Multicentre studies need a shared view of data quality without reducing quality to one score. Useful indicators make a specific risk visible, use a stable definition and lead to an agreed action. Thresholds should be study-specific and interpreted alongside recruitment, workload and protocol context.
1. Completeness
Completeness asks whether expected forms or fields have been received. Define the expected set carefully: enrolled participants, required visits and applicable forms. Report both numerator and denominator, and distinguish forms that are late from those legitimately not applicable.
2. Timeliness
Timeliness can measure time from visit to entry, entry to review, or a scheduled milestone to completion. Pick the interval that reflects the operational risk. Median and percentile measures often reveal more than an average when a small number of records are very late.
3. Query ageing and resolution
Count open queries by age band and severity, not only the total. Track time to first response and time to closure where those states are reliable. A falling open-query count can be misleading if sites are receiving fewer reviews or queries are being closed without adequate resolution.
4. Reconciliation status
Safety, laboratory, randomisation and other external sources may need reconciliation. Define the records expected on each side, matching keys, allowed timing differences and exception ownership. Report unresolved mismatches and records that could not be matched, not just successful transfers.
5. Site-level variation
Compare sites using rates with appropriate denominators. Differences in recruitment, visit schedules, staffing or participant populations may explain variation, so dashboards should support investigation rather than automatic blame. Small-cell disclosure risks must also be considered.
Turn indicators into workflows
For every indicator, document its definition, data source, refresh frequency, owner, threshold, escalation and closure evidence. Version changes when the protocol or database changes. Reconcile dashboard results against source extracts before relying on them.
A monitoring dashboard is valuable when the study team knows what it means, trusts the calculation and acts consistently when a threshold is reached.