From Recruitment to Retention: Closing the Leaks in Clinical Trials

Some participants leave because the study was hard to take part in. Others leave because the protocol asked more than anyone could sustain. Reporting a single dropout rate hides which is which, and guarantees the wrong response.

Where the money goes

Around 85 percent of clinical trials fail to retain enough patients, the dropout rate across trials runs at approximately 30 percent, and between 32 and 40 percent of a trial's budget is dedicated to recruitment.

Read those together. A third or more of the budget goes to bringing participants in, and roughly a third of them leave before the end. The industry is running a process with substantial leakage and responding primarily by increasing the inflow.

That response is not irrational. Recruiting a replacement is a known process with a known cost. Diagnosing why the last participant left is neither.

Two different losses, one number

Almost every organisation reports retention as a single figure, and that figure merges two problems with nothing in common.

The first is avoidable friction. A participant missed a visit because the reminder never arrived, or arrived by a channel they do not use. They could not reach anyone with a question between visits. They did not fully understand what the next assessment involved. Their travel stipend took eleven weeks to process. They lost the information sheet and were too embarrassed to ask for another.

None of that is a protocol problem. It is operational, it is fixable within a study rather than between studies, and the participant would have stayed if the experience had been better run.

The second is designed burden. The schedule of assessments genuinely asks more than a working adult with a chronic condition can sustain for eighteen months. No reminder cadence solves that. The decision that caused the dropout was made in protocol development, months before the participant was recruited.

Merging the two into one dropout rate means neither gets the right response. Friction gets treated as inevitable, and burden gets treated as an engagement failure.

The burden half is real and growing

Between 2015 and 2025, procedures in phase 3 pivotal trials rose by more than 60 percent, from 187 to 301. Tufts CSDD analysis covering several thousand protocols found a near doubling of endpoints and a tripling of data points collected between 2010 and 2020, and found this inversely related to trial performance through longer cycle times, more amendments and higher dropout rates.

That last clause matters. The relationship between protocol complexity and dropout has been measured, not assumed.

It accumulates through individually defensible additions. A translational team requests an extra biomarker sample. A regulatory function adds a safety assessment to pre-empt a question. A commercial function requests a patient-reported outcome to support a future label claim. A statistician adds a secondary endpoint. Each is reasonable, each is approved by someone with legitimate authority, and nobody owns the total. The participant experiences the total.

There is rarely a role in protocol development whose explicit job is to argue for removal. Every other function has a reason to add.

Closing the friction half

The avoidable losses are worth attacking first, because they can be fixed inside the current study and the fix does not require anyone to reopen a protocol. Six capabilities in Cloudbyz Patient Recruitment map directly onto them.

Missed Visits: Most are logistical rather than motivational. The Appointment Calendar generates study-specific slots and sends branded confirmations, reminders and follow-ups automatically by email or text. The mechanism matters more than the message: the reminder fires from the same system that holds the visit schedule, so it cannot drift out of sync when an appointment is moved. Reminders sent from a separate tool go stale the moment a site reschedules, which is precisely when the participant most needs one.

Comprehension Gaps: The Participant Portal carries patient education content, information sheets, eBooks and FAQs alongside eConsent, ePRO, lab reports and the participant's own screening results. A participant who can see what is coming, in language written for them, is materially less likely to withdraw ahead of a visit they do not understand. Comprehension is also measurable here rather than assumed, because eConsent records where participants hesitate.

Isolation between Visits: The platform includes a moderated forum supporting communication between participants, sites and researchers, as public posts or private messages. This is the least discussed retention lever and one of the cheapest. A participant who can ask a question on a Tuesday and get an answer behaves differently from one whose only contact is the next scheduled visit.

Payment Delay: Stipends and reimbursements are processed directly from the volunteer database rather than through a separate finance cycle. Sponsors underestimate this consistently. A participant waiting three months for a travel reimbursement is receiving a clear signal about how much the study values their time, and they act on it.

No proactive contact. Recruitment Call Centre Management places inbound and outbound calls inside the application, with call recording and per-recruiter metrics covering call volume, duration and time between calls. The same capability used to convert candidates works for retention outreach, and because calls are logged against the participant record, a pattern of unanswered contact becomes visible rather than anecdotal.

No early warning: Reports and Dashboards cover recruitment, screening and retention metrics across sites and studies, and the Volunteer Database segments participants by demographics and medical history. Together these show which cohorts and which sites are losing people, which is the difference between a retention programme aimed at the study and one aimed at the seventeen participants actually at risk.

None of this reduces what the protocol asks. It removes the reasons people leave that have nothing to do with the protocol at all, which is what makes the remaining dropout interpretable.

Instrumenting the burden half

The second problem needs different work, and this is where most retention programmes stop short.

Predictive retention modelling has advanced and it does work. Models scoring participants monthly against demographics, clinical profile, site history and live operational signals identify at-risk participants earlier than missed-visit monitoring alone, giving site teams a window to intervene.

But prediction is treatment, not prevention. It finds people the protocol is failing and prompts an effort to keep them. Deployed without a feedback loop into design, it is needed permanently, at rising intensity, as complexity keeps climbing.

The version that changes anything is the one where retention data reaches the next protocol. That requires something specific to be possible: visit-level and assessment-level burden data from completed studies has to be available, queryable, to the people drafting the next schedule of assessments. Which visits had the highest missed rates. Which assessments correlated with subsequent discontinuation. Which instruments saw completion fall away after month four.

In most organisations that information technically exists and is practically unreachable. It sits inside a closed-out study's data capture system, summarised in a clinical study report protocol authors do not read, in a format nobody can query across studies.

Because Cloudbyz ePRO and eCOA capture completion as structured records on the same platform as study management and recruitment, assessment-level patterns from prior studies are a report rather than a research request. Reports and dashboards cover recruitment, screening and retention metrics across studies, which is what makes the comparison possible at all.

The feedback loop stops being a good intention and becomes a query someone can run in an afternoon.