← Back to Playbook

OPS-SOP-030 — Gradient AI Onboarding Risk Screening

Document IDOPS-SOP-030 Version1.0 Effective Date[TBD — upon adoption] Next ReviewAnnually OwnerOperations / Underwriting
Prepared ByOperations

1. Purpose & Scope

This SOP governs the use of Gradient AI for pre-screening all new employer groups before quoting and onboarding into the Unity Care Member Plan Master Trust. Gradient AI provides predictive risk analytics including SAIL scores, morbidity indices, predicted annual claims, high-cost drug flags, and individual claimant projections. This SOP also covers annual re-screening for existing groups at renewal.

This SOP applies to: Operations, Underwriting, and any staff who submit data to or retrieve outputs from the Gradient AI platform. This SOP does not authorize the submission of PHI to Gradient AI until a Business Associate Agreement (BAA) is fully executed.

2. CRITICAL — BAA Requirement

STOP — HIPAA BAA NOT YET EXECUTED WITH GRADIENT AI

As of the writing of this SOP, a Business Associate Agreement (BAA) between Unity Care Solutions and Gradient AI has not been executed. No Protected Health Information (PHI) — including claims data, diagnosis codes, member identifiers, or any individually identifiable health information — may be submitted to Gradient AI until a fully executed BAA is in place.

Submitting PHI to a vendor without a BAA is a HIPAA violation that may trigger mandatory breach notification obligations. Operations must not proceed beyond the census/non-PHI step of the screening process until Legal Counsel (Dickinson Wright PLLC) confirms BAA execution.
TBD — Action Required: Legal Counsel to negotiate and execute BAA with Gradient AI. Operations Director to track BAA status in the BAA Registry (OPS-SOP-033). Do not submit PHI until BAA is confirmed executed and filed.

3. Definitions

TermDefinition
SAIL ScoreGradient AI's composite risk score. A score above 1.5 indicates elevated actuarial risk and requires Davies actuarial review before quoting.
Morbidity IndexA measure of the group's disease burden relative to a standard population. Higher index = higher predicted claims cost.
Predicted Annual ClaimsGradient AI's projected total claims spend for the group over the next 12 months, expressed as aggregate and PMPM.
High-Cost Drug FlagIndicator that one or more members are actively using a high-cost specialty medication (e.g., biologics, oncology drugs). Triggers stop-loss facultative review.
Individual Claimant ProjectionGradient AI's prediction of claims spend for individual high-risk members. Any member projected above $250K triggers HLRA notification.
Census (Non-PHI)Demographic data without individually identifiable health information: employee count, age bands, gender split, zip code, industry code. This data may be submitted before BAA execution.
PHIProtected Health Information — any individually identifiable health information. Requires executed BAA before submission to Gradient AI.
HLRAHigh-Level Reinsurance Agreement — Unity Care's reinsurance carrier arrangement via Jason Strain (Brown & Brown). $350K per-person attachment point.

4. Screening Process — New Employer Groups

Step 1Census Submission (Non-PHI Phase)
Operations receives the employer group census from the broker or employer HR contact. Before BAA execution, submit only non-PHI census data to Gradient AI: employee headcount, age bands, gender distribution, zip codes, and industry/SIC code. Do not include member names, SSNs, member IDs, diagnosis codes, or claims history until BAA is executed.
Step 2BAA Confirmation Check
Before submitting any PHI or claims data, Operations verifies that the Gradient AI BAA is fully executed by checking the BAA Registry (OPS-SOP-033). If BAA is not executed, pause at Step 1 and escalate to Legal Counsel for BAA completion. Document the pause in the group's onboarding file.
Step 3PHI/Claims Data Submission (Post-BAA Only)
Once BAA is confirmed executed, Operations submits the full risk screening dataset to Gradient AI per their data intake format. This may include: claims history (de-identified or identified per BAA terms), Rx utilization, diagnosis codes, and enrollment data. All submissions are logged with date, submitter, and data set description.
Step 4Retrieve Gradient AI Outputs
Operations retrieves the full Gradient AI risk report including: SAIL score, morbidity index, predicted annual claims (aggregate and PMPM), high-cost drug flags, and individual claimant projections. The report is downloaded and saved to the PHI-secure storage location (see Section 7).
Step 5Red Flag Triage
Operations reviews the Gradient AI output against the decision matrix in Section 5. Each red flag trigger is evaluated and routed to the appropriate party before quoting proceeds.
Step 6Decision & Quote Authorization
Once all red flags are resolved (actuarial sign-off, stop-loss facultative review, HLRA notification as applicable), Operations documents the final risk determination and authorizes the underwriting team to proceed with quoting. The Gradient AI report and all triage decisions are filed in the group's onboarding record.

