Healthcare data comes in two main types: structured and unstructured. Structured data includes clear, organized information stored in specific fields, like lab results, medication orders, diagnosis codes, or smoking status. Electronic Health Records (EHR) systems have used structured data for a long time to record patient health, making it easy to access and standard.
Unstructured data makes up about 80% of all healthcare information. It includes clinical notes, doctor’s stories, imaging reports, scanned handwritten notes, and other free-text documents. This type holds important details that are often not in structured fields, like detailed symptom descriptions, social history, doctor observations, and context that can help find patients for clinical trials or special care.
Using only structured data can miss many potential patients because important clinical details are inside unstructured formats. This is a big problem in recruiting patients for clinical trials and matching patients correctly.
For example, in lung cancer screening, Vanderbilt University Medical Center found that an AI tool using natural language processing (NLP) to look at both structured smoking data and unstructured clinical notes found 73.8% more eligible patients than using just structured data. The tool also found 119% more patients from groups like Black/African American communities. This shows that looking beyond regular fields is needed to find all qualified patients.
Finding patients for clinical research is a big problem. About 80% of clinical trials in the U.S. do not meet recruitment deadlines, and 15% to 20% don’t get enough patients to finish the study. These delays can add many months and cost sponsors between $600,000 to $8 million every day. This makes new treatments slower to reach patients.
Traditional recruiting relies heavily on structured data, which limits potential candidates and slows the process. Often, healthcare workers must manually search through records, including unstructured notes, to find patients who qualify. This is slow and prone to mistakes. Because of this, many patients are missed and studies get delayed.
To fix this, healthcare leaders are using AI tools that combine and analyze both structured and unstructured data in real time. This gives a full picture of patient history and eligibility. It speeds up recruitment and lowers the workload on clinical sites.
Artificial intelligence, especially through methods like natural language processing (NLP) and deep-learning neural networks, can turn unstructured data from clinical notes and handwritten records into searchable information. New models based on BERT (Bidirectional Encoder Representations from Transformers) help machines understand language and medical terms.
For example, BEKhealth, an AI company that works on patient matching, made a platform that pulls information from both structured and unstructured parts of electronic medical records (EMRs). It looks at over 24 million medical terms and synonyms to create a detailed patient profile that covers three times more trial criteria than usual methods. This helps find ten times more qualified patients and enrolls three times more candidates, all at twice the speed of older methods.
At Memorial Cancer Institute, BEKhealth’s platform helped doctors pre-screen more than 10 new lung cancer patients in three weeks, finding patients who were missed before. Working with Areti Health, this AI system found 200 eligible patients and scheduled eight appointments within one hour for their first client.
Carta Healthcare showed similar results after buying Realyze Intelligence, which uses AI trained by clinicians. Cancer centers matched seven times more patients to trials and doubled enrollment rates. Their method quickly processes lab values and unstructured notes, reducing manual work and speeding up patient matching by finding over 80% of criteria hidden in text notes.
This AI change is important for finding more patients faster and more accurately while easing the workload on medical staff.
In a healthcare system with limited resources, these benefits help manage clinical research and special patient services better.
Medical administrators and IT managers know that using technology is not just about data but also making workflows better. AI tools that analyze both data types are designed to work with automation systems to speed up patient recruitment and management.
The partnership between BEKhealth and Areti Health is an example. BEKhealth finds eligible patients using AI, while Areti Health automates tasks like pre-screening questionnaires, appointment scheduling, and reminders. Together, they create a complete system that speeds up enrollment and lowers the workload at clinical sites.
Automation plus AI helps by:
In the U.S., practices combining AI and automation see clear improvements in recruitment speed and lower costs. Finding 200 eligible patients and scheduling eight appointments in one hour is a good example of this change.
Multimodal AI means using algorithms that work with different types of data at once. Since 80% of healthcare data is unstructured, combining it with structured data gives a fuller patient profile.
Using the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM), AI tools standardize both data types. This allows U.S. healthcare organizations to share and analyze data securely, improving research and trial networks across multiple centers.
Companies like IOMED help bring data together through efforts like the European Health Data Space (EHDS). Although EHDS focuses on Europe, similar data-sharing happens in the U.S. with federated EHR systems by companies like TriNetX and Flatiron.
By using multimodal AI and shared standards, U.S. healthcare can reduce delays in patient recruitment, cut costs, and improve data quality to support study design and patient grouping. For example, a multi-hospital thyroid cancer study using NLP and patient data showed better clinical results and faster feasibility checks.
One important result from using AI on unstructured data is better identification of underserved groups often left out of clinical trials. The lung cancer screening study at Vanderbilt University found that NLP on unstructured notes increased finding Black/African American patients by 119%. This shows how detailed data analysis can help reduce gaps in healthcare by including more diverse patients.
Given healthcare outcome differences among racial and economic groups in the U.S., medical practices should work to open clinical trials to diverse groups. AI systems that use all patient information help find these candidates better.
Being able to analyze both structured and unstructured healthcare data is becoming very important for clinical practices and research organizations in the U.S. AI platforms that do this can find patients faster, with better accuracy, cut operational costs, and improve clinical results.
By adopting advanced data analysis tools and pairing them with workflow automation, medical administrators and IT leaders can make clinical trial recruitment and patient care more efficient. This helps deliver healthcare better and gives more patients access to new treatments.
As healthcare creates more complex data, using AI that understands all kinds of information will be more important for keeping medical practices competitive and following rules.
BEKhealth’s AI platform is designed to identify clinically qualified patients for clinical research efficiently, enabling organizations to speed up feasibility and detect more protocol-eligible candidates.
The platform leverages AI-powered patient matching by extracting structured and unstructured data from electronic medical records (EMRs) to enhance the identification of eligible participants.
BEKhealth claims to identify 10x more patients and enables enrollment at twice the speed, helping researchers find suitable candidates more quickly.
BEKhealth analyzes both structured data and unstructured data, including handwritten notes and medical records, to create a comprehensive patient profile.
The platform utilizes deep learning neural networks based on BERT to analyze and interpret unstructured texts from EMR records, identifying key medical entities.
BEKhealth employs a human-in-the-loop feedback mechanism to refine its models, maintaining a 93% accuracy rate in interpreting EMRs.
Feasibility reports provide researchers with real-time insights into patient populations, facilitating better decision-making and identification of untapped trial opportunities.
BEKhealth and Areti Health have partnered to streamline clinical trial recruitment by combining AI-driven patient identification with automated engagement and scheduling.
By utilizing BEKhealth, clinical trials can achieve smarter recruitment processes, faster enrollment, and stronger trial outcomes through efficient patient matching.
Within 60 minutes of implementation, BEKhealth can identify hundreds of clinically eligible patients, significantly speeding up the recruitment process for trials.
Simbo is using AI agents to automate all phone related workflows. Connect now to know more.
Schedule AI Agents Demo Now© Since 2019, Simbo AI.
