🏘 Social Risk Screening Clinical Workflow Integration

Bringing Social Risk Screening Into the Visit Itself

For decades, social determinants of health (SDOH) — housing, food, transportation, utilities, safety — were treated as background context clinicians might informally ask about. CMS's Accountable Health Communities (AHC) Health-Related Social Needs (HRSN) screening tool and the PRAPARE (Protocol for Responding to and Assessing Patients' Assets, Risks, and Experiences) instrument turned that informal awareness into a structured, billable, EHR-native clinical workflow step — administered at check-in, alongside vitals and intake forms, for every eligible patient.

• 5: AHC-HRSN core domains (housing, food, transport, utilities, safety)
• 21: PRAPARE item count (core + optional supplemental modules)
• 28: CMS AHC model sites (2017-22) (bridge organizations, 4M+ screened)
• 4-7 min: Typical completion time (tablet or paper at check-in)

PRAPARE and AHC-HRSN — two validated instruments, one goal

PRAPARE, developed by NACHC and partners for Federally Qualified Health Centers, uses 21 core questions spanning personal characteristics, family and home, money and resources, and social and emotional health — plus optional modules on safety, incarceration history, and refugee status.

The CMS AHC-HRSN tool, built for the 2017 Accountable Health Communities Model, is shorter (10 core items) and explicitly maps to 5 domains CMS considers actionable within a clinical encounter: housing instability, food insecurity, transportation problems, utility help needs, and interpersonal safety (IPV screening via HITS or similar).

Both instruments are validated, forward-scored (higher score = higher risk), and designed to be self-administered on a tablet or completed with staff assistance — critical for workflow speed, since screening that takes 20 minutes never survives contact with a busy primary care visit.

Where screening lives in the visit workflow

Successful implementations embed screening at one of three points, each with tradeoffs:

• Pre-visit (patient portal, 24-48h before): highest completion when patients are engaged online, but excludes low-digital-literacy populations and misses walk-ins • Check-in (front desk / MA rooming): captures nearly all patients physically present, but adds 5-7 minutes to intake and requires staff workflow redesign • During the visit (clinician-administered): highest clinical context and immediate discussion, but competes directly with the 15-minute visit clock and is the least scalable

Most mature programs use a hybrid: portal pre-screen when possible, MA-administered tablet at check-in as the safety net, ensuring near-universal coverage regardless of patient digital access.

Screening cadence and population targeting

Screening every patient at every visit is rarely feasible or necessary. Most health systems tier cadence:

• Annual screening for general primary care populations, tied to wellness visits • Every-visit screening for high-risk panels: Medicaid, dual-eligible, pediatric well-child, prenatal, and chronic disease management populations • Event-triggered re-screening after a hospitalization, ED visit, or major life change (job loss, eviction filing, new diagnosis)

The screening frequency slider in this simulation reflects this tradeoff directly: annual-only screening minimizes staff burden but misses rapidly-changing needs like eviction or a shut-off utility notice; every-encounter screening catches more but multiplies front-desk workload and screening fatigue.

From Raw Answers to Actionable Domain Scores

A completed questionnaire is not yet clinically useful — it is 10-26 individual answers that must be algorithmically scored into discrete, independently-actionable risk domains. Domain-level scoring, rather than a single composite score, is what allows a positive housing flag to trigger a housing referral without waiting on food or transportation questions, and is central to how AHC-HRSN and PRAPARE were designed to plug into existing care management workflows.

• 5: Core scored domains (food · housing · transport · utilities · safety)
• 25-40%: Typical positive-screen rate (at least one domain, safety-net settings)
• ~1 in 3: Multi-domain positive patients (of all positive screens)
• 15-20%: Housing instability prevalence (most common single domain nationally)

How each domain is scored

Each domain uses its own validated sub-instrument, not a generic Likert average:

• Food insecurity: the 2-item Hunger Vital Sign ("worried food would run out" / "food didn't last") — either affirmative response is a positive screen, matching USDA's validated food security screener • Housing instability: living situation today (owns/rents/temporary/shelter/street) plus a forward-looking worry item about losing housing in the next 2 months • Transportation: lack of reliable transportation that has caused a missed medical appointment, work, or other necessary activity in the past 12 months • Utilities: threat of shutoff for electric, gas, heating oil, or water in the past 12 months • Interpersonal safety: adapted HITS (Hurt, Insulted, Threatened, Screamed) intimate partner violence screen, scored separately given its distinct clinical and legal handling requirements

Domain independence matters clinically: a patient can be housing-stable but food-insecure, or vice versa, and each triggers a different downstream referral pathway.

Multi-domain positivity compounds risk disproportionately: patients positive on 2+ domains show significantly higher ED utilization and worse chronic disease control than the sum of each domain's individual effect — which is why most risk-stratification models weight domain count, not just presence.

