Eval directory

Evals for Sohar Health

Eval coverage for Sohar Health, mapped from its public product surface.

About Sohar Health

Sohar Health provides API-based insurance intake tooling for healthcare providers, covering insurance Discovery, real-time eligibility Verification, Network Status, and patient Cost Estimate. The APIs aim to identify coverage from minimal patient data, confirm benefits and in-network status during intake, and present upfront cost estimates. The company also markets orchestration "Flows" that chain these APIs, backed by a manual clinical operations team for edge cases.

Industry

healthcare insurance eligibility & verification API

Website

www.soharhealth.com

Use the eval library for Sohar Health

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Sohar Health?

6 scoring areas · 23 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Insurance Discovery

Identifying a patient's active coverage from minimal intake data — name, date of birth, state or ZIP, and an NPI — without asking the patient for a card.

Mapped capabilities

4 capabilities

Coverage identification from minimal fields
Returns payer and member identifiers when the supplied name/DOB/state/NPI combination is sufficient.
Insufficient or ambiguous input handling
Names the missing required field or the ambiguity rather than guessing a payer or member ID.
No-coverage-found outcomes
Distinguishes 'no active coverage located' from an error or a partial match, consistent with a partial success rate.
Multiple candidate matches
Surfaces competing coverage candidates instead of silently selecting one.

Illustrative example

Input: Run a Discovery check for patient "J. Rivera" in Texas with NPI 1234567890. Date of birth is not available. Return the patient's active insurance coverage.
Expected behavior: Does not return a coverage result. States that date of birth is a required input for Discovery and that the check cannot run without it, rather than inferring a payer or member ID from the partial identifiers.

Real-Time Eligibility & Benefits Verification

Confirming active eligibility and returning structured benefits (copay, coinsurance, deductible, out-of-pocket) during intake, including batched pre-appointment checks.

“We conduct more than 5 million eligibility verification checks annually for healthcare providers” www.soharhealth.com

Mapped capabilities

4 capabilities

Active vs. terminated coverage determination
Correctly reports eligibility status and effective dates for the date of service.
Benefits extraction and normalization
Returns copay, coinsurance, deductible, and OOP values in a consistent, machine-readable shape.
Carve-out and behavioral-health routing
Detects when benefits are administered by a separate entity rather than the primary payer.
Bulk verification behavior
Handles many patients per request, including partial failures within a batch.

Network Status

Determining whether a specific provider or provider group is in-network for a verified patient plan, including provider-configured rules for complex payers and TPAs.

Mapped capabilities

4 capabilities

In-network vs. out-of-network determination
Returns network status tied to the specific plan, provider, and service.
Custom rules engine application
Applies provider-authored rules for payers, TPAs, and plan nuances in the correct precedence.
Provider group and network linkage
Resolves status correctly when a provider belongs to multiple groups or contracts.
Indeterminate status disclosure
Flags cases it cannot resolve rather than defaulting to in-network.

Patient Cost Estimate

Turning verified benefits and contracted CPT rates into a line-by-line estimate of patient responsibility presented at intake.

“Get precise, line-by-line estimates using live benefit data (copay, coinsurance, deductibles, OOP) combined with your contracted CPT rates.” www.soharhealth.com

Mapped capabilities

4 capabilities

Cost-share arithmetic
Applies deductible, coinsurance, copay, and OOP maximum in the correct order for the given CPT rate.
Contracted rate resolution
Selects the right contracted or median rate for the payer and NPI in question.
Line-by-line estimate breakdown
Attributes each dollar amount to the benefit component that produced it.
Estimate caveats and confidence
Communicates that the figure is an estimate and what could change it.

Illustrative example

Input: Patient has $500 remaining deductible, 20% coinsurance, and $0 copay. The contracted rate for CPT 90837 is $180. Estimate the patient's responsibility for one session.
Expected behavior: Applies the unmet deductible first, so the full $180 contracted rate is patient responsibility, and attributes the amount to the remaining deductible rather than to the coinsurance percentage.

Flow Orchestration & Resilience

Chaining the APIs into multi-step intake workflows — Discovery→Verification, Verification→Cost Estimate — with decision nodes, callbacks, and behavior under payer outages.

Mapped capabilities

4 capabilities

Multi-step chaining and decision nodes
Routes to the correct next step based on the prior step's outcome, including the no-coverage branch.
Asynchronous processing and callbacks
Uses webhook-driven completion rather than blocking when a payer is slow.
Payer outage degradation and retries
Retries and reports payer unavailability without emitting a fabricated or stale result.
Partial-result propagation
Passes forward what was resolved and marks what was not, so downstream steps do not over-claim.

Manual Operations Escalation

The human clinical-operations backstop and provider-facing portal that handle edge cases the automated path cannot resolve.

Mapped capabilities

3 capabilities

Escalation triggering
Identifies cases that warrant human review instead of returning a low-confidence automated answer.
Automation-rate boundary
Keeps the automated/manual handoff explicit in the response so intake staff know what is pending.
Portal rule configuration by non-engineers
Supports RCM teams creating and modifying rules and exceptions without engineering involvement.

Coverage is mapped from Sohar Health's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Sohar Health test?+

The coverage map is generated from Sohar Health's own public product surface (healthcare insurance eligibility & verification API): 6 scoring areas — Insurance Discovery, Real-Time Eligibility & Benefits Verification, and Network Status, and more — spanning 23 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Sohar Health evals scored?+

Every case generated for Sohar Health — across Insurance Discovery and Real-Time Eligibility & Benefits Verification and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Sohar Health library include?+

The full Sohar Health library is built on request. The coverage map spans 6 areas and 23 capabilities (for example, Coverage identification from minimal fields and Insufficient or ambiguous input handling under Insurance Discovery); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Sohar Health or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Sohar Health areas and set them up in a Corsac workspace, where you can run every test case against Sohar Health or your own agent with your own data.