Getty Images/iStockphoto

How augmented analytics in healthcare improves patient outcomes

High quality data has become essential to quality healthcare. Explore how augmented analytics is aiding healthcare organizations' efforts to provide better care for patients.

Healthcare has always been a data-oriented business. Doctors and specialists analyze images, patient data and medical literature in their efforts to improve patients' health. Over the years, healthcare has become increasingly digital with the advent of electronic medical records (EMRs), networked devices and, more recently, IoT.

Nonprofit healthcare system Community Health Network is so data-driven, the CIO reports to vice president and chief analytics officer Dr. Patrick McGill, who also maintains a family practice. Community Health Network uses predictive modeling as part of a network-wide screening program that examines how social determinants impact health.

Augmented analytics improves patient care

The predictive models use the clinical data in EMRs and publicly available social determinant data gathered by healthcare AI company Jvion. Community Health Network used the data to develop at-risk patient personas, such as people who face transportation challenges or who might delay mammograms. Community Health Network also uses predictive models to improve patient outreach.

"A lot of organizations talk about the last mile of a patient's journey, and everyone's last mile looks different," said McGill. "When you try to put people down the same journey, it doesn't work very well, so we've leveraged some predictive models to automate our appointment reminders, for example."

This year, Community Health Network's two goals are data literacy and digital transformation. Several analytics classes are already available, the most popular of which is self-service tools.

"We did all our new dashboard development in Power BI so that end users can have all the data at their fingertips, on their desktop, on their phone," said McGill. "The real opportunity lies in some of the operational workflows so we can predict volumes, financial performance and other operational-type business processes."

Augmented analytics helps meet success criteria

Healthcare practitioners and patients agree that the healthcare system needs to be more efficient, which is a great use case for analytics.

"The industry is changing in terms of the way [health] plans are measured. Satisfaction and experience are going to be king from 2023 onward, and the best way of doing that is to start using data more intelligently than in the past and [be] connected though the whole ecosystem of the consumer rather than it being a traditional hospital provider plan kind of approach," said Michael Wood, chief strategy officer at personalized member engagement platform provider Insightin Health.

Already, natural language understanding, an element of augmented analytics, is being used to analyze text and voice data to understand parts of speech such as keywords, entities, syntax and sentiment to understand whether patients are angry or happy, what concerns they have and the problems they're facing. Natural language generation, another element of augmented analytics, outputs text descriptions and summaries.

"Augmented analytics] makes the story[telling] much easier and it also helps uncover the answer to your question," said Wood.

Shufang Ci, chief data scientist at Insightin Health, said the combination of EMRs and IoT data will help improve patients' health, the healthcare experience and care coordination while lowering the costs of healthcare.

One impact of COVID is the increase in the number of virtual visits. Obviously, the data captured in a virtual session is richer than doctors' notes, which is great from a healthcare analytics standpoint -- just don't forget about HIPAA compliance.

Meanwhile, pharmaceutical companies and healthcare providers are partnering to bring people into clinical trials and they're using augmented analytics to identify the right populations. They're also tracking who responds to a drug and who doesn't with the goal to provide more individualized treatment.

"The responder to non-responder ratio for the top three drugs is 40/60 which tells you that 60% is a waste in some cases," said Gurpreet Singh, partner and U.S. healthcare services sector leader at PwC. "What if you could understand when the patient walked in the door that they're on the 60% side? Another use case for augmented analytics is to be able to personalize the approach to medicine and the apply the right regimen."

Singh also said pharmaceutical companies spend $100 million and $400 million per year tracking adverse events. Machine learning algorithms and advanced analytics could reduce that cost by as much as half, he said.

Challenges with adoption and use

Augmented analytics democratizes analytics by simplifying tasks that used to require technical skill. For example, natural language queries replace SQL queries and data visualizations now include a natural language narrative explaining the data visualization. As a result, the platform providers will sell directly to departments as well as enterprises. However, in the healthcare sector, data scientists tend to help create and deploy the self-service tools that lesser skilled people will use. But even then, tools are not a substitute for analytical thinking so data literacy will continue to be important.

Another challenge is data quality, since the analytics and machine learning depend on it. For example, in the pharmaceutical sector clinical trial participants are often not as diverse as the potential population that could use a drug. That means the benefits and risks of a blood pressure drug observed among a population of primarily white males may not transfer to a wider population that includes women and people of color.

Pharmaceutical companies are partnering with healthcare providers to identify clinical trial candidates so they can observe more diverse populations. It's not that pharmaceutical companies have behaved in a deliberately racist or sexist manner in the past. For example, socioeconomic differences between men and women, as well as between Caucasians and people of color, could prevent some individuals from participating in trials.

"A lot of emphasis is on the algorithm, the model and the analytics, but don't forget, if you are getting into a non-deterministic, unsupervised kind of mode, you need to have very strong training data sets," said Sumitro Sarkar, healthcare and life sciences analytics practice leader at business and consulting firm West Monroe. "There's also a literacy level that needs to be imparted because it's not just models and algorithms. I think what they miss is [the need] for this whole data-driven culture in an organization."

Next Steps

Healthcare device maker boosts production with data quality

Dig Deeper on Data science and analytics

Data Management
Content Management