How Northwell Health's multi-pronged approach to data cleansing aims to improve care
Northwell Health has taken several steps over the past two years to reduce both the backlog and creation of potential duplicate records within the health system.
First, Northwell contracted with Just Associates to work as an extension of the centralized business operations team to oversee resolving the worklist queue for the health system, which comprises 23 hospitals and more than 500 physician practices.
Under the leadership of Rebecca Way, director of revenue cycle business operations at Northwell, Just Associates helped resolve more than 200,000 potential duplicates. This took roughly a year to whittle down.
Second, around the same time, Northwell Health implemented InterSystems' HealthShare Patient Index to leverage probabilistic and deterministic logic to link records across Northwell's various information systems.
Based on a weighted algorithm, the patient index automatically links those records that meet defined criteria. The patient index works in conjunction with Northwell's master patient index from Allscripts to reduce potential duplicates from 700 to a little more than 300 a day.
Third, automated scripts were developed to ensure Northwell's core systems could process an HL7 merge message, reducing the manual process of merging records.
"You can be an organization of world-class clinical care, but when a clinician cannot see the entire continuum of care for that patient, there can be dire consequences."
Keely Aarnes, Northwell Health
That brings Northwell to its current state: Business operations are now down to roughly 10,000 potential duplicates on the queue and a portion of the team is resolving potential duplicates same-day.
"With that said, Northwell still needs to do better to not end up back in the same situation," said Keely Aarnes, director of IT at Northwell Health.
So Aarnes and her team began conversations with patient matching vendor Verato this past year, soon deciding to move forward with a proof of concept based around Verato's Auto-Steward technology. It went live June 13.
With the Auto-Steward system, Northwell sends all potential duplicates daily via an application programming interface. After comparing to their referential dataset, a result is returned of match, no match or inconclusive.
"Currently we are using these results to focus on the potential duplicates Northwell is unable to resolve on our own," Aarnes said. "It is too soon to offer metrics as we only went live June 13. However, prior to implementing Auto-Steward, we completed a proof of concept with static files of our patient population in HealthShare Patient Index and of the master patient index resolve queue."
This approach yielded an 87 percent match, she explained, and gave Northwell enough information to know it wanted to move the system into production for onboarding new physicians as well as acquired hospitals or practices.
Correctly indexing and cleansing data is key to health IT set-ups. There are many vendors of master patient indexes and data cleansing technologies. These vendors include ARGO Data Resource, Epic, Iatric Systems, Information Builders, NextGate and Occam Technologies.
The goal of the MPI proof of concept was to verify the lift the vendor's technology could add to resolving the potential duplicates and be a tool for the business operations team in merging records that were the same patient or confirming the records belong to different patients, she added.
With close to 350 million unique lives, including correct and incorrect information about a person, Northwell is able to resolve potential duplicates that it was incapable of doing previously. As an example, a quick admission process in the emergency room can often create a duplicate with little information gathered about a patient, and Northwell uses the tech to fill in those blanks, she said.
The daily process with Verato will help Northwell resolve potential duplicates quickly and greatly reduce the risk of having duplicates in the active patient population, she said.
"I have worked in healthcare technology for 25 years and during that time I have had a duplicate patient record twice at two different healthcare entities," she recalled. "Unbeknownst to me, I had unnecessary and expensive tests following a radiology appointment because the radiologist didn't know I had previous exams. I was registered as a new patient and had two different patient IDs."
These tests were declined by her insurance and it took six months of work to clear up the duplicate and get the claim paid.
"The second time I had lab tests run unnecessarily with the same results – denying my claim as not medically necessary," she explained. "I find this ironic, as patient identification has always been one of my passions. Being the patient behind the duplicate has made me far more empathetic to impact. Most patients don't know what a duplicate record is or that they may have a duplicate."
These implications she faced did not have clinical consequences, but she knows that is not the case for many. As patients are becoming more involved in their care and driving a new digital transformation where access to their information is now expected, healthcare organizations cannot afford to get that wrong.
"You can be an organization of world-class clinical care, but when a clinician cannot see the entire continuum of care for that patient, there can be dire consequences," she said. "We have to be better. We have to get this correct."
Big Data & Healthcare Analytics Forum
The Boston forum to focus on effective pop health management, AI and precision medicine Oct. 22-23.