We identified the study populations that were entered into the National Trauma Data Bank (NTDB version 7.1). This NTDB version contained data from over 900 United States trauma centers. There were over 2.7 million patient records from the years 2002 to 2006 that were entered into the NTDB. The NTDB contains patient data compiled from medical records before, during, and after admission. This information is submitted to the American College of Surgeons for quality control and maintenance.
The NTDB reports all injuries by using the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnostic codes. The data was imported and merged into a single dataset from the 13 NTDB files using SAS version 9.2 (SAS Institute, Cary, NC). Abbreviated Injury Scale (AIS) scores for lower extremities were linked to a pelvic fracture code (see later). Linking to the pelvic fracture code ensured that AIS values were recorded for only pelvic injuries.
The initial NTDB population consisted of over 2.7 million entries. For the purpose of this study, all burn or penetrating injuries were excluded reducing the group to1.7 million patients. If a patient's multiple ICD-9 DCODE (diagnosis codes) entries contained at least one of the following codes, the patient was considered to have a pelvic fracture: 808.2, 808.3, 808.4, 808.41, 808.42, 808.43, 808.49, 808.5, 808.51, 808.52, 808.53, 808.59, 808.8, and 808.9. Acetabular fractures were excluded from the study. If the ICD-9 codes were 808.2, 808.41, 808.42, 808.43, 808.49, or 808.8, the pelvic fractures were closed pubis, ilium, ischium, multiple, other, or unspecified, respectively. If the ICD-9 codes were 808.3, 808.51, 808.52, 808.53, 808.59, or 808.9, the pelvic fractures were open pubis, ilium, ischium, multiple, other, or unspecified, respectively. Those entries without a pelvic fracture were removed, giving a total of 54,459 patients with pelvic fractures.
In order to analyze the elderly population and compare it to the adult population, children aged 17 years and younger were excluded yielding a study population of 45,081 patients. This final study population was subdivided into octogenarian, defined as all patients older than 79 years, elderly, defined as ages 65 to 79 years old, and adults, defined to be ages 18 to 64 years old. This subdivision was performed because of many previous studies that evaluated strength, performance, and outcomes in groups younger or older than 80 years old [10
The main outcomes of interest were mortality and severe complication. A severe complication was defined as having renal failure, acute respiratory distress syndrome (ARDS), or a pulmonary embolism recorded in the NTDB. Analysis of prehospital risk factors was considered to determine the association with the main outcome variables of mortality and severe complications. These risk factors included sex, race, age, arrival in shock (systolic blood pressure less than 90
mmHg), head injury (a positive head computed tomography), injury severity score (ISS), and mechanism of injury.
2.1. Statistical Analysis
All statistical analyses were conducted using SAS. Descriptive statistics were performed on the entire study population. To determine associations between risk factors and the main outcomes of interest, bivariate analysis was conducted between the main end points and each pre-hospital risk factor. Continuous covariates were dichotomized. Risk factors that were not already dichotomous were transformed into two mutually exclusive categories. For example, ISS became greater than or equal to 16 and less than 16. For this assessment, open fractures and unknown mechanisms of injury were excluded, yielding a study group of 31,475.
A subpopulation of severe pelvic injuries was also created to determine if any octogenarian subgroup has better outcomes. For this population, an AIS of less than 3 and all patients with lower extremity fractures besides pelvic fractures were excluded. This created a study subgroup of 3,101 patients with isolated severe pelvic fractures.
For both the main study population and the subgroup of severe pelvic injury, prehospital risk factors were determined to be significant using the Mantel-Haenszel test. This method produced crude odds ratios and 95% confidence intervals, and those variables of significance were included for multivariate analysis. Mechanism of injury was not included in the final multivariate analysis due to the inconsistency and unreliability of the data.
Logistic regression analyses including those variables deemed to be of significance were performed to determine the association between the prehospital risk factors and the two main outcomes (death and severe complication). In order to assess the importance of age and outcome after sustaining a pelvic fracture, each model included a specific age group (octogenarian or elderly) as compared to adults as referent. The Hosmer-Lemeshow goodness-of-fit test was performed on each model to determine whether or not the observed event rates matched the expectant event rates. The P value must be greater than 0.05 to indicate a good fit.