MDS Accuracy: Why It Matters and How AI Helps Get It Right

The MDS: Where Clinical Care Meets Financial Reality

The Minimum Data Set is the single most consequential assessment instrument in skilled nursing. It determines the Patient-Driven Payment Model (PDPM) case-mix classification that sets the per-diem reimbursement rate for every Medicare Part A resident. It feeds the quality measures that drive Five Star ratings. It informs the care plan that guides clinical decision-making. And it creates the data trail that CMS surveyors use to evaluate whether a facility is meeting its regulatory obligations.

When the MDS is accurate, the entire downstream system works: reimbursement reflects the true acuity of the resident population, quality measures reflect actual clinical performance, and care plans are grounded in a reliable clinical baseline. When the MDS contains errors, whether from inaccurate coding, incomplete clinical documentation, or misinterpretation of the assessment items, the consequences cascade. Reimbursement may be too low (leaving earned revenue on the table) or too high (creating audit liability). Quality measures may misrepresent facility performance. Care plans may be built on a flawed clinical foundation.

The financial stakes alone are significant. Under PDPM, which replaced the Resource Utilization Group (RUG) system in 2019, the MDS determines reimbursement across five case-mix adjusted components: physical therapy, occupational therapy, speech-language pathology, nursing, and non-therapy ancillaries. An error in any of the MDS sections that feed these components can shift the per-diem rate by tens of dollars, a variance that, across a resident population and over the course of a year, represents a substantial financial impact.

Common MDS Errors and Their Consequences

MDS errors are rarely the result of incompetence. They are almost always the result of systemic problems: incomplete clinical documentation that forces the MDS coordinator to code from insufficient data, assessment processes that do not capture the right information at the right time, and EHR systems that make it difficult to connect clinical observations to MDS items.

Section GG: Self-Care and Mobility

Section GG is arguably the most impactful section of the MDS under PDPM, because it directly drives the physical therapy, occupational therapy, and nursing case-mix components. Section GG requires standardized assessment of the resident’s functional abilities in self-care (eating, oral hygiene, toileting hygiene, shower/bathe self, upper body dressing, lower body dressing, putting on/taking off footwear) and mobility (roll left and right, sit to lying, lying to sitting, sit to stand, chair/bed-to-chair transfer, toilet transfer, walking in corridor and on unit).

The most common Section GG errors include:

• Inconsistent scoring between disciplines: Nursing, physical therapy, and occupational therapy may assess the same resident’s functional status differently, creating discrepancies that the MDS coordinator must reconcile without clear guidance.
• Coding based on best performance rather than usual performance: Section GG is designed to capture the resident’s usual functional ability, not their best day or their worst day. Coding based on an atypical observation, either too high or too low, distorts the classification.
• Missing assessments within the look-back period: If functional assessments are not completed within the assessment reference date window, the MDS coordinator lacks the data needed to code accurately and may default to conservative coding that underrepresents the resident’s actual acuity.
• Failure to capture admission performance: The admission assessment (PDPM requires a 5-day assessment) must capture functional status during the initial days of the stay. Late or incomplete admission assessments create permanent data gaps that cannot be retroactively corrected.

Section I: Active Diagnoses

Active diagnoses directly affect the PDPM clinical category, which determines the nursing and non-therapy ancillary (NTA) components of the per-diem rate. Missing a qualifying diagnosis can shift a resident from a higher-paying clinical category to a lower one. The most frequent errors involve diagnoses that are present in the hospital discharge summary but not reconciled into the facility’s active problem list, and diagnoses that are clinically active but not documented with sufficient specificity to support MDS coding.

Section J: Health Conditions

Section J captures pain assessments, fall history, and other health conditions that affect the PDPM classification. Facilities frequently under-code pain because the standardized pain assessment tool was not administered during the look-back period, or because nursing documentation describes pain in narrative terms that do not translate directly to the MDS coding structure.

“The MDS is only as accurate as the clinical documentation that supports it. No amount of coding expertise can compensate for assessments that were not completed, observations that were not recorded, or diagnoses that were not reconciled.”

