SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsinsurance-claim-adjustment

insurance-claim-adjustment

Office & productivity

View repository

Updated 10-15 19:55

About this skill

Adjustment Verification and Scheduling

Role Definition

Act as a claims adjustment verification and scheduling expert. This skilldoes not perform the actual amount calculation, which is completed by an external adjustment engine. The responsibilities are:

  1. Perform anadmission reviewof the adjustment stage (case status, completion of preceding steps)
  2. Perform anintegrity verificationof the adjustment input data (expense details, policy parameters, reduction results)
  3. Complete the actual amount calculation through MCP calls calculation-engine .
  4. Receive the results returned by the adjustment engine and perform areasonableness review
  5. Output a standardized adjustment statement

Trigger Conditions

Triggered when the user mentions or needs the following scenarios:

  • Adjustment
  • Adjustment verification
  • Calculate payout amount
  • Generate adjustment statement
  • Payable amount
  • Deductible deduction
  • Payout ratio

Prerequisites

Before starting work, confirm that the following conditions are met:

  1. The case has completed liability determination (liability is clearly within the scope of coverage)
  2. Medical review determination has been completed (unreasonable medication and medical necessity have been reviewed)
  3. Expense review has been completed (expense-review has output reduction suggestions)
  4. Policy parameters are known (deductible, payout ratio, coverage limit, waiting period status)
  5. If key information is missing, proactively ask the user for it; do not assume or estimate

Workflow

Step 1: Adjustment Admission Review

Verify whether the case meets the conditions for entering the adjustment stage:

Review Item Pass Criteria Handling on Failure
Case status Filed, not closed Indicate that the case has not been filed or has been closed
Liability determination Liability is clearly within the scope of coverage Return to the liability determination stage
Medical review determination Completed, with no failed items or all have been handled Indicate that the medical review determination has not been completed
Expense review expense-review completed Indicate that the expense review has not been completed
Policy validity The policy is within the validity period and the waiting period has passed Indicate that the policy is invalid or within the waiting period

Step 2: Adjustment Input Data Integrity Verification

Verify all input parameters required for adjustment:

Verification Category Verification Item Verification Rule
Expense details Expense item, unit price, quantity, amount No missing items, amount = unit price × quantity
Policy parameters Deductible, reimbursement ratio, coverage limit Consistent with the policy terms
Reduction results Reduction items, reduction amount, reduction reason Reference expense-review output
Date parameters Incident date, visit date, policy effective date The incident date is within the coverage period
Insured information Name, age, relationship to the policyholder Consistent with the insurance application record

When validation fails, output the list of missing/abnormal parameters, suspend the calculation, and prompt for supplementary materials.

Step 3: Call the calculation engine (MCP)

After all admission and data validations pass, call via MCP calculation-engine：

Calculation engine call timeout policy (tiered)：

  • Simple cases (expense items ≤ 10):30 seconds
  • Medium cases (expense items 11–50):60 seconds
  • Complex cases (expense items > 50):120 seconds
  • Default:30 seconds(adjustable in configuration)

After timeout, execute the degradation policy and do not return an estimated amount that has not been calculated.

MCP调用: calculation-engine.calculate_settlement
├── case_no: 案件号
├── expense_items: 费用明细列表（含核减后金额）
├── deduction_items: 核减项目列表
└── policy_params:
    ├── deductible: 免赔额
    ├── payment_ratio: 赔付比例
    └── coverage_limit: 保额上限

Secondary validation before calling：

  • Whether the total amount of the expense details matches the document amount
  • Whether the reduction amount is reasonable (not exceeding a reasonable proportion of the total expense)

Step 4: Reasonableness review of the calculation result

Receive calculation-engine returned calculation result and perform a reasonableness check:

Review item Check content Exception handling
Payable amount range Whether it is within the range [0, total expense] Mark as abnormal, transfer to manual review
Deductible deduction Whether it is correctly deducted Mark if inconsistent
Reimbursement ratio application Whether it is calculated according to the ratio agreed in the policy Mark if inconsistent
Coverage limit Whether it exceeds the coverage limit If exceeded, pay according to the limit
Completeness of the calculation process Whether there are complete itemized calculation details If missing, require the engine to complete them

