Cognitive Assessments
Signant's Central Quality Review program has been built on a straightforward conviction: that expert human review of assessment recordings is one of the most powerful tools available for protecting endpoint data quality. CQRAssist doesn't replace that conviction. It makes it economically viable to act on it at scale.
Central Quality Review has always been the right approach. Independent, doctorate-level clinical scientists reviewing audio recordings of site-administered assessments - evaluating interview technique, probing adequacy, anchor point alignment, and scoring accuracy - and providing direct, targeted remediation to raters whose work doesn't meet the standard. That is rigorous, human-centred quality oversight. And for the studies where Signant has deployed it, it works.
The problem has never been with the approach. It's been with the economics.
The cost constraint that CQRAssist was built to solve
Across the industry, the standard for Central Quality Review is somewhere between 10% and 20% of assessments. That figure reflects a deliberate trade-off, not a technical ceiling. Full 100% human review is operationally achievable - Signant has the clinical staff to do it - but the cost per study makes it impractical for most sponsors. The 10-20% sampling model emerged as a pragmatic balance: if errors are infrequent, random sampling will likely catch enough of them; if errors are frequent, sampling should surface the pattern and trigger intervention. It is a reasonable approach within a constrained budget.
The errors that random sampling misses introduce variability into endpoint data that impact on study outcomes.
CQRAssist changes the economics of this trade-off. The question is no longer 'how much coverage can we afford?' but 'how do we direct expert review time where it will have the highest impact?'
What changes with CQRAssist
CQRAssist transcribes 100% of assessment audio recordings, applies speaker diarization to separate rater, patient, and caregiver voices, and runs a trained AI model across every transcript - evaluating the same administration and scoring dimensions that Signant's expert Central Quality Reviewers have always assessed.
The output is a structured quality signal for every recording in the study - not a sample, every one. That signal then determines the human workload. Assessments that are identified as high risk are automatically routed to the Central Quality Review queue.
Expert reviewers, the same doctorate-level clinical scientists who form the core of Signant's Central Quality Review programme conduct their review independently, with no visibility into the AI's output. Their findings, alongside CQRAssist's outputs, feed into PureSignal Analytics: a longitudinal record of rater and site performance across the full 100% of the data.
The validation behind the claim
CQRAssist was validated against Signant's own expert Central Quality Reviewers, who reviewed 100% of a retrospective assessment set to establish confirmed ground truth. CQRAssist was then run across the same data, blinded, and its outputs compared against that expert-established baseline.
Across thousands of assessments spanning key CNS and cognitive scales - PANSS, MADRS, HAM-D, NPI-C, ADAS-Cog, MMSE - the results consistently showed that CQRAssist detects the substantial majority of confirmed issues, with high specificity that keeps unnecessary reviewer burden low. The positive predictive value is several times higher than random sampling, meaning expert reviewers spend their time where quality risk is real.
Where CQRAssist sits within the integrated data quality approach
CQRAssist operates at the at-visit layer of Signant's integrated data quality architecture, alongside eCOA Edit Checks that fire logic in real time during administration. Every finding from both flagged and clean assessments feeds into PureSignal Analytics, Signant's blinded data analytics engine. When PureSignal Analytics surfaces a pattern, it connects with everything else known about that rater and site. An at-visit finding that is identified as risk doesn't sit in a review queue and close - it contributes to a study-wide quality picture that supports both immediate remediation and longer-term performance tracking.
CQRAssist adds the piece that has always been structurally constrained by cost: comprehensive, evidence-based triage of the entire assessment population, so that expert human reviewers are always focused where quality risk is highest.
Regular PureSignal Analytics meetings and direct access via the self-serve Insights Portal from first patient first visit onwards mean sponsors can engage with the full quality picture in real time - not retrospectively, when intervention is no longer straightforward.
The continuity that makes this credible
There is a version of AI-augmented quality review that treats human expertise as overhead to be reduced. CQRAssist is not that. The Central Quality Reviewer remains the endpoint of the process for every assessment that is identified as high risk. The remediation conversation and direct feedback between reviewer and site rater still happens.
From random sampling to evidence-based focus
CQRAssist replaces probability-based sampling with evidence-driven selection. Expert reviewers are guided to the highest-risk data, making previously unexamined assessments visible - not by replacing human review, but by directing it precisely where it matters.
CQRAssist is a logical extension of what Signant's Central Quality Review programme has always done: put expert human oversight at the centre of endpoint protection. The approach is the same. The coverage is complete. And the economics are finally workable.
About the authors
Helen Brooker is a specialist in cognitive test development and aging, with over 18 years of experience in academic and clinical research. She leverages her expertise in neuropsychological assessment, clinical trial delivery, and digital health solutions as a Senior Product Manager at Signant, where she oversees the company's proprietary computerized cognitive test solution.
Dr. Alan Kott is the Practice Leader for Data Analytics at Signant Health, with both academic and industry experience in clinical trials. He has led the development of Signant’s Data Analytics Program, overseeing data analytics in over 200 clinical trials across multiple indications. Prior to joining Signant, Dr. Kott was an Assistant Professor at Charles University and a house officer in psychiatry at General
