High-charge Providers — Who & Where
Project-defined high-charge flags by service, supplemented by IQR and z-score screens, with a descriptive cross-dataset provider overlap. Inpatient is FY2023 and outpatient is CY2023. The method is inspired by OIG's work but is not the OIG cohort.
OIG / HHS, "Medicare Hospital Outlier Payments Warrant Increased Scrutiny" (OEI-06-10-00520, 2013), used hospital-level outlier-payment percentages and a 75th-percentile-plus-1.5-IQR threshold, then compared cohorts using actual CCR and other measures. This project instead selects provider-service rows at or above the submitted-charge 75th percentile. OIG's CCR and payment-dollar findings are historical context only; this output does not reproduce them.
Three methods, compared
Each provider × service row in a service with at least 30 providers is screened by up to three project methods. IQR and z-score screens are two-tailed billing-anomaly screens; only the first method is the directional high-charge cohort:
- Project-defined high-charge cohort — at or above the within-service 75th percentile of
avg_submitted_charge, inspired by OIG screening but not equivalent to its outlier-payment cohort - IQR — outside
[Q1 − 1.5×IQR, Q3 + 1.5×IQR]on billing - Z-score —
|z| > 2
The three-way agreement matrix below compares provider flag sets. Agreement is a descriptive screening result, not validation against an external outcome or payment-based truth.
Project-defined cohort comparison
For each eligible service, we compare the project-defined high-charge cohort against the remaining rows on submitted-charge level, the charge-to-payment proxy, and volume. The inpatient proxy is submitted charge divided by average total payment; the outpatient proxy is submitted charge divided by average allowed amount. Because selection uses submitted charge and the proxy uses submitted charge as its numerator, proxy premiums are post-selection descriptive summaries, not independent evidence of cost or markup.
The median submitted-charge premium is 145.3% inpatient and 142.7% outpatient within these project-defined cohorts. The corresponding charge-to-payment proxy premiums are 111.8% and 135.5%. These are post-selection, within-project descriptive summaries; they are not CCR, cost, markup, overcharging, or decade-deviation findings and are not compared numerically with OIG.
Concentration of project-defined high-charge flags
This output ranks services by counts of flagged provider-service pairs. It does not count payment dollars. OIG's historical 41% benchmark is a payment-dollar share, so the units are not comparable:
Inpatient
Top 16 DRGs share of flag pairs: —
Smallest N accounting for 40% of flag pairs: —
Current project unit: provider-service flag pairs
Outpatient
Top 16 APCs share of flag pairs: —
Smallest N accounting for 40% of flag pairs: —
Current project unit: provider-service flag pairs
Current values: 23.23% of inpatient flag pairs and 51.17% of outpatient flag pairs are in the top 16 services. These values describe this flag unit only; they do not establish a change in behavior over time.
Cross-dataset high-charge flag overlap
We join project-defined high-charge flags on provider_ccn and count providers appearing in both settings. The current result is 1,134 of 1,304 inpatient flagged providers (86.96%) and 1,134 of 1,702 outpatient flagged providers (66.63%). The flag-overlap opportunity denominator contains 2,902 eligible inpatient providers, 2,993 eligible outpatient providers, and 2,808 providers in both; this is distinct from the 2,812-provider proxy denominator. Shared-universe and opportunity-breadth-matched conditional-independence baselines are shown below; neither is evidence of institutional behavior. No comparable join was found in the sources reviewed, which is not an absolute novelty claim.
Shared-universe independence baseline: - expected providers, observed/expected ratio -. Breadth-matched independence baseline: -, observed/expected ratio -. Both are descriptive nulls, not institutional evidence.
Top 10 providers by repeated high-charge flags in both settings. The ranking favors providers with more eligible services and more opportunities to be flagged; it does not support academic, institutional, or generalist classifications.
High-charge flag profile — where do repeated flags occur?
Descriptive profile of selected high-charge flags across geography and service breadth:
Inpatient — by census region
Inpatient — Urban / Rural
Outpatient — by census region
Outpatient — Urban / Rural
Breadth findings on this page:
← insights · next: Predictive Modeling →