- As value-based reimbursement takes an increasingly firm hold on the healthcare industry, provider organizations are leaning heavily on their EHR and big data analytics vendor partners to equip them with the tools, health IT infrastructure, and business intelligence they need to succeed in a risk-based environment.
The trend towards integrated, longitudinal population health management has led many organizations to choose a similarly streamlined complete suite of health IT tools from the vendor – an approach that may offer enhanced patient management, revenue cycle insights, and quality reporting tools without the headaches of joining up a patchwork of disparate systems.
But choosing the right value-based care “operating system” is a difficult process for most providers, who face innumerable choices in a crowded, fragmented, and quickly changing marketplace.
Organizations must balance their interest in innovation from nimble new players with the stability and experience of a larger vendor with numerous implementations under their belt.
Luckily, the process of picking a big data analytics partner is getting a little easier, says a new report by Chillmark Research.
Now that vendors are successfully bringing their products into better alignment with the data analytics needs of value-based care participants, organizations can start to move beyond the basics of quality reporting for meaningful use and MACRA into a more holistic approach to population health management and coordinated patient care.
“We found a common product approach among all the vendors surveyed [in this report],” wrote author and Chillmark Research analyst Brian Murphy. “In short, all of their analytics solutions aggregate a longitudinal patient record based claims and EHR data.”
“This provides the basis for an analysis to determine which patients to focus on for performance improvement programs. It also enables reporting for the purposes of quality and cost improvement efforts. Finally, these applications support the ability of healthcare organizations to refer patients to care management programs.”
How are healthcare big data analytics companies adjusting their approaches to meet the demands of the modern provider organization, and how are they working to support the value-based care environment to maximize revenue and reduce risk?
Meeting the demand for risk stratification and patient segmentation
“The most basic question for [healthcare] organizations is which patients should be the focus of which population health management efforts,” the report states simply.
Ensuring that the right patients are identified early in the cycle of chronic disease management, that they are matched to programs that best suit their needs, and that they stick with their recommended protocols, are the fundamental challenges of population health.
They are also the biggest areas of opportunity for vendors to step in with actionable clinical intelligence. Patient segmentation, also known as risk stratification, requires providers to have access to trusted, timely, and accurate patient data that combines as many high-integrity data sources as possible to produce a comprehensive portrait of risk and behavior.
Without a robust health IT solution geared towards creating risk scores, providers often struggle to access the necessary data sources, including information buried in their own electronic health records.
Adding critical claims data to the mix in order to correctly assign risk scores to patients is another challenge – one that vendors have started to become rather good at.
“EHRs and claims data are the fuel for analytics,” says the report. “Combining the two a big picture view of patient/provider interactions augmented with the sort of detailed clinical observations found in an EHR.
“Every vendor profiled in this report has experience combining EHR and claims data — in contrast to their healthcare organization customers, many of whom do not have straightforward access to EHR data from other healthcare organizations and/or may not have experience dealing with multiple payer or plan claims formats.”
Compared to their capabilities in 2016, a number of high-profile big data analytics companies have significantly upped their game when it comes to risk scoring and patient segmentation using these data sources, the report details.
Arcadia Healthcare Solutions has added bulk patient outreach tools and at-a-glance views of coding opportunities for risk adjustment within disease registries, while Cerner Corporation has bolstered its pre-fabricated algorithms for risk scoring, normalization and quality measures.
Meanwhile, CareEvolution has enhanced its analytics related to changes in population risk scores, while athenahealth is focused on using data analytics to guide users towards improved care management strategies.
The majority of companies included in the report have cohort building features as well as patient registries for specific diseases or conditions.
“Leading vendors tend to offer more filters and those filters are applicable in multiple contexts within the application. Nonetheless, the current capabilities have limitations,” Murphy explains. “The available filters may not always allow users to be as specific as they need to be, though analysts can use query tools to compensate.
“Increasingly, HCOs want clinical users to be able to build cohorts based on complex combinations of conditions, medical history, risk, and other factors. As the number and variety of value-based contracts increases, this requirement will grow in importance.”
Combining numerous data sources to generate actionable insights
While providers can certainly rely on risk scores derived from a mix of EHR and claims data, organizations are starting to demand more detailed insights into the socioeconomic and lifestyle challenges of their chronic disease populations.
The social determinants of health (SDOH) have become a major concern for organizations that are now financially responsible for everything a patient does, even outside of the clinic’s direct control. But collecting, normalizing, and analyzing these nebulous novel data sources is still a difficult proposition.
“Given the rising importance of risk-adjustment to reimbursement, SDOH data will ultimately provide a more complete picture of each patient’s risks than past claims and EHR-based data alone,” says the report.
“A wider variety of data types will also make cohort identification more effective. Providing this will involve more extensive use of internal data sources that may not be reflected in analytics data stores, such as clinician notes. It will also require more comprehensive data about patients from external data sources.”
Currently, very few major analytics vendors are offering products that successfully integrate SDOH data at scale, but Chillmark anticipates that this will change very quickly.
By 2022, vendors are likely to be able to integrate social determinants data, patient-generated health data, Internet of Things devices, financial data, and even patient retail spending patterns alongside more traditional EHR and claims information.
Source: Chillmark Research
“Grocery purchase histories could augment a clinical history to put health risks in a new light,” the report suggests. “Retail purchasing of over-the-counter medical products could point to undocumented healthcare risks, at either a patient level or a zip code level.”
