Definition, deliverables, and system boundaries
Electronic data capture system development and maintenance configure, validate, and continuously operate an environment in which authorized users can enter, check, correct, review, freeze, export, and retain protocol-required participant-level data. The service belongs to WSIC 6.5.1, Clinical Trial Services. Deliverables commonly include user requirements, data flows and source-data location, eCRFs and visit matrices, fields and controlled terminology, edit checks, calculations and query rules, role permissions, validation evidence, release and training, support and incident handling, change control, database lock, archive, and exit handover. The system must be fit for intended use, with controls proportionate to participant safety, blinding, and the importance of data to trial conclusions.
EDC is not a collective name for all clinical technology. An electronic health record or paper record may be the source, while an observation entered directly into EDC may itself be electronic source data. ePRO or eCOA, randomization and supply systems, CTMS, safety databases, central laboratories, image or sleep-signal repositories, and statistical environments retain distinct responsibilities and records. An interface does not automatically change ownership, provenance, or legal retention duties. This service receives project-level requirements and oversight from the preceding entry but concentrates on the system life cycle. It does not replace protocol design, medical judgment, data-management decisions, statistical analysis, privacy legal advice, or regulatory authorization.
Requirements, data flow, and eCRF design
Build begins by extracting testable requirements from the approved protocol, schedule of assessments, data-management plan, statistical plan, safety processes, and source-data list. Each field should have a reason for collection, originator, timing, source status, correction route, and destination. A data-flow diagram covers sites, EDC, laboratories, ePRO, DHTs, randomization, central reading, and safety databases. It identifies participant codes, interface keys, timestamps and time zones, units, dictionary versions, transformations, derivations, reconciliations, and downstream consumers. Requirements distinguish regulatory or protocol need, critical-to-quality value, operational convenience, and optional features so that low-value fields and duplicate transcription do not accumulate in forms.
An eCRF is organized around events and visits and represents required, conditionally displayed, unknown, not-done, repeating, withdrawal, loss-to-follow-up, deviation, and early-termination states. Field type, length, precision, unit, permitted value, partial date, medical coding, and cross-form relationships should match how information actually arises. Free text should not replace analyzable structure, while a rigid field must not force invented precision. Completion guidance explains origin, calculation, window, and exception handling. Each edit check needs a business rationale, severity, trigger, message, and closure route. Excessive low-value checks cause alert fatigue and can pressure a site to replace a true outlier with a superficially compliant value.
Configuration, validation, testing, and controlled release
Vendor assessment covers platform capability, quality system, personnel, development and hosting locations, subcontractors, security, continuity, portability, and inspection support. Development, test, training, and production environments should be appropriately separated, while configuration items, code, dictionaries, forms, rules, reports, and interfaces are version controlled. Risk-proportionate validation traces user requirements to specifications and tests. It covers both the standard platform and trial-specific configuration, automated calculations, critical edit checks, permissions, blinding, signatures, imports, exports, and interfaces. Generic vendor testing cannot replace sponsor confirmation that the implemented system is fit for the particular trial use.
Test data should exercise normal, boundary, erroneous, missing, duplicate, concurrent, cross-time-zone, excessive-privilege, recovery, and high-impact business scenarios. Records retain expected and actual results, evidence, deviations, and approval. User acceptance is performed by people who understand the protocol and real work; a fully passed script is not a substitute for risk judgment. Unresolved defects require severity assessment, rationale, and mitigation. Release occurs only after applicable approvals, validation summary, production-configuration review, accounts, training, support, and rollback readiness are complete. Form changes caused by a protocol amendment must also align with each site's authorized effective date.
User access, security, and electronic signatures
Every user has a unique identity and receives a role based on duties, organization, site, blinding, and least privilege. Request, approval, provisioning, modification, periodic review, suspension, and prompt removal leave records. Investigators should be able to see and supervise their team permissions. Elevated system-administrator and support rights should be segregated, time limited where appropriate, and monitored; shared accounts and borrowed signatures are unacceptable. Authentication, sessions, passwords or other credentials, network protection, encryption, malware controls, patching, vulnerability handling, and log monitoring are maintained according to risk, with emergency access and continuity arrangements preserved.
