Cervical Spine Fracture and Other Diagnoses Associated with Mortality in Hospitalized Ankylosing Spondylitis Patients

Abstract

Objective

Little data exist regarding mortality in Ankylosing Spondylitis (AS). We assessed diagnoses associated with in-hospital mortality in AS using a population-based inpatient dataset.

Methods

Data were abstracted from the Healthcare Cost and Utilization Project- Nationwide Inpatient Sample (HCUP-NIS) between 2007–2011. We identified AS admissions using International Classification of Diseases-9 Clinical Modification (ICD9-CM) code 720.0. In-hospital mortality was the primary outcome. Logistic regression was used to evaluate the association between top diagnoses and in-hospital mortality. We performed a secondary analysis from the same years in all patients (with and without AS) with cervical spine (c-spine) fracture to determine whether AS was an independent risk factor for mortality.

Results

Between 2007–2011 we identified 12,484 admissions and 267 deaths in AS patients. C-spine fracture with spinal cord injury and sepsis had the highest odds (OR) of death, 13.43 (95% Confidence Interval (CI) 8.00–22.55; p<0.0001) and 7.63 (95% CI 5.62–10.36; p<0.0001) respectively. In the same time period, there were 53,606 C-spine fracture admissions of which 408 were coded with AS. Among all c-spine fracture hospitalizations, an AS diagnosis was associated with in-patient death (OR 1.61; 95% CI 1.16–2.22; p=0.004).

Conclusion

In AS patients admitted to the hospital, c-spine fracture is a leading cause of in-hospital mortality. Other diagnoses associated with mortality include sepsis, pneumonia, CVD and comorbid illnesses. Among all hospitalizations with c-spine fracture, AS was associated with an increased odds of death. C-spine fracture-associated mortality warrants further study to elucidate risk factors in order to prevent this devastating event in AS patients.

INTRODUCTION

Ankylosing Spondylitis (AS) is a chronic inflammatory disease of the sacroiliac joints and spine. Extra-articular manifestations and comorbidities are common and include uveitis, inflammatory bowel disease, and psoriasis. Other complications include cardiovascular disease, osteoporosis and fracture. Studying mortality in AS is difficult given the low disease prevalence and because the disease is not usually the primary cause of death.

During the radiation treatment era for AS, Weiss et al. demonstrated that radiation-treated patients had an increased risk of death attributed to cancer, notably leukemia. Patients who did not receive radiation therapy, however, had mortality rates similar to the general population (1). The majority of studies in the post-radiation treatment era have shown increased mortality rates in AS when compared to the general population with standardized mortality ratios between 1.32–3.07 (2–7). Using hazard ratios (HR), a study of Swedish AS patients found increased mortality compared to controls HR 1.60 (95% CI 1.44–1.77) (8). The etiologies of this excess mortality are not uniform, though commonalities include cardiovascular and cerebrovascular disease, pulmonary disease, infections and malignancy (2–5,8,9). There are few contemporaneous studies on mortality in AS, and no population-based data in the United States. In addition, most studies have been small, and were published in the pre-Tumor Necrosis Factor inhibitor (TNFI) era.

Vertebral fracture is a known complication of AS carrying an estimated lifetime risk of 14% based on a study surveying patients with AS (10). A review in the spinal trauma literature found a mortality rate of 6.4–11.3% in AS patients admitted to the hospital with spinal fracture (11). In a study of c-spine fractures in the general in-patient population, mortality was found to be 5.3% using the same HCUP-NIS dataset over the years 2002–2010 (12).

Because AS flares rarely require hospital admission and the disease is almost entirely managed in the ambulatory setting, little focus has been placed on studying hospitalized AS patients and there is limited data on outcomes in this population. Hospitalization data, and specifically diagnoses associated with mortality, may elucidate disease-associated outcomes in AS. Identifying conditions which are unique to the AS population may provide the foundation for further studies and an improved focus on risk factor modification to prevent adverse outcomes.

Our objective was to evaluate diagnoses associated with death in hospitalized AS patients using a large population-based dataset. As c-spine fracture was found to have a strong association with in-patient mortality, we secondarily aimed to study whether a comorbid diagnosis of AS increased the risk of death in all patients hospitalized with c-spine fracture.

PATIENTS AND METHODS

Data were derived from the Nationwide Inpatient Sample (NIS), part of the Healthcare Cost and Utilization Project (HCUP) sponsored by the Agency for Healthcare Research and Quality (AHRQ). HCUP-NIS is the largest publicly available inpatient healthcare database in the United States (US). It contains data on approximately eight million hospitalizations each year from nearly 1,000 hospitals, and is a 20% representative sample of all US hospital discharges (13). Each observation in the NIS represents a single de-identified hospital discharge record with information including patient demographics, a primary discharge diagnosis and up to 24 secondary diagnoses, whether the hospitalization was elective or urgent, and whether the patient died during the hospitalization. Diagnoses are based on International Classification of Diseases-9 Clinical Modification (ICD9-CM) codes. Clinical Classification Software (CCS) codes were developed by the AHRQ to group the greater than 14,000 unique ICD9-CM codes into 296 clinically meaningful diagnostic categories (Example in supplementary table S1) (14). CCS codes have been used in the study of other rheumatic disease states when analyzing the HCUP-NIS dataset (15,16).

