Too Sick to be True? Evaluating Potentially Problematic Diagnosis Coding Practices in Medicare's Patient-Driven...

To use a quasi-experimental design to quantify changes in skilled nursing facility (SNF) diagnosis documentation associated with Medicare's Patient-Driven Payment Model (PDPM). PDPM aims to promote patient-centered care in skilled nursing facilities (SNFs) by matching reimbursement to patient characteristics, including clinical complexity, which is captured in part through documentation of diagnoses.

We used a difference-in-differences design to estimate PDPM's effects on SNF diagnosis documentation, including the number of diagnoses and clinical complexity scores via the Elixhauser comorbidity index. Hospital claims served as a non-equivalent dependent variable control. Triple interaction terms in fixed effect linear models assessed variation by SNF profit status. Changes in the probability of recording five documentation-sensitive conditions were estimated via marginal effects from generalized linear models.

Secondary analysis of 100% Traditional Medicare claims (2018-2021), comprising over 4.8 million hospital-to-SNF episodes.

Compared against hospital claims from hospital-SNF episodes, PDPM announcement was associated with 0.83 additional diagnoses on SNF claims, representing a relative increase of 7.1%. Similarly, Elixhauser scores increased by 0.88 points (relative 13.6%). We observed significant variation by profit status; when accounting for anticipatory behavior, profit status was associated with an additional relative 2.8% in diagnoses and 4% in Elixhauser points. PDPM was also associated with increased probability of documenting all five documentation-sensitive conditions: 3.9 percentage points (pp) for chronic pulmonary disease, 5.0 pp for complicated diabetes, 2.8 pp for heart failure, 7.3 pp for obesity, and 9.8 pp for weight loss (all reported p < 0.001).

PDPM was associated with increased coding intensity across multiple measures-and more so in for-profit SNFs-highlighting the need to further evaluate whether SNFs are accurately documenting or falsely inflating clinical complexity. Sustaining Medicare's payment accuracy will require continued monitoring of diagnosis coding behavior and its alignment with actual clinical complexity.