# Predictive Enrollment Engineering™

Canonical page:
<https://dyno.clinicaltrialscan.ai/brochures/predictive-enrollment-engineering>

Predictive Enrollment Engineering™ is Clinical Trial Scan's methodology for
forecasting clinical trial enrollment outcomes from a protocol's own text and
from population data, rather than estimating them from historical study
averages.

## The core idea

Most recruitment plans price a study by analogy: a similar trial in a similar
indication cost a certain amount per patient, so this one should too. That
approach hides where patients are actually lost. Predictive Enrollment
Engineering instead decomposes enrollment into a sequence of independent
probabilities, each of which is modeled from a specific, identifiable driver in
the protocol or the local population.

## The seven-stage funnel (P1-P7)

| Stage | What it measures | Primary driver |
|---|---|---|
| P1 | Reach and response to advertising | Media cost, geography, indication awareness |
| P2 | Phone-screen eligibility | Criteria a patient can answer over the phone |
| P3 | Appointment booking | Site responsiveness and scheduling friction |
| P4 | Show rate (booked to attended) | Age mix of the eligible population, indication burden, caregiver support, condition urgency |
| P5 | Site records review | Criteria that require medical records, labs, or imaging the patient cannot self-report |
| P6 | Screen pass | Procedural and threshold criteria measured on site |
| P7 | Randomization | Remaining protocol gates and patient consent |

Multiplying the stage probabilities converts a cost per lead into a cost per
randomized patient. Because each stage is modeled separately, the output shows
*which* stage is destroying the budget, not only that the budget is too small.

## Inputs to the model

- Protocol inclusion and exclusion criteria, extracted verbatim and classified
  as phone-askable or site-confirmable.
- Eligible age range, used to weight the age distribution of the target
  population instead of assuming one universal age mix.
- Indication prevalence and co-indication subset rates.
- US Census demographics for the selected metro areas, states, or the full
  United States.
- Observed advertising performance benchmarks from comparable indications and
  geographies.

## Why it changes the answer

A protocol with many record-verifiable exclusions can look inexpensive at the
lead level and be extremely expensive at the enrollment level. A protocol with
a narrow eligible age range in a metro area whose population skews outside that
range needs a materially larger media budget than population size alone
suggests. Both effects are invisible to average-based budgeting and explicit in
a staged probability model.

## Related pages

- [Enrollment foundation](https://dyno.clinicaltrialscan.ai/brochures/enrollment-foundation)
- [The milestone fallacy](https://dyno.clinicaltrialscan.ai/brochures/milestone-fallacy)
- [The demographic fallacy](https://dyno.clinicaltrialscan.ai/brochures/demographic-fallacy)
- [Screen pass](https://dyno.clinicaltrialscan.ai/brochures/screen-pass)