Data — The Core Bottleneck to Digital Health (Part 3)
This is Part 3 of a 5 part series. Here’s Part 1, Part 2, Part 4, and Part 5.
This post was originally published on Medium.
This is Part 3 of a series. Here’s the Overview, Part 1, Part 2, Part 4 and Part 5.
With the digital health industry seeing such steady growth as a result of tailwinds from industry trends and increased demand from a global pandemic, a core set of questions arises: can the industry maintain the growth and meet the demand? What challenges lie in the way?
The Problem with Healthcare Data
The answer to these questions lies in the titular bottleneck. Let’s take a second to understand the context surrounding healthcare data and how it creates a bottleneck for the digital health industry.
Largely speaking, healthcare data lacks interoperability and usability. For the sake of simplification, I’ll observe healthcare data in two large buckets: health data and administrative data. Health data will refer to all data that’s relevant to a patient’s health, and administrative data refer to all data that’s relevant to the administration of care, such as provider data, insurance coverage data, claims data, etc.
Health Data
Prior to 2009, a vast majority of hospitals still operated through outdated processes, leaning on fax, paper, and landlines. The HITECH Act (2009) incentivized the adoption and utilization of health information technology (to the tune of $38B from the CMS by 2018), resulting in an explosion in adoption of EHRs (electronic health records) from 8% across hospitals and health systems in 2008 to 96% in 2017. The problem? These EHRs are unsatisfactory to say the least.
The largest EHR players at the time of the HITECH Act (namely Epic, Cerner) started out in the pre-World Wide Web-era building software for particular hospital record keeping functions, such as billing, and tacked on features over time to accommodate the growing list of demands from client health organizations. They eventually became one-size-fits-all solutions for all of health systems’ software needs, despite initially being designed just for records.
To paint the picture of the antiquated nature of these solutions, many of the main EHR players use MUMPS, an outdated programming language invented in 1966 (Epic, MEDITECH, GE Healthcare, the list goes on). Not only did the obsolete software and design practices lead to a burnout-causing, unusable product, the costs surrounding EHR contracts combined with implementation and change management overhead are sky-high. To only make matters worse, the dollars and hours attributed to these contracts mean that healthcare executives are emotionally committed to their systems’ EHRs. Many health systems are stuck with their EHR for the foreseeable future.
As enormous as these downfalls are, they may not even be the worst part — that honor goes to the lack of data interoperability. EHRs store massive amounts of information, and are the primary repositories for patient health data, but aren’t designed towards using or sharing the data in a meaningful way. Due to this, less than half of hospitals electronically integrate data from other hospitals outside their system, and many hospitals still resort to printing and faxing to share data, despite having spent billions of dollars implementing EHRs.
The above flow diagram (from another great read, The Fragmentation of Health Data by Datavant) displays only some of the steps where healthcare data is created and exchanged in a patient’s interaction with the healthcare system (notably, the complexities of provider-payer data exchanges are not included). For many health systems, EHRs are the owners of most of these steps and significantly involved in others. And, if you recall, they are poorly designed one-size-fits-all products that are not conducive to sharing or analysis.
At each of these points, multiple data points are generated by multiple entities, perhaps incorrectly or duplicatively, then exchanged. Thus, as each transaction increases potential for additional complexity / fragmentation, the usability of health data decreases.
This has materially impacted patients, consumers, and physicians. In fact, in the time of COVID-19, it’s a matter of life and death. CMS Administrator Seema Verma said, “We’re seeing this with the situation with coronavirus and patients coming into a hospital and providers are stressed with dealing with sick patients and they have no time to gather patient information to have a complete medical record.”
In a great breakdown of the impact of medical record fragmentation, Ilana Yurkiewicz, Stanford physician and medical journalist, elaborates, “When I meet a new patient, I have to gather slips of these histories from various sources — electronic records, paper documentation, outside faxes, notes in wallets, family members — to piece together a meaningful narrative. … When a patient with a complex medical history … arrives under my care, it’s like opening a book to page 200 and being asked to write page 201. That can be challenging enough. But on top of that, maybe the middle is mysteriously ripped out, pages 75 to 95 are shuffled, and several chapters don’t even seem to be part of the same story.”
Meanwhile, access to medical records is nearly impossible for a consumer. To cite a high-profile scenario, former Vice President Joe Biden said, “I was stunned when my son for a year was battling stage 4 glioblastoma. I couldn’t get his records.”
Part of the reason EHR companies viewed data as proprietary (as opposed to belonging to the patient), and thus were prohibitive to other EHRs and healthcare software companies in accessing patient data, causing a deeply fragmented data landscape.
Particularly, Epic (the market leader) has taken a “walled garden stance” as it seeks to be the only EHR on the market. It’s been historically uncooperative in data interoperability, charging fees for data exchanges with non-Epic providers, impeding data access, and lobbying against interoperability rules.
Though Epic has responded to interoperability criticisms through a variety of ways, these efforts are half-hearted. For instance, gaining access to AppOrchard, an app store-type platform for housing third-party tools on Epic, has been known to be an expensive, time-consuming, unnecessarily complicated process that requires playing by Epic’s rules.
