A Nationwide Analysis of Geriatric Odontoid Fracture Incidence, Complications, Mortality, and Cost

Ram Alluri, MD; Gabriel Bouz, MD; Samantha Solaru, BS; Hyunwoo Kang, MD; Jeffrey Wang, MD; Raymond J. Hah, MD


Spine. 2021;46(2):131-137. 

In This Article


The National Inpatient Sample (NIS) database is a national hospital database which is part of the Healthcare Cost and Utilization Project (HCUP). The NIS is the largest publicly available all-payer administrative database and samples approximately 8 million inpatient hospital stays annually in the United States.[24] This sample is designed to approximate 20% of United States community hospitals. Patient information such as demographic variables, hospital charges, mortality, discharge data, and International Classification of Disease, Ninth and Tenth Revision, Clinical Modification (ICD-9-CM and ICD-10 CM) diagnosis and procedure codes are available within the database.

Patients between ages 65 and 90 years with odontoid fractures who underwent operative or nonoperative management between the years 2003 and 2017 were included in this study (ICD codes listed in Appendix Table 1, http://links.lww.com/BRS/B651). Operative management included anterior osteosynthesis or posterior arthrodesis, while nonoperative management included application of a cervical collar or halo device (ICD codes listed in Appendix Table 2, http://links.lww.com/BRS/B651). Patients who received both a halo device or cervical collar and operative management were considered to only have received operative management, similar to a prior study.[5] We excluded patients who did not have an assigned ICD procedure code (operative or nonoperative) associated a diagnosed odontoid fracture. Additionally, patients with concomitant diagnoses of spinal cord injury, pathologic fractures, stress fractures, infection, and malunion or nonunion were excluded (ICD codes listed in Appendix Table 3, http://links.lww.com/BRS/B651). ICD-9-CM codes were used from 2003 through the third quarter of 2015. ICD-10-CM codes were used for the fourth quarter of 2015 through 2017.

Year of injury, demographic variables, comorbidities, inpatient complications, mortality, length of stay, inpatient cost, and hospital characteristics were analyzed. Specific inpatient complications assessed included pulmonary, cardiac, renal, infectious, neurologic, venothromboembolic events (VTE), dysphagia, and wound (ICD codes listed in Appendix Table 4, http://links.lww.com/BRS/B651). The primary outcome was to compare temporal trends in operative and nonoperative management. Secondary outcomes included comparison of comorbidities, inpatient complications, hospital characteristics, LOS, and cost between treatment groups.

Statistical analysis comparing the two treatment groups was performed using STATA 13.0 (StataCorp LP College Station, Texas, United States). Stratified sample weights provided by NIS were applied to raw data to create national estimates. Treatment trends over time were analyzed using linear regression, and the P-value of the slope from the line of best fit was calculated. Chi-square analyses were performed to compare categorical data between treatment groups and populations treated at different hospital types. For continuous or discrete variables, a Student t test was used. When assessing inpatient complications, odds ratios (OR) were calculated. A P-value ≤0.05 was considered significant.