healthcare-claims-analytics

Overview

Deliver deterministic, copy-paste-ready Python and SQL for healthcare claims data processing and billing-integrity screening: X12 837/835 parsing, procedure and diagnosis code manipulation, provider billing profiling, Evaluation and Management (E&M) upcoding detection, National Correct Coding Initiative (NCCI) edit validation, duplicate claim identification, Benford's Law charge analysis, and impossible-day detection. Each task has a self-contained reference file holding the working code, its parameters, and its known pitfalls. Use this skill when a user needs correct code for one of these tasks, optionally run against their own data.

Workflow

Task	Trigger examples	Reference file
X12 837 parsing	parse claim file, read 837P, extract claims	references/x12-parsing.md
E&M provider profiling	upcoding, 99214/99215 rate, z-score vs peers	references/em-profiling.md
NCCI edit validation	code pair edits, bundling, modifier 59	references/ncci-validation.md
Duplicate claim detection	exact and near duplicates, resubmissions	references/duplicate-detection.md
Benford's Law analysis	first-digit test, charge distribution	references/benfords-law.md
Impossible day detection	>24 hours billed, time-based service sums	references/impossible-days.md
Parameter tuning	thresholds, defaults, sensitivity	references/parameters-and-pitfalls.md
Domain pitfalls	why a signal is a false positive	references/parameters-and-pitfalls.md
</Definition - Analysis Tasks>

<Workflow - Deliver Claims Analytics Code description="Identify the requested claims analysis task, load its reference code, and deliver a working example, optionally running it against the user's data." tools=[file_read, run_python] triggers=["User asks to parse X12/837/835 claims", "User asks to profile provider billing or detect upcoding", "User asks to validate NCCI edits", "User asks to find duplicate claims", "User asks to run Benford's Law on charges", "User asks to flag impossible billing days", "User asks about claims analytics thresholds or pitfalls"]

1. [Decide] Match the request to a task in <Definition - Analysis Tasks>. Validate: Exactly one task is selected. If fails: If the request maps to more than one task, handle them in sequence. If it maps to none, [Ask user] which of the supported tasks they want.

2. [Agent] Read the reference file for the selected task with file_read. Validate: File content loaded and contains the code block for the task. If fails: Re-read using the path from <Definition - Analysis Tasks>. If still missing, tell the user the reference is unavailable and stop.

3. [Ask user] Confirm the input the code expects: file path for X12 parsing, or the DataFrame columns for the tabular tasks (each reference file lists the columns it needs). Skip if the user already stated their inputs. Validate: The user has confirmed inputs, or the task needs none. If fails: Ask once more, naming the exact fields the code requires.

4. [Agent] Deliver the answer using the template: confirm inputs, then the working code, then the key parameters, then the pitfalls that apply to this task (drawn from the reference file and ). Validate: Output leads with code and names every environment-specific parameter the user must set. If fails: Re-order so code comes first and add any missing parameter.

5. [Decide] Does the user want the code run against their own data now?

   • Yes -> continue to step 6.
   • No -> state the liability disclaimer from Rule 7 and stop.

6. [Agent] Run the code against the user's data with run_python, keeping all data in the session per Rule 1. Return the resulting rows or summary. Validate: The code runs without error and returns a result. If fails: Report the error and the offending input, correct the code or ask the user to fix the data, then re-run once.

7. [Agent] Close with the liability disclaimer from Rule 7. Validate: The disclaimer is present in the final message. If fails: Add it before ending.

</Workflow - Deliver Claims Analytics Code>