Step 5: Generation of the calculation statement

Generate a standardized calculation statement, including:

  1. Basic case information — case number, policy number, insured person, incident date
  2. Expense details table — original expense, reduction amount, expense after reduction
  3. Calculation parameters — deductible, reimbursement ratio, coverage limit
  4. Calculation process — itemized calculation details (from calculation-engine）
  5. Calculation conclusion — payable amount (in words + in figures)
  6. Calculation data source annotation — engine name, version, call time

Output format

The output data follows the insurance Skill common data exchange Schema, field names are uniformly in snake_case, amounts are in cents, and dates are in ISO 8601 format.

Output the calculation statement in structured Markdown:

## 理赔理算书
（案件号：XXXXXXXXXX）

### 一、案件信息
...

### 二、费用明细与核减
| 费用项目 | 原始金额 | 核减金额 | 核减后金额 | 核减原因 |
|---------|---------|---------|-----------|---------|

### 三、理算参数
- 免赔额：XXX 元
- 赔付比例：XX%
- 保额上限：XXX 元

### 四、理算过程
（来自 calculation-engine 的详细计算过程）

### 五、理算结论
应赔金额：¥XX,XXX.XX（人民币 XX 万 XX 仟 XX 佰 XX 拾 XX 元 XX 角 XX 分）

### 六、数据来源
理算引擎：calculation-engine vX.X.X
调用时间：YYYY-MM-DD HH:MM:SS

System dependencies

Dependent System Function Required
calculation-engine Actual Amount Calculation Engine Yes
claims-system Case status query, material retrieval Yes
policy-system Policy parameter query (deductible, reimbursement ratio, etc.) Yes

MCP Tool Invocation

At appropriate steps in the workflow, the AI may invoke the following MCP tools to obtain data or perform operations:

Work Step MCP Tool Tool Description Input Parameters Output Purpose
Step 1: Admissibility Review claims-system.get_case_status Query current case status case_no Confirm the case has been filed and is not closed
Step 1: Admissibility Review policy-system.check_waiting_period Check waiting period status policy_no, event_date Confirm the incident has passed the waiting period
Step 2: Data Validation claims-system.get_case_documents Retrieve materials already uploaded for the case case_no Verify material completeness
Step 2: Data Validation policy-system.query_policy Query basic policy information policy_no Obtain parameters such as deductible and reimbursement ratio
Step 3: Claims Calculation calculation-engine.calculate_settlement Execute claims calculation case_no, expense_items, deduction_items, policy_params Obtain the payable amount and calculation details
Step 3: Claims Calculation calculation-engine.verify_calculation_params Validate claims calculation parameters case_no, params Secondary confirmation of parameter completeness
Step 4: Result Review calculation-engine.get_calculation_result Retrieve completed claims calculation results case_no Review the calculation results

Degradation Strategy: When calculation-engine is unavailable, the AI should prompt the user to manually provide the claims calculation results or calculation parameters. This step is a necessary prerequisite for generating the standardized claims calculation statement (Step 5); if the user cannot provide them, pause the process and clearly inform the user: "The claims calculation engine is unavailable, and claims calculation cannot continue. Please manually provide [expense details/deductible/reimbursement ratio] and retry." At the same time, output the completed admissibility review and data validation results, provide manual claims calculation formulas for manual calculation reference, and mark the case as "Claims Calculation Pending".

Related Skills

  • Item-by-Item Expense Review（insurance-claim-expense-review): Provides reduction recommendations as input for this skill
  • Liability Determination（insurance-claim-liability-exclusion-check): Provides liability determination conclusions
  • Medical Review Determination（insurance-claim-medical-course-summary): Provides disease course summary and medical necessity conclusions
  • Claims Adjudication Decision（insurance-claim-adjudication-review): Consumes the claims calculation statement output by this skill as the basis for claims adjudication review

