Challenges in Clinical Trial Recruitment: Decoding the Data Types

One of the things that makes clinical trial recruitment challenging for researchers is the broad variety of data sources needed to effectively identify eligible patients. For certain, being able to comb through multiple, diverse types of data can absolutely help research teams gain more precise understandings of the kinds of patients they need to target, but the wide range of data types and the massive volumes of data they produce make analyzing and deriving value from that data a complex process.

The Value of Structured and Unstructured Data Sources?

To understand how AI solutions can aid in patient recruitment, it’s essential to first recognize the various types of data that are utilized in this process. Data within these broader categories can be either structured, or unstructured – structured data being information that is formatted and standardized and, thus more easily analyzed, and unstructured data being information that does not have a predefined format, such as hand-written clinical notes, images, and audio recordings, which require more advanced processing techniques to interpret and utilize effectively. 

Whether representing structured, unstructured, or a combination of both, each type of data provides unique insights that can help identify suitable candidates for clinical trials. Here are some of the most common data types used for patient recruitment:

Electronic Health Records (EHR) / Electronic Medical Records (EMR) 

EHRs are digital versions of patients’ paper charts. They contain comprehensive medical histories, including diagnoses, medications, treatment plans, immunizations, allergies, imaging, laboratory tests, medical procedures and surgeries, clinical observations, various types of notes, and more. EHRs include both structured data (e.g., medical and billing codes such as ICD or CPT) and unstructured data (e.g. physician’s notes and reports).

EHR data provides a holistic view of a patient’s health status and treatment history, making it invaluable for identifying patients likely to meet study eligibility criteria.

Claims Data

Claims data encompasses information submitted by healthcare providers to insurance companies for reimbursement. This includes such data as diagnoses, long and short-term medication usage, and procedures performed. Claims data can help identify patient populations based on their healthcare utilization patterns and the treatments they have received, which is useful for targeting specific groups for recruitment. Claims data is almost exclusively structured in nature and de-identified at the patient level. Due to these factors, claims data is usually used for real world evidence (RWE) analyses around populations of patients as well as site selection as opposed to direct patient recruitment. 

Genomic Data

Personalized medicine is becoming increasingly common, even approaching the standard of care for many conditions. As a result, genomic data is becoming more available for clinicians and researchers alike. Genomic data includes information about an individual’s DNA sequence, genetic variations, potential mutations/variants. This information can come from both commercial vendors or core facilities at large hospitals. In addition, these can cover a variety of genetic sequencing offerings (e.g., targeted panels and whole “X” sequencing such as exomic or genomic coverage). For trials involving personalized medicine or therapies targeting specific genetic profiles, genomic data is necessary for study teams to create targeted recruitment strategies to find needles in haystacks for rare genetic variation – identifying patients who meet extremely specific criteria.

Social Determinants of Health (SDoH)

SDoH data includes information on the conditions in which people are born, grow, live, work, and age. This can include data on socioeconomic status, education, neighborhood and physical environment, access to transportation, employment, and social support networks. Understanding SDoH is important for addressing health disparities and ensuring diverse and representative trial populations. It is also necessary to tailor recruitment strategies toward underrepresented groups that have traditionally been less willing to participate in research.

Social Media and Online Forums

Social media data includes information posted by individuals to social media platforms. This is not generally used directly to identify candidates but can be used to understand how patients are discussing their medical situation with others. When analyzed effectively, this can help target recruitment engagement (e.g., adwords purchasing). Doing this well also requires ongoing sentiment analysis and monitoring. 

Registries, Health Information Exchanges, and Health Data Integrators

Health data integrators include information collected across healthcare organizations for streamlining the sharing of information between healthcare providers, for example a vaccine immunization registries maintained by states. In addition, this information can be used for ongoing observational analysis such as to search for clinical outcomes in larger patient populations. 

The Role of AI in Leveraging Patient Data

With the vast amounts of data available, the challenge becomes effectively integrating and analyzing this information to identify suitable trial candidates. This is where AI platforms, like BEKplatform, can help.

Data Integration and Harmonization

AI can aggregate and harmonize data from multiple sources, including EHRs, SDoH, genomic data, and more. This allows study teams to derive value from the data faster, enabling informed next steps. This is also necessary for effective master data management for deduplication as well as relationship mapping between patients, providers, and healthcare organizations.

Advanced Analytics and Predictive Modeling

AI algorithms can analyze large datasets to identify patterns and predict which patients are most likely to be eligible and willing to participate in clinical research. These predictive models consider a wide range of factors, including medical history, genetic information, demographics, lifestyle factors, and even characteristics of the research team.

Natural Language Processing (NLP)

NLP technologies, like the one developed by BEKhealth, can extract relevant information from unstructured data sources such as clinical notes, pathology reports, patient narratives, and even hand-written notes. This ensures that critical pieces of information concerning target patient populations are not lost. 

Efficiency and Speed

By automating aspects of the recruitment process like data ingestion and standardization, AI can reduce the time and resources required to identify patients. This not only accelerates the start of the trial but also reduces costs, improves overall efficiency, and ensures comprehensive evaluation of all candidates. 

Conclusion

The successful integration of data from broadly diverse data types is vital for effective clinical trial recruitment for current and future clinical trials. AI solutions like the BEKhealth patient recruitment platform give clinical researchers the ability to harness the full potential of all available data. By leveraging AI, researchers can streamline the recruitment process, ensure more accurate patient profiles, and ultimately enhance the success of clinical trials.