Other promising non-traditional data sources include information from remote monitoring devices such as internet-connected scales and glucose monitors, patient satisfaction data, social media records, and environmental stats such as weather patterns, air quality, and water quality data.
Using natural language processing and machine learning
Artificial intelligence, natural language processing (NLP), and machine learning have become some of the most sought-after technologies in healthcare as providers look to dig deeper into their unstructured data assets.
Free text clinical notes, call center voice recordings, imaging studies, and PDFs of lab reports and old hand-written documents contain a wealth of information that requires specialized extraction.
“Free text presents a number of intriguing possibilities for analytics and could supplement structured data for multiple purposes,” writes Murphy.
“Most of the vendors in this report have an interest in further developing natural language processing capabilities and some have embarked on a variety of exploratory efforts. These efforts are, for now, focused on finding uncoded risk factors, reimbursement opportunities, and care gaps.”
While none of the vendors profiled in the report currently have NLP functionalities commonly available to their customers, automated analytics of unstructured data is seen as a near-future replacement for manual chart reviews and abstraction.
“The vendor with the loftiest ambitions is Health Catalyst, which aspires to extract maximum value from clinical observations, regardless of how they are recorded,” the report states.
“IBM Watson Health has also made early moves to apply its cognitive capabilities to free text. Currently, its Care Manager offering uses NLP on notes in the patient’s record. Epic Systems also extracts data from clinician notes and presents the result in its EHR.”
At the moment, NLP and machine learning are still subject to concerns over the integrity, quality, and reliability of information distilled from unstructured data sets. While some studies and test cases are starting to produce clinical decision support and diagnostic results that approach human-level accuracy, artificial intelligence has a long way to go before it can truly augment population health decision-making at scale.
Open APIs, FHIR, and collaborative development
Application programming interfaces (APIs) and internet-based technologies like FHIR are changing the way vendors source new creative ideas and integrate innovative offerings into their product suites.
“Open APIs are fundamental to better support for developers,” Murphy says. “FHIR is the obvious candidate for an industry-wide standard to underpin open APIs.”
Most of the top EHR and analytics vendors, including Epic Systems, Cerner Corporation, and Allscripts, have started to rely heavily on FHIR and the blossoming API ecosystem to supplement their traditional development cycles.
“Distinct application modules responsible for a single operation within a larger application ecosystem – have become an important way for enterprise and consumer developers to think about applications,” says the report. “They support code reusability in a way that monolithic and prior service-oriented approaches to application development did not.”
“The approach delivers a collection of independently usable small functions distributed over loosely coupled organizations and computing environments. IT organizations access microservices through APIs wherever they reside: on-premise, remote-hosted, cloud-hosted, or across hybrid environments.”
Epic Systems and athenahealth are among the notable vendors to open up app stores so their customers can shop for new functionalities that connect with their existing systems, offering flexibility and customization to help individual organizations to meet their unique patient care needs.
Many of these tools rely on the Fast Healthcare Interoperability Resource (FHIR) to easily share data sets and leapfrog some of the most intractable interoperability issues that have plagued providers for years.
“Most of the major HIT vendors have ratified FHIR in various ways,” Murphy explains. “Prominent EHR vendors have announced programs that point to FHIR as a way for developers to access EHR-resident data. Several of the vendors in this report have announced that they intend to provide some level of support for FHIR. But vendor deliverables remain modest as of mid-2017.”
Improving quality reporting workflows to address gaps in care
Ultimately, success with value-based care requires providers to eliminate gaps in care and bring patients into alignment with best practices in chronic disease management as early as possible.
While quality metrics and performance reporting are intended to help providers understand their opportunities for improvement in pursuit of this goal, the process of attesting to regulatory programs in order to avoid penalties or accrue incentives has taken on a life of its own for many organizations.
Source: Chillmark Research
“In general, today’s analytics products support a collection of evidence-based quality metrics and their associated care gaps. Revenue assurance remains a high priority for most healthcare organizations and these products provide a way for them meet that goal by monitoring performance so they can optimize revenue.”
At the moment, data analytics companies provide relatively robust support for monitoring quality metrics and highlighting care gaps, though fewer are equipped to help providers with bundled payment arrangements.
Caradigm and Epic Systems lead the industry in the ability to drill down into the details of quality metrics, especially those for MSSP and PQRS reporting.
Chillmark also tapped Arcadia, athenahealth, and CareEvolution for their ability to flag evidence-based gaps in care, including metrics vital to success under the Medicare Shared Savings Program.
“The differences between vendors are often the result of their data expertise and history,” explains the report. “Vendors with early payer experience may be better at claims-based quality reporting while vendors with early provider experience may be better at EHR-based quality reporting.”
“The reality of the current market is that quality reporting and care gaps identification requires well aggregated EHR and claims data combinations.”
As providers gain more experience with big data analytics and start to get a handle on the financial realities of value-based care, choosing the right vendor partner to support them along their specific journey will only become more important.
“The variety of different goals and approaches to care management (e.g., disease specific, complex comorbid patients, risk buckets, risk bands, by practice) makes it difficult for analytics vendors to be all things to all people,” The report concludes. “The development of robust data analytics capabilities that can provide comprehensive solutions for population health management and care coordination are still likely several years away.”
“But leading institutions are beginning to meld clinical, operational, financial, structured, and unstructured data to understand the costs and outcomes of care delivered across the larger community they serve. The amount and variety of data, as well as its value, is set to grow dramatically in the coming years.”