Falls With Major Injury Quality Measure Benchmarking in Skilled Nursing Facilities

Falls with Major Injury quality measure benchmarking in skilled nursing facilities is a test of reporting completeness as well as resident safety. The national benchmark is depressed by systematic underreporting, and the move to claims-assisted measurement corrects only part of that problem. SNF leaders need a more complete internal account of falls than the published score provides to interpret performance and manage regulatory exposure.

How the FMI measure is constructed and what it actually counts

Under the original specification, a fall enters the Falls with Major Injury (FMI) numerator only when two MDS assessment items are triggered on the same assessment. J1800 must show a fall, and J1900C must show major injury. Both must appear together on that assessment. A documented fall without the corresponding major-injury indication does not qualify, even if injury information exists elsewhere in the chart. The measure reports the percentage of patients or stays with one or more qualifying events during the measurement window.

The original injury definition is also specific. It includes bone fractures, joint dislocations, closed-head injuries with altered consciousness, and subdural hematoma. Other harm from a fall does not automatically count in this numerator. The distinction is clinical as well as administrative, because the measure describes a defined subset of falls rather than the full burden of fall events and their consequences.

CMS maintains two parallel measures for different populations. The long-stay measure is part of the Five-Star Quality Rating System, while the Medicare Part A SNF-stay measure is part of the SNF Quality Reporting Program (QRP). Both appear on Care Compare and were built from MDS assessment data, although the QRP measure is transitioning to a hybrid model. Their populations and program roles are not interchangeable. Under the original MDS-only design, neither measure has an independent verification mechanism, so the assessment determines what becomes visible in the reported rate.

Why the national benchmark is built on systematically incomplete data

Provider reporting operates within a direct financial and reputational disincentive. Each reported qualifying event worsens a publicly displayed rate that feeds quality ratings and now reaches into payment. That incentive creates a structural weakness when the same facilities being evaluated supply the data used to evaluate them, though it does not establish why any individual event was omitted.

The HHS Office of Inspector General (OIG) quantified that weakness in its September 2025 report OEI-05-24-00180. OIG reviewed assessments from July 1, 2022 through June 30, 2023 and found that nursing homes did not report nearly half of falls with major injury and hospitalization among Medicare-enrolled residents. The omissions were not concentrated among the weakest reported performers. Facilities with the best ratings on this measure underreported most, while those with the worst ratings underreported least. A good score alone therefore cannot show that care is safer.

OIG also found patterned differences across facilities and residents. For-profit, chain-affiliated, and larger facilities had the highest non-reporting rates, while short-stay residents were underreported far more often than long-stay residents. Younger residents and male residents were also omitted more often. Geography added another source of distortion. Washington D.C., California, and Nevada had the highest non-reporting rates, while South Dakota, Vermont, and North Dakota had the lowest, making state comparisons vulnerable to different degrees of missing data.

OIG concluded that MDS assessment data alone is insufficient for quality measurement at these levels of underreporting. Missing qualifying events systematically push that average downward, and uneven omissions distort comparisons within it. A facility with complete reporting may be comparing an accurate local rate against a certainly-deflated national reference.

What the respecified measure changes — and what it leaves untouched

CMS addressed the data-source weakness in its respecified SNF QRP measure. Its November 2025 Technical Specification Report used feedback from the cross-setting Technical Expert Panel held May 12 and 14, 2025. The new approach adds Medicare fee-for-service claims from inpatient hospital stays, emergency department visits, and observation stays to the MDS data already used. Claims supply a second source of injury evidence. They can reveal severity that the assessment did not record.

When an assessment documents a fall but does not indicate major injury, a subsequent claim with a qualifying injury diagnosis can trigger the numerator. A fall previously assigned a low-severity MDS code can therefore become a counted event when an inpatient fracture claim establishes major injury. Facilities that understated injury severity on assessments face greater exposure when the second source is introduced.

The injury definition also becomes materially broader. It includes traumatic bone fractures, joint dislocations/subluxations, internal organ injuries, amputations, spinal cord injuries, head injuries, and crush injuries, without being limited to that list. A rate increase can consequently reflect better ascertainment or broader eligibility even when safety has not worsened. Accurate prior documentation reduces exposure to the first effect, but does not eliminate the consequences of the expanded definition.

The revision applies only to the SNF QRP measure. The long-stay FMI measure used in Five-Star still uses its MDS-only methodology, so the most visible public rating still relies on self-reported assessments. The two measures will therefore have different evidentiary foundations. SNF leaders interpreting a change in performance need to distinguish a clinical change from a change in what the relevant measure can recognize.

How the FMI measure now reaches into VBP payment and survey selection

For FY 2027, the baseline period is FY 2023 and the performance period is FY 2025. FMI joins seven other measures in that program year. Falls performance is thus both a public reporting item and part of a multi-measure payment calculation, and the measure entering that calculation remains the long-stay version, distinct from the respecified QRP version.

CMS sets the FY 2029 performance standard benchmark at 0.99951 and the achievement threshold at 0.95455. These are payment performance standards, not the national incidence average discussed earlier. The benchmark concentrates the strongest incentive opportunity among the best-performing facilities. A facility near average performance is defending against a payment penalty rather than competing for the top incentive, although FMI remains only one component of the overall calculation.

