The Hidden Layer

The European push toward structured health information has produced, over the past five years, a remarkable transformation of how clinical data is exchanged. National programmes from the United States to the Nordic countries have adopted HL7 FHIR as the de facto framework for interoperability, and the European Health Data Space is now framing the supranational regulatory context in which Member States must align their digital infrastructures. Italy is part of this trajectory: Fascicolo Sanitario Elettronico 2.0, the National Telemedicine Platform, and the forthcoming Health Data Ecosystem are all steps along a clear direction of travel.

Mental health has not been absent from this evolution. Brief screening instruments — the Patient Health Questionnaire (PHQ-9), the Generalized Anxiety Disorder Scale (GAD-7), the Mini-Mental State Examination — are by now adequately represented in the major terminologies. LOINC provides pre-coordinated codes for the PHQ-9 panel and its individual items; SNOMED CT mirrors the procedural concepts; the United States Behavioral Health Profiles Implementation Guide and the parallel work of the Gravity Project on social determinants of health are gradually articulating FHIR-native representations of these data.

And yet, for the clinician who actually produces a specialist psychological report, the picture is starkly different. The diagnostic instruments that constitute the core of clinical, neuropsychological, and forensic psychology — the Wechsler intelligence scales (WAIS-IV / WAIS-V, WISC-V), the Minnesota Multiphasic Personality Inventory in its current third edition (MMPI-3), the Rorschach Performance Assessment System (R-PAS), to mention only the most established — are either absent from the major terminologies or represented only at the most generic procedural level. The result is a paradox that deserves more attention than it currently receives: the portion of the psychological record on which significant clinical, educational, and forensic decisions are taken is, paradoxically, the portion least readable by the wider electronic health record.

A gap that is not merely terminological

It is tempting to read this gap as a temporary lag in terminology development, the kind of thing that gets resolved with the next LOINC release. That reading is too optimistic. The seminal coverage analysis published by Ranallo and colleagues in the AMIA proceedings more than a decade ago already identified, with surgical precision, the structural absence of standardized psychological assessment instruments from SNOMED CT, LOINC, and the QS terminology. Twelve years on, the architectural side of the problem has been transformed by FHIR — but the terminological side has barely moved.

The persistence of this gap reflects a convergence of three different orders of constraint, none of which can be dissolved by goodwill alone.

The first is legal. Modern psychometric instruments are not in the public domain. They are commercial products distributed by a small number of publishers — Pearson, Psychological Assessment Resources, Hogrefe, Multi-Health Systems — under restrictive licences that prohibit the reproduction of items and, in some cases, of normative tables. A pre-coordinated LOINC panel for the MMPI-3 analogous to the PHQ-9 panel, if it published item text, would infringe those rights. The PHQ-9 is freely usable because Pfizer placed it in the public domain in 2010; the MMPI-3 is not, and its publisher has no incentive to follow suit. This is not a defect of the standards bodies; it is a feature of the market within which professional psychometric instruments are produced and distributed.

The second is deontological. The professional codes of the American Psychological Association and the equivalent European associations treat item content as protected information whose disclosure threatens the construct validity of the test for future examinees. A standardized item-level representation, even stripped of normative tables, materially increases the risk of item exposure if widely deployed. This is not paranoia: a single MMPI item that becomes well known in a litigation community can compromise the test’s diagnostic value for years. The Italian Codice Deontologico degli Psicologi reflects the same concern.

The third is methodological. Profile-based instruments are not interpreted additively. An elevation on MMPI-3 Restructured Clinical scale 2 is read in the context of the full validity scale configuration, the higher-order scales, and the demographic profile of the examinee. A Rorschach Complexity score is interpreted against the Card Pull and the entire structural summary. A Wechsler Full Scale IQ that masks a thirty-point discrepancy between the Verbal Comprehension Index and the Processing Speed Index is a different clinical document than the same FSIQ resting on a homogeneous cognitive profile. A representation that captures items or even individual scale scores without preserving their configurational relationships loses much of what makes the assessment a clinical document in the first place.

The gap is not a technical accident waiting to be repaired. It is the residue of three distinct logics — proprietary, deontological, and psychometric — that the standards machinery has not yet learned to accommodate.

