Category: eClinical Systems
Patient-reported entries arrive late, incomplete, or not at all once novelty wears off — and a missed diary window is data lost for good.
Patient-reported outcomes and diaries have an unforgiving property: once an entry window closes, the data is gone — it can't be queried into existence later. Engagement predictably decays over a study's life, and by the time low compliance shows up in a listing, weeks of endpoints may already be missing. Compliance-monitoring tools track engagement patterns participant by participant and flag the ones who are slipping while the data is still recoverable, so sites can intervene early instead of documenting losses late.
Compliance tooling works on the engagement data the ePRO platform already produces: when entries happen, how completeness trends, and how each participant's pattern is changing. Rather than reporting last month's compliance rate, predictive approaches flag participants whose recent pattern — later entries, thinner responses, lengthening gaps — looks like the run-up to disengagement, and route those flags to the site while outreach can still change the outcome. Some platforms pair this with configurable reminders and scheduling that adapt to a participant's demonstrated rhythm. What the tooling cannot do is supply the reason or the fix: whether a participant is struggling with the app, the schedule, or the trial itself is a conversation between the participant and the site, and retention decisions are clinical ones. The honest framing is earlier visibility — sites hear about a slipping participant while the entry windows ahead still matter, rather than after the endpoint data is unrecoverable.
Flag quality is the first test: on a completed study's engagement data, do the participants the tool would have flagged correspond to the ones who actually became non-compliant — and how early would the flag have come? A flag that arrives after the site could already see the problem in a report adds process without adding time. Probe the intervention pathway too: a flag only matters if it reaches the site in a workflow they act on, so look hard at how alerts land and what sites are expected to do with them. Because this involves participant behavior data, review the privacy and ethics posture: what the tool infers about participants, what sites and sponsors see, how that use is covered by consent, and whether engagement profiling stays within what participants agreed to. Blinding also needs a check — engagement patterns must not leak information that could unblind site staff or sponsors.
A retrospective look at a finished study is the cheapest evidence: replay the engagement data, see which participants the tool flags and when, and compare against the compliance outcomes the team already knows. If the flags are early and accurate, a live pilot on a subset of sites — flags on, with a defined outreach playbook — measured against comparable sites without the tooling shows whether earlier visibility actually changes completeness. The comparison has to be honest: better dashboards that don't change outcomes are cost, not value.
It watches each participant's engagement pattern and flags the ones trending toward disengagement while intervention can still recover future entries. Sites then do what tooling can't — talk to the participant. The gain is timing: outreach happens before the missing data accumulates, not after.
Most patient-reported instruments are valid only within their entry window — a pain score recalled a week later is a different measurement than one recorded that evening. Once the window closes, the protocol treats the entry as missing, which is why prevention is the only real remedy.
It uses behavioral data — when and how participants interact with the app — so it must stay within what consent covers and what ethics review approved. Ask what is inferred, who sees it, and how it is protected. A credible vendor treats those as design constraints, not afterthoughts.
Test it retrospectively first: on a completed study, check whether its flags identify the participants who really became non-compliant, and early enough to have mattered. Then pilot live against comparable sites without the tooling and compare data completeness. Earlier alerts that don't change outcomes aren't worth the spend.
The European Medicines Agency issued a qualification opinion on PROCOVA, the prognostic covariate adjustment methodology underpinning Unlearn's digital-twin approach, supporting its use to increase statistical efficiency in randomized controlled trials.
Vendor: Unlearn.AI · regulatory filing (European Medicines Agency) · 2022-09-01
Medable launched a partner program to help contract research organizations and technology partners stand up its decentralized and hybrid trial platform faster, with greater control and transparent pricing.
Vendor: Medable · press release (Medable) · 2025-06-10