Estimated reading time: 2 minutes
Prescreening candidates can typically be a limiting process with oncology clinical trials. 
It requires manual efforts towards collecting qualifying data, scanning charts, and confirming eligibility. This process is a hassle that creates delays and increases costs that which cannot find enough patients within the clinically-controlled settings to evaluate such productive results. AI has played a large role in the healthcare realm, but now as cancer is taking the lives and livelihood of so many, AI could really come in hand before the worst of it monopolizes.
The Problem with Patient Pre-Screening
Identifying individuals who might make the appropriate candidates for specific oncology treatments and trials requires an understanding of the disease’s state of being. This is needed so that healthcare professionals can meet the increasingly complex and specific eligibility requirements. EMRs were designed to facilitate clinically kept billing processes and the concept of tracking assessments, activities, and treatment plans. This is known as structural diagnostic data.
80-90% of medical data is unstructured when read by humans, not machines. This highly consists of pathology reports, lab reports, and clinically-recorded images. The Non-Machine-Readable Nature coupled with the fact that the most useful clinical data is unstructured, however since it is recorded and analyzed primarily by humans, unstructured data requires manual efforts. This process is slow, costly and can skip over the most eligible patients, while bringing an overload of challenges to clinical centers and hospitals.
Essential Factors for Clinically-Driven Automation
Data Extraction: AI needs to possess a data extraction skill, so it can efficiently sift and pull through the mass of unstructured data. Extracting is not understanding, so that’s where NLP comes into the situation.
Natural Language Processing (NLP): Once the data has been extracted it goes onto the next step, Natural Language Processing. Clinical language is just as vague, context-wise, as ordinary language seems to be. 
AI must be constructed specifically to a certain therapeutic area, including oncology. This way it is better defined. The more general AI is, the weaker it will perform within specific parameters. Only domain-centered AI tools can complete such specialized tasks.
Clinical Staff Guidance: AI is still just a tool with the capability of augmenting human skills, not replacing them. Nurses, doctors, and other clinicians are the only ones fully experienced in oncology studies. These staff members can refine and guide recommendations made by AI applications.
Formulating the Relationship: AI needs large amounts of patient data in order to perform properly perform pre-screenings. This data is sometimes difficult to sift through and can stored in separate locations. 
On top of that, this data is typically guarded by a wide range of access rules. Software developers need to formulate relationships with hospitals and clinics to acquire that content efficiently enough.
Experience in securing, acquiring, transferring, storing, and processing data is a vital tactic to possess. It shows reliability on both ends. When it comes to AI, integrating contracts and processes with external hospital operations are a key strategy to be mindful of. AI tools working in collaboration with oncology patient pre-screening trials, but cooperation from both parties is needed and necessary.
