HI545 Unit 4 Audit Sampling Plan Example

Thresholds Before Charts: A Unit 4 Audit Sampling Plan for Office Visit Levels Billed by Six Acquired Pulmonologists

[Student Name]

Purdue University Global

HI545: Health Care Compliance

Unit 4 Audit Sampling Plan

[Instructor Name]

[Date]

Esterline Health, its physicians and every count are composite; the guidance and tools cited are real.

Purpose

This plan sets up quarterly monitoring of evaluation and management office visit levels for the pulmonology and sleep medicine practice acquired by Esterline Health. The practice had no coding audit in the three years before acquisition. Office visits are its largest source of professional revenue, and the level of service chosen for each visit is the coding decision most open to drift. The plan is a monitoring tool, not a statistically valid extrapolation for repayment; if findings suggest a repayment obligation, a separate statistical sample would be designed with counsel. The general compliance guidance counts routine auditing and monitoring among the seven pillars of a working program (Office of Inspector General [OIG], 2023), and this plan is the practice's first instance of it.

Population

The population for each quarter is every established-patient office visit, codes 99212 through 99215, billed by one of the six physicians with a date of service in the quarter, paid by any payer, and not reversed by the time the extract is run. New-patient visits, consultations, sleep study interpretations and procedures are excluded here and audited under a separate plan. The extract is pulled from the billing system on the fifteenth day after quarter end, so most claims have paid. For the first quarter, the population held 4,388 visits, from 512 to 961 per physician.

Sample Design

A stratified random sample is drawn with each physician as a stratum, ten visits per stratum, for 60 visits per quarter. Equal allocation is used on purpose: the question is whether each physician codes accurately, not what the practice's overall error rate is, and a physician with fewer visits deserves the same scrutiny. Within each stratum, every eligible visit has an equal chance of selection. The draw uses a short script that reads the extract, sorts visits by claim number, sets a random seed, and selects ten claim numbers per physician without replacement. The seed, the extract's file name and row count, and the script version are recorded in the audit log, so anyone can rerun the draw and get the same sample. The OIG's RAT-STATS software can generate an equivalent selection if a reviewer prefers an outside tool (OIG, n.d.). Recording the seed and the extract before the draw means the sample cannot be quietly redrawn until it shows a result someone prefers.

What Counts as an Error

Each sampled visit is reviewed against the documentation in the record, using the office visit guidelines in effect since 2021, under which the level rests on either medical decision making or the clinician's total minutes spent on that patient that day (American Medical Association, 2020). The reviewer records the level supported by the documentation and compares it with the level billed.

1. Overcoded: the billed level is higher than the level supported.

2. Undercoded: the billed level is lower than the level supported.

3. Unsupported: the documentation does not support any billable visit, for example a missing signature or an absent note.

4. Correct: the billed level matches the supported level.

Where the level was billed on total time, the time statement must give total minutes for the date and must be supported by the note's content. A time statement that conflicts with the record, such as forty minutes for a two-line note, is reviewed as if no time were documented, and the level is judged on decision making alone. Undercoding is counted and reported, but it does not count toward the error threshold below, because the purpose of the threshold is to find claims paid in error.

Thresholds, Set Before Review

1. Pass: zero or one of ten visits overcoded or unsupported. The physician receives a written result and returns to the next quarterly sample.

2. Education: two of ten. The physician receives a one-on-one review of the errors with the coding educator within thirty days, and is sampled again next quarter.

3. Expanded review: three or more of ten. Thirty further visits from the same quarter are drawn by the same method, a new seed recorded, and reviewed within thirty days. The compliance officer is informed.

4. Expanded review result: if eight or more of thirty are overcoded or unsupported, the compliance officer refers the matter for a corrective action plan and for counsel's assessment of any repayment obligation.

The thresholds were set by the compliance committee before the first sample was drawn, and they are written in this plan so that no one can raise or lower them after seeing a physician's results.

Reviewers and Independence

Reviews are done by two certified coders from Esterline's central coding audit team who do not code for the practice. Each reviews half the sample, and ten percent of each reviewer's visits are reviewed blind by the other. Disagreements are settled by the coding audit manager and logged. Physicians may submit a written response to any finding within ten business days, and the response is kept with the audit.

Reporting

Each quarter's report gives, by physician, the number of visits sampled, overcoded, undercoded, unsupported and correct; the threshold result; and the action taken. A practice summary shows the trend across quarters. Results are reported to the compliance committee, and physician-identified results go to the practice's medical director and the physician only. The Medicare Program Integrity Manual describes how contractors design statistical samples for overpayment estimates (Centers for Medicare & Medicaid Services, n.d.), and that chapter will guide any repayment sample the plan triggers.

Limits

Ten visits per physician per quarter can detect a pattern but cannot estimate an error rate with precision. A physician with a true overcoding rate of 20 percent would show two or more errors in ten visits about six times in ten, so a pass in one quarter is not a clean bill of health. Repeating the sample every quarter, and expanding it when results warrant, is how the plan makes up for its small size.

References

American Medical Association. (2020). CPT evaluation and management (E/M) office or other outpatient services and prolonged services code and guideline changes. American Medical Association.

Centers for Medicare & Medicaid Services. (n.d.). Medicare program integrity manual, chapter 8: Administrative actions and statistical sampling for overpayment estimates. U.S. Department of Health and Human Services.

Office of Inspector General. (2023). General compliance program guidance. U.S. Department of Health and Human Services.

Office of Inspector General. (n.d.). RAT-STATS statistical software. U.S. Department of Health and Human Services.