The Centers for Medicare and Medicaid Services (CMS) two midnight rule: policy at odds with reality

Introduction: To reduce costs, the Centers for Medicare and Medicaid Services (CMS) implemented new policies governing which patients are automatically admitted as inpatients (staying greater than “two midnights”) and which require additional justification with physician documentation to be admitted. This study examines procedures missing from the Medicare Inpatient Only (MIO) list and uses national data to evaluate its appropriateness. Methods: Non-MIO procedures were identified from the current MIO list. Utilizing relevant billing codes, procedures were queried in the National Surgery Quality Improvement Program database for length of stay (LOS), percentage requiring >2 day stay, and inpatient status from 2005 to 2012. In addition, a separate analysis was performed for patients 65 years old or older who would qualify for Medicare. Results: A majority of patients stayed more than 2 days for several procedures not included on the MIO list (% staying >2 days, mean LOS), including component separation (79.1 %, 5.9 ± 12.3 days), diagnostic laparoscopy (64.2 %, 5.5 ± 11.9 days), laparoscopic splenectomy (60.0 %, 9.0 ± 13.6 days), open recurrent ventral hernia repair (58.2 %, 6.3 ± 9.0 days), laparoscopic esophageal surgery (46.4 %, 5.3 ± 13.3 days), and laparoscopic ventral hernia repair (24.7 %, 2.5 ± 8.8 days). In patients ≥65 years, the average LOS was longer than the general population; for example, 40.2 % of laparoscopic appendectomies and 38.7 % of laparoscopic cholecystectomies in this older group required more than two nights in the hospital. In 92.3 % of procedures examined, patients ≥65 years required greater than two nights in the hospital with an average LOS of 2.5–10.7 days. Conclusion: Commonly encountered non-MIO surgical procedures have national precedents for inpatient status. Before enacting policy, CMS and other regulatory bodies should consider current data to ensure rules are evidence-based and applicable.