Not long ago, I was sitting across from a nursing home leadership team as they stared at a quality measure report that didn’t make sense to them.

“We’re not a below-average building,” one of them said, tapping the score. “This doesn’t reflect who we are.” They weren’t being defensive. They were confused.

Clinically, they felt strong. Their survey history was stable. Staff turnover was improving. But their QM domain was dragging, and they couldn’t reconcile the score with the care they believed they were delivering.

So we did what more facilities should do when a QM score feels off: We stopped looking at the rating and started looking at the math. That’s where the real story surfaced. It wasn’t about poor care, but about exclusions not captured, covariates not fully coded, and pieces of the clinical picture that never made it into the structured data fields the Centers for Medicare & Medicaid Services reads.

Because QMs don’t interpret intent or goodwill or even care plans. They interpret documentation.

Exclusions are not technicalities

Take the Long-Stay Antipsychotic Medication measure (CMS ID: N047.01). Under this measure, residents are excluded if specific conditions are present and properly supported. These include:

• Schizophrenia
• Tourette’s syndrome
• Huntington’s disease
• Hospice enrollment
• Certain Medicare or Medicaid enrollment criteria

Beginning with the re-specification effective January 2026, CMS no longer relies on MDS coding alone. The measure now incorporates Medicare and Medicaid claims and encounter data in addition to MDS data. That means something operationally significant. Now, for diagnoses such as schizophrenia to exclude a resident, the condition must be:

1. Coded on the MDS, and
2. Supported by corresponding diagnosis codes in Medicare or Medicaid claims data within the defined look-back window

If the MDS says schizophrenia but claims data does not support it during the required time frame, the exclusion does not apply, and the resident remains in the denominator. And if antipsychotics are present, the resident is in the numerator and the QM triggers — not because the medication was necessarily inappropriate, but because the MDS documentation and billing systems were not aligned.

We have officially entered the era where exclusions are no longer just MDS decisions. They are system-level documentation decisions.

Hospice is not a check box; it’s a claims event

The same complexity applies to hospice. Under the long-stay antipsychotic measure, a resident is excluded if they receive Medicare Part A or Medicaid-covered hospice services or are enrolled in hospice during any month between the beginning of the target period and the end of the episode.

Simply coding hospice on the MDS is not enough if it is not reflected in claims or eligibility data during the appropriate time frame. Hospice is not an intent; it is a documented and billable status. If social work knows it, nursing codes it, but claims do not reflect it, and the algorithm does not see it. And the QM will reflect that.

Risk adjustment only works if the story is complete

Exclusions remove residents from the denominator; covariates adjust the risk. They are not the same.

Consider the Long-Stay Pressure Ulcer measure (CMS ID: N045.02). This measure is risk-adjusted using covariates such as:

• Impaired mobility
• Bowel incontinence
• Peripheral vascular disease or diabetes
• Low BMI
• Malnutrition
• Infection indicators
• Hospice status

CMS acknowledges that residents with greater clinical complexity carry greater risk. But here’s the problem:

• What if bowel incontinence is present but undercoded?
• What if malnutrition is treated but not captured?
• What if diabetes is documented in a hospital summary but never reconciled into Section I?

In all of these cases, the expected risk score drops, the observed rate stays the same, and the adjusted QM worsens. Risk adjustment is not magic. It is math based on what you code.

The interdisciplinary blind spot

In that conference room, we discovered something telling. Housekeeping had noticed trays coming back untouched weeks before weight loss was triggered. CNAs had documented behavioral changes, but these were never reviewed. Families had expressed comfort-focused goals based upon prognosis shared by the physician, but there was no documentation of limited prognosis. Hospice was desired, but enrollment was delayed.

No one was negligent. But everyone was siloed.

As my colleague Karen Welsh, SLP, RAC-CTA, put it recently: “Quality measures don’t belong to the MDS nurse. They belong to the entire building. If housekeeping sees decline before nursing does, and dietary knows intake before weights change, and families understand baseline better than anyone, then we either bring them into the quality conversation, or we accept that our data will always lag behind our reality.”

QMs should not simply reflect what was coded on the MDS. The process must move upstream to capture the voices of the family and interdisciplinary team, the covariates and exclusions, and the potential QM triggers.

The real work

QMs are not about optics. They are about alignment: between clinical reality and structured data; between MDS coding and claims submission; between what the interdisciplinary team knows and what the algorithm can actually see. When those align, the score stabilizes. When they don’t, the measure exposes the gap — not as punishment, but as feedback.

Steven Littlehale is a gerontological clinical nurse specialist and chief innovation officer at Zimmet Healthcare Services Group.

The opinions expressed in McKnight’s Long-Term Care News guest submissions are the author’s and are not necessarily those of McKnight’s Long-Term Care News or its editors.