Severity tiers and clinical urgency

Beyond binary positive/negative, mature implementations assign severity tiers that drive triage speed:

• Tier 1 (urgent/safety): active IPV disclosure, imminent eviction (<7 days), utility shutoff notice already issued — same-day warm handoff to a social worker or crisis line • Tier 2 (high need): housing instability without imminent loss, chronic food insecurity, transportation barrier affecting medication adherence — care coordinator outreach within 1-3 business days • Tier 3 (moderate need): single lower-severity domain, patient declines immediate assistance — resource information provided, added to routine care coordinator queue

This tiering is what keeps referral queues from becoming an undifferentiated backlog — the same failure mode that undermines paper-based social work referrals when volume increases with universal screening.

Structured Data, Z-Codes, and the Clinical Decision Support Trigger

A positive screen that lives only on a scanned PDF changes nothing. The workflow-defining step is writing results back into the EHR as discrete, queryable data — populating the problem list with ICD-10 Z-codes for social determinants (the Z55-Z65 range), firing a best-practice advisory (BPA) visible to the clinician in the same encounter, and making the finding available to population health dashboards and value-based care risk models.

• Z55-Z65: ICD-10 Z-code range used (social/environmental circumstances)
• <2%: Z-code capture rate (national) (of eligible encounters, pre-2020 baseline)
• 2023+: CMS Z-code reporting incentive (quality measure & HRSN adjustment pilots)
• 55-70%: BPA fire-to-action rate (when embedded in rooming workflow)

Z-codes: the coding layer that makes social risk visible

ICD-10-CM Z-codes Z55 through Z65 ('Persons with potential health hazards related to socioeconomic and psychosocial circumstances') are the coding mechanism that lets a positive screen travel beyond the visit note:

• Z59.0-Z59.9: homelessness, housing instability, inadequate housing, discord with neighbors • Z59.4-Z59.48: food insecurity, lack of adequate food • Z59.6: low income • Z59.7: insufficient social insurance/welfare support • Z75.3: unavailability/inaccessibility of health care facilities (proxy for transport barriers) • Z65.x: other problems related to psychosocial circumstances

Historically these codes have been billed at under 2% of encounters nationally despite screening positivity rates of 25-40% — a documentation gap CMS is now actively closing by tying Z-code capture to value-based payment models and, in several state Medicaid programs, direct HRSN service reimbursement.

Best-practice advisories and structured discrete fields

Workflow design determines whether a positive screen becomes clinical action or invisible data:

• Discrete field capture: each domain result stored as a structured flowsheet row or SmartData element (not free text), enabling registries, dashboards, and automated referral triggers to query it directly • Best-practice advisory (BPA): a pop-up or banner surfaced to the clinician during the same encounter, summarizing positive domains and offering one-click order-set access to a social work consult or referral order • Problem list integration: chronic/persistent positive domains (e.g., ongoing housing instability) added to the active problem list so every future encounter has visibility, not just the screening visit • Registry population: positive screens feed population health registries used for outreach campaigns, risk stratification, and HEDIS/quality measure reporting

Alert fatigue is the central design risk here — a BPA that fires on every visit for a chronic, already-being-addressed need trains clinicians to dismiss it reflexively, which is why most systems suppress re-firing within a defined follow-up window.

Systems that route the BPA to a care coordinator's queue instead of (or in addition to) the physician see 2-3x higher referral completion, because the physician's 15-minute visit rarely has room to also place and track a community referral — the trigger needs a workflow owner, not just a data flag.

Routing Positive Screens to Community-Based Organizations

A positive food-insecurity flag is clinically meaningless without a functioning pathway to an actual food pantry, SNAP enrollment assistance, or medically-tailored meal program. Closed-loop referral platforms — Unite Us, findhelp (formerly Aunt Bertha), NowPow, and health-system-built equivalents — sit between the EHR and a curated directory of vetted community-based organizations (CBOs), turning a clinical flag into a trackable referral with a defined resource, contact method, and expected response window.

• 3-4: Closed-loop referral platforms (major) (Unite Us, findhelp, NowPow, Aunt Bertha)
• 500k+: CBO directory listings (large platforms) (community resources nationally)
• 50-75%: Referral acceptance rate by CBO (varies sharply by resource capacity)
• 2-3x: Warm handoff vs. self-directed uptake (higher completion for warm handoffs)

Referral pathway types, from lightest to heaviest touch

Not every positive domain needs the same referral intensity — matching intensity to need and available staffing is a core workflow design decision:

• Self-directed resource list: patient receives a printed or texted list of nearby resources with no active follow-up — lowest cost, lowest completion rate (often <20%) • Electronic referral (e-referral): a structured referral is sent through the platform to a specific CBO, which can accept, decline, or request more information — trackable but not personally facilitated • Warm handoff: a care coordinator personally calls or walks the patient through contacting the CBO, sometimes during the same visit — highest completion, highest staff time cost • Co-located/embedded services: CBO staff (e.g., a food bank navigator or legal aid partner) physically present in clinic on scheduled days for immediate connection

The care-coordinator staffing slider in this simulation reflects exactly this tradeoff: richer staffing ratios shift more referrals from self-directed toward warm handoff, which is the single strongest lever on downstream closed-loop completion.