An electronic signature must remain inseparably linked to the signer, time, meaning, and signed record and satisfy applicable jurisdictional requirements. Logging in is not necessarily signing, and bulk signature must not conceal content that was not reviewed. Privacy by design minimizes direct identifiers and prevents unnecessary disclosure through screens, exports, notifications, screenshots, and support tickets. A security event records discovery, scope, containment, forensic preservation, recovery, affected data, and notification decisions. A vendor or administrator must not silently edit data in the back end; any necessary intervention needs authorization, a traceable record, and impact assessment.
Data entry, audit trails, and query management
The system preserves the initial entry and the old value, new value, operator, role, date and time, and applicable reason for every later modification or deletion. It also records relevant metadata for accounts, permissions, workflows, imports, freezes, signatures, and locks. Ordinary users cannot disable or overwrite audit trails and critical logs. Records should be understandable at data-point level and exportable in a searchable form. Audit-trail review is a planned, risk-based quality activity prioritizing critical data, unusual changes, bulk actions, blinding, and privileged activity. It is not an indiscriminate log review immediately before inspection, and automatic capture alone does not prove that anyone reviewed meaningful signals.
Entry and query workflows preserve attribution. A site answers and corrects against contemporaneous source records while the system preserves the original value and reason. A sponsor or vendor cannot rewrite investigator or participant data without authorization for tidiness, timeliness, or a preferred analytic result. Issues arising from automated checks, manual review, medical coding, and external reconciliation need distinct origin, responsibility, status, escalation, and closure evidence. Repeated queries should prompt analysis of rule or training causes. Investigators require timely access to their participants' data and relevant metadata and should endorse reported data at predetermined important milestones rather than provide a single retrospective signature before lock.
Sleep data, external interfaces, and reconciliation
Sleep trials may connect PSG, HSAT, actigraphy, wearable or contactless sensing, electronic diaries, central scoring, and laboratory results to EDC. System design states where raw waveforms or device files reside and whether EDC holds original data, reviewed results, derived endpoints, or only file references. It preserves device and firmware, algorithm and scoring-rule version, recording start and end, time zone, units, validity, repeat status, scorer, and adjudication provenance. Large waveforms are often better held in a controlled signal repository, but that is not a reason to discard original files, metadata, or the association between a signal and its structured result.
Each interface defines format, schema version, frequency, encryption, send and receipt confirmation, duplicate and late handling, retransmission, quarantine, error reports, and ownership. Validation uses challenging examples to test mappings, codes, precision, units, time zones, nulls, reversals, and identity matching, followed by reconciliation through record counts, critical values, checksums, or other suitable methods. A displayed value from an algorithm dashboard does not become an endpoint without definition. Device or scoring updates require compatibility and impact assessment. A technically successful interface proves that a transfer completed; it does not prove that data are clinically correct, complete, or suitable for analysis.
Operations, change, lock, archive, and retirement
Operations require a tiered help desk, incident and defect register, service levels, trend review, escalation, and CAPA, with continuing surveillance of capacity, performance, interfaces, security, backups, and critical jobs. Change control records rationale, risk, affected requirements and data, tests, approval, deployment, communication, and rollback, assessing already collected data separately from active sites. A small configuration adjustment, dictionary update, or vendor platform release still needs impact assessment. Backups must be paired with restore exercises. A downtime procedure explains safe temporary capture, identification of provisional records, post-recovery entry, and reconciliation; an untested backup is not evidence of continuity.
Interim snapshots, freezes, soft locks, hard locks, unblinding, and unlocks require distinct permissions, prerequisites, approvals, and audit evidence. Final lock follows planned checks, query resolution, coding, external and safety reconciliation, and investigator endorsement. A necessary unlock records reason, scope, changes, and relock. The archive includes data, relevant metadata, audit trails, configuration, dictionaries, signature meaning, validation, and operational documentation and remains retrievable, readable, and protected throughout retention. Migration or retirement exercises exports, verifies completeness and readability, transfers access and decryption conditions, and only then disposes of the old environment under authorization. EDC services cannot promise no missing data, no queries, absolute security, inspection without findings, or product approval.