Model System:


Reference Type:


Accession No.:


Archives of Physical Medicine and Rehabilitation

Year, Volume, Issue, Page(s):

, Epub, ,


Objective: To examine whether commonly used comorbidity indexes (Deyo-Charlson comorbidity index, Elixhauser comorbidity index, the Centers for Medicare and Medicaid Services [CMS] comorbidity tiers) capture comorbidities in the acute traumatic and nontraumatic SCI inpatient rehabilitation population.

Design: Retrospective cross-sectional study.

Participants: Data were obtained from the Uniform Data System for Medical Rehabilitation from October 1, 2015 to December 31, 2017 for adults with spinal cord injury (SCI) (Medicare-established Impairment Group Codes 04.110-04.230, 14.1, 14.3). This study included SCI discharges (N=66,235) from 833 inpatient rehabilitation facilities.

Main Outcome Measures: International Classification of Diseases–10th Revision–Clinical Modifications (ICD-10-CM) codes were used to assess 3 comorbidity indexes (Deyo-Charlson comorbidity index, Elixhauser comorbidity index, CMS comorbidity tiers). The comorbidity codes that occurred with >1% frequency were reported. The percentages of discharges for which no comorbidities were captured by each comorbidity index were calculated.

Results: Of the total study population, 39,285 (59.3%) were men and 11,476 (17.3%) were tetraplegic. The mean number of comorbidities was 14.7. There were 13,939 distinct ICD-10-CM comorbidity codes. There were 237 comorbidities that occurred with >1% frequency. The Deyo-Charlson comorbidity index, Elixhauser comorbidity index, and the CMS tiers did not capture comorbidities of 58.4% (95% confidence interval, 58.08%-58.84%), 29.4% (29.07%-29.76%), and 66.1% (65.73%-66.46%) of the discharges in our study, respectively, and 28.8% (28.42%-29.11%) of the discharges did not have any comorbidities captured by any of the comorbidity indexes.

Conclusion: Commonly used comorbidity indexes do not reflect the extent of comorbid disease in the SCI rehabilitation population. This work suggests that alternative measures may be needed to capture the complexity of this population.


Donna Huang, MD; Sameer Siddiqui MD; Chloe S.Slocum MD, MPH; Richard Goldstein, PhD; Ross D. Zafonte, DO; Jeffrey C. Schneider, MD

Participating Centers: