Luis Louro - Fotolia
Patient data is more complex than it was even a few years ago. For a better transition to value-based care, providers now need to access data of different kinds from various sources, which would be difficult without having a standard data dictionary across systems.
The data sources may include care quality databases, hospitals' financial databases and cost reporting. Meanwhile, patient data can be transmitted from wearables, home health devices and other mobile or web-based apps.
"In today's healthcare world, you have to have a strong backbone of not just quality analytics, but also clinical operation analytics and financial analytics to bring all of those things in line with delivering the best healthcare," said Jason Smith, chief medical officer at University of Louisville (UofL) Hospital in Kentucky. "From our standpoint, we were standing up a new organization. We had a lot of different IT systems."
New database architecture integrates claims data with EMR
In order to transition to value-based care, providers need to take different aspects of patient data into account. One essential practice is to normalize massive data from different data systems and output with one source, which providers would feel easy to navigate through.
The UofL Hospital introduced Data Operating System (DOS), developed by Health Catalyst, into their workflow this September, so they can pull massive patient data into one large enterprise data repository. As providers, they need to think about which part of the data source system they need to bring over, while the actual process of loading and normalizing data into the data warehouse is taken care of by the automated DOS environment.
"This solution [DOS] gives us that ability to marry things into a coherent report for not just the administrators, but also the providers that are delivering care to patients," Smith said.
Facing the same challenge of normalizing patient data for meaningful use, Health Quest, a nonprofit hospital system based in Hudson Valley, is converting their existing database to DOS and bringing in claims data as an important index to support the transition to value-based care.
Health Quest uses claims data to fill the gap of information that has not been recorded into the EHR. Claims data refers to the information in medical claims, which are bills submitted by physicians and hospitals for payment. Following a consistent format and using pre-established codes that describe specific diagnoses, procedures and drugs, claims data help providers better evaluate the quality and cost of healthcare.
"We have a series of queries that we have built and are bringing [them] into the DOS system, where we can standardize the way in which we're looking at our 835 remit data and 837 claims data to give a complete holistic perspective of how all that works, from when a claim is dropped to when a claim is paid," said Ray Pankuch, director of business intelligence and data services at Health Quest.
Additionally, from a clinical perspective, one project Health Quest has applied DOS into is VTE prophylaxis. Provided with the integrated database and the capability of data analytics of DOS, they are able to do audits to ensure patient safety by checking that all the clinical interventions providers scheduled are being done by nurses, in which way they can keep track of the care delivery and start getting better outcomes for their patients, according to Pankuch.
Deep data and AI play key roles in data analytics
DOS positions data analytics in innovative use by operating on top of open health data sources and applying APIs to ensure the interoperability among different patient data systems.
"EMRs are great data collection tools. But we don't have a data platform or a data infrastructure in healthcare to allow you to use that data, in ways that is transformative to healthcare," said Bryan Hinton, senior vice president at Health Catalyst.
After collecting massive data from the mainstream EHR and bringing in underrepresented health information, the next step is data analytics, which could provide clinical insights for providers. And here is where deep data comes into play.
Similar to big data, deep data is a method of processing data in large amounts. More importantly, it supports the transition to value-based care by putting raw data into context and makes further comparisons and statistic analytics. By navigating through deep data, providers are able to compare personal health records with those of the overall population, but also figure out what the personal health trend is.
"One of the things that Health Catalyst can help us do within our own system is to look at longitudinal care of the patient, both pre- and post-acute admission to the hospital," Smith said. "We can look and make sure that [patients] have the appropriate pre-hospital care in our outpatient system by looking at the different markers, for example, for diabetes management or hypertension management prior to their coming to the hospital for the acute episode."
Overall, deep data prompts and promotes actions to assist providers with clinical decisions, which aids the transition to value-based care. For instance, if a patient with a higher-than-average A1C score has been detected with data analytics, it would lead a care management team to intervene and reach out to the patient with additional educational materials, or to invite the patient for a doctor visit to talk about health practices, so that the diabetes doesn't go out of control.
Ray Pankuchdirector of business intelligence and data services, Health Quest
Another important feature of DOS is the application of AI and machine learning, which provide a data-first application platform, where better interoperability among applications is in place for the developer teams in hospitals to compute new insights and surface them in web or mobile apps through APIs.
"The best part of the whole process with DOS is that the technology is there to help us move out of repeatable steps that we need to do from a development process," Pankuch said. "And they provide many more APIs that allow nondevelopment users to create data-driven applications. And there was one application in particular that we're just implementing now, called Population Builder. That's a key tool that uses the DOS platform."
A step further to the future: Challenges and expectations
The DOS UofL Hospital applied is hosted in the Microsoft Azure cloud computing service, where patient information has been stored in a public cloud. Smith said he was positive about the data security within DOS and the Azure service, which is HIPAA certified. But there is a question about giving people access to patient data and controlling the scale of data that certain people could access.
With EHRs, providers look at a patient chart one-on-one, which makes it easier to keep the scale of accessible data. When the data environment brings in different dimensions of patient information at once to give providers a view of the whole care plan, the data accessibility can be tricky. But it might be worth trying as long as patient privacy is being protected by the security system.
In addition, from an organizational perspective, the future use of DOS lies in real-time care delivery, which requires the healthcare system to use the enhanced machine learning capabilities to bring patient data and clinical insights right to the bedside and integrate with the EHR, according to Hinton.
"Using actionable data and seeing what's happening to the patient or other possible treatment types is where we're going to start to see the shift," Hinton said. "And that's where the real-time data, alongside with the analytics data that we already have in the warehouse, is going to be so important."