CMS' skilled nursing facility 30-day all-cause readmission measure (SNFRM) Design and analyses
Approximately 20% of Medicare beneficiaries who enter a hospital are readmitted within 30 days, 90% on an emergent or unplanned basis (Jencks, Williams & Coleman, 2009). This results in Medicare expenditures exceeding $17.4 billion annually. Furthermore, hospital readmissions impact patients’ care transition by interrupting therapy/care plan, escalating discomfort and increasing the risk of hospital-acquired adverse events. Under contract with CMS, RTI designed a Medicare claims-based SNF hospital readmission measure (SNFRM) using 2009 MedPAR claims (n>2.5 million). The SNFRM is defined as the percentage of patients admitted to a SNF who experience an all-cause, unplanned, hospital readmission within 30 days of discharge from their prior proximal hospitalization. The SNFRM is harmonized with CMS’ hospital-wide readmission measure used in Medicare’s Hospital Readmissions Reduction Program under the Affordable Care Act. In this session we detail SNFRM specifications including numerator, denominator, and exclusions. We describe risk adjustment analyses based on hierarchical logistic regression and detail seven cohort models developed using the Agency for Healthcare Research and Quality’s Clinical Classification System. Findings indicate cohort modeling did not improve efficiency, evidenced by the C-statistic for non-cohort model (0.67) equal or greater to the C-statistic from any cohort-specific model. Furthermore, the unadjusted mean readmission rate among SNFs with at least 25 index stays was 22.2% (SD 7.6%) compared to the adjusted mean readmission rate of 23.2 (SD of 2.9%). In conclusion, we discuss measure harmonization with Long Term Care Hospitals and Inpatient Rehabilitation Facilities and explore implications for future SNF quality public reporting and payment policy.
Smith, L., West, S., Ingber, M., Feng, Z., & Reilly, K. (2013). CMS' skilled nursing facility 30-day all-cause readmission measure (SNFRM): Design and analyses. Gerontologist, 53(Suppl. 1), 579-580.