Designing around the constraints, not against them

If the three constraints are accepted as features rather than obstacles, the design move that follows is not to push standardization downward toward the item, but upward toward the unit of analysis that already carries clinical meaning. The principle is the same that medicine has long applied to laboratory data: the standardized observable is the analyte value, not the reagent formula. Reporting a sodium of 142 mmol/L does not require disclosing how the ion-selective electrode was calibrated. The score is data about the patient; the test is intellectual property.

An HL7 FHIR Implementation Guide for structured psychological assessment can be organized along four design principles that take this premise seriously.

Scale-level Observations as the default unit of structuring. Each named scale of an instrument is represented as a FHIR Observation profile carrying the raw score, the standardized score (T-score, scaled score, or percentile), and the interpretive band. The binding is to LOINC where a code exists, and to an instrument-specific code system maintained by the publisher or by the standards community where it does not. This is the layer where the IG can deliver immediate clinical value without touching item content.

Profile-level DiagnosticReport as the wrapper that preserves configurational meaning. A DiagnosticReport profile for each instrument groups the scale Observations and exposes the configuration through coded interpretive concepts: MMPI-3 validity pattern, Wechsler index discrepancy, R-PAS complexity range. This permits downstream systems — case management software, longitudinal patient summaries, decision support modules — to retrieve the profile as a clinical whole rather than as an unstructured collection of numbers. The configural logic of psychometric interpretation is preserved at the resource level.

A gated, optional item-level layer using QuestionnaireResponse. For instruments and jurisdictions in which item-level capture is licensed — typically through institutional digital scoring agreements with the publisher — the IG defines a QuestionnaireResponse profile that references a private Questionnaire resource held inside the licensed scoring system. Only the canonical URL and the version of the Questionnaire are exchanged; the item text never crosses the licensing perimeter. This separates data interchange from item distribution and provides a controlled pathway for those use cases — institutional research, multi-centre studies, regulated digital scoring platforms — in which item-level granularity is both lawful and operationally appropriate.

PlanDefinition for administration protocols. Psychometric validity depends on standardized administration. A PlanDefinition resource encodes the protocol: administration order, time limits, examiner qualifications, environmental conditions. The IG thus carries not only the result but the procedural context required for its interpretation — a non-trivial requirement for forensic and educational uses, where the conditions under which a test was administered can determine whether its results are admissible at all.

What this looks like in practice

Consider how an MMPI-3 protocol could move through such an architecture. The examinee completes the administration inside the publisher’s licensed digital scoring platform; the platform scores the protocol and produces the 52 standardized scale values, along with the validity pattern interpretation. From that point onward, the data crosses into the FHIR layer: 52 Observation resources, each bound to an MMPI-3 scale code, are issued; a DiagnosticReport aggregates them and exposes a coded validity-pattern concept; a QuestionnaireResponse, if institutionally licensed, references the canonical URL of the MMPI-3 v1.0 Questionnaire without ever transmitting the item text. A PlanDefinition records that the protocol was administered in supervised mode, in two hours, by a qualified clinician.

From the receiving system’s point of view, the protocol is fully structured, fully interoperable, and fully respectful of the publisher’s licence. The clinician opening the chart sees the validity pattern; the case management system can plot the elevations longitudinally; the research database can extract scale-level data for cohort studies without ever accessing item content. The architecture has not solved the copyright problem by ignoring it; it has solved it by separating the data interchange from the content distribution.

The same architecture extends, with minor adaptations, to the Wechsler scales — where the structural Observations are the subtest scaled scores, the index composites, and the Full Scale IQ — and to the Rorschach R-PAS, where Complexity, Synthesis, Form Quality, and the Ego Impairment Index are the structural Observations and item-level layer is not applicable, the instrument being performance-based rather than questionnaire-based.

The Italian context: FSE 2.0 as a national instantiation

The architecture sketched above is, deliberately, jurisdiction-agnostic. But it is worth asking what its adoption would mean for the Italian Fascicolo Sanitario Elettronico 2.0, both because Italy is the immediate context for any practitioner reading this and because the FSE 2.0 architecture offers a particularly clean test case.

The current FSE production environment relies predominantly on HL7 Clinical Document Architecture R2 for document exchange. FHIR is adopted natively within the National Telemedicine Platform and is indicated as the target evolutionary standard for the forthcoming Health Data Ecosystem (Ecosistema Dati Sanitari). Mental health services within the Servizio Sanitario Nazionale — Dipartimenti di Salute Mentale, consultori familiari, neuropsichiatria infantile units, prison psychology services, district psychology services — produce a substantial volume of structured psychometric data. At present, that data is either preserved as PDF documents inside the FSE Taccuino, or held only inside departmental information systems that do not communicate with one another, let alone with the national infrastructure.

