Will the Data Walls Come Tumbling Down?
Stories that caught our attention this week
It was the summer of 2017, and Sanjay Mishra, MD, was on his way home to Indianapolis with his two children. They were in the airport when he collapsed and went into cardiac arrest. Quick-thinking bystanders administered CPR, and first responders called his wife in search of medical information that might save his life.
His wife Seema Verma — yes, the administrator of the Centers for Medicare & Medicaid Services — spent the next few hours on the phone frantically trying to piece together Mishra’s health record. As Fred Schulte and Erika Fry report in a joint Kaiser Health News/Fortune Magazine investigation, “her husband survived the episode, but it laid bare the dysfunction and danger inherent in the existing health information ecosystem.” Even now, two years after the health scare, Mishra’s family is missing a complete health record for him. They have “a few papers and a CD-ROM containing some medical images — but missing key tests and monitoring data.”
We tend to think about health care as a singular system, but the lack of centralization and standardization of patient data show that it is anything but. As Hong Truong, senior program investment officer for the CHCF Health Innovation Fund, explains in a blog post, “compiling a patient’s complete health record still requires a herculean effort involving multiple web portals, reams of data files in different formats, and — more often than we’d like to admit — fax machines.”
The Rise of the Silos
The advent of electronic health records, or EHRs, was supposed to help with this, taking the US health care system into the 21st century while improving quality of care. It has been positive in some ways. A study of 3,249 US hospitals that adopted EHRs between 2008 and 2013 found that the more EHR functions a hospital adopted over the baseline functions, the greater the reduction in 30-day mortality rates. And a survey of over 500 primary care physicians (PDF) conducted by The Harris Poll and Stanford Medicine found that the majority of physicians (63%) think EHRs have led to improved patient care. However, 67% of physicians also said the biggest priority in long-term EHR development should be improving interoperability so that EHRs can talk across hospital systems.
Interoperability deficiencies exist because there is no centralized authority for EHRs, meaning that each vendor — Schulte and Fry report that there are more than 700 EHR vendors — can use a unique format for collecting patient information. The type of information collected can also vary from vendor to vendor.
These data silos are further isolated from one another by business incentives that work against interoperability. “A free exchange of information means that patients can be treated anywhere,” Schulte and Fry write. “And though they may not admit it, many health providers are loath to lose their patients to a competing doctor’s office or hospital.” But this also makes treating patients in emergency situations, like Mishra’s, more complicated. When hospital systems can’t easily share health records, patients can face harm.
In addition to problems with inter-hospital data exchange, hospitals struggle with data challenges within their own systems. Pew Charitable Trusts cites data that “as many as one in five patients may not be matched [with the correct health records] at a health care facility they’ve previously visited.” This can lead to delays in care or improper and potentially dangerous treatment.
Lola Butcher reports for the digital science magazine Undark that there are two major types of patient matching problems. The most common type is the creation of duplicate records for a single patient — for example, when Jennifer Sarah Doe is alternatively identified as Jen Sarah Doe and Jennifer S. Doe at her physician and ob/gyn’s offices.
The less common but more dangerous type of patient-matching problem occurs “when records for different patients are mistakenly combined,” sometimes leading to horrific medical mistakes. Butcher recounts that two patients with the same name had kidney scans at a Worcester, Massachusetts, hospital on the same day in 2016. One patient’s scan showed healthy kidneys while the other’s showed a tumorous kidney that needed to be removed. Due to poor patient matching, the health records got mixed up, and the patient with healthy kidneys had a kidney removed by mistake.
The Interoperability Dream
Don Rucker, MD, the head of the Office of the National Coordinator for Health Information Technology (ONC), writes in Health Affairs that a bipartisan majority of Congress passed the 21st Century Cures Act of 2016 to improve the interoperability of health information. Since then, the ONC has been working with health industry stakeholders and federal agencies to develop a trusted exchange framework that “will set common principles, terms, and conditions that facilitate trust between disparate health information networks. It will seek to scale interoperability nationwide and enable participating networks to work together to provide an on-ramp to electronic health information.”
In the meantime, individual health systems are trying to crack the patient-matching problem with a variety of approaches. Butcher reports that Northwell Health, the largest health care provider in New York, began tackling its backlog of over 220,000 possible duplicate records in 2016. First, Northwell conducted a manual review of duplicates that were almost certainly for the same person but required a spot-check. Next, the health system used probabilistic matching, an algorithmic matching technique that assesses the likelihood of two records belonging to the same patient, to check newly created health records against those in its master patient index. Northwell also introduced referential matching, an approach that employs third-party demographic data like US Postal Service address information to match patients’ records.
Pew examined four approaches — including referential matching — to addressing patient matching challenges. It specifically called out referential matching as a promising approach given that “it has generated among the highest match rates currently published.” Verato, a health startup (and a CHCF Health Innovation Fund portfolio company) that developed a referential matching technology, reports that its technology can achieve match accuracy rates of 98%. However, Pew cautioned that “many of the data sources used for referential matching do not contain information on children, homeless individuals, or other subpopulations.”
Pew also studied the development of a unique identifier system to link patients to their records; patient-empowered approaches that would distribute some of the responsibility of record matching to patients; and standardization of demographic data to make it easier to match patients to their records. Ben Moscovitch, Pew’s project director for health information technology, told Butcher, “all the opportunities we examined to address matching have benefits and drawbacks — and none of the opportunities alone can solve this problem.”
In the long term, Pew recommended establishing a single, national entity “to help advance the identification and adoption of standards to improve patient matching as technology develops.”
The government is playing a growing role in the regulation of patient data. In February, the Department of Health and Human Services proposed new rules “to support seamless and secure access, exchange, and use of electronic health information.” In California, Senator Cathleen Galgiani introduced SB 441, which would enact the California Interoperability Enforcement Act to regulate EHR vendors operating in the state. Additionally, the California Department of Health Care Services is rolling out its new California Health Information Exchange Onboarding Program, which includes up to $50 million to help Medi-Cal providers improve their ability to exchange data.