Tech-Enabling Evidence Generation: Four Types of Insights Gleaned from Unstructured Narratives

Key takeaways:

• Artificial intelligence (AI), machine learning (ML) and clinical natural language processing (CNLP) are being utilized to analyze unstructured clinician notes and results summaries, eliminating the need for time- and labor-intensive manual chart reviews.
• CNLP technology has the potential to read more than two million documents per hour.
• Data collection and shifting regulations have opened new opportunities for real-world data (RWD) and real-world evidence (RWE) utilization.

Healthcare leaders and life sciences companies are constantly looking for ways to optimize real-world data (RWD) for meaningful insight generation to inform and support prospective research and clinical trials. Using artificial intelligence (AI), machine learning (ML) and clinical natural language processing (CNLP) can accelerate efforts in a more cost-effective manner. PINC AI™ Applied Sciences (PAS) is using a combination of CNLP and ML technology, through their partner Clinithink, to deliver the fastest, most accurate analysis of unstructured data across patient populations. This technology has the potential to read more than two million documents per hour and utilizes clinical CNLP to recognize the myriad of ways clinicians describe conditions in real life.

With patient-clinician discussions and cancer staging data locked in narrative form, AI and CNLP help process the unstructured information contained within clinician notes and pathology reports and deliver insights into details such as biomarkers and early signs of disease that aren’t easily accessible through the electronic medical records (EMRs). These insights can then be leveraged to streamline clinical trials and generate real-world evidence (RWE) that is increasingly needed to secure regulatory approvals for medical therapies and other treatment interventions.

Whether seeking to better understand a vulnerable population, identify early signs of disease, pinpoint patients with stage 4 cancer or identify a suitable trial site, PAS’s solutions provide health systems and clinical decision makers with actionable insights gained in a fraction of the time that manual processing would require.

AI, ML and CNLP are helping boost research efforts, automate candidate selection for trials and generate actionable insights through the analysis of the written and spoken word contained in the unstructured text of clinicians’ notes and results reports.

Here are four types of insights PAS can help glean from unstructured data using Clinithink’s CLiX technology:

1. Identification of Clinical Trial Sites and Participants

According to the National Institutes of Health (NIH), more than 80 percent of clinical trials in the U.S. fail to meet their patient recruitment timelines. Typically, there’s a heavy burden on trial staff to manually review charts to identify patients who meet complex inclusion/exclusion (I/E) criteria. This I/E criteria defined in the study protocol and developed based on a research hypothesis can easily consist of 10-20 eligibility criteria that participants must meet. With unmatched access to structured and unstructured patient data, PAS can help identify potential participants outside of a site’s known population in accordance with narrow eligibility criteria and desired clinical and diverse demographic characteristics, which could boost diversity of recruited patients, a key goal of the Food and Drug Administration (FDA).

For instance, if researchers are conducting a study and seeking patients with metastatic ovarian cancer, unstructured data can rapidly reveal critical details such as tumor type, tissue type and substages to help researchers target patient populations that can meet the disease study’s narrow eligibility criteria. PAS’s technology-backed approach can help study sites quickly and accurately locate study subjects and increase the number of participants eligible for a clinical trial. In the past, traditional clinical trial sites were selected based on experience and relationships. CNLP is changing that by estimating eligible patient population volumes with a new level of granularity and accuracy previously unattainable when screening sites to match trial protocols.

This tech-enablement allows trial developers to assess the suitability of a site based on investigator availability, experience in therapy area and historical performance metrics. Based on the assessment, the sites that have the best chance to outperform against expected site metrics for each trial can be selected.

2. Identification of Early Disease for Intervention

“Early detection is the best detection,” is a statement patients and clinicians hear often. Unfortunately, there are many diseases that go undiagnosed until symptoms appear due to significant organ damage – for instance, even the subtle cognitive decline noticed by Alzheimer’s patients and their families may only become obvious years after disease onset. Once patients are at this stage and seeking care, it's too late to reverse organ damage or for intervention with potential disease modifying therapies.

Life sciences organizations and clinicians are continuously looking for ways to address this gap between disease diagnosis and treatment with medical solutions developed for clinical intervention earlier in patients’ diseases stages. However, this requires much earlier identification of specific disease states by looking for very subtle signs that are often found in the unstructured clinician notes in patient charts.

CNLP technology is well suited for learning these details with contextual relevance in narrative expression. By correlating these details with subsequent development of disease, it's possible to identify which risk factors and clinical signs and symptoms are most predictive of subsequent disease development. These can then be applied prospectively to create highly enriched cohorts of patients with very early disease for potential disease modifying therapies.

CNLP can rapidly scan millions of unstructured EMR notes and PDF reports, such as genomic results, to identify these patients for early intervention, further study participation or evaluation.

3. Patient Journey Mapping

Mapping a patient’s journey as they navigate a disease from early signs and symptoms through diagnosis to treatment and monitoring is bringing greater understanding of the care patients receive and how they access that care.

