Representativeness of the Spinal Cord Injury Model Systems National Database

Study design Secondary analysis of prospectively collected observational data. Objectives To assess the representativeness of the Spinal Cord Injury Model Systems National Database (SCIMS-NDB) of all adults aged 18 years or older receiving inpatient rehabilitation in the United States (US) for new onset traumatic spinal cord injury (TSCI). Setting Inpatient rehabilitation centers in the US. Methods We compared demographic, functional status, and injury characteristics (nine categorical variables comprising of 46 categories and two continuous variables) between the SCIMS-NDB (N = 5969) and UDS-PRO/eRehabData (N = 99,142) cases discharged from inpatient rehabilitation in 2000-2010. Results There are negligible differences (< 5%) between SCIMS-NDB patients and the population for 31 of the 48 comparisons. Minor differences (5-10%) exist for age categories, sex, race/ethnicity, marital status, FIM Motor score, and time from injury to rehabilitation admission. Important differences (> 10%) exist in mean age and preinjury occupational status; the SCIMS-NDB sample was younger and included a higher percentage of individuals who were employed (62.7 vs. 41.7%) and fewer who were retired (10.2 vs. 36.1%). Conclusions Adults in the SCIMS-NDB are largely representative of the population of adults receiving inpatient rehabilitation for new onset TSCI in the US. However, users of the SCIMS-NDB may need to adjust statistically for differences in age and preinjury occupational status to improve generalizability of findings.