CBO capacity — the referral platform's hidden constraint

Closed-loop referral technology solves the routing and tracking problem, but not the underlying capacity problem: many CBOs are underfunded relative to referral volume, especially for housing and utility assistance where waitlists commonly run weeks to months. A well-designed referral system surfaces CBO capacity and typical response time transparently so care coordinators and patients can set realistic expectations, rather than referring into a black box.

Some health systems address this by contracting directly with high-volume CBOs (flexible housing funds, food-as-medicine programs) to guarantee capacity for their referred patients, effectively purchasing throughput rather than relying purely on best-effort community partnership.

Data sharing and consent across the referral boundary

Referral platforms must navigate consent and data-sharing rules that differ from standard clinical HIE (health information exchange), since CBOs are typically not covered entities under HIPAA. Best-practice implementations:

• Obtain explicit patient consent before sharing PHI with a specific named CBO, not a blanket 'may share with any partner' consent • Limit shared data to the minimum necessary — domain and contact info, not full chart access • Use platform-level business associate agreements (BAAs) where the CBO does handle PHI • Allow patients to decline the referral itself while still receiving general resource information

Closing the Loop — Verifying That the Social Need Was Actually Resolved

The single biggest gap between social risk screening's promise and its measured impact is the "open loop": referrals sent but never confirmed as resolved, declined, or failed. Closed-loop tracking means the referring clinical team receives a structured status update back from the CBO or care coordinator follow-up call — resolved, in-progress, declined, or unreachable — and that status feeds back into the patient record and program-level outcome metrics, not just the referral count.

• 40-60%: Referrals with any status update (even on "closed-loop" platforms)
• 35-55%: Need actually resolved (of tracked) (varies widely by domain)
• 30/60/90d: Standard follow-up cadence (coordinator check-in schedule)
• ~6-12mo: Re-screen interval after resolution (or next eligible visit)

What closing the loop actually requires operationally

True closed-loop tracking requires three linked capabilities most legacy referral processes lack:

• Bidirectional data flow: the CBO (or the platform on its behalf) sends a status update back to the referring EHR, not just a one-way referral fax • A defined follow-up owner: a care coordinator (or an automated outreach system — text/IVR check-in) responsible for contacting the patient at set intervals if no CBO status arrives • Outcome documentation standards: consistent categories (resolved / in progress / declined by patient / CBO unable to serve / lost to follow-up) so program-level reporting is comparable across domains and time

Without all three, "referral rate" becomes a vanity metric — a number that looks like program success while telling you nothing about whether anyone's food or housing situation actually changed.

The Accountable Health Communities Model's national evaluation found that navigation-supported closed-loop referrals achieved roughly double the resolution rate of self-directed referrals — the strongest evidence to date that staffing the follow-up step, not just the initial referral, is what converts screening into outcomes.

Re-screening and the longitudinal risk record

Social risk is not static — a resolved housing crisis can recur, and a stable food-secure household can become insecure after a job loss. Mature programs treat closed-loop resolution as the start of a re-screening clock rather than a terminal state:

• Confirmed-resolved domains are re-screened at the next annual wellness visit or a defined interval (commonly 6-12 months) • Unresolved or declined domains remain flagged and are surfaced again sooner, often at the very next encounter • Longitudinal domain history (not just current status) is retained so care teams can see recurrence patterns — e.g., a patient with three prior housing-instability episodes warrants a different intervention than a first-time flag

Measuring program impact beyond referral counts

The metrics that matter for program evaluation and payer reporting go well past "number of screens completed":

• Screening completion rate: eligible encounters actually screened, stratified by patient population • Positive screen rate by domain: tracks community need prevalence over time and geography • Referral rate: share of positive screens that generate an actual referral (not just a resource list handout) • Closed-loop completion rate: share of referrals with a confirmed final status, resolved or not • Resolution rate: share of tracked referrals confirmed resolved — the outcome measure payers increasingly want tied to HRSN-adjusted payment models

Health systems reporting HRSN Z-code data to CMS and state Medicaid programs are increasingly required to report on this full pipeline, not just the screening step, as HRSN services move from voluntary quality improvement toward reimbursable, auditable care.