Correcting Section GG Scoring to Restore Accurate PDGM Functional Levels
Home health agency (~640 active patients) · Midwest U.S.
76% → 41%
Periods at Low Functional Level
+8%
Avg Case-Mix Weight
94%
Score-to-Narrative Consistency
Both ways
Corrections Applied
An agency whose caseload was scoring almost entirely at the low functional impairment level was recording capability instead of usual performance. Retraining and a documentation-consistency check restored accurate functional levels — and the case-mix weight that came with them.
The Challenge
A Midwest home health agency serving roughly 640 active patients came to us with a distribution that did not match its own clinicians' description of its caseload. More than three-quarters of its 30-day periods were landing at the low functional impairment level under PDGM, yet its nursing and therapy staff consistently described a demanding, high-acuity population — many patients living alone, several with recent hospitalizations and mobility limitations. The agency assumed the problem was its case-mix mix or its referral sources. It was neither. When we reviewed a sample of completed OASIS assessments against the corresponding visit documentation, the pattern was immediate and consistent: functional items were being scored on what patients could do rather than what they usually did. Clinicians — most of them therapy-trained — were recording best observed performance, because that is how rehabilitation goals are framed. Several charts showed a patient scored as needing only supervision for transfers while the same visit's narrative described the clinician physically steadying them. The assessments were not fraudulent; they were answering a different question than the one Section GG asks. Two things made the error invisible to the agency. First, it was systematic rather than random — every misread pushed in the same direction, toward independence, so it never averaged out across the caseload the way a scattered mistake would. Second, nothing in the agency's workflow ever compared the functional score against the clinical narrative, so the contradiction sat inside individual charts without anyone reading them side by side. The agency had no internal signal that anything was wrong, and the understatement compounded quietly across every period it billed.
Our Solution
We treated this as a process failure rather than a training gap alone, and rebuilt functional scoring in three layers. First, we retrained clinical staff on the usual-performance standard using the agency's own de-identified charts rather than generic examples — showing clinicians the specific cases where their score and their narrative disagreed. Seeing the contradiction in their own documentation did more than any slide deck. Second, we added a documentation-consistency check before submission. A certified reviewer compared every functional item against the visit narrative, and any item where the score and the story disagreed went back for clarification rather than being corrected upward on assumption. Where the record was genuinely ambiguous, a query closed the gap before the assessment was locked. Third, we established a monthly review of functional impairment level distribution, so drift would surface as a trend rather than being discovered a year later. Crucially, the objective throughout was accuracy in both directions — the same review flagged a small number of periods that had been scored as more impaired than the documentation supported, and those were corrected downward. That symmetry is what makes the result defensible under medical review rather than merely favourable. The distribution moved within two assessment cycles and then held, which mattered more than the initial correction: a one-time cleanup would have drifted straight back once the reviewers stopped watching. Because the fix lived in the submission workflow rather than in a training memo, new clinicians inherited the standard automatically. The agency also gained something it had not asked for — a functional-level distribution it could actually explain to a reviewer, backed by visit documentation that agreed with the assessment it accompanied.