AI for Skilled Nursing Facility Billing: MDS Assessment and RUG Classification

The MDS Drives Everything in SNF Billing

In skilled nursing facilities, the Minimum Data Set (MDS) assessment is the cornerstone of both clinical care planning and reimbursement. Under the Patient-Driven Payment Model (PDPM), MDS responses determine how each resident is classified across five case-mix components: physical therapy, occupational therapy, speech-language pathology, nursing, and non-therapy ancillary services. Each component has its own classification, and together they determine the daily payment rate for that resident.

The financial stakes are significant. A single MDS item coded differently can shift a resident from one classification to another, changing the daily rate by $50 to $200 or more. Over a 20-day skilled stay, that difference adds up to $1,000 to $4,000 per resident. For a 100-bed facility with regular admissions, the cumulative impact of MDS accuracy on annual revenue is measured in millions of dollars.

Where MDS Completion Goes Wrong

MDS completion involves hundreds of items covering everything from the resident cognitive function and mood to their physical capabilities and clinical conditions. The assessment is performed by a registered nurse MDS coordinator, often with input from therapy staff, physicians, and other clinicians. The sheer volume of items creates opportunities for error and inconsistency.

Common problems include incomplete section completion, inconsistency between MDS responses and the clinical record, missed diagnosis codes that affect the non-therapy ancillary component, and timing errors where assessments are not completed within the required window. Each of these issues can result in either underpayment (when the MDS understates the resident care needs) or compliance risk (when the MDS overstates them).

How AI Assists MDS Coordinators

AI-driven MDS assistance works alongside the MDS coordinator during the assessment process. The system cross-references MDS responses against the clinical documentation in real time, flagging inconsistencies and potential errors before the assessment is finalized and transmitted.

If the clinical notes document that a resident requires extensive assistance with bed mobility but the MDS codes them as needing limited assistance, the system highlights the discrepancy. If the medication administration record shows a resident receiving IV medications (which affects the nursing classification) but the MDS does not capture this, the system alerts the coordinator.

The system also tracks the PDPM classification implications of each MDS response. As the coordinator completes each section, the system shows how the responses will affect the resident case-mix classification. This is not about gaming the assessment. It is about ensuring that the MDS accurately reflects the clinical reality so the facility receives appropriate payment for the care it is providing.

Diagnosis Code Optimization

Under PDPM, diagnosis codes play a larger role in payment than they did under the previous RUG system. The primary diagnosis assigned on the MDS determines the clinical category for several PDPM components. Secondary diagnoses contribute to comorbidity adjustments that can significantly affect the non-therapy ancillary payment.

AI systems review the resident clinical record and identify all active diagnoses that should be captured on the MDS. They flag situations where a diagnosis documented in the clinical record is not reflected in the MDS, particularly diagnoses that would result in a higher-paying clinical category or comorbidity adjustment. The coordinator then verifies the clinical documentation and determines whether the diagnosis should be added.

Assessment Scheduling and Tracking

MDS assessments follow a strict schedule defined by CMS. The 5-day assessment must be completed by day 8 of the stay. The interim payment assessment (IPA) occurs when there is a significant change in condition. Discharge assessments have their own timing requirements. Missing these windows results in default payment rates that are almost always lower than what the facility would have received with a timely assessment.

AI systems track every resident assessment schedule and alert the MDS team well in advance of upcoming deadlines. They also monitor for significant changes in condition that would trigger an interim assessment, using data from nursing notes, medication changes, and therapy evaluations to identify residents who may need reassessment.

Audit Preparation

SNFs are subject to audits from Medicare Administrative Contractors, the OIG, and private audit firms working for Medicare. These audits compare MDS responses against the clinical record to verify that the assessment accurately supports the resulting payment classification.

AI systems prepare for audits proactively by maintaining a mapping between every MDS response and the clinical documentation that supports it. If an auditor questions a particular MDS item, the system can immediately produce the relevant nursing notes, therapy documentation, and physician orders that support the coded response. This documentation reduces audit exposure and speeds up the response to audit inquiries.

For skilled nursing facilities, MDS accuracy directly translates to revenue accuracy. AI assistance during the assessment process improves both clinical documentation and financial performance while reducing the compliance risk that comes with inaccurate assessments. Explore how AI supports healthcare facility billing at FirmAdapt.