Real-time, actionable health information exchange requires an accurate, reliable approach to matching patients with their electronic health records across different organizations and information systems. The Michigan Health Information Network Shared Services needed to ensure its patient-matching capabilities were as accurate as possible.

PROPOSAL
Michigan's approach was to create a unique system-to-system attribute – not human-readable – called a "common key" that its participating stakeholders could include in messages to ensure dependable, inter-organizational patient-matching.

In Michigan, the health information network model is designed to encourage both decentralized, peer-to-peer communication and centralized services that channel through the Michigan Health Information Network. So, it needed a solution that would let two organizations sharing data directly with each other – not through the statewide network – to also benefit from the common key.

"For example, we wanted two organizations using Epic EHRs to improve their patient-matching rates sharing with each other, while at the same time a hospital and physician organization sharing patient information with an accountable care organization could also benefit," said Shelley Mannino-Marosi, senior director of state and national programs at the Michigan Health Information Network.

The Michigan Health Information Network also wanted an agile approach that was not "all-in" or "all-out." It wanted organizations to be able to come onboard incrementally. So, as they gain trust and confidence in the service, they can add more weight to the common key to inform their own match process.

"While we needed to jumpstart the process with the last four digits of the SSN, we hoped over time to seed an additional attribute that had greater stability and reliability than an SSN."

Shelley Mannino-Marosi, Michigan Health Information Network Shared Services

"Finally, we wanted an approach that might help organizations across the state move away from depending on social security numbers as an identifier over time," Mannino-Marosi said. "While we needed to jumpstart the process with the last four digits of the SSN, we hoped over time to seed an additional attribute that had greater stability and reliability than an SSN."

Michigan has been fortunate with a healthcare ecosystem that has largely embraced health information exchange, so the HIE has millions of messages and notifications flowing through its statewide network each week. But it needed a robust, scalable master patient index that could quickly process millions of requests weekly and respond in near real time to verify the HIN had accurately matched its records to its patients.

"We also took a fairly strict approach to patient-matching – we must be sure we have the right patient so we can accurately alert the patient's entire care team – so we needed a solution that could support the rigorous match rules we put into place to ensure our stakeholders were confident in the reliability of the common key," Mannino-Marosi explained.

The HIN did quite a bit of research into the master patient index sector to identify indexes that could meet its needs. It tested and evaluated several master patient indexes for accuracy of common key assignment, impact to data stewardship workflows, performance, scalability and cost. It selected the master patient index of 4medica.

MARKETPLACE
There are many master patient indexes on the market today. Master patient index vendors include Dedalus, Iatric Systems, Interoptex, InterSystems, Just Associates, NextGate, Occam and Verato.

MEETING THE CHALLENGE
Its Common Key Service is Michigan's approach to improve patient-matching across organizations through the addition of a unique attribute. Multiple factors contributed to this solution.

First, nearly every healthcare organization – hospitals, physician organizations, health plans, Medicaid/Medicare – has its own unique identifier for patients. Beneficiary IDs, medical record numbers, patient ID numbers – all these numbers are floating around and saved in different databases, but in the absence of a national health identifier, there isn't necessarily one common thread that links them all to one another.

"Social security numbers are unique identifiers, but they weren't really an option because they're so often targeted for identity theft, they're human-readable and very easily discoverable, and a large population of citizens, such as newborns or some tribal populations, don't always have one," Mannino-Marosi said.

The Michigan HIN has always taken a "one and done" or "connect once" approach to replace the need for multiple connections between organizations, or multiple submissions of data, with a single connection to the statewide network. Further, it is not trying to build another giant pile of data with the statewide network. There are plenty of those data piles already, she remarked.

"Instead, our approach is to establish services that are shared across the state to be sure the data is moving between organizations accurately – exactly when and where it's needed," she said. "We wanted to strictly protect the integrity of any data introduced into our master patient index and create a statewide process for improving the quality of demographic data in everyone's system."

All of these pre-conditions led to the Common Key Service. With this approach, a common key is attributed to each unique patient, stored in the statewide master patient index, and then shared with all participating organizations to keep as an external identifier in their own electronic health records systems.

"We work through a patient-matching process with every participating organization to ensure they associate the common key to the same patient in their system, so for every organization exchanging health information, the common key will help ensure they are talking about the same individual," Mannino-Marosi explained.

This closed loop cycle allows the Michigan Health Information Network and healthcare organizations across the state to work together to determine who they can match and who they cannot. They can then collectively determine how to clean up the data in those scenarios where they cannot make a match.