Some notable efforts have arisen to combat these issues, listed below:
The 21st Century Cures Act (2016) — policy that, among other items, creates clarity around interoperability and prevents information blocking
Fast Health Interoperability Resources (FHIR) — the most recent and popular data API standard created by HL7, structured in programming-friendly languages
The CommonWell Alliance is a non-profit collaboration between EHR companies, led by Cerner seeking to define standards and best practices around data interoperability. Notably, Epic is not a member, going so far as to publicly identify the Alliance as a “weapon” against market share.
Another collaboration, the CARIN Alliance, includes industry giants such as Apple, Microsoft, Google Humana, Walgreens, Blue Shield of California, Salesforce, etc., and its members have pressed for expedience in interoperability.
Though these efforts have been steps in the right direction, they have not yet resulted in definitive progress.
Concurrently, healthcare data is increasing in volume and variety: the quantity is expected to be 15 times greater in 2020 than in 2013, and consumers are increasingly using digital health tools to complement or even supplement their healthcare. With the proliferation of consumerized genomic data, wearables, and other forms, healthcare data is an expanding gold mine with severe obstacles around mining.
Administrative Data (Provider and Claims Data)
Healthcare administrative activity is reliant on back-and-forth between providers and payers, in data-heavy processes such as claims (submission, inquiry, payment), prior authorization, eligibility verification, contract management, referrals and credentialing. However, as mentioned before, administrative inefficiencies consist the largest portion of healthcare’s exorbitant costs. According to a16z general partner Julie Yoo, healthcare revenue cycle/billing is “roughly 8X less productive than other sectors when it comes to the number of employees affiliated with a certain level of revenue.” This, too, is largely due to the core issues of healthcare data: interoperability and usability.
VBC-initiatives are driving the growth of payer-provider partnerships. These partnerships heavily rely on data-sharing to enable their pursuits of improved care coordination, better interventions and better analytics around population health measures. This is reflected in the astonishing volume of administrative data transactions — according to a 2017 CAQH Index Report, “an estimated 925 million manual transactions and nearly 13 billion electronic transactions were conducted by medical plans”, which was a 38% increase over the prior year. Another CAQH Whitepaper found that “a typical practice … must manage 8,400 data points” of provider data for contract management. Utilizing claims and provider data is a critical function in healthcare.
Sadly, healthcare claims have problems similar to health records. For one, many claims processes are still excruciatingly manual. A McKinsey report finds that automation of payer data activities could achieve cost savings between 15–40%, but, in order to automate, a solution needs to integrate with the EHR and payer claim data sources, which we know is an incredibly tall order. An example of these problems is prior authorizations. 83% of physician survey respondents said prior authorizations are “very” or “extremely” burdensome, one reason being that 84% of medical necessity attachments are manually exchanged, often containing the wrong amount or wrong type of information.
Even when automated, electronic claims systems were designed without consumer needs in mind, and are difficult to manage. As Mario Schlosser (Oscar CEO) puts it, “the biggest purpose for these data files is for claims to get paid. This data wasn’t built for driving better clinical outcomes, attaching better clinical payloads to the data transmissions, for having more real-time insights and things like that. It’s a payments system first and foremost, and that’s it.” Strikingly, 835 files (electronic claim payment information file) could have a 30–40% error rate.
This all leads to 1 in 5 claims being mishandled, a problem which compounds over multiple complex processing steps that occur between the provider and payer.
On the other hand, provider data is woefully inaccurate and difficult to maintain. 52.2% of provider directories have at least one inaccuracy and the data changes frequently: 2–3% every month. To put this into context, this consistently inaccurate data worsens provider communications, causes delayed provider payments, impedes patient access, and all of the administrative processes shown in the below diagram.
Most critically, poor provider data creates severe challenges in care coordination and fluid payer-provider partnerships. Though health systems’ networks are growing through partnerships and mergers, poor provider data quality (for ex. lack of granularity into specialties) limits optimal patient-to-provider matching and provider-to-provider referrals, leading to more back-and-forth and inappropriate care. 75% of specialists received a clinically inappropriate referral in the last 12 months, while 62% of referring physicians say they lack reliable information for in-network specialists.
In sum:
All healthcare data suffers from two key problems — interoperability and usefulness.
The administration of healthcare is highly complex with multiple touchpoints and data transactions between many stakeholders, including patient, provider, and payer. As fall off can occur at each step, the problems of inaccurate data (patient, provider, claims) and inefficient data sharing processes snowball into severe administrative waste, a multi-billion dollar problem
Current solutions are poor fits to solve for these problems. Manual processes are unscalable, and incumbent software solutions do little to provide operational lift.
Health systems widely adopted expensive EHRs that are cumbersome products designed around logging information rather than to create meaningful use of healthcare data.
EHRs are historically uncooperative with each other, with other software, and even internally. This creates multiple, disparate data silos throughout the healthcare landscape. Efforts to encourage interoperability have been so far unimpactful.
Claims systems and provider directories suffer from similar problems, making provider-payer transactions difficult and slowing the progress of value-based care.
Physicians are paying for this with burnout, and patients have worse outcomes.
Health data is continually increasing in quantity and variety, but the healthcare industry is unable to take advantage because of these issues.
These issues clearly are detrimental to the healthcare system — how do they impede digital health?