# SDOH Assessments in FHIR

## Design goals (functional requirements)

From recent discussion on [Jan 2022 US Core ballot content](http://hl7.org/fhir/us/core/2022Jan/) for SDOH assessments, it's important to reach agreement on functional requirements. That is, based on ONC's objectives, we need to decide whether the following are requirements, or nice-to-haves, or out of scope entirely.

##### Scope Questions
1. Should every US Core system be **able to represent multi-question SDOH surveys** (e.g., instruments like PRAPARE)? I'll take it for granted that no *specific* instruments are going to be required, so this question is just about the capability to represent multi-question surveys in general.

2. Should every US Core system be **able to represent "check all that apply" SDOH questions** (e.g., "*In the past year, have you or any family members you live with been unable to get any of the following when it was really needed? Check all that apply.*")

I think the answer to both questions should be "yes". This would require a slight expansion to the US Core ballot content as part of reconciliation, since the ballot content stops shy of (1).

Such an expansion could take several forms. The following proposal is intended to build **minimally and transparently** on the existing ballot profiles.

(See [FHIR-35364](https://jira.hl7.org/browse/FHIR-35364) for background.)

## Proposal Supporting survey instruments in US Core

### Proposal A: FHIR US Core should define two SDOH Observation profiles that implementers would be required to support**.

- [x] "SDOH Item" such as an individual question about food security, housing security, or emotional stress. This essentially already exists; the vocab bindings need slight refinement, but the basic requirements are aligned with `us-core-observation-screening-response` [from the current ballot](http://build.fhir.org/ig/HL7/US-Core/StructureDefinition-us-core-observation-screening-response.html). I'd suggest renaming this to `us-core-observation-sdoh-item`
- [ ] "SDOH Panel" that aggregate individual questions belonging to a single panel. This is would be a new profile like `us-core-observation-sdoh-panel` with the following MustSupport elements:
* `.category` with a fixed Coding that means "This is an SDOH Panel"
* `.effectiveTime` indicating when the survey was administered
* `.code` indicating the panel (e.g., https://loinc.org/93025-5 for PRAPARE)
* `.hasMember[]` with references to SDOH Items or other SDOH Panels collected as part of this SDOH panel

### Proposal B: FHIR US Core should focus on QuestionnaireResponse for SDOH Assessments

- SDOH assessments are backed by a FHIR Questionnaire where `Questionnaire.code` indicates `sdoh-assessment` (in addition to more specific codes like https://loinc.org/93025-5)
- SDOH assessments are represented by a FHIR QuestionnaireReponse pointing to the relevant Questionnaire
- Clients can search for SDOH assessments based on one of the following
- QuestionnaireResponse tags (e.g., `GET QuestionnaireResponse?_tag=sdoh`)
- *note, this can work even if the upstream Questionnaire is not explicitly written with US Core queries in mind*
- Questionnaire.code or .usageContext (`GET QuestionnaireResponse?questionnaire.questionnaire-code=sdoh-assessment`)
- *note, this would require upstream constraints on the Questionnaire, to ensure the US Core specified codes are present*
- ? (If needed) Clients can search for SDOH assessments by item (e.g., "Do you feel physicially and emotionally safe..." via `GET QuestionnaireResponse?questionnaire.item-code=https://loinc.org|93026-3`)

### Pros and cons

|Criterion|A|B|
|---|---|---|
|Keep related data together|:heavy_plus_sign:|:heavy_plus_sign::heavy_plus_sign:|
|Allow queries for "All SDOH" responses|:heavy_plus_sign:|:heavy_plus_sign:|
|Allow queries about specific items|:heavy_plus_sign::heavy_plus_sign:|:heavy_plus_sign:|
|Convey details about language, source, author|:heavy_minus_sign:|:heavy_plus_sign:|
|Communicate question text, all answer choices|:heavy_minus_sign:|:heavy_plus_sign:|
|Minimizes changes from Jan 2022 US Core ballot|:heavy_plus_sign:|:heavy_minus_sign:|
|Good basis for non-survey assessments|:question:|:question:|

### Proposal C: Represent SDOH responses in QuestionnaireResponses *and* Observation hierarchies -- i.e. Require (A) and (B)

This has some of the advantages of (A) and (B) but introduces substantial complexity. It might be better to consider this after something simpler is shown to work.