Master Person Indexes: Key Factors to Consider
Master Person Indexes (MPIs) help maintain consistent, accurate person data across various health care and community organizations, making them a powerful tool for population health management. To generate a better understanding of how MPIs are used in practice, All In: Data for Community Health hosted a recent webinar where experts shared considerations for using and developing MPIs. We’ve summarized some of their key lessons and insights.
Why use an MPI?
The purpose of an MPI is to properly integrate multiple data sources and prevent data disintegration. MPIs improve the accuracy of an individual’s data and can help enhance care coordination by providing a more complete view of a person, both at the individual and population level. The following two case examples demonstrate how improving patient matching can have a major impact on data quality.
Example 1: Camden Coalition for Healthcare Providers
Camden Coalition for Healthcare Providers runs a city-wide Health Information Exchange (HIE) in Camden, NJ which includes a vendor-managed MPI. On a daily basis, Camden Coalition imports MPI data into their internal performance and care coordination tracking system, which allows them combine records that share subsets of identifier fields and manually review the data to feed corrections back into the HIE. Stephen Singer, Senior Program Manager for Data Analytics at Camden Coalition, commented:
“It allows us to chain together different sets of identifiers, build a richer picture of patients, and move towards more accurate, complete, and correct identification of individuals.”
Singer shared an illuminating example that demonstrates why improving the person matching process was critical. Through the MPI, Camden Coalition identified a person who had visited four hospitals over five years and was assigned fifteen distinct medical record numbers, five different dates of birth, multiple first and last names, and an array of social security numbers. If their matching process had been too simplistic, they wouldn’t have been able to catch these errors. Singer explained:
“It’s not just clerical or data entry error. At any point when the data is entered, manipulated, transferred, or edited, there’s an opportunity for error. Even if you’re doing this entirely within a single health system or organization, you don’t have control over all those inputs, so errors will continue to crop up.”
Example 2: San Diego Health Connect
San Diego Health Connect, a regional health information exchange (HIE), also uses an MPI to improve the reliability of person matching. When implementing the HIE, their governance board required a 100% match on six variables—first name, last name, middle initial, gender, date of birth, and social security number. When these factors didn’t match, the records were put into an exception queue, which grew very quickly as more identification errors occurred. Dan Chavez, Executive Director at San Diego Health Connect, commented:
“We cannot achieve the triple aim or get the desired care coordination for our population unless we properly measure, identify, match, and link patient records. Patient matching and linking must be correct to enable proper health information exchange.”
By augmenting their MPI with Verato’s cloud-based referential matching technology, which includes a variety of data sources in San Diego, they were able to match all the identities properly in the HIE. After implementing referential matching, San Diego Health Connect saw a 110% improvement in patient matches in their MPI.
Chavez emphasized that patient identification is a community-based process and that maintaining relationships is critical to ensure the sustainability of an MPI. San Diego Health Connect has found community participation to be especially helpful in the data cleaning process. They have made an effort to understand the different identity data governance models of participant organizations and create standards for data cleaning. Every month, they publish scorecards measuring the performance of each participant, which has greatly improved their patient matching. Chavez noted:
“Even though we could do the data cleanup, if folks do to the data cleanup on their back end MPIs, the problem recreates itself. We had to prove that we could not only make it work for one cleanup, but that we could develop a process at the community level to fix this problem going forward.”
Questions to Consider When Developing an MPI
Deciding whether to buy or build
Depending on the specific needs of an organization and/or coalition, data teams may decide to buy an MPI system, build a customized solution, or combine both of these options as Camden Coalition did. Singer offered a list of questions projects should ask when considering these possibilities.
How soon do you need it? If you need to add data quickly, buying an MPI may be the best option.
How expensive is it? Your budget will help guide your decision about whether to buy or build an MPI and inform your process for doing so.
How flexible does your solution need to be? Building your own MPI can allow you to create a system that’s adaptable and customized to your unique needs.
How stable are your use cases? If patient data is constantly changing, it may be useful to have a self-built system.
How interoperable does it need to be? If you have a diverse coalition with several different systems, it may be valuable to buy a solution.
How accountable does it need to be? Self-built systems can help you independently verify data quality.
Vetting potential vendors
Both Camden Coalition and San Diego Health Connect engaged third party vendors to help build and augment their MPIs. Based on their experience reviewing different vendors, the speakers suggested asking the following questions during the selection process.
Can I get all my data back when I need it? Make sure you don’t get locked in to a vendor and that doesn’t allow you to access the data after it is input, edited, and changed.
How does the vendor do it? Get as much technical detail about the back end as possible so that you can ask the right questions and be aware of the strengths and limitations of the system.
Who can you talk to (outside of sales and marketing)? Before you sign a contract, make sure to establish a connection with someone from the technical team who has the right expertise to answer your questions.
How responsive is technical support? Understanding the process and turnaround time for fixing errors will be crucial in planning for successful implementation.
Can you flag records by linkage quality? This question can help facilitate an initial conversation about data quality.
Are there any other technologies or vendors that augment what they do? Every solution involves some tradeoffs, so vendors should be candid about their capabilities and what other technologies can be used to enhance their system.