World Sleep Industry ENCYCLOPEDIA

PSG reporting workstations

2.2.3

Definition, purpose, and system boundaries

A PSG reporting workstation is a clinical computing system that turns quality-reviewed and scored polysomnography records into diagnostic reports that can be reviewed, interpreted, signed, and distributed. It commonly receives patient and study identity, technical comments, scored events, sleep architecture, respiratory and oxygen statistics, cardiac and movement findings, graphics, and selected segments, then produces structured results and formatted documents. Its defining task is not placing numbers into a template; it is ensuring that every reported result is traceable to the correct study, score-set version, and calculation rule.

This entry has a deliberate boundary with the sleep lab data workstation in entry 751. The data workstation retains source waveforms, annotations, and the review environment, whereas the reporting workstation performs summarization, clinical interpretation, attestation, and release. It is also more than a word processor, an electronic-record inbox, or a PDF print utility. Integrated PSG software may combine acquisition, scoring, reporting, and archive functions, but each function still needs its own sources, permissions, states, and failure analysis.

The complete system may include a report editor, rules and calculation services, a template library, terminology and reference ranges, electronic signature, version storage, an interface engine, print or PDF services, and electronic-health-record connectors. Deployment documentation should state supported study types, adult or pediatric scope, diagnostic or treatment modes, available algorithms, data sources, and intended users. A template configured for an adult diagnostic PSG cannot be extended without validation to infants, MSLT, MWT, split-night titration, or home respiratory testing.

Study identity, source, and report state

Every report should bind to a stable patient identifier, study identifier, study date and location, request context, procedure type, recording system, recording interval, and the final score-set version used. A repeat examination, same-night segments, reacquisition, treatment titration, and outside import must retain distinct provenance even when they concern the same patient. Similar names and dates are not sufficient matching controls. Critical identity should be prominent when a report is opened, saved, signed, and sent so an interpretation is not entered into another patient's window.

Sources should distinguish manually entered content, PSG-populated values, algorithmic calculations, outside-device imports, and physician interpretation. If scoring changes after report generation, the workstation should mark the report stale or require renewed review rather than silently refreshing selected numbers. Missing source data, unavailable video, poor channel quality, incomplete treatment logs, or early study termination should remain visible in technical comments and interpretation. An empty field must not automatically become zero, normal, or no abnormality detected.

States should at least distinguish draft, preliminary, awaiting signature, final, appended, corrected, withdrawn, and entered in error, with institutional policy controlling who can advance each state. Previewing, printing, or exporting a draft must not make it appear finally attested, and signing should not proceed while unresolved validation errors remain. Cancelling an unperformed study is different from withdrawing a previously released report; the latter has implications for received copies, decisions already made, and notification duties.

Calculation rules, denominators, and consistency

Total sleep time, sleep efficiency, stage percentages, arousal index, respiratory-event indices, oxygen-desaturation index, limb-movement index, and treatment-pressure summaries all depend on explicit definitions. The workstation should retain the scoring rules, software version, event thresholds, included intervals, and denominator. Values based on total sleep time, monitoring time, or total recording time must not all receive the same unexplained per-hour label. A study without EEG, in particular, cannot have recording time automatically relabeled as sleep time.

Units, rounding, decimal places, missing values, and true zeros should agree across structured fields, narrative, tables, and plots. After rescoring, both numerator and denominator need recalculation. A manual override should retain its reason and prior value. Combining nights, splitting a study, comparing baseline and treatment periods, or analyzing body position and sleep stage requires explicit intervals. Overlapping intervals must not be counted twice, and a whole-night average must not hide a short but important episode of hypoxemia or rhythm disturbance.

Reasonableness checks can compare total duration with the sum of stages, event totals with indices, the oxygen nadir with its trend, treatment settings with their intervals, and measured channels with populated fields. These checks identify contradictions; they should not automatically repair clinical data. When source records and summaries disagree, the workflow should return to the source to investigate scoring, artifact exclusion, unit conversion, or interface mapping instead of changing the final number merely to make the table balance.

Report content, graphics, and readability

According to study type, a report should identify the recording date and location, measured variables, technical quality, behavioral factors, relevant medication or treatment conditions, scoring definitions, and denominator. A full PSG commonly also presents sleep continuity and architecture, arousals, respiratory events, oxygenation, heart rate or rhythm, movements, and position. Treatment and split-night studies should clearly separate baseline and intervention periods. Not measured, not applicable, inadequate quality, and a genuine zero result require different representations.

