Ask a behavioral health program running PROMIS across five clinics whether the short-form or the computerized adaptive test (CAT) version is the better fit, and the honest answer is that both work, but the operational picture underneath them is very different. Fixed short forms like PROMIS-29 or the PROMIS Depression 8a are a static set of items every patient sees. CAT selects the next item based on the previous answer, drawing from an item bank against a calibrated Item Response Theory model. In a multi-site FHIR deployment, that choice reshapes the Questionnaire runtime, the extraction pipeline, and the interoperability story across sites.
For FHIR resources for behavioral health teams, the broader desk covers the surrounding stack. This piece narrows to the fixed-vs-CAT split as it plays out across a multi-site behavioral health program on FHIR.
What Fixed Short Forms Buy You Operationally
The short form is a canonical FHIR Questionnaire. It has a fixed itemset, a fixed order, and a static scoring rule. Every site renders the same instrument. The QuestionnaireResponse is comparable across sites without qualification, because the item set is identical.
That predictability shows up in three concrete places. Extraction is straightforward: the item linkIds are fixed, the scoring rule collapses to a sum, and the Observation extraction is deterministic. Auditing is straightforward: a compliance review can walk a randomly sampled QuestionnaireResponse and reason about it without pulling the item bank. Multi-site normalization is straightforward: the T-score conversion table is the same for every response.
SDC's ExtractDefinition is the standard mechanism, and whether the runtime is built in-house or lives inside an engine like Formbox, the extraction contract stays the same across sites. That contract stability is what a multi-site program leans on when it needs to run a cross-site trend report and be sure every site is contributing comparable data.
What CAT Adds and What It Costs
CAT delivers a shorter test with a comparable precision estimate. In practice, a PROMIS CAT visit often runs four to eight items instead of the fixed eight or twenty-nine, and the T-score standard error is meaningful. Patients complete faster, adherence tends to hold up better, and the burden on longitudinal PRO programs drops noticeably.
The cost is that the runtime is doing much more work. The next item is selected by an IRT engine against a live item bank, so the FHIR Questionnaire is no longer a static definition; it is a shell that names a bank, and the actual items are chosen at run time. Extraction has to record which items were administered, in which order, with which theta estimates, before a downstream service can even begin to trust the T-score.
The Multi-Site Interoperability Angle
In a multi-site FHIR program, the interoperability question is whether a QuestionnaireResponse from Site A is meaningfully comparable to one from Site B. For fixed short forms, the answer is yes without qualification. For CAT, the answer is yes if both sites are running the same item bank, at the same calibration release, with the same stopping rules. The teams that make this work put the CAT engine behind a shared service so every site sees the same behavior. Teams building on native FHIR often reach for engines like Formbox at this layer, because the item-selection rules and the extraction into Observation resources sit in the same runtime, which spares the program a second scoring service on every site.
If you want to walk through this before touching a production runtime, form-builder.aidbox.app runs standard FHIR Questionnaire JSON in the browser.
Where the Choice Actually Lands
Multi-site programs running an established fixed short-form catalog usually keep it. The comparability across sites is worth the extra items. Multi-site programs where patient burden is the binding constraint, particularly in oncology PROs or longitudinal chronic-care follow-up, tend to adopt CAT and centralize the runtime.
For teams still picking a form builder before this question is even in scope, Top 6 SDC form builders for behavioral health clinics in 2026 covers the shortlist. For screening-specific comparisons, Top 4 FHIR Questionnaire engines for PCL-5 and PTSD screening sits in the same silo. The right choice comes down to whether cross-site comparability or patient burden is the harder constraint in your program.
