Data Analytics

How Are Predictive Analytics Applications Changing Oncology?

Intuitive models are helping providers determine the best course of cancer treatments and assess a patient’s odds of readmission.

Kevin Joy

Kevin Joy is the managing editor of HealthTech. He previously worked in content marketing for a major university health system and as a features reporter for the Columbus Dispatch.

A decade ago, when Dr. Debra Patt was starting her professional career, the oncologist would use her medical training — plus paper, a pen and a calculator — to determine the best course of treatment when writing chemotherapy orders.

Now, she enters details such as patients’ height and weight and the characteristics of their cancers into an electronic health record. Patt then uses analytics tools to process that information and guide her next steps with data-driven precision.

Such automated guidance is complementary, not a substitute.

“Oncologists today see many different diseases, and the field is changing so rapidly. Predictive analytics helps us be compliant with the most recent evidence-based guidelines,” Patt, an executive vice president at Texas Oncology and editor-in-chief of JCO Clinical Cancer Informatics, tells HealthTech.

“Having treatment plans integrated within the structured data elements and supported by clinical decision reduces error because there’s variability in humans writing orders,” she adds.

The tools also stand to provide added benefit to clinicians as value-based care — reimbursement tied to quality of care rather than the number of patients seen — gains traction.

Now more than ever, “we are laser-focused on care management and reducing ER visits and hospitalizations for the patients we serve,” says Patt, also a former clinical practice committee chair for the American Society of Clinical Oncology.

The movement shows no sign of slowing. With a current estimated value of $14 billion, the global healthcare analytics market is expected to reach $50.4 billion by 2025.

Growth of Predictive Analytics in Oncology Care

In a May 2019 article published in the ASCO Educational Book, Patt and five industry colleagues examine the challenges and capabilities of predictive analytics in oncology. Citing an “immense potential” to improve risk stratification, the writers identify three main functions:

Population health management: Predictive algorithms can be applied to identify high-risk cancer patients with a higher chance of readmission after surgery or chemotherapy. Such data can prompt crucial preventive care while reducing costs and strain on a patient. Patt’s team, for example, reviews detailed readouts during staff huddles.
Radiomics: The field of computer-assisted texture analysis uses quantitative data from scans to study tumor characteristics. Recently, computers examined differences in lesions of 125 lung cancer patients treated with platinum chemotherapy at the Cleveland Clinic. The patterns could help predict which future patients might benefit.
Pathology: Inaccurate biopsy reads can lead to excessive or inappropriate treatment, the authors note. Artificial intelligence algorithms are offering deep insight on biopsy reads — Google claims its AI tool has 99 percent accuracy in metastatic breast cancer detection — and give oncologists more time to focus on other aspects of care.

Future uses of predictive analytics, the paper notes, could entail a wide range of skills, including more routine decision-making at the point of care and applying machine learning models to targeted next-generation sequencing panels that otherwise are too expensive as a blindly administered screening approach for an entire population.

Having treatment plans integrated within the structured data elements and supported by clinical decision reduces error because there’s variability in humans writing orders."

Dr. Debra Patt Executive Vice President at Texas Oncology

Such benefits aren’t just appealing to practitioners. According to research conducted last year by the Deloitte Center for Health Solutions, 84 percent of the 56 health system executives surveyed said analytics will be important to their organizational strategies in the next few years.

Likewise, a March survey conducted by the Society of Actuaries found 89 percent of health care executives plan to use predictive analytics in the next five years — 4 percent more than 2018.

Partnerships Boost Data Collection for Oncology Predictive Analytics

To glean greater insights that can inform care via machine learning, an immense amount of data is needed.

A major boost came in 2013, when ASCO launched CancerLinQ, a subsidiary created to give oncologists a robust monitoring system that collects and analyzes data from all patient encounters. Built using SAP’s HANA platform, the system uses data lake infrastructure to store high-volume data in a centralized repository for Big Data and real-time analytics.

With more than 100 organizations sharing data, the effort “is rapidly evolving into the largest and most robust source of real-world evidence ever assembled in oncology for comparative effectiveness research and discovery,” Dr. Robert Miller, the medical director of CancerLinQ, told the Journal of Clinical Pathways in 2018.

And earlier this year, the Food and Drug Administration made news after announcing it would partner with Syapse, a provider of cloud-based software for precision medicine, for a multiyear research project using real-world evidence to support regulatory decision making in cancer care.

The joint effort will evaluate multiple sources — including clinical data from EHRs and molecular data from testing labs — to explore conclusions beyond the findings of retrospective clinical trials, which stakeholders say can be limited in context and patient diversity.

It aims to “make prospective trials more efficient and more reflective of how care is delivered in the real world,” former FDA commissioner Dr. Scott Gottleib said in a January 2019 speech.

How Analytics Improve the Treatment Experience for Cancer Patients

Not only can predictive analytics help guide a course of cancer treatment, they can streamline workflows to boost staffing efficiencies and improve the patient experience.

The oncology infusion center at Wake Forest Baptist Health in North Carolina uses analytics tools to anticipate peak utilization times and adjust scheduling accordingly. That enables “a much easier way to work than the steep incline and decline we used to deal with,” Karen Craver, a clinical practice administrator for the health system, tells HealthITAnalytics.com.

A similar notion is in play at Stanford Health Care. Several years ago, the busy medical complex turned to iQueue, a predictive analytics platform, to better understand factors that affect traffic.

“We look at extensive modeling of patient types and impose real-world constraints,” LeanTaaS president Sanjeev Agrawal, whose company developed iQueue, explained to MedCity News. “How many doctors are available? How many nurses? How long is a shift? Are they using these beds for something outside of a chemo unit?”

After one year using iQueue, Stanford was able to accommodate 25 percent more patients without sacrificing quality or safety, Agrawal says. It also saw wait times decrease by 30 percent at peak hours.

Challenges of Predictive Analytics in Oncology

Still, the practice of predictive analytics isn’t yet widespread, especially among oncologists.

Compared to analytics-driven strides to address heart failure, pediatric asthma and other conditions, oncologists’ use remains “in the middle of the pack,” says Patt, who cites a number of hurdles that must be addressed.

Among them: HIPAA laws, which she says prevent “optimal data sharing.” Patt also cites the potential of incomplete patient data — a platform might contain outpatient data but not hospitalizations — and EHR tools that lack clinical decision support functionality.

Healthcare organizations must also support the necessary infrastructure to put predictive analytics into practice, says Dr. Parsa Mirhaji, director of the Center for Health Data Innovations at the Albert Einstein College of Medicine and Montefiore Medical Center in New York.

It’s “analytics that can consume these kinds of data service lines and workflow processes that are informed by a combination of care providers as well as automated systems collaborating with each other,” Mirhaji recently told HealthTech.

Patt considers the tools to be valuable for all specialties, even if startup costs seem daunting.

“Sometimes these investments in data infrastructure that are critical to improve in care delivery may be delayed because of the challenge in demonstrating return on investment,” she says, adding that resulting efficiencies and error reductions may satisfy those concerns.

“At the end of the day, as healthcare organizations, we all want to take terrific care of patients.”

JazzIRT/Getty Images