When sponsors ask about pre-screening burden, they usually mean: how many hours does the coordinator spend per screened patient? That is a reasonable question, but it misses the more important one: who is actually doing the work, and under what conditions?
Over several months, we talked with research coordinators and sub-investigators at 12 sites ranging from academic medical centers to independent oncology practices. We were not trying to produce a generalizable study. We were trying to understand the actual mechanics of how protocols get matched to patients before the formal screening visit. What we found was a wider range of practice than we expected, and several patterns that sponsors and CROs consistently underestimate.
The Principal Investigator Bottleneck
In the most common workflow we encountered, the process looks roughly like this: a coordinator pulls a list of potentially eligible patients from the EHR or from an existing disease registry, reviews charts against the protocol's inclusion and exclusion criteria, and then flags candidates for the PI to review and approve before contacting the patient. The PI review step is where things slow down.
At sites with high trial volumes, PIs are doing this review in pockets of time between clinic responsibilities. One coordinator described the workflow as: "I put the list in a folder, and Dr. [name] looks at it when he can, usually Sunday nights." That rhythm introduces a structural latency that has nothing to do with coordinator availability or EHR access speed. The review slot is rationed, and it competes directly with patient care obligations.
What sponsors often do not realize is that the PI review step is not optional at most sites. Institutional policies and IRB protocols at many centers require PI sign-off before any patient contact related to a trial, even at the pre-screening stage. The coordinator cannot reach out to a potential participant without that approval. So the entire pipeline stacks up behind a single-person gate that is typically uncharged time for the PI.
Variation Across Site Types
The 12 sites split into roughly three operational patterns, and the differences were more structural than we anticipated.
At the larger academic sites, dedicated research nurse practitioners or sub-investigators had been credentialed to run pre-screening with only intermittent PI review. These sites handled the volume most efficiently, but they had also invested years in building those roles and the associated SOPs. The coordinator at one such site told us they could typically complete an initial eligibility pass on a new protocol within two to three weeks of receiving the protocol document. That is genuinely fast compared to the other end of the spectrum.
At smaller community sites, the PI was frequently doing the chart review personally, not just approving a list. In some cases, that meant a physician reading through free-text notes to determine whether a patient's prior therapy history was close enough to an exclusion criterion boundary to disqualify them. That kind of judgment call requires clinical training, and there is no way to delegate it to a coordinator without more guidance than most protocols provide. The result is that these sites enroll slowly not because they lack motivated staff, but because the work is physician-density-limited.
A third pattern appeared at sites that had participated in many trials in a specific indication. These sites had built internal eligibility checklists that went beyond the protocol's own criteria document. They had seen ambiguous criteria produce screen failures, so they had added internal decision rules. This made their pre-screening faster and more accurate, but it also meant their eligibility pass was filtering against a slightly different set of criteria than the sponsor expected. Discrepancies sometimes only surfaced at the formal screening visit.
What Coordinators Are Actually Doing With the Protocol Document
Several coordinators described a step we had not anticipated: translating the protocol's eligibility section into a working checklist before beginning any chart review. The protocol document is often written for regulatory and clinical review audiences, not for operational use. Compound criteria like "no prior treatment with any agent targeting X receptor within 6 months of enrollment" require some unpacking before a coordinator can check them against a chart.
This translation step varied from a quick informal markup to a documented SOP that the site kept on file and updated when the protocol was amended. The quality of this translation directly affected how consistent the pre-screening was across different coordinators at the same site. At sites without a documented translation process, we heard about cases where two coordinators reviewing the same patient had reached different conclusions about eligibility. This is not a failure mode that shows up in any sponsor-facing metric until a screen failure occurs.
We are not saying that sponsors should write operational checklists on behalf of every site. Site autonomy in protocol interpretation is both legitimate and practically necessary. But there is a real cost when protocol language requires interpretation at the site level that could have been resolved earlier in the document drafting process.
The EMR Search Problem
Most sites we spoke with were using their EHR's native search or reporting tools to generate candidate lists. A few had access to a third-party cohort discovery tool connected to their EHR. The gap in capability was significant.
Sites relying on native EHR queries were typically building searches around ICD codes, diagnosis fields, or problem list entries. This captures structured data reasonably well for diagnoses, but misses patients whose relevant history is recorded primarily in clinical notes. For many criteria, particularly prior therapy history or procedures documented during inpatient stays at other facilities, the relevant information simply does not exist in a queryable structured field.
The practical consequence is that sites using ICD-based queries often produce a candidate list that looks complete but is actually missing a significant fraction of potentially eligible patients. These patients are invisible to the pre-screening process not because they do not exist, but because their records are not structured in a way the query can find.
What we found at sites with better tooling was that they were not simply running faster queries. They were retrieving different patients: people whose eligibility depended on information that only existed in discharge summaries, oncology notes, or pharmacy records. The yield difference on initial candidate lists could be substantial, though it varied considerably by indication and protocol.
Where Sponsor Assumptions Break Down
Sponsors and CROs tend to model site pre-screening as a capacity problem: if we give the site enough time and resources, they will work through the eligible population. What we observed is that it is often a process design problem. The site does not have a queue that needs draining. It has a series of handoffs between roles, each with its own latency, and some of those handoffs are governed by institutional policies that cannot be accelerated by adding coordinator hours.
When a sponsor builds an enrollment timeline assuming a site can begin pre-screening at full pace the week after protocol receipt, they are typically underestimating the setup time for those internal translation and SOP processes. Sites that have run trials in the indication before are faster, but they are faster partly because they have already done that setup work. First-time sites in an indication are doing it fresh, and that takes time that does not show up in any feasibility questionnaire.
The variation in what sites actually do is worth taking seriously at the protocol design stage. Criteria that require chart abstraction from unstructured notes, that have narrow windows requiring precise date calculations, or that depend on information that may have been recorded at a different institution are going to create disproportionate burden at sites without the tools to handle those lookups systematically. That burden lands on the PI and coordinator team, and it eventually shows up as enrollment rate underperformance.
Understanding the actual mechanics of how sites work is a prerequisite for designing protocols and enrollment plans that hold up in practice, rather than on paper.