Survival After Surgical Treatment of Axis Fractures in Elderly

The Swedish National Patient Registry (NPR) is hosted by the Swedish Agency of Health and Welfare and contains all patient contacts within Sweden with a coverage of >90% for orthopaedic diagnoses. Registered are main diagnosis and co-morbidity using the ICD-9 code until 1996 and from 1997 onward ICD-10 codes. Treatment is coded since 1996 using the Swedish classification of surgical procedures. Furthermore, information on hospitalisation time is available from the registry.

In the Swedish Cause of Death Registry (CDR) all deaths in Sweden are registered with date of death and cause of death. While date of death coverage is complete, the cause of death has only 46% agreement with the final hospital diagnosis.

All patients with the main diagnosis of C2 fracture (ICD-10: S12.1) treated between January 1st 1996 to December 31st 2014 are extracted from the NPR and merged with the CDR for incident deaths. Prior to data transmission, the Swedish Agency of Health and Welfare anonymised the individual personal identification numbers using a key which remained with the Agency.

Invalid age data will be excluded either due to missing entries or mismatch with date of death. Multiple admissions (cases with same identification number but more than 12 months between admissions) and duplicate entries (cases with same identification number) will be removed from the dataset, after valuable co-morbidity data from the duplicate entries were stored in the original record. Duplicate entries originate from separate recordings from each hospital if the patient is referred to a specialised hospital. Furthermore, primarily non-surgically treated cases with a change of treatment modality are registered as surgical patients. The registered hospitalisation time for these patients is the combined non-surgical and surgical in-hospital treatment period.

From each record the Charlson Comorbidity Index (CCI) is calculated. Using Swedish surgical procedure codes for spinal fusion ("NAG") and spinal osteosynthesis ("NAJ") patients receiving surgical treatment were marked, and the cohort divided into two groups receiving non-surgical or surgical treatment.

The age distribution differences of patients with C2-fractures treated surgically and non-surgically was visualised with a density distribution plot. A logit logistic regression analysis identified covariates of surgical treatment assignment and was presented with 95% confidence intervals (C.I.) and statistical probability p. Goodness-of-fit of the model was presented with pseudo-r2 according to McFadden.

Using the Kaplan-Meier method for non-surgical and surgical treatment mean survival, as well as survival rate at 3 months, 1 year, and 2 years was determined, and visualised using a Kaplan-Meier plot with 99% C.I. With the Cox proportional hazards regression method multiple covariates contributing to survival are entered and hazard ratio (HR) presented with 95% C.I. and probability p. As relevant covariates for survival besides surgical treatment age, gender, spinal cord injury, and CCI were identified. Since the assumption was that the medical advancement improved patient survival in general during the last decades, even year of admission was included as covariate in the model. In order to identify a cut-off age where a treatment modality is not associated anymore with greater survival the adjusted HR of a certain age and above was plotted over age with 95% C.I.