CMS also changed how Special Focus Facility candidates are chosen in January 2026, telling state survey agencies to weigh falls along with survey scores. Facilities chosen face more frequent surveys and are publicly labeled. MDS reporting requirements now also cover intercepted falls and falls from external forces, so the reported count can rise even when the number of falls has not changed. Facilities must check how they code and document those categories again, since a rise in counts need not mean that care has worsened.

CMS introduced Validation Audits for MDS items supporting QRP and VBP measures, with up to 1,500 SNFs randomly selected annually. Documentation accuracy is now directly auditable. CMS issued the Risk-Based Survey system under QSO-26-14-NH on July 16, 2026, with nationwide implementation beginning September 8, 2026, providing a faster recertification path for demonstrated continuous compliance and using falls performance as an indicator. OIG’s active 2025–2026 workplan also names falls as a high-risk area where serious injuries occur and records can be wrong.

What the measure cannot see — the underreporting that claims data still won't catch

Claims can correct an assessment only when relevant claims evidence exists. A major injury managed within the facility, without a qualifying hospital, emergency department, or observation encounter, supplies no such cross-check. If the assessment also omits the injury, the hybrid measure cannot recover it. Fractures or head injuries treated conservatively on-site can therefore still go unseen by the measure.

MDS data also lacks the detail needed to reconstruct events reliably. It records whether falls happened during an assessment period, but leaves out fall dates. That limits sequencing and trend analysis from assessments alone. Misclassification creates a separate problem, because an event can be coded incorrectly in type, severity, or timing without producing a discrepancy that claims will reveal.

A 2026 JAMDA study developed a natural language processing (NLP) algorithm to find falls in nursing home clinical notes, using VA notes. The study described the limitations of MDS fall capture, including missing dates, underreporting, and misclassification. Comparing narrative extraction with MDS exposed events that assessments had not recorded. The gap extended beyond injury severity coding. Some events were present in clinical text but absent from the structured measure altogether.

Nursing-note text-mining research likewise found actionable fall-risk information recorded only in narrative form. Few studies have examined registered nurses’ notes systematically for this purpose, leaving the scale of missed information incompletely understood. A published rate therefore reflects the reach of its sources as well as true incidence. Two facilities with the same underlying fall rate can report different results because their documentation captures different portions of it.

How continuous documentation from existing clinical notes can close the internal gap

Nurses already record fall events, contributing factors, changes in resident condition, and responses to interventions in the EHR. The missing step is systematic extraction and synthesis of existing data across residents. Individual readers encounter those notes episodically. A continuous monitoring approach instead treats existing narrative documentation as a source of event and risk information that structured fields may not surface.

The JAMDA study demonstrated the value of that approach by extracting fall dates and exposing omissions and misclassification in MDS data. NLP-based extraction can also identify fall events, near-misses, and documentation gaps that might otherwise emerge only after hospitalization or during a survey. This is an internal reconciliation function. It gives a facility evidence with which to examine its own structured reporting, rather than relying on a national benchmark to reveal missing events.

Researchers have also developed an Extraction-Based Language Model Classification technique that distills clinical H&Ps into task-relevant categories before producing a concise AI Risk Snapshot. The technique offers an approach applicable to falls documentation workflows, rather than proof that every such workflow has been validated. Extraction and prediction serve different purposes. The first identifies what the record says, while the second uses available information to estimate changing risk.

A 2024 longitudinal study validated a dynamic, landmark-based fall prediction model across thousands of permanent residents in residential aged care facilities. Monthly prediction windows produced a more accurate picture than static approaches that fail to capture evolving risk factors. This supports the feasibility of updating risk from existing data. Related models predicting hospitalization or death among SNF patients have also been developed without requiring additional staff time.

Embedding such tools within EHR workflows can use charting nurses already complete without creating new documentation tasks. Better coding is a consequence of that fuller internal picture. When event evidence and changing risk are visible together, facilities have a stronger basis for complete MDS reporting and fewer unexpected discrepancies when claims are used to cross-check assessments.

What operational reliability looks like between surveys, not just before them

CMS is using falls performance as an indicator of operational reliability, including the capacity to detect risk, intervene consistently, and sustain performance between surveys. VBP evaluates a full fiscal year. The Risk-Based Survey approach also offers faster recertification to facilities that show continuous compliance, so the record of their monitoring becomes part of how their performance is judged.

For nursing managers, the practical response is to connect event review with recurring patterns. Tracking falls by shift, location, and unit can show where they cluster in ways that reviews of one incident at a time can miss. Post-fall huddles need a specific causal hypothesis. QAPI minutes can then document closed-loop learning by showing the change made after analysis and how its effects were monitored.

Documentation and workflow failures can leave sound clinical actions invisible where surveyors look. Yet assigning more manual chart review and report assembly also consumes nursing managers’ attention, reducing the time available for resident-facing supervision. Tools that work from existing documentation offer a way to recover capacity without additional data entry or headcount. A reliable falls program uses that freed-up time to keep event records complete, keep risk assessments up to date, and put what staff learn into daily care.