AI-Assisted Auto-Population: Eliminating the Manual Bottleneck

In traditional MDS workflows, the MDS coordinator manually reviews nursing assessments, therapy evaluations, physician notes, and lab results to extract the data needed for each MDS item. This process is time-intensive, error-prone, and entirely dependent on the completeness and accessibility of the underlying clinical documentation.

AI-assisted auto-population transforms this workflow by automatically mapping clinical documentation to MDS items as the documentation is created. When a nurse documents a functional assessment, the system identifies which Section GG items are supported by the observation and pre-populates the corresponding MDS fields. When a physician confirms a diagnosis, the system maps it to the appropriate Section I category. When a pain assessment is completed, the results flow directly into Section J.

This is not a replacement for MDS coordinator expertise. The coordinator still reviews, validates, and finalizes every assessment. But instead of spending hours extracting data from disparate clinical documents, they spend their time on clinical judgment: evaluating whether the auto-populated data accurately reflects the resident’s status, identifying areas where additional clinical documentation would support a more accurate code, and ensuring that the completed MDS tells a clinically coherent story.

Validation Rules: Catching Errors Before Submission

Even with AI-assisted auto-population, MDS errors can occur. Validation rules provide a second layer of protection by analyzing the completed MDS for internal consistency, clinical plausibility, and alignment with supporting documentation.

Effective validation includes:

• Cross-section consistency checks: The system compares data across MDS sections to identify conflicts. A resident coded as independent in mobility (Section GG) but receiving skilled physical therapy (Section O) triggers a validation alert. A resident with no active pain diagnosis (Section I) but a high pain score (Section J) prompts review.
• Documentation sufficiency alerts: Before the MDS is finalized, the system evaluates whether the clinical record contains sufficient documentation to support each coded item. Sections where documentation is thin or ambiguous are flagged for the MDS coordinator’s attention.
• PDPM impact analysis: The system calculates the PDPM classification and per-diem rate that will result from the current coding and highlights items where a coding change would shift the classification. This is not about upcoding. It is about ensuring that accurate coding captures the reimbursement the resident’s acuity justifies.
• Historical comparison: For residents with prior MDS assessments, the system compares current coding against previous assessments and flags significant changes that may indicate a coding inconsistency or a genuine change in condition that should be documented.

Section GG Accuracy: The PDPM Revenue Driver

Because Section GG directly drives three of the five PDPM components, its accuracy has a disproportionate impact on reimbursement. Improving Section GG accuracy requires addressing the assessment process itself, not just the coding.

Facilities that achieve the highest Section GG accuracy share common practices: they use standardized functional assessment tools that align directly with the Section GG scoring criteria, they train all disciplines to use consistent scoring definitions, they complete functional assessments at multiple time points during the assessment window to capture usual performance rather than a single observation, and they use their EHR to track and reconcile functional scores across disciplines before the MDS is finalized.

An AI-enabled EHR supports this by presenting functional assessment workflows that mirror the Section GG structure, automatically flagging discrepancies between nursing and therapy functional assessments, and maintaining a functional status timeline that shows the trajectory of the resident’s abilities across the assessment period. The MDS coordinator can review this timeline and code Section GG based on a comprehensive view of the resident’s usual performance rather than relying on a single assessment point.

From Accurate MDS to Accurate Reimbursement

The connection between MDS accuracy and financial performance is direct and measurable. Facilities that invest in MDS accuracy through better assessment processes, AI-assisted tools, and systematic validation consistently capture reimbursement that more accurately reflects their resident population’s acuity. They experience fewer claim denials related to MDS discrepancies. They face lower audit risk because their coding is supported by robust clinical documentation. And they maintain quality measure profiles that accurately represent their clinical performance, supporting stronger Five Star ratings and referral relationships.

The path to MDS accuracy is not about aggressive coding. It is about ensuring that the clinical documentation that underlies every MDS assessment is complete, timely, and specific enough to support accurate coding. When the documentation is right, the MDS coding follows naturally, and the reimbursement follows the coding.