4medica's master patient index is where the HIN matches and stores the information it needs to correctly identify each patient, along with the common key associated with that patient. The common key augments master patient index capabilities by protecting the integrity of the data being introduced, essentially acting as an external "guard" for data prior to being linked in the master patient index.

Together with the master patient index, the Common Key Service lets the HIN's wide range of stakeholders across the state communicate with each other and be sure they're talking about the same patient.

"Basically, when we receive a message through the network, we run it through the master patient index to check if we know who the patient is with certainty," Mannino-Marosi said. "There are three possible responses."

A "Yes" response means the HIN absolutely knows who it is, and here's the common key already assigned to that patient. With this response, the HIN adds the common key to the message and sends it along its way so the receiving organization can better identify the patient referenced in the message.

A "No" response means the master patient index can tell with certainty that this person is not in the master patient index, the HIN has never seen the person before, and there is no common key. With this response, the HIN adds that patient to the master patient index, assigns a new common key, and adds it to the message, and now the HIN has the common key for next time.

And the third response, a "Maybe," means the master patient index doesn't know because it's too close to call. Maybe the message references Nick Smith and the master patient index has Rick Smith, and both have the same address and date of birth, but the last 4 digits of their social security numbers are off by one digit. Are they twins or is it a typo?

"We don't know, and we won't add this patient to the master patient index unless we're absolutely sure," she said. "With this response we send an error message back to the sending organization and say we're unable to say with certainty what the common key is for that patient. This distributes the data clean up to the source of the information and strengthens the reliability of information passing through our network for everyone."

RESULTS
"Patient-matching is an interesting business because you'll see companies claiming 100 percent or near 100 percent matching, but they don't explain what criteria they're using to match, so you don't know how good the matches really are," said Mannino-Marosi.

For instance, if one just matches on the first letter of the first name, one can get a 100 percent match rate every time. But those matches will be nearly worthless. If one matches on full name, date of birth, last four digits of social security number, zip code, etc., it's much harder to achieve perfect matching.

"That being said, our goal is to correctly notify care team members when an event occurs," she said. "We want to be able to identify more members of care teams, so they can receive notifications and coordinate care amongst themselves for that patient, so the metrics we focus on measure outbound messages flowing through and out of the network."

One measurement the Michigan HIN cares about is the number of uniquely identified patients. Michigan's population is right around 10 million. In 15 months, the HIN has gone from zero uniquely identified patients to 7.6 million as of the last week of August 2018.

"That has translated to us routing more than 1.3 million more admission-discharge-transfer notifications per week – allowing healthcare providers to coordinate care for those patients, improve those outcomes, reduce costs, and reduce burdens on the industry," Mannino-Marosi said.

Another metric the HIN boasts is that all major health systems in the state of Michigan have either onboarded or are currently onboarding to participate with the Common Key Service. Some of those organizations have even adopted the common key into their own patient-matching processes because they've found it so reliable, she added.

ADVICE FOR OTHERS
"We've already been asked by another state to help them implement the common key service and are willing to share the approach with others who may be interested," she offered.

The HIN is encouraged by the matching it has been able to achieve so far with common keys, the Common Key Service, and 4medica's master patient index. It is processing millions of messages weekly and watching its ability to inform care team members improve with each additional participating organization.

"A couple of lessons we've learned are that it's a relatively low lift to participate," Mannino-Marosi said. "There was some trepidation among participating organizations at first, but it's ended up being pretty straightforward for organizations to join the service."

Another lesson: As the common key has taken hold in the state, two of the major stakeholders have now decided to use it internally to clean up their own data as well. They trust the common key so much now that they have tweaked their algorithms to say if the common keys match then give the record a higher weight, she said.

"And finally, above all else, the Common Key Service is a mechanism to harness the collective work across multiple organizations to share in the clean-up of the demographic data for patients, and for everyone to benefit from the work done by others," she said. "The goal is to get as many organizations participating and including common keys in their data so correct patient data is exchanged when it is sent to another participating organization."

Obtaining buy-in from participating organizations, especially the larger ones, is critical as they have and continue to make substantial investments in data clean up, Mannino-Marosi added.

"This crowdsourced approach is the 'key,' pun intended, to achieving a workable patient-matching solution for health information exchange," she concluded.

Twitter: @SiwickiHealthIT
Email the writer: bill.siwicki@himssmedia.com