11.3 SOI/ROM as CDI Program Performance Metrics

11.3 SOI/ROM as CDI Program Performance Metrics

Quick Answer: Under All Patient Refined Diagnosis Related Groups (APR-DRGs), severity of illness (SOI) and risk of mortality (ROM) are four-level subclasses (1 minor, 2 moderate, 3 major, 4 extreme) scored independently. Secondary diagnoses—with principal diagnosis, age, and certain procedures—feed those scores. That is a different severity engine from MS-DRG CC/MCC payment tiers. Programs may track SOI/ROM distributions as performance metrics; they must not treat a higher subclass as a query quota.

IPPS payment for most Medicare acute inpatient stays uses MS-DRGs. Many hospitals still run APR-DRG grouping in parallel for quality profiling, some Medicaid or commercial contracts, and internal severity dashboards. CCDS Domain IV expects you to use SOI and ROM as program metrics without confusing them with CC/MCC. (Chapter 20 returns to SOI/ROM in mortality-index work; this section stays on how the scores are built and how a CDI program should read them.)

Four levels, two independent subclasses

APR-DRG assignment gives each stay three descriptors: a base APR-DRG, an SOI subclass, and an ROM subclass.

SOI denotes the extent of physiologic decompensation or organ-system loss of function. ROM denotes the likelihood of dying. They are related clinical ideas and they are not the same number.

The four SOI subclasses and four ROM subclasses are numbered 1 through 4. A stay can be SOI 4 and ROM 2, or SOI 2 and ROM 4. A program that reports “severity” as a single digit has already collapsed the model.

APR-DRG logic uses principal diagnosis, secondary diagnoses and their combinations, age, sex, and certain procedures. CDI specialists do not key a 1–4 into the grouper. The grouper infers subclasses from coded data. Unspecified, unlinked, or missing secondary diagnoses produce subclasses that understate how sick the patient was. Manufactured secondary diagnoses produce subclasses that will not survive audit.

Independence is the exam trap. A patient can consume large hospital resources (high SOI) without a high short-term likelihood of dying (lower ROM)—for example, a severe but usually survivable acute illness with intensive monitoring. Another patient can carry high mortality risk (high ROM) without the same burst of acute organ-failure resource use (lower or moderate SOI). If a stem says SOI and ROM “always move together,” that statement is false.

How secondary diagnosis documentation feeds the scores

Uniform Hospital Discharge Data Set (UHDDS) secondary diagnoses are conditions that received clinical evaluation, therapeutic treatment, diagnostic procedures, extended length of stay, or increased nursing care or monitoring. When those conditions are specified and linked where linkage matters, they become coded secondary diagnoses. The APR-DRG methodology then asks not only “is a CC present?” but how those diagnoses interact for that base APR-DRG.

That is why “add any secondary diagnosis” is not a severity strategy. A symptom that is integral to the principal diagnosis may add no SOI. Two moderate conditions together may raise SOI more than either alone. Age can modify ROM. A procedure can change the base APR-DRG before subclasses are applied.

CDI’s program-metric job is to make the coded severity match the clinical severity that was already present, by clarifying type and acuity (systolic versus unspecified heart failure; acute versus unspecified kidney injury), linkage when the classification requires it (organ dysfunction due to sepsis), conditions that were treated but named only in ancillary notes until a provider documents them, and clinical validation when a high-severity label has no support.

Moving a population’s average SOI from 2.1 to 2.4 can be appropriate capture or it can be label inflation. As with CMI, pair the metric with validation and with mix. A month heavy with elective low-risk surgery will suppress average SOI even if medical CDI is excellent. A software or grouper-version change can move averages without any chart changing. Domain IV analysis names those confounders instead of declaring a CDI victory or collapse.

Contrast with MS-DRG CC/MCC tiers

A case can have no MCC for MS-DRG purposes and still reach a high APR-DRG SOI if multiple interacting diagnoses meet that methodology. The reverse also happens: a single MCC can lift MS-DRG payment while APR-DRG ROM stays moderate. Domain IV items that treat “we captured an MCC, therefore ROM is 4” are testing that confusion.

Do not report APR-DRG subclass changes as IPPS dollar impact unless the hospital is actually paid on APR-DRG for that stay. Mixing the engines is how a program double-counts value. CC and MCC are not the names of APR-DRG subclasses; they are CMS severity categories inside MS-DRG logic.

Using SOI/ROM as performance metrics

Programs commonly track mean SOI and mean ROM for the hospital, a service, or a high-frequency APR-DRG; the percent of discharges at SOI 3 or 4 and, separately, ROM 3 or 4; mortality by SOI/ROM cell (deaths in extreme ROM should not shock anyone; deaths in ROM 1 should prompt review); and change after a documentation initiative compared with a control service.

These are lagging, mix-sensitive metrics, like CMI. They are useful when you hold the base APR-DRG or service line roughly constant and you have a clinical-validation backstop. They are harmful when leadership sets “average SOI must be 3.0” as a specialist goal. That goal recreates the query-quota problem from section 11.1: specialists hunt for extreme labels instead of for accurate labels. Putting “please document extreme mortality risk to improve our ROM metric” on a query also fails the 2026 ban on quality-outcome language on the query.

Scenario: high SOI, lower ROM. A 62-year-old is admitted with severe acute pancreatitis, large fluid resuscitation, and a week of intensive monitoring, then goes home. Resource use is high, so SOI can be high, while inpatient death risk for that trajectory may be lower, so ROM need not match. If the record only says “abdominal pain, unspecified,” both subclasses understate the stay. If CDI queries for pancreatitis acuity and treated complications that indicators already support, SOI/ROM can rise because the chart now matches the care. If CDI queries “please document extreme mortality risk” to move a dashboard, the query is leading and the metric is corrupted.

Scenario: ROM moved, capture did not. A monthly report shows average ROM up 0.2 while MS-DRG CC/MCC capture is flat. Honest explanations include APR-DRG-sensitive diagnoses that are not MCCs being specified, age mix changing, a high-ROM service growing, or a grouper version changing. A dishonest explanation is “ROM is the same thing as MCC capture.” Analyze the engines separately.

Track SOI/ROM to see whether coded secondary diagnoses reflect physiologic severity and mortality risk. Do not use them as a fourth query quota, and do not substitute them for MS-DRG CMI when the question is IPPS payment.