In a perfect world, patients would receive evidence-based care from internationally recognized evidence-based guidelines throughout their entire patient journey, but alas, we aren’t perfect. Using CNLP, clinicians and researchers can follow patient journeys in the real world and track changes as standards of care develop – a trend we expect to continue accelerating as personalized medicine continues to make care pathways more complex.

As part of this journey mapping, clinicians and researchers can gain an intimate understanding of which doctors are taking the lead in arranging diagnostics and initiating therapies, and which doctors are referring to specialist centers.

As an example, for oncology patients, we observe the impact of genomic diagnostics as inflexion points in a patient’s care and see current genetic mutations, and how this is directing choice of therapeutics. Because CNLP searches the entire patient record, it's possible to search within radiology and pathology reports as well as the oncologist’s records to build an accurate chronology of cancer staging and grading without having to rely on clinical coding, which doesn’t allow for the granularity required for deep insight.

Because the unstructured clinical note contains the actual thoughts of the doctor at the time of the patient consultation, it's also possible to see some of the reasons for why patient care may deviate from evidence-based guidance.

Many times, this is due to medical issues such as intolerance of medicines or the presence of other medical conditions; however, it's also possibly due to social barriers of care.

4. Predictive Modeling

With rapid digital transformation comes huge amounts of valuable data that can help reveal important medical trends. Clinicians and life sciences organizations are turning to predictive analytics to help improve healthcare, reduce pressure on clinicians and reduce costs. These data science informed predictions can help identify patient populations in critical moments along their healthcare journeys.

Predictive modeling is a form of advanced analytics that relies on tech-enabled data mining to detect previously unrecognized trends, correlations and patterns. Based on the trends in data, predictive modeling can generate actionable recommendations that can help clinicians and researchers provide better care for improved outcomes, improved operational efficiency and reduced costs.

PAS is utilizing AI, CNLP and ML to mine the structured data of the PINC AI™ Healthcare Database (PHD) and unstructured notes and results reports to help clinicians and life sciences organizations gain national-level insights into populations’ potential for disease diagnosis, progression or probability for serious medical events.

“Innovation in healthcare right now is very exciting and we’re focused on what the future holds for AI and CNLP technology,” said Donna Sabol, Senior Vice President and Chief Quality Officer at St. Luke’s University Health Network. “This technology’s ability to mine the unstructured narrative of clinicians’ notes and results reports is key to helping our clinicians identify patients with early signs of disease for intervention and implementation of evidence-based treatment pathways at scale across our health system.”

Real-time data analysis can identify populations at risk before they show signs of disease. Based on the possibility of a chronic condition, clinicians and life sciences organizations can focus their efforts on preventive interventions or development of therapies to help reduce the risk of future complications.

Predictive modeling also plays a role in clinical trials through analyzing patient‑specific data to help predict a therapy’s impact on patients, calculate the trial design, optimize dosing and help predict potential safety concerns and adverse effects.

In addition, CNLP technology can help extract social drivers of health data from clinicians’ notes such as proximity and access to health services, social isolation and support, helping to develop analytics that can predict social needs, overall health risks of a population and healthcare trends.

Premier is using its diverse and deep relationships with life sciences organizations, hospitals and health systems to dive into social drivers of health (SDoH) data to help illuminate diversity in patient communities and disparities in healthcare. Insights derived from this type of tech-enabled data mining can help inform appropriate actions to advance health equity by closing gaps in care or making an impact within an alternative payment model.

Taking things further, Premier’s Health Equity Collaborative will leverage patient-specific social needs data, analytics and benchmarking, individual assessments and peer-to-peer learning to guide life sciences and healthcare organizations as they work to improve the health of communities.

Advances in AI are enabling the collection and analysis of RWE to yield a deeper understanding of how therapies can benefit specific patient populations.

PAS’s CNLP expertise is helping life sciences and healthcare organizations search large amounts of data and quickly translate bits and bytes into actionable insights to help advance the development and delivery of medical solutions and evidence-based patient care.

With these tech-enabled tools, organizations can better predict likelihood, progression and severity of disease, conduct studies with the right trial sites and reach the most eligible populations more efficiently and effectively.

It’s time to leverage unstructured narratives to inform evidence-based care decisions and medical solution development.

Learn how your organization can potentially gain time and cost savings through utilizing CNLP technology to rapidly reveal the unstructured narrative to drive value and patient benefit and, most importantly, get the right solutions to the right populations at the right time.

For more:

• Learn about how PINC AI™ Applied Sciences (PAS) and PINC AI™ Healthcare Data (PHD) can drive your research and medical solution development strategy.
• Discover how CNLP is harnessing data to solve healthcare’s toughest challenges.
• Read our new e-book to learn more about how RWD, technology and innovative approaches are changing the life sciences landscape.
• Read more about unlocking the intel in population healthcare and patient data and using CNLP to take clinical trials to the next level.

The insights you need to stay ahead in healthcare: Subscribe to Premier’s Power Rankings newsletter and get our experts’ original content delivered to your inbox once a month.