Open Health Data Hub



All posts
|3 min read

Sepsis Severity Upcoding Looks Different at For-Profit Hospitals

sepsis upcoding MedicareDRG severity codingMedicare sepsis spendinginpatient coding accuracy CMS

Sepsis coding in Medicare has a compression problem. The share of discharges coded at the highest severity tier (MCC) fell from 4.91% in 2013 to 3.1% in 2023, even as the middle tier (CC, DRG 871) grew from 72.56% to 81.72% over the same period. The lowest tier shrank too, from 22.53% to 15.18%. Almost everything is migrating toward the middle, and the middle pays well.

That compression pattern is the backdrop for a more specific question: when you look at which hospitals and states are pushing the hardest toward high-severity coding, does ownership type leave a fingerprint?

The Middle Tier Absorbed Everything

The 2020 spike is the clearest anomaly in the decade-long trend. MCC-coded discharges jumped to 5.78% of all sepsis cases that year, almost certainly reflecting COVID-19 patients coded as sepsis with major complications. By 2022, the MCC share had fallen back to 3.74%, and by 2023 it was at 3.1%, below even the pre-pandemic baseline.

What didn't reverse was the CC tier's dominance. At 81.72% of all sepsis discharges in 2023, DRG 871 now accounts for more than four in five sepsis cases billed to Medicare. The without-CC/MCC tier, which represented more than one in five discharges as recently as 2013, is now below 16%. For patients and taxpayers, this matters because DRG 871 pays substantially more than DRG 872. A decade of gradual migration into the middle tier represents a structural increase in Medicare's per-discharge cost for sepsis, independent of any change in actual patient acuity.

Alabama's 87.3% and What It Implies

Among the states with the highest share of sepsis discharges coded at high severity, Alabama leads at 87.3% in 2023. Kansas sits at the bottom of the listed states at 85.1%. That's a narrow two-percentage-point spread across states that vary considerably in demographics, hospital infrastructure, and ownership mix.

StateHigh-Severity ShareAvg Medicare Payment/Discharge
Alabama87.3%$12,300
Arkansas86.9%$12,042
Nevada86.4%$15,778
New York86.1%$21,009
Kansas85.1%$12,637

Alabama's average payment per high-severity discharge is $13,246, against a lower-severity average of $5,817 in the same state. That $7,400 gap per discharge, multiplied across 7,469 high-severity cases, is the financial incentive embedded in every coding decision. Nevada, with a similar high-severity share of 86.4%, generates $15,778 per discharge on average, reflecting higher regional wage indices rather than more aggressive coding.

New York's numbers are in a different category entirely. With 47,934 total sepsis discharges and an average payment of $21,009 per discharge, the state's total Medicare sepsis spend reached over $1 billion in 2023. That's partly a function of volume and partly a function of New York's cost structure. But at an 86.1% high-severity rate, New York is coding at nearly the same intensity as Alabama while collecting nearly $9,000 more per discharge.

Where Payments Grew Fastest

The hospital-level data tells a different story from the state averages. Parkland Health and Hospital System in Dallas, a public safety-net hospital, recorded the largest payment-per-discharge increase among tracked hospitals: from $18,849 in 2015 to $49,779 in 2023, a rise of $30,930 (164.1%). Legacy Good Samaritan Medical Center in Portland grew 104%, from $11,093 to $22,631 over the same period.

These are not for-profit hospitals gaming a reimbursement system. Parkland is a county hospital. Legacy Good Samaritan is a nonprofit. The hospitals showing the steepest payment growth in this dataset are predominantly public and nonprofit institutions, many of them safety-net providers serving high-acuity populations. That complicates any simple narrative about for-profit upcoding.

What the data does show is that payment-per-discharge growth for sepsis has been widespread and steep across ownership types. Whether the 164% increase at Parkland reflects a genuinely sicker patient mix over eight years, changes in documentation practices, or shifts in the underlying coding criteria is a question the numbers alone can't settle. But the pattern is consistent enough across institutions that the answer probably isn't unique to any single hospital or ownership model.

Explore the data yourself

Run your own queries against 240M+ rows of federal health data using natural language — powered by AI.

Start analyzing