Ask any clinical research coordinator how they spend their first two weeks after a protocol activates and the answer is almost always the same: pulling charts. Not analyzing them. Not making eligibility decisions. Just pulling charts and moving through them one by one, looking for information that may or may not be in a findable place.
In 2026, with all the tooling available to clinical teams, this is still where enrollment dies. Not in the consent conversation. Not at screening. It dies in the hours before any of that, when a coordinator at a 15-bed academic oncology unit is reviewing a queue of 300 potential patients and has no structured way to rule anyone in or out at scale.
The actual shape of the problem
When we talk with coordinators about their pre-screening workflows, we hear a consistent pattern. The site receives a protocol with 25 to 40 eligibility criteria. Some of those criteria are checkable in structured fields: age range, primary diagnosis code, current enrollment in another trial. Those take seconds per chart.
But a meaningful share of the criteria live in free-text notes. Prior therapy lines. Disease stage documentation from a consult note three visits ago. Creatinine trends that require reading through lab narratives rather than pulling a clean value from a structured field. For those, a coordinator is doing something closer to reading comprehension under time pressure, and the cognitive load compounds when the criteria list is long.
We are not saying chart review is avoidable in every case. Some eligibility decisions genuinely require clinical judgment that no automated system should make. The real issue is the ratio: across the protocols we work with, roughly 60 to 70 percent of the chart review time goes toward criteria that are either clearly met or clearly not met and could be resolved much faster with better tooling. The remaining 30 to 40 percent is where human judgment legitimately belongs.
Why previous tools have not solved this
Electronic data capture systems made protocol execution easier but pre-screening harder in one specific way: they gave sites a structured place to put data that had been collected, without doing much to help identify which patients should be collected on in the first place. EDC and CTMS tools are enrollment-management systems. They are not candidate-identification systems.
Query-based tools that run against the EHR have been more promising. The problem is precision. A query that captures every patient who has received a prior line of therapy for a given diagnosis will return a list that is still too long to manually review at most sites. The coordinator still has to go through that list chart by chart, because the query can identify the broad population but cannot verify the specific combination of criteria the protocol demands.
The gap between "patients who might be eligible" and "patients who have passed a genuine first-pass screen" is where chart review time lives. Closing that gap requires reading clinical notes, not just querying structured fields.
A real example of how time compounds
Consider a Phase 2 solid tumor trial with 28 eligibility criteria. Of those, 12 require structured data: ECOG performance status, prior lines of therapy count, absence of certain comorbidities in the problem list. The remaining 16 require reading: the actual treatment dates from infusion notes, the response assessment language from radiology reads, the toxicity grading from oncology progress notes.
At a site with 40 potentially eligible patients in the queue, a coordinator can move through the structured-criteria pass in roughly 20 minutes. The free-text pass takes four to six hours. If the site is running three concurrent trials, that free-text burden stacks. Most sites we talk with report that coordinators have between 30 and 50 percent of their weekly hours consumed by this kind of review before any consent activity starts.
That is not a staffing problem. That is a tooling gap, and it has a direct effect on enrollment pace.
What changes when the free-text layer is covered
When we run Enrollvue against a protocol, the system reads the free-text criteria and maps them against clinical note content for each candidate in the queue. The output is not a yes/no eligibility decision. That is not appropriate given the clinical stakes. The output is a prioritized candidate list with evidence pointers: which note, which passage, which criterion is met or flagged for review.
The coordinator still reviews. But instead of starting from a blank chart and spending 10 to 15 minutes per candidate before reaching a judgment, they start from a pre-populated summary and spend 2 to 4 minutes confirming or overriding the system's read. The cumulative effect across a 40-patient queue is a full day of coordinator time returned per protocol.
One early site we worked with ran a Phase 3 cardiovascular outcomes trial where the coordinator was spending roughly 18 hours per week in pre-screening review. After integrating with Enrollvue, that figure dropped to under 5 hours. The trial itself consented its first patient 6 weeks ahead of the sponsor's projected timeline.
The organizational dynamics that keep manual review in place
Even when teams understand the tooling gap intellectually, chart review workflows are difficult to change for reasons that have nothing to do with technology. Sites have developed their own rhythm around chart review. CRCs have tacit knowledge about which parts of the chart matter for which trial types. IRB-reviewed procedures often reference "manual chart review" explicitly, and teams are cautious about workflow changes that could raise compliance questions during monitoring visits.
There is also a real concern about accountability. If a coordinator is making an eligibility determination, the reasoning is visible and defensible. If a tool flags a candidate as eligible and the coordinator consents them without reading the chart carefully, any protocol deviation traces back to a workflow that a monitor might challenge.
We are not saying these concerns are wrong. They are legitimate. The way we address them is by positioning Enrollvue as an input to the coordinator's review, not a replacement for it. The system produces evidence, not verdicts. That framing matters for adoption and it matters for regulatory defensibility.
What teams can do right now
Even without new tooling, there are workflow changes that reduce chart review burden. The most impactful is building a tiered review protocol: apply structured-data criteria first and build a ranked list before touching free-text at all. Sites that do this reduce their free-text review volume by 40 to 50 percent before any additional technology is involved, because many candidates are ruled out in the first pass.
The second is documentation quality at the site level. Trials where the site has previously run similar protocols develop a muscle for documenting the specific data points that eligibility decisions rely on. New trials in familiar therapeutic areas start with better note quality, and review is faster as a result.
Chart review is not going away. It is part of how sites confirm eligibility and maintain audit trails. But the version of chart review that exists at most sites today, unstructured, unassisted, and uniformly applied to every candidate regardless of how obviously ineligible they are, is a significant and correctable drag on enrollment pace.