Hospitals and health plans are increasingly turning to data to inform patient care improvement efforts, but gathering and managing actionable data can prove a tiresome and resource-intensive challenge.
This content is sponsored by NTT Data
Today's healthcare organizations are inundated with data — from EHRs to medical claims to genomics — and hospital and health plans want to put their wealth of information to good use. In its 2018 Health Trends Report, Stanford Medicine in California found healthcare organizations are actively seeking new tools that rapidly and accurately interpret medical data — yet, the industry still hasn't reached the "promised land of digitally enabled healthcare," as these tools continue to lag behind those available in other industries.
"With the convergence of a lot of different segments — provider, pharmacy, health plans — there's more and more data becoming available across the continuum," said Karen A. Way, MHA, global practice lead for analytics at NTT DATA. "There's a driving force of wanting to bring data together, so that you really have that true holistic picture of a patient. Then, we can use data to predict and improve the outcomes for that patient."
But before healthcare organizations can reap the rewards of an advanced analytics project, they must first establish the right foundation — a technical process that requires serious resource and staff expenditure if done in-house or using tools that are no longer cutting edge.
This executive briefing provides an overview of the challenges that healthcare organizations face when working with multiple data types, from ingesting datasets to drawing out actionable insights. This article also provides two examples of how a next-generation data and intelligence platform can help hospitals and health plans leverage non-traditional datasets to improve patient care.
Laying the groundwork: It's not just collecting data
Predicting adverse events and prescribing interventions for patients with complex conditions are two examples of data-driven projects healthcare organizations are interested in, according to Ms. Way.
To lay the groundwork for these types of patient care projects, hospitals and health plans may assume the first step is simply collecting data — whether from providers, payers, pharmaceutical companies or population-level datasets on social determinants of health. But even with a wealth of data on hand, it takes significant organizational and infrastructure lift to get these projects off the ground. This is one of the reasons that traditional enterprise data warehouse or data-lake development efforts have been costly and time-consuming. With the advent of new and emerging technologies and advanced analytics, solutions such as the NTT DATA Business Insights Engine have lessened the burden on healthcare organizations.
"The biggest challenge, and it's been out there for several years, is interoperability," Ms. Way said. Interoperability, as proposed by the Healthcare Information and Management Systems Society, refers to the "ability of different information systems, devices or applications to connect, in a coordinated manner, within and across organizational boundaries to access, exchange and cooperatively use data amongst stakeholders, with the goal of optimizing the health of individuals and populations." This is a critical concept for organizations that want mature their data management and analytics processes.
A foundational step that healthcare leaders must consider when bringing data together from disparate sources is applying master data management and data governance principles to the data that they're interested in using. This means combining datasets in a cohesive way that minimizes redundant data points and creates an integrated record of information, while at the same time ensuring data integrity between related data.
"It's making sure that everybody has the same definition of what something means," Ms. Way explained. "It's making sure that a diagnosis code for a provider means the same thing as a diagnosis code for a health plan and as a diagnosis code for a pharmacist. It's making sure that data can flow cleanly between one system and another without losing integrity or meaning."
There are manual ways to manage and manipulate data in-house; the most common method is extract, transform and load processes, or ETL. However, these methods can deplete valuable time and resources that are already stretched thin within many hospitals and health plans. To support patient care improvement efforts without exhausting resources, organizations need a platform that can ingest and manage data in near real-time, such as the NTT DATA Business Insights Engine. These functions are embedded within the solution and are known collectively as "harmonization."
"Healthcare organizations in today's world need to think about the cost-benefit of building a data platform on their own," explained Dr. Suman De, global practice lead for analytics at NTT DATA. "To do this, they would need to bring data from all of their existing applications together. [This type of program] becomes very difficult and complex for healthcare organizations to manage, both from a resources and skills standpoint, without a next-generation solution like the Business Insights Engine."
How to bring data together in real-time
Many healthcare organizations are increasing their efforts to integrate data from disparate sources. One way to alleviate the work involved with these projects is to implement a solution capable of ingesting data "as-is" from various sources.
Ms. Way shared how NTT DATA's Business Insights Engine, a cloudbased data and analytics platform, serves as a central repository from which healthcare organizations can access and analyze structured, unstructured and semi-structured healthcare data. The platform can ingest data from various sources in near real-time, allowing healthcare organizations to move from reactive to preventative or proactive interventions with their members and patients.