For a regional initiative such as the Calabrian Parliamone insieme programme — a 24/7 sociopsychological support service being rolled out across the five provincial health authorities under the joint governance of the regional Welfare Department and the local ASPs — the difference between document-based and scale-level structuring is not academic. A PDF report stored in the Taccuino cannot be queried, cannot be aggregated, cannot feed a regional dashboard, and cannot support longitudinal follow-up across the network. A scale-level DiagnosticReport, exposed to the FSE through the IG profiles, can do all of this without any leakage of item content and without compromising the licensing agreements that the regional service holds with the test publishers it uses.

Equivalent considerations apply, with national variations, to the German Medizininformatik-Initiative core data set, to the French Mon Espace Santé, and to the Nordic FHIR implementations. The Italian case is illustrative rather than unique.

Three implications and an honest acknowledgement of limits

If the gap is real and the design move is sound, three sets of actors are in a position to do something about it.

For terminology developers, the priority is the systematic coding of the named scales of established instruments as LOINC observations, with bindings to SNOMED CT procedural concepts. This is feasible without item disclosure, and it parallels the way laboratory analytes are coded independently of the proprietary assay platforms that produce them. A coordinated request to the LOINC committee from a multi-stakeholder group — clinicians, publishers, standards bodies — would be the appropriate vehicle. The work is unglamorous but tractable.

For test publishers, the IG offers a route to participate in standardization without ceding control of item content. By holding the canonical Questionnaire resource within their licensed digital scoring platforms and exposing only scale-level Observations, publishers preserve their intellectual property while making their instruments interoperable. This is not a concession; it is an alignment of commercial interest with the public health imperative for structured psychological data. The publisher whose instrument can be cleanly integrated into a national EHR is the publisher whose instrument national procurement will prefer.

For national programmes — FSE 2.0, EHDS, and the equivalent infrastructures across the EU — the IG provides a template for incorporating specialist psychological assessment into the structured record without waiting for full terminological coverage of every instrument. Scale-level coding can begin with the most widely used instruments — Wechsler, MMPI-3, R-PAS, BVN, ENB, MOCA, and a small set of others — and expand incrementally. This consideration is particularly relevant for regional welfare and mental health initiatives that operate at the intersection of social and health services, where the absence of structured psychological data fragments the longitudinal view of the person across the boundary of the two sectors.

Several limitations of the approach should be acknowledged. The IG sketched here has not been implemented or balloted through HL7; what is articulated above is a set of design principles, not a conformance package. The treatment of performance-based instruments such as the Rorschach raises further questions about how to encode the response process itself — the location and determinant coding of individual percepts — that are not fully resolved by the scale-level approach. The legal landscape varies across jurisdictions in ways that may require national adaptations: what is permissible under the Italian Codice Deontologico is not necessarily permissible under the German Berufsordnung or the British Psychological Society code. These are not objections to the framing; they are next steps in a programme of work.

A modest proposal, an immodest stake

It would be a mistake to read the proposal above as a niche concern of the standards community. The stake is much broader. Specialist psychological assessment is the layer at which significant individual decisions are taken — about treatment plans, about educational placements, about custody and capacity, about disability eligibility, about therapeutic suitability. The fact that this layer is, today, structurally invisible to the electronic health record means that those decisions are taken on the basis of documents that cannot be aggregated, audited at scale, or longitudinally followed. It is, quite literally, a quality-of-care problem.

It is also, increasingly, a regulatory problem. The European Health Data Space will require Member States to expose structured clinical data for primary and secondary use. National infrastructures will have to declare what they expose and what they do not. A national health service that exposes laboratory data and imaging reports to the EHDS but cannot expose its psychological assessments is not exposing the patient; it is exposing an incomplete person.

The technical solution does not require a revolution. It requires a coordination — between standards bodies that have not yet coded the scales, publishers that have not yet been asked to expose their canonical Questionnaires, national programmes that have not yet specified the IG, and clinicians who have not yet pushed for it. The architecture is available. The constraints are real but workable. What is missing is the multi-stakeholder conversation that turns design principles into a balloted Implementation Guide.

That conversation deserves to start.