to determine if records are an automatic match, a potential match, or are distinct.
Deterministic algorithms, which compare or look for exact matches, are also used to identify candidate records or identify specific situations where a rule should override the probabilistic score. Common examples for this logic are twins with similar demographics or father/son and Senior/Junior relationships.
Flexible Data Model and Customizable Algorithms - Every organization is unique and will need to store and match records based on the fields in their environment and the prevalent characteristics of their patient population. An EMPI data model must be able to accommodate those requirements. Similarly, the matching process must be configurable, both in the fields that are used to match as well as the algorithms and weight given to a given field.
When records are indexed together, the EMPI creates a merged record, a "Golden Record", which represents the best single view of the patient's demographics and related information. How this single view is composed and the survivorship rules used to determine which values comprise the record must also be configurable.
Rich Application Interfaces - An EMPI should integrate easily into the IT infrastructure. Application interfaces, or APIs, allow records from external systems to flow into and out of the EMPI using standards-based, real-time or batch interfaces.
Batch interfaces are used during implementation to perform the initial matching and loading of records into the EMPI, and for periodic or incremental loads of records for those systems that don't have real-time feeds. These interfaces must scale to process millions of records and must not require downtime to be performed.
An EMPI must support standard HL7 inbound and outbound messages as well as the PIX/PDQ IHE Integration Profiles as methods to receive records as they are created or updated, and notify other systems of changes in the composition of the "Golden Record". The EMPI will typically be a consumer and producer that interacts with an organization's interface engine and must have a standards-compliant messaging layer that can work with engines from any vendor.
An EMPI must also support standards-based Web services for building interactive applications or portals that will make use of the information in the EMPI.
Intuitive User Interfaces -The EMPI will automatically match as many records as possible according to its configuration, but some records will still be flagged as potential matches and require review by a data steward. The user screens must be intuitive and easy to use, providing a quick view of the relevant records and a side-by-side comparison of the matching and differing fields.
Configurable workflow management is an important feature, allowing an organization's processes to be captured and implemented. For example, the workflow configuration can assign tasks to appropriate personnel, escalate stagnant tasks, and require approvals where situations dictate.
The user interfaces must also generate and display reports on operational statistics and performance of the EMPI.
Security - Since an EMPI stores sensitive healthcare information, it must be secured properly. Role-based access control is an essential EMPI feature in order to limit access to functions, systems, records, and fields. Security should be based on configurable criteria that include context sensitive rules. For example, certain users should not be able to see the SSN field, while other users are able to see it unless a VIP flag is set, in which case they are not able to see it.
Sharing healthcare data in today's environment requires a sophisticated and flexible EMPI. As a mission-critical application, the EMPI must have features to address site-specific characteristics, logic to accommodate variances in data quality, and solid security to protect data content. With an enterprise master patient index in place, organizations can be more confident in the accuracy of their data exchange and gain a reliable, consistent resource for patient demographics across the enterprise.
Kevin Schmidt is director of Product Management, NextGate.