Within the Business Insights Engine, the data is harmonized into the Fast Healthcare Interoperability Resources canonical data model, an interoperability standards framework better known as FHIR. "The Business Insights Engine can ingest raw data in its original format, whether it's an X-ray, a claims record, a voice recording of clinical notes — all of that data can be ingested as it exists," Ms. Way said. "Healthcare organizations don’t have to do anything. The FHIR model facilitates the sharing of data across the healthcare continuum and promotes interoperability."
The Business Insights Engine not only provides healthcare organizations with a central location to manage data from different sources, but also serves as a foundation for healthcare organizations to build their analytics strategy. Such strategies might include efforts to combine various types of data to predict an adverse patient outcome or to prescribe a potential intervention.
Through application programming interfaces that are included in the Business Insights Engine, advanced analytics and artificial intelligence tools are connected to the Business Insights Engine, allowing healthcare organizations to elicit new insights from data stored within. A hospital or health plan can also set up messaging processes to direct data from the platform to external applications like member portals, case management systems and provider dashboards.
"The platform gives data scientists and business experts the freedom to work on data without any hesitance," Dr. De said. "With the Business Insights Engine, users can create predictive models with machine-learning techniques in just a few days … We are democratizing the use of data for every person in the organization to become a citizen data scientist."
The next step: Weaving data together to create actionable insights
From creating operational reports to risk-stratifying patients, there are many ways healthcare organizations can leverage a data and analytics intelligence platform like NTT DATA's Business Insights Engine to improve patient care.
Here are two scenarios Ms. Way shared with Becker's Hospital Review:
1. Predicting asthma flare-ups. Say a hospital or health plan is working with a man who has asthma. By reviewing the patient’s clinical data — such as test results, claims and emergency room visits — the organization can determine what the triggers for his asthma attacks may be. In this patient's case, dust is identified as a major trigger, as the patient lives in an area that is prone to dust storms.
The team subsequently engages a physician to prescribe the patient a rescue inhaler, although this doesn't ensure the inhaler is available when he needs it.
"As a next step, we could combine non-traditional data sources, like weather data, with claims, X-rays, spirometry testing, all things that tie into the patient's asthma," Ms. Way suggested. "Then, ahead of the next dust storm, a predictive model can see that the patient hasn’t refilled his rescue inhaler in nine months — and the inhaler is only good for six months. With the inclusion of socioeconomic data in the model, we can also see that the patient doesn’t have easy access to a pharmacy. [The program] could then deliver a message to a provider requesting a refill."
The Business Insights Engine would then deliver an alert to the healthcare organization's care manager, who would contact the patient with information that their rescue inhaler is being delivered to their home, and to suggest he set up an appointment for follow up with his primary care physician. "It's that kind of proactive scenario that healthcare really needs to move toward," Ms. Way said. "The only way we can remove all the barriers to care is by using data to create a holistic picture of that specific patient."
2. Directing next steps in diabetes care. Ms. Way also described a potential scenario involving a 65-year-old female patient who suffers from diabetes. By tracking her healthcare data, a hospital or health plan can see that she has been attending medical appointments with her physician, filling her prescriptions and completing all the tests she needs. She is doing everything to close gaps in her care.
However, a predictive model within the Business Insights Engine also flags that she presents at the emergency room three times a year. By integrating socioeconomic indicators like purchasing data, the patient is identified as eating out at restaurants five nights a week. By layering in data related to her neighborhood, it finds that she lives in an area with poor public transportation and that lacks access to a supermarket and fresh produce.
"It's very likely that those five nights a week she's not necessarily making healthful food choices, and that's why she's still having episodes of diabetic ketoacidosis or anaphylaxis that send her to the ER," Ms. Way said. "Using the multimodal dataset, we can decide that the next best action — and most precise treatment answer — for this patient is to get her in front of a nutritionist to help her manage her diet, since she seems to be doing everything else right so as to prevent further ER visits."
Advanced analytics projects aren't out-of-reach
While using data and analytics to achieve the patient care detailed in these two scenarios may seem beyond reach, it's not — healthcare organizations just need to lay the right groundwork by creating a centralized source of harmonized, actionable data.
"The more data there is, the better your analytics outcome is, the better your predictive modeling," Dr. De said. The caveat? This information must be organized and managed in a way that can communicate in concert with other datasets. "Bringing in all the data helps you paint that picture of the holistic view of [your patients]."
An as-a-service offering like NTT DATA's Business Insights Engine helps healthcare organizations achieve these data-driven goals by taking on the hard work of harmonizing and managing the data. This empowers healthcare leaders to instead focus on patient care activities — such as the creative side of developing new uses for data.
"There are tons of use cases," Dr. De said. "And the more data is available, the more precise these can be."