Restricting Claims Data to 25 ICD-10-CM Codes Leads to Underestimation of Elixhauser Comorbidities

The Centers for Medicare and Medicaid Services (CMS) inpatient claims data include up to 25 International Classification of Diseases Clinical Modification (ICD-10-CM) diagnosis codes per encounter, which are relied upon for comparing hospitals' risk-adjusted outcomes. Because hospital billing frequently includes more than 25 codes, we aimed to assess the extent of underestimation of Elixhauser comorbidities in risk-adjusted hospital outcomes.

A retrospective cohort study of all Medicare inpatient encounters across our health system from January 2019 to December 2025. We calculated number of codes per encounter and prevalence of Elixhauser comorbidities. We compared Elixhauser prevalence when limiting to 25 codes versus collecting all codes using chi-squared and Kruskal-Wallis tests and calculated the percent of Elixhausers that go uncaptured by the 25-code restriction. We assessed variation ICD-10 capture across all US hospitals using the 2024 CMS inpatient claims data, using codes in the 25th position as a proxy for percentage of encounters potentially having more than 25 codes.

148,929/397,301 (37.5%) inpatient encounters had more than 25 ICD-10-CM codes in our health system. 16/38 (42.1%) Elixhausers were significantly more prevalent when using all codes. There were 25 ICD-10 codes documented in 1,470,064/8,096,044 (18.2%) Medicare inpatient encounters across all US hospitals.

More than 1/3 of inpatient encounters at our institution, and as many as 1/6 across the United States, have ICD-10-CM codes which are not included in CMS risk-adjustment models due to the limit of 25 ICD-10-CM codes per claim. CMS should expand claims data to include all ICD-10-CM codes.