Introduction
Imagine you are presented with two options to keep your home or business safe. The first is a traditional camera, which takes a photo of the property every week. The second is a video recorder, which enables 24/7 around-the-clock video surveillance of the property. The choice is laughably obvious.
The same dynamic exists today in screening patients for clinical trials. While traditional solutions offer only point-in-time snapshots, generative AI-powered platforms like Triomics allow for continuous monitoring of patients and trials. In this whitepaper, we’ll explore why continuous screening matters and how Triomics Prism accomplishes the task.
The Status Quo: Point-in-Time Monitoring
At any point in time, a typical U.S. cancer center may have hundreds of ongoing trials and thousands of oncology patients. When evaluating each potential patient-trial match, dedicated clinical research staff must consider dozens of precise inclusion and exclusion criteria that can (dis)qualify a patient for the trial.*
Infographic highlighting how step-wise productivity gains compound.**
However, as the above infographic shows, successfully matching a patient and a trial is only one part of the puzzle, with timing an implicit, yet crucial, third variable. Consider:
1. Patient Interest in a Trial Varies Over Time If a patient has been stable for a long time on a given treatment regimen, the timing may not be right. Depending on specifics, it may not be ethical to switch to an experimental treatment to begin with, but even if it is, what appetite will the provider and the patient have to switch from what’s working?
2. Patient Eligibility Varies Over Time Over time, patients may have different results for key trial inclusion/exclusion criteria, such as age, functional status (e.g., ECOG score), or disease stage. Even if a patient is ineligible for a study today, they may be eligible in a week’s time, and vice versa.
3. Trial Logistics Take Time Compounding the above problems, navigating the administrative hurdles of a clinical trial takes time, for patients, providers, and researchers alike. This logistical friction can prevent patients from being enrolled in trials timely. For example, patients may undergo imaging or laboratory tests a few days in advance of their oncology appointment, in order to discuss the results with their oncologist at that time. Even when these scans show disease progression which may make the patient a good trial candidate, will the radiology department have entered their formal interpretation in the EHR? Will the clinical research team then have enough time to screen that patient, and communicate potential trial matches to the provider – all in this narrow window before the appointment?
Where Point-In-Time Monitoring Falls Short
We’re all constrained by the technological limitations and the system in which we operate. While doctors today aren’t necessarily smarter than doctors from prior generations, doctors today are far more effective at preventing and treating disease, simply because we have better tests, drugs, and scientific knowledge at our disposal. Similarly, though clinical research teams are highly devoted professionals, today’s tools fail them in three ways:
1. Computational Burden Returning to the above infographic, no human could screen one million potential data points at once – and were they to try, there would likely be another million data points’ worth of new patients and trials by the time they finished, years later. Necessarily then, research teams must take computational shortcuts: perhaps they only screen patients with noted disease progression or an upcoming appointment within a certain timeframe, and so on. While these shortcuts are reasonable and correct, given the technological constraint, they are not perfect, and potential patient-trial matches are inevitably missed.
2. Downskilling Much of trial coordinators’ time is spent on rote, objective tasks, such as evaluating whether patients meet trial criteria. This crowds out time for higher-level “human touch” tasks, such as empathically communicating with patients and providers about trial eligibility, and consenting and managing patients throughout trials. Additionally, both our experience and common sense suggest that most research staff find searching in EHRs for key data points less rewarding than human interaction, and as a result, job dissatisfaction and turnover in clinical research teams is often high.
3. The Timing Factor Finally, even if an industrious research team could screen every potential patient-trial match, they can do so only at discrete points in time, rather than continuously. Returning to our analogy of a snapshot vs. a livestream video, if the patient is “snapshot” on Wednesday, but then becomes study-eligible on Thursday, this potential pairing may well be missed.
How Triomics Prism Can Help
Prism is Triomics’ trial enrollment solution that addresses each of the above limitations.
1. Continuous Screening Triomics Prism leverages generative AI to continuously screen potential patient-trial matches, rather than provide moment-in-time snapshots. Additionally, our software specifically highlights patients with disease progression and upcoming appointments, empowering staff to focus upon the most time-sensitive work to boost trial enrollment.
2. Low Cost It constrains cost at pennies - dollars per screened pair. That’s far cheaper than human experts and 35 times cheaper than ChatGPT.
3. Low Labor Though Prism has been shown to beat the accuracy of trained experts, the platform enables human-in-the-loop verification of key data points by citing source documents, which we consider to be best practice in a no-fail space like oncology for a generative AI solution. Still, asking a human to verify cited patient information – and directly linking to the relevant documentation – is far more efficient than forcing research coordinators to search through the patient chart to find and input these data points.
4. Upskilling of Labor In turn, as data entry and verification is now quicker, research staff are now freed to screen more patients, as well as spend more time interacting directly with patients and providers. Over time, our intuition is that job satisfaction will increase and turnover will decrease as a result.
Finally, we believe that improving the single task of patient-trial matching can have department and institutional-level impact:
5. From Reactive to Proactive First, due to higher throughput, clinical trial teams can pivot from the current backlogs of patients to be screened, trials to be filled, and physician requests to be addressed. As the queues clear, the model can invert, and clinical trial teams can become proactive communicators – pre-emptively suggesting to physicians patients who may benefit from a clinical trial – rather than merely reacting to clinician requests.
6. A Win-Win Increasing clinical trial productivity drives hospital revenue, rankings, and prestige. Most importantly though, clinical trials also benefit patients. Trials can provide patients who may not have other options with an alternative therapeutic pathway – all while advancing the search for the next scientific breakthrough.
Footnotes
* As the science of diagnosing and treating cancer progresses, so does the complexity and computational burden of these inclusion/exclusion criteria.
** These estimates are intentionally conservative, and are round numbers meant to directionally illustrate the overall concept, rather than precisely quantify effect size.