5. Red Flag Decision Matrix

Trigger ConditionRequired ActionOwner
SAIL Score > 1.5Davies actuarial review required before quoting. Do not issue quote until Davies confirms risk acceptability and recommended rate adjustment.Operations → Davies Actuarial
High-cost specialty drug active (any member)Stop-loss facultative review required. Operations contacts Jason Strain (Brown & Brown / HLRA) to request facultative reinsurance coverage assessment for the specific drug/member.Operations → Jason Strain (Brown & Brown)
Individual claimant projected > $250,000HLRA notification required. Operations notifies Jason Strain of the projected high-cost claimant prior to enrollment. Attach point and specific reinsurance terms must be confirmed before the member is enrolled.Operations → Jason Strain (HLRA)
Morbidity index > 1.3 (elevated but below SAIL threshold)Flag for Operations Director review. Enhanced monitoring recommended for first plan year. Document decision rationale in onboarding file.Operations Director
Multiple red flags simultaneouslyEscalate to Leadership before proceeding. Document all flags and triage steps. Leadership makes final accept/decline decision.Operations Director → Leadership

6. Annual Re-Screening (Existing Groups)

All existing employer groups are re-screened through Gradient AI at each annual renewal. The re-screening process follows the same steps as new group screening (Section 4), using the current plan year's claims data.

Re-screening must be completed no later than 60 days before the renewal effective date to allow adequate time for rate adjustment, stop-loss negotiations, and actuarial review if triggered.

Renewal Re-Screening Outputs: The annual re-screen report is compared against the prior year's report to identify members with materially increased risk profiles. Any member whose individual projection increases by more than 50% year-over-year is flagged for individual review.

7. PHI-Secure Storage of Gradient AI Outputs

TBD — Action Required: Operations Director and IT/Legal to confirm the designated PHI-secure storage location for Gradient AI output reports. Storage must meet HIPAA Security Rule requirements (access controls, encryption at rest, audit logging). Until confirmed, store outputs in the most secure system currently available and restrict access to authorized Operations staff only.

Regardless of storage location, the following access controls apply to all Gradient AI output files:

8. Gradient AI Access & Credentials

TBD: Gradient AI contact name, support email, account manager, platform login URL, and access credentials to be documented here upon vendor onboarding completion. Store credentials in the designated secure credential vault (not in this document).

9. Roles & Responsibilities

RoleResponsibilities in This SOP
Operations DirectorOwns this process; approves data submissions; reviews red flag triage decisions; escalates to Leadership when required
Legal Counsel (Dickinson Wright)Negotiates and executes BAA with Gradient AI; advises on HIPAA compliance for data submissions
Davies ActuarialPerforms actuarial review when SAIL score exceeds 1.5; provides rate recommendations
Jason Strain (Brown & Brown / HLRA)Performs stop-loss facultative review for high-cost drug flags; receives HLRA notification for projected claimants exceeding $250K
Underwriting / Operations StaffSubmits census and screening data; retrieves and files Gradient AI reports; flags red flags for Director review
LeadershipFinal decision authority on groups with multiple simultaneous red flags or borderline risk profiles