Overview of HealthUnity Master Person Index (MPI) Server
The HealthUnity MPI system was specifically designed for HIE environments. The system has repeatedly proven itself as highly effective in demanding clinical environments. HealthUnity has deep experience in multiple HIE environments. In our customer’s regions, the MPI systems work seamlessly in the background, and the attention of the HIE stakeholders is appropriately focused on using HIE to improve clinical care instead of worrying about MPI issues.
The HealthUnity MPI can be effective as a stand-alone system, but has also been optimized to perform as a core component of an end to end HIE solution provided by HealthUnity. Microsoft selected HealthUnity’s MPI for native integration with Microsoft’s Amalga Unified Intelligence System specifically because of its full functionality, standards-based approach, and significant overall value compared to other providers.
Built and Optimized for Healthcare
The HealthUnity MPI system has been designed from the beginning by applying our knowledge and experience specifically to the HIE domain. We have developed our MPI by focusing on the needs of stakeholders in healthcare environments, including hospitals, integrated delivery networks, ambulatory practices, payer organizations and patients and their families. Requirements for an MPI in healthcare can be unique and often differ from those associated with record linking problems prevalent in other industries. For example, MPI linking in healthcare in most cases is expected to render virtually zero false positives (i.e. bad matches). False positives can result in the wrong patient data appearing in another patient’s chart, with potentially serious consequences.
Designed for Performance in an HIE Deployment Model
The HealthUnity MPI architecture is ideally suited for deployment in a IHE/HIE configuration. Specifically, the architecture provides:
- Support for hierarchical instances of MPI functionality (e.g., locally within a hospital, within an IDN, and at the state level)
- Support for multiple parallel HIE instances (e.g., one instance that matches between providers; another instance that maps between provider medical record number and patient record ID in a compatible PHR, and support for a “Health Record Bank” model)
Rules-based with an Emphasis on Patient Safety
Our solution exploits unique characteristics of different patient demographic fields to yield near zero false positives and still results in very low levels of false negatives, a quality which is critical for patient record linking. We use configurable rules based matching. This approach provides predictability even when using probabilistic matches based on incomplete data, and provides the ability to match with increasing certainty as additional data becomes available. The MPI system includes administrative tools that allow for visual inspection of every record relationship in the database, and the ability to identify the rule that affected the grouping of records. And the system provides workflow tools for manual processes such as inspection of cases that are close matches but do not meet an automated threshold for matching. This way the HIE Operator may, at its option, implement services for manual de-duplication and record merging at the HIE level, or can refer cases back to source data providers to provide the manual labor required to investigate cases requiring de-duplication. For optimal results, our MPI comes with a default configuration of 11 record grouping rules that we have identified as being highly effective in our previous successful MPI implementations.
Tolerance for Incomplete Data
Our rules processing engine not only uses fuzzy logic to calculate similarity between fields but also utilizes equivalent strings such as abbreviations, synonyms, and diminutives for common names and words in use. This approach makes our MPI tolerant to common verbal and typographical mistakes and results in highly improved record matching. Quality of data is one of the major challenges in healthcare demographics. And by adding a powerful data cleansing and standardization engine, again based on our healthcare experience, we are able to preprocess the data before it hits the core MPI for patient indexing and grouping to improve results.
Predictive and Self-Tuning
Another very important aspect of our product is the capability to apply predictive logic to cases that may be false-positives and false-negatives. The HealthUnity MPI system has the ability to effectively identify potential false positives as well as negatives and then surface these cases in the workflow tools that provide the ability to manually merge or unmerge patient records. An administrator or analyst can look at the potential false positives and false negatives and resolve the false behavior by confirming the record linkage or non-linkage using our intuitive MPI administration console. Manually altered record relationships are audit logged, and treated as permanent, and can only be altered manually.
Standards Compliant
Our MPI system, like all of our HIE software components, is standards compliant. We have passed all relevant PIX/PDQ test cases released under the pre-Connectathon 2010. We are closely tracking the developments on the NHIN front, and due to our focus on healthcare, we are committed to comply with the standards requirement of NHIN as they emerge. Our flexible architecture is designed to allow compliance with current standards and rapid adaptation to the standards of the future.
Scalable
Our solution is lean and our architecture supports distributed, load-balanced computing that results in support for large data volumes with outstanding performance. We have already run models to demonstrate the capability of the solution to comfortably support the HIE data volume for large states and multi regions.
