Maximizing the value of AI and machine learning in cancer treatment

iPad held over a Coffee Table
Oncology Informatics
Feb 21, 2023 6 minute read

Information is the engine that drives progress. And nowhere is that more important than in the delivery of cancer care. Today, more data is being generated than ever before: health records, diagnostic images, genomic and molecular data, pharmacological data, patient-reported data.

Hand iPhone

Artificial Intelligence (AI)-powered tools are transforming the landscape of oncology care data – and investing in AI and machine learning solutions has become a high priority for healthcare executives. AI is already contributing to cancer care and treatment, especially in radiation therapy, and AI can optimize workflow and unify operations to increase efficiency, profitability and effectiveness of care.

However, there is a missing element in most oncology AI solutions – real-time patient experience data.

To predict treatment outcomes more accurately for individual patients, and truly realize the full value of AI and machine learning in the oncology sector, providers need AI solutions that can learn from huge volumes of real-time data from patients, oncologists and cancer centers.

At Elekta, we recognize the potential of AI solutions to change the oncology sector – and are at the forefront of developing tools that can harness the crucial missing link of real-time patient data.

The benefits of real-time patient data – the missing link

Most of the data used within healthcare AI solutions is from past experiences or records. This makes it very challenging to truly be proactive and responsive to current patients.

A 2021 report supported by IBM Watson Health (now Merative) identified the benefits of using data from patient-reported outcome measures for enhancing the predictions of oncology AI solutions and improving patient outcomes. These included accurate estimates of a patient’s risk for experiencing a host of outcomes, including rehospitalization, cancer recurrence, treatment response, treatment toxicity and mortality – and concluded that patient-reported outcome measures (PROMs) represent an ideal mechanism to collect standardized data early in the process directly from the patient.

At Elekta, we’re working to bring together our patient engagement solution that provides personalized digital health interventions for cancer patients, based on self-reported data – with MOSAIQ® Plaza*, our certified health information system that manages documentation and workflow for oncology healthcare settings. This will create an integrated solution that uses real-time PROMs and realizes the true value of AI and machine learning for improved outcomes.

Recording and using real-time patient data also helps healthcare executives to expand their data sets and develop stronger, more versatile data strategies for their organization. As discussed in a recent Deloitte data strategy report, an effective data strategy helps chief data officers to engage stakeholders, plan and implement strategic projects, and emphasize successes – potentially driving a strong competitive advantage.

It’s not just having the data – it’s how you collect it

The 2021 study supported by IBM Watson Health identified that the demand to collect more patient data could exacerbate clinician burnout. This demonstrates that solutions that allow patients to input data themselves are poised to have a significant and lasting impact.

Because Elekta’s patient engagement solution enables patients to report data in real time through an easy-to-use app, the benefits are manifold: patients, oncologists and other caregivers can have greater peace of mind in the moment; clinicians are not burdened with the need to manually document data; and the machine learning algorithm can use these large data sets to improve the accuracy of its predictions.

Real-time data for improved patient outcomes

Many systems that currently use patient data take their data from patient surveys, rather than from real-time self-reports. This doesn’t allow for truly actionable insights because the data is gathered too late in the treatment experience. Real-time patient-reported data also allows for real-time interventions – equipping clinicians to make the right treatment recommendations at the right time.

Kaiku

A 2020 review of the outlook of oncology informatics led by Putora P.M. et al. found that collecting high-quality data through electronic patient-reported outcomes was feasible in the general population and had high acceptance rates from both patients and healthcare providers, demonstrating that patient self-reporting can simultaneously work to the benefit of patients, providers and algorithms.

So how does it work?

Using real-time patient data from diagnosis, to treatment, to follow up, the platform’s algorithms consider each item of self-reported data and provide the care team with an easy-to-read dashboard on the patient’s symptom development. The platform also triggers automated alerts to both care teams when symptoms get worse.

From diagnosis, patients who use the app have access to relevant education materials and resources.

During treatment, patients can report their own symptoms in real time and receive timely education on symptom self-management, or early intervention from HCPs. Based on previously reported symptoms, the app can also accurately predict future symptom development and provide guidance on how to ease the symptoms at home.

Once active treatment is completed, the system captures longitudinal outcomes, and can detect late-emerging symptoms, and alert users and HCPs to possible signs of relapse.

Real-time data, real impact. Predicting patient responses to ICI therapies

In a collaborative study with Oulu University Hospital (OYS, Oulu, Finland), a team demonstrated the power of machine learning to predict patients’ responses to oncology treatment.

The study investigated whether machine learning abilities would be able to predict the objective response rate (ORR) of patients undergoing immune checkpoint inhibitor therapies (ICI) using clinical and patient-reported data.

Kaiku Open on 3 Devices

ORR was defined as the proportion of patients for whom partial or complete responses were viewed as the best overall response.

The results showed that machine learning was able to predict ORR in ICI patients with 75 percent accuracy. The study received remarkable attention from the European Society for Medical Oncology (ESMO) when they were presented at the ESMO IO 2020 Virtual Congress.

Whole picture. Limitless possibilities.

Within MOSAIQ Plaza, the patient engagement solution can be used to enhance the comprehensive, multidisciplinary picture of oncology care by collecting ePRO data for any patient receiving treatment.

MOSAIQ Plaza’s patient engagement solution has already enabled us to support the following therapies, and inspires a clear vision for how machine learning algorithms can be developed for future therapy areas:

  • Breast and prostate cancer radiotherapy

    Predicting the occurrence and severity of treatment-related symptoms in breast and prostate radiotherapy patients for the upcoming week has enabled earlier intervention from care teams, proactive and personalized guidance and support for patients, and precise and timely follow-up with additional questionnaires.

    Engaging the patient in post-treatment follow-up, the care provider will be able to gather data from the period in which the sub-acute and late adverse effects of radiotherapy may emerge. This is important not only from the patient’s safety perspective but also allows for quality assurance of the treatment.
  • Immune Checkpoint Inhibitor (ICI) therapy

    Adverse effects caused by ICI toxicities can be life threatening, but most are reversible if detected and treated early enough. Prediction and early detection of symptoms related to ICI toxicities can lead to earlier intervention from care teams, proactive patient support, and precise and timely follow-up with additional questionnaires. Combining ePRO data with laboratory measurements, machine learning can also predict the onset of immune-related adverse events during ICI therapy with high accuracy.

Real-time patient-reported data integrated into MOSAIQ Plaza is the best way to equip clinicians with timely data to inform critical treatment decision-making.

MOSAIQ Oncology Analytics helps turn the right data into a “single source of truth,” and when used alongside our patient engagement solution, enables healthcare providers to maximize their operations by truly leveraging the value of AI and machine learning.

Learn more about our AI and machine learning software solutions.

*MOSAIQ Plaza is not available in all markets.

LARMSQ23015

Enjoy the article?

Feel free to share.