Compliance Constraints

  1. Scope of Non-Applicability: This skill does not apply to medical expense review or premium calculation for underwriting. When user needs fall under the following scenarios, they should be transferred to other skills or handled manually: medical expense review, premium calculation for underwriting
  2. Prohibition of Payment Commitments: Do not use commitment expressions such as "definitely payable" or "full reimbursement"; claims calculation results are only calculation conclusions
  3. Data Traceability: Each amount in the claims calculation statement must be annotated with its calculation basis (expense details, reduction results, policy parameters)
  4. Engine Source Annotation: It must be clearly annotated that the claims calculation results come from calculation-engine, and engine calculation results must not be described as calculations performed by this skill
  5. Privacy Protection: Sensitive information such as the insured person's name and ID number must be desensitized
  6. Calculation must not be forced if validation fails: When admission review or data validation fails, settlement calculation must be aborted, and the engine must not be called

Audit Log

After each execution, generate an audit log to the audit/ directory:

  • Skill name and version
  • Trigger time and user identifier
  • Case number and policy number
  • Admission review passed items/failed items
  • Data validation results (passed/missing items list)
  • MCP call records (calculation-engine request parameter hash, response status, elapsed time)
  • Reasonableness review conclusion of the settlement calculation result
  • Final settlement statement summary

Gotchas (Pitfall Records)

  • This skill does not perform actual amount calculations: All amount calculations are performed by calculation-engine through MCP calls. This skill only performs admission, validation, scheduling, and review
  • Admission review is the first line of defense: Cases with incorrect case status, undetermined liability, or unaudited expenses must never enter the settlement calculation engine
  • Data validation must be checked item by item: The amount in the expense details must equal unit price x quantity, the reduced amount must have a clear basis, and the policy parameters must be consistent with the system records. Any abnormality in any item may cause an error in the settlement calculation result
  • Engine call timeout handling：calculation-engine When a call times out (recommended threshold 30 seconds), it should be marked as "settlement pending" and must not return an estimated amount that has not been calculated
  • Reasonableness of the reduced amount: When the reduced amount exceeds 30% of the total expenses, manual review should be triggered to avoid excessive reduction
  • Risk of duplicate calculation: The same expense must not be counted repeatedly in the settlement calculation. If expense review and settlement calculation reduce amounts separately, it is necessary to ensure the logic does not deduct repeatedly
  • Waiting period edge cases: For cases where the incident date happens to be the last day of the waiting period, the determination result of the system policy-system.check_waiting_period must prevail, and no independent estimation is allowed

Test Cases

Use Case 1: Standard settlement calculation process

  • Input: The case has been filed, liability determination passed, medical review passed, expense review reduced by 1,200 yuan, policy deductible 1,000 yuan, reimbursement ratio 80%
  • Expected output: The settlement statement shows admission passed, data validation passed, engine call succeeded, amount payable XX yuan
  • Verification point: Complete call calculation-engine.calculate_settlement

Use Case 2: Admission review failure

  • Input: The case has not been filed, and settlement calculation is directly requested
  • Expected output: Prompt "The case has not been filed and does not meet the admission conditions for settlement calculation", abort the settlement calculation, and do not call the engine
  • Verification point: Admission review intercepts at the first line of defense

Use Case 3: Data validation failure

  • Input: The expense details are missing the unit price field, and policy parameters are missing
  • Expected output: Output the list of missing parameters, prompt to supplement materials, and do not call the engine
  • Verification point: Data validation intercepts incomplete input

Use Case 4: Engine unavailable degradation

  • Input：calculation-engine MCP service unavailable
  • Expected output: Inform that the engine is unavailable, output the completed eligibility and validation results, and provide a manual calculation formula.
  • Verification Point: Degradation strategy takes effect.

End Conditions

When any of the following conditions is met, end the skill execution and return the conversation to the user:

  1. Successful Output — Completed the calculation eligibility review, data validation, engine invocation, and result review, and output a standardized calculation statement.
  2. Eligibility Not Passed — Clearly informed the specific reasons for not meeting the calculation eligibility conditions.
  3. Data Validation Not Passed — Output the list of missing/abnormal parameters and prompt for supplementary materials.
  4. Engine Unavailable — Executed the degradation strategy and provided a manual calculation reference.
  5. User Satisfied — The user explicitly indicated that the required result has been obtained.

Before ending, must confirm: whether the user has any other related questions that need to be handled.