Reusable and Expandable
Our solution is based on the industry-standard Microsoft server platform. Microsoft SQL Server RDBMS is a value and performance leader with excellent scalability. The HIE Operator can potentially reuse the hardware and software platform as new services are added (e.g. HIE services) to the the HIE Operator Health HIE infrastructure, particularly if the HIE Operator chooses to deploy the HIE infrastructure provided by HealthUnity and its partner, Microsoft Corporation.
Flexible Deployment and Low Total Cost of Ownership
We feel that the HIE Operator may find the business need to install multiple instances of the MPI system over time due to privacy concerns and the lack of effective MPI services among some source system providers. For example, some providers may seek the HIE Operator MPI services at the single hospital level, some at the IDN level, some at the provider-public health gateway, and some at the patient-public health gateway. HealthUnity provides a very flexible licensing model that will allow the HIE Operator to install our solution at multiple sites at a very reasonable cost, saving its resources for its other important endeavors.
Data Stewardship
We think that the main data stewardship issue in HIE MPIs are around security, privacy, availability, tracking, structure, nomenclature, relevance, and quality of the patient demographic data. Our solutions to these issues are described below.
Security
We have taken utmost care while addressing security of the MPI data. Our MPI interfaces and storage are secured using multiple coordinated facilities. Our .NET APIs can be invoked only from the local machine by an authorized account. The web service interfaces use SSL/TLS with client certificate for authentication. The IHE interfaces require approved VPN connections. The MPI database leverages standard SQL server security features. Access to the MPI administration application is restricted via authentication and authorization. In addition, every functional transaction in the MPI is logged for auditing purpose.
Privacy
HealthUnity MPI provides the ability to control the visibility of patient data to participating entities. We enable every entity to control the visibility of its patient data to other entities during patient search as well as during PIX notification. With reference to PHR, every participating patient can control the visibility of its data to different providers. A patient can grant or revoke the consent and thereby control the data flow to different entities using HealthUnity’s patient consent portal.
Availability
Our architecture supports failover and load balancing using a bank of load balanced servers, which provides high system availability. In addition, HealthUnity MPI comes with HealthUnity’s network management console (HUMan), which has the ability to manage and monitor network resources remotely. HuMan can alert the administrators when the system reaches critical stage or any of the services become unavailable, enabling the administrators to take preventive or corrective action in time.
Tracking
HealthUnity MPI maintains all the original messages coming from a patient record source. We have tools to track every message received by the MPI system, and their current state whether they are waiting to be processed, processed, failed to process, or successfully processed.
Structure
We standardize the incoming messages into a HL7 standard message using Microsoft BizTalk server’s message transformation capability. BizTalk server is an integral part of our solution offering. By utilizing a standard internal message structure, the consumers need to deal with only one standard message structure. Since we have passed all the relevant pre-Connectathon test cases corresponding to PIX/PDQ, we feel confident that any PIX/PDQ compliant patient record source or consumer will be able to communicate with the HealthUnity MPI.
Nomenclature
It is understandable that different participating systems might use different coding schemes or nomenclature in their data. Again, we translate the terminologies into our HIE wide standard using BizTalk server.
Relevance
We return patient data that is current and relevant. While we maintain historical values internally for better record linking, we return logically the latest patient demographic information as received from a patient record source. This takes care of the situation when messages are received out of order, or the situations when a patient update message comes before patient create. HealthUnity MPI also has the ability to do an intelligent overlay of patient data and synthesize a master patient record using our message superimposition algorithm.
Quality
HealthUnity proposes a two-fold solution for ensuring quality of data. First, our MPI has powerful features for data cleansing and duplicate detection that controls the quality of data getting into the MPI. Second, our licensing terms allow you to use free licenses for IDNs. IDNs can use the data cleansing and the duplicate detection capability of our MPI to cleanse the data before sending it to the central MPI. This is also a unique value proposition that the HIE Operator can offer its hospital constituents and thus move towards the path of a financially sustainable HIE.
As for data ownership, we believe that the MPI data is jointly owned by the source that creates the data and the patient. We, therefore, give complete power to these two actors to control the way data should be distributed in the HIE ecosystem. In our MPI solution, a patient record source can control data distribution by specifying which entities should see patient data either during patient query or during PIX notification. The patient can control which providers should see their data using HealthUnity’s patient consent portal. A patient can use the portal to grant or revoke the consent given to a provider.