Resources

SDOH screening instruments: a working comparison

The US ecosystem of SDOH screening tools is heterogeneous. Different instruments cover different domains, target different settings, and were validated on different populations. This is a side-by-side reference for the instruments most commonly encountered in emergency medicine, primary care, and community health settings, with notes on when each is the better operational fit.

For program leaders selecting an instrument, the right starting point is to write down what domains your program will respond to, then pick the shortest instrument that covers them. Domain coverage you cannot act on is documentation overhead with no clinical return.

AHC-HRSN

Accountable Health Communities Health-Related Social Needs
Steward
Centers for Medicare & Medicaid Services (CMS)
Items
10 core items, 16 supplemental items (26 total)
Domains covered
Housing instabilityFood insecurityTransportationUtility help needsInterpersonal safety
Setting
Patient-administered or clinician-administered. Designed for use across clinical and community settings.
Validation context
Developed and tested as part of the AHC model. Used widely in CMS-aligned reporting.
When to choose it
The de facto US baseline for hospital and ED-based screening. Shorter than PRAPARE; favored when integration into a structured EHR module is the goal.

PRAPARE

Protocol for Responding to and Assessing Patients' Assets, Risks, and Experiences
Steward
National Association of Community Health Centers (NACHC)
Items
21 items across 16 core measures plus 5 optional measures
Domains covered
Personal characteristicsFamily and homeMoney and resourcesSocial and emotional healthOptional: incarceration, refugee status, safety, domestic violence
Setting
Designed for federally qualified health centers and primary care; adopted in some EDs.
Validation context
Field-tested across hundreds of community health centers. Has standardized FHIR-based mappings.
When to choose it
Stronger domain coverage than AHC-HRSN; longer, more burdensome to administer in a single ED encounter.

AHC core 10

Accountable Health Communities core 10
Steward
CMS (subset of AHC-HRSN)
Items
10 items
Domains covered
HousingFoodTransportationUtilitiesSafety
Setting
ED-friendly short form. Patient self-administered on a tablet or kiosk in under three minutes.
Validation context
The same psychometric base as AHC-HRSN, applied to the core domains only.
When to choose it
Most ED programs that want a baseline screen settle here for time reasons. Loses the supplemental domain coverage but is operationally tractable.

HVS

Hunger Vital Sign
Steward
Children's HealthWatch
Items
2 items
Domains covered
Food insecurity (only)
Setting
Two-question screen, often deployed alongside broader instruments to catch food insecurity rapidly.
Validation context
Validated against the USDA 18-item Household Food Security Survey Module with high sensitivity and specificity in pediatric and adult populations.
When to choose it
Single-domain screen; not a substitute for a multi-domain tool but a useful add-on for high-throughput settings.

WE CARE

WE CARE
Steward
Boston Medical Center / academic groups
Items
Variable, typically 7 to 10 domain-specific items
Domains covered
FoodHousingChildcareEducationEmploymentHeatOther psychosocial
Setting
Pediatric primary care; some emergency department implementations.
Validation context
Multiple studies of feasibility and downstream referral effects in pediatric populations.
When to choose it
Strong evidence base in pediatrics. Less common in adult ED settings.

Reporting an instrument's performance, in your population

Published validation data is necessary but not sufficient. Every instrument behaves differently in a different population. A program that wants to know whether the screen it is running is performing as expected should report sensitivity, specificity, positive predictive value, and negative predictive value against a reference standard within its own population.

For multi-instrument comparisons, percent-positive rates differ across instruments even on the same patient sample. Inter-instrument agreement should be reported with Passing-Bablok or Deming regression for continuous outputs, or with confusion matrices and Cohen's kappa for categorical outputs.

Subgroup analysis is the part most published implementation papers skip. Completion rates, positivity rates, and operating characteristics frequently differ across demographic subgroups in ways that change which instrument is right for which population. The diagnostic sample size question for these subgroup analyses is non-trivial; sample sizes that are adequate for the overall mean are routinely underpowered for the subgroup.

Match the instrument to the response

The most common implementation failure mode is screening for needs the program cannot respond to. If your program has no transportation referral pathway, screening for transportation insecurity creates documentation burden and erodes trust without producing a connected service. The right discipline is to start with the response side, then choose the screening instrument whose domains match.

For programs partnering with a home-based care vendor, the response-side checklist is concrete: which Z-codes can the vendor act on, what is the dispatch latency, and how does the outcome flow back to the ordering EHR. The closed-loop referral anatomy reference covers the architecture.

Selecting an instrument for your program?

The conversation should start with what your program will respond to. We help with both sides.