The problem isn't recruitment — it's infrastructure
Go to ClinicalTrials.gov right now and search for any Phase II or Phase III trial with enrollment status "Recruiting." Nearly 80% of them will miss their estimated completion date. Not because the science failed. Not because the investigators are incompetent. Because they couldn't find enough patients.
That number has been consistent for over a decade. Despite millions of dollars in recruitment spend, sophisticated digital ad platforms, and armies of site coordinators, the industry keeps hitting the same wall.
The human cost of that delay is not abstract. Every day a trial runs late is a day patients who could have benefited from an experimental treatment are waiting — on standard of care, on placebo, or simply unaware a trial existed. In oncology and rare disease in particular, that delay is measured in lives.
"Every protocol amendment means re-screening patients who were already ruled out — a process that used to mean printing charts and sitting in a hallway for hours."
Why traditional recruitment keeps failing
The industry has been fighting the wrong battle. Most recruitment strategies focus on attracting more patients to trials — more ads, more outreach, better awareness. But the bottleneck is not at the front of the funnel. It's buried in the middle.
Fragmented EHRs. Patient data is scattered across incompatible systems — Epic here, Cerner there, a legacy SQL server in the basement. A site coordinator trying to find eligible patients has to manually query multiple systems, each with different data schemas and access protocols. The search isn't even searching across a patient's full record — it's searching across whatever happened to be digitized in whatever system they can access that day.
Manual chart review, circa 1990s. Clinical Research Coordinators (CRCs) spend an enormous portion of their time manually reviewing patient records to determine eligibility. They cross-reference protocol criteria against lab values, ICD codes, medication histories, and clinical notes. This is slow, error-prone, and doesn't scale. One coordinator can review maybe 50–100 charts per week. A trial that needs 2,000 enrolled patients might need 200 sites, each with multiple coordinators doing this work.
CRC burnout. The Tufts Center for the Study of Drug Development (Lamberti et al., 2021) found that 55% of clinical research staff report increased burnout since COVID. This isn't just a people problem — it's an enrollment problem. Burned-out coordinators make more screening errors, miss more eligible patients, and leave. Replacing a coordinator mid-trial costs 6–9 months of lost institutional knowledge.
Protocol amendments compound everything. Phase III protocols now average 5.96 million data points, according to the TransCelerate/Tufts CSDD benchmark (September 2025). That's not a typo. Every substantial protocol amendment requires manual re-matching of already-screened patients. Sites that thought they had a viable cohort suddenly find themselves back at square one, with no additional staff to run the re-screening.
Bottom line: The enrollment crisis is structural, not motivational. Throwing more budget at patient acquisition won't fix a broken screening process.
Four shifts that move the needle
Based on what we're seeing in trials that consistently hit enrollment targets, four practices separate the best performers from the rest.
1. Continuous screening, not periodic review. The standard model is batch screening: coordinators review charts in sprints before enrollment milestones. But eligibility changes as a patient's condition evolves — new lab results, new diagnoses, new medications. Continuous AI-powered screening runs 24/7 across EHR records, flagging patients the moment they meet inclusion criteria. This matters most for rare disease and oncology trials where eligibility windows are narrow.
2. Predictive risk scoring for sites. Enrollment forecasting at the site level is still largely intuition-based. Sites that will miss their enrollment targets often don't show visible warning signs until 90 days before the deadline — when there's nothing left to do. Predictive models that ingest site-level historical performance, protocol complexity, and patient pool characteristics can identify at-risk sites 6+ months in advance, giving sponsors time to add resources, open backup sites, or adjust inclusion criteria proactively.
3. Multi-site coordination at scale. For large multicenter trials, coordinating outreach across 50–200 sites is a logistical nightmare. Manual coordination introduces delays, inconsistencies, and gaps. Automated coordination — synchronizing screening protocols, outreach cadences, and enrollment reporting across all active sites simultaneously — reduces the coordination overhead that burns out site staff and slows enrollment velocity.
4. Protocol-change responsiveness. When a protocol amendment occurs, the manual re-screening process takes weeks to months. An AI agent that has been continuously screening can re-match the entire patient pool against the updated criteria in hours, generating a new ranked cohort immediately. This isn't just faster — it means patients who would have aged out during the manual re-screening window get enrolled in time.
What an autonomous screening agent looks like in practice
The model we're building toward at Syncra is an agent that continuously monitors EHR records against active protocol criteria — running in the background, not constrained to business hours. When a patient becomes eligible, the agent scores their fit against the full protocol, surfaces high-confidence matches to the site coordinator, and logs the reasoning so it's auditable.
For multilingual sites or diverse patient populations, the agent can generate outreach in the patient's preferred language. For patients who didn't respond to initial outreach, it can schedule re-engagement with updated messaging. This isn't a replacement for the coordinator's judgment — it's a layer that handles the mechanical work of scanning, scoring, and sequencing, so coordinators spend their time on the patients who need human conversation.
We're not there yet on all of it. But the infrastructure to do this exists today. The question is whether the industry will keep patching the old model or commit to building the new one.
"The trial that finishes enrollment on time isn't the one with the cleverer ad. It's the one that found every eligible patient before anyone knew they were eligible."