Tables and figures should expose units, time axes, legends, statistical definitions, and missing intervals. Whole-night trends may show relationships among position, stage, treatment setting, and events, but a thumbnail cannot replace review of source waveforms. Automatically selected representative events may omit a preceding arousal, artifact, or sensor fault; the author should be able to verify and replace them. Color should not be the sole encoding for normality or abnormality, and the report must remain understandable in monochrome printing, on different displays, and with accessibility needs.

Templates should support completeness without producing boilerplate noise. A test that was not performed must not leave behind a default normal sentence, adult and pediatric reference text must not be mixed, and branded explanations must not conceal study limitations. A conclusion for a referring clinician should be clear without losing precision; a patient-facing form may explain terminology but should preserve material uncertainty. Headers, page numbers, patient identity, report state, and version should remain associated with every page or detachable section to prevent page mixing after print, scan, or fax.

Clinical interpretation, review, and signature

The report should separate measured facts, technical limitations, clinical interpretation, and recommendations. An index beyond a threshold does not automatically explain the entire cause or severity; symptoms, comorbidity, stage and position distribution, hypoxic burden, and study quality still matter. The workstation may supply prompts, phrase libraries, and decision support, but it should not insert a diagnosis when required information is absent, nor should template recommendations replace the responsible physician's review of source data and clinical context.

Before signing, the authorized interpreter should check patient and study identity, procedure type, source score version, key statistics, technical comments, graphics, conclusion, and recipients. Technologist scoring, automated analysis, physician interpretation, and administrative release should keep distinct authorship. Collaborative editing needs locks, conflict handling, handoff, and review. The last user to save must not silently overwrite completed review or attribute another person's draft text to the signer without presenting the full content for attestation.

An electronic signature should bind an identifiable signer, time, intent, and exact report version. A pasted signature image or typed name supplies appearance but does not by itself prove authentication, integrity, or non-repudiation. A digital signature can add cryptographic evidence of integrity, although its validity also depends on certificates, keys, time, and the verification environment. Any post-signature content change should visibly invalidate the prior signature state or create a newly signed version rather than preserving the old signature appearance on modified content.

Revision, distribution, and privacy protection

After a final report is released, new information may be appended, an error may require a corrected version, and a wrong-patient or invalid report may need controlled withdrawal. Every state should retain the prior report, editor, time, reason, changed content, and relationship between versions. Correcting critical identity, diagnosis, severity, or treatment advice cannot stop at replacing a PDF on the server. The system should identify copies already sent to the health record, referrer, patient portal, print queue, or external interface and support policy-based notification.

Before distribution, the recipient, purpose, minimum necessary content, route, and attachment should be checked. A patient portal, secure message, health-record interface, print, fax, and removable medium each has different confirmation and failure behavior. Interface success may only mean that a server received a message, not that it entered the correct patient chart; retries can also create duplicates. The workstation should record the version sent, destination, time, status, and necessary acknowledgment so drafts and withdrawn versions do not continue circulating.

A report contains diagnostic conclusions and may embed whole-night trends, event snapshots, and sensitive behavior descriptions. Role permissions, individual accounts, automatic locking, audit logs, transmission protection, print collection, and remote support therefore require governance. Copies used for quality improvement, teaching, or software testing should be minimized and de-identified for the purpose, but rare events, dates, images, and free text can still re-identify a patient. Retention, patient access, correction requests, and secure disposal should follow applicable law and institutional policy.

Acceptance, change, and quality management

Commissioning should use representative cases with known answers, including normal and abnormal findings, poor signals, missing channels, split-night studies, treatment titration, adults and any other supported population, multiple revisions, and namesake patients. Source scoring should be compared field by field with identity, durations, event counts, indices, units, denominators, plots, and narrative in the report. PDF, print, structured interface, and electronic-record display should agree. Draft watermarks, signature permissions, post-signature edits, withdrawal, resend, interface failure, and audit retrieval also need testing.

Templates, formulae, scoring rules, reference language, interface mappings, and software upgrades can all change clinical output and should be version and change controlled. A baseline test set and rollback conditions should exist before an update; after it, critical results should be compared and old reports should remain readable and verifiable. Visual browsing alone will miss rounding, unit, denominator, and conditional-logic errors. Bulk migration should sample different eras and study types and verify that attachments, signatures, version chains, and prior delivery states survive.

Continuing quality management can monitor identity corrections, post-signature revisions, wrong recipients, duplicate sends, calculation discrepancies, interface backlogs, unsigned reports, template defects, and user near misses. Investigation should trace a final document back to its source data, score-set version, calculation rules, template, operators, and delivery record. A reliable reporting workstation is not a marketing claim of automatic reports; it is a controlled clinical system that keeps results correct, interpretation accountable, signatures trustworthy, revisions transparent, and delivery directed to the right recipient.