Smart Data Quality (SDQ) Query Anomaly Detection (QAD)

Query Anomaly Detection uses generative AI to find contextual, cross-domain data anomalies — and automatically generates the queries to resolve them.

Find what rules can’t anticipate

Query Anomaly Detection (QAD) uncovers unknown data patterns and anomalies that manual and rule-based review routinely miss. Rather than checking only what a rule anticipates, QAD uses generative AI to evaluate relationships within and across clinical domains and detect contextual inconsistencies.

QAD ships with roughly 50 out-of-the-box scenario categories spanning safety domains, and sponsor-specific scenarios can be added for any domain — with actionable queries generated automatically.

Why teams choose Query Anomaly Detection

Find the
unknown

Detect patterns and anomalies that rule-based and manual review never anticipate.

Cross-domain validation

Catch contradictions between treatments, symptoms, labs, and findings.

Automated query generation

Turn anomalies into actionable queries, including cascading flags across dependent domains.

Customizable

Layer study-specific requirements on top, with dynamically evolving rules.

~50

Out-of-the-box anomaly scenarios

Across safety domains, with sponsor-specific scenarios on top — plus automated query generation

How Query Anomaly Detection works

QAD applies layered AI analysis to incoming data, then drafts the query.

Validate logical relationships

QAD checks inter- and intra-domain relationships to detect contextual anomalies.

Detect domain-level anomalies

AI pattern analysis evaluates lab results and adverse events, including timeline validation.

Check cross-domain consistency

QAD validates relationships across domains and flags treatment/symptom contradictions.

Generate the query

Actionable queries are created automatically, with cascading flags across dependent domains.

Tuned to your study. Out-of-the-box scenarios cover the common cases; sponsor-specific scenarios and dynamically evolving rules adapt QAD to your protocol, and ongoing fine-tuning reduces false negatives over time.

Features

~50 out-of-the-box scenarios

Covering adverse-event grade and outcome anomalies, AE-treatment consistency (AE-CM), ECG versus adverse-event assessment (AE-EG), clinical measurement and lab-value anomalies, medication-indication mismatches, and data-entry or coding errors.