The Committee on Human Research at the University of California, San Francisco determined that this study was exempt from review as the dataset was de-identified.

Primary Analysis

Identification of Ankylosing Spondylitis Hospitalizations

We identified admissions with AS ICD9-CM code 720.0 listed as a discharge diagnosis between the years 2007–2011. This ICD9-CM code, 720.0, has been previously validated for use in administrative data using a Veteran Health Administration dataset (17). It has also been used in the study of AS patients in several other administrative datasets (18,19). Patients under 18 years of age or those who were dual-coded for systemic lupus erythematosus (710.0) or rheumatoid arthritis (714.0) were excluded to improve specificity. We excluded hospitalizations where data on age and the variable describing in-hospital death were not coded (Figure 1).

Figure 1.

Flow chart depicting inclusion and exclusion criteria and characteristics of the primary and secondary analyses.

Variables

We used CCS codes to identify primary diagnostic categories associated with mortality in AS patients. The primary CCS code is based on the “principal diagnosis” which is the first ICD9-CM code listed on each discharge summary. We identified the top 5 primary diagnostic CCS codes associated with death and created corresponding variables based on validated ICD9-CM codes (Supplementary Table S2) (9,17,19–23). Statistical modeling was based on variables using all discharge ICD9-CM codes rather than CCS codes as ICD9-CM codes have increased specificity for each diagnosis.

Two of the top five CCS codes were “spinal cord injury” and “other fractures” which were broad categories largely driven by spinal fracture. The CCS code “other fractures” contained ICD9-CM codes for non-spinal fractures that were not associated with mortality in the univariate analysis and therefore were dropped from our analysis. To represent the CCS categories “spinal cord injury” and “other fractures” we created variables based on previously validated definitions- spine fractures by level (cervical, thoracic and lumbar spine fracture) with and without spinal cord injury (20).

We did not include “respiratory failure” in our multivariable model as it only included ICD-9CM codes for acute respiratory failure and ventilator dependence, a final common pathway to mortality. ICD-9CM codes for restrictive lung disease and pulmonary fibrosis, diseases that have been associated with AS, were not represented as a top diagnosis in patients who died. Cardiovascular disease (CVD) was included in the multivariable model even though it was not one of the top categories associated with in-patient mortality in AS patients, as it has been suggested to be a top cause of mortality in AS in the literature (8,9,18).

Measures

Outcome

In-hospital mortality. We assessed diagnoses associated with AS patients who died during their hospitalization.

Independent Variables

Demographic and hospitalization characteristics included age, gender, race (White, Black, Hispanic, Asian and other) and whether or not admission was elective. Diagnoses included spinal fracture by level (cervical, thoracic, lumbar) with and without spinal cord injury, sepsis, pneumonia and CVD (ischemic heart disease, atherosclerosis, peripheral vascular disease, congestive heart failure and cerebrovascular disease). To address the influence of comorbidity on mortality in AS, we used the Charlson Comorbidity Index (CCI). This is a validated measure which enables prediction of 1-year mortality based on composite comorbid conditions (24). We created a Modified CCI for our multivariable analysis by removing CVD variables so that CVD could be analyzed as an independent variable (25). AS is not included in the CCI.

Statistical Analysis

Because the HCUP-NIS does not provide information on readmissions, we were unable to perform analyses that would accommodate for this within person variability. We therefore examined only characteristics that were unique to each hospitalization such as age, hospital diagnoses and whether the hospitalization was elective or not. Static characteristics such as race were only used to describe the cohort.

We examined the association of each variable with in-hospital mortality using univariate logistic regression. Variables significant at the p=0.05 level in univariate analyses were included in a multivariable logistic regression model to identify independent associations with in-hospital mortality. Data were analyzed using Stata/SE 13.0 (StataCorp LP, College Station, TX).

Secondary analysis

Because c-spine fracture was associated with mortality in the primary analysis, we performed a secondary analysis over the same period comparing c-spine fractures in patients with and without AS to determine if AS was an independent risk for in-patient mortality. We identified all admissions coded for c-spine fracture (ICD-9 CM 805.0, 805.1, 806.0, 806.1) (20). We used the same exclusions as the primary analysis (Figure 1).

Identification of Variables

To aid in describing our secondary dataset, we sought to determine etiologies of c-spine fracture. External cause-of-injury codes (E-codes) were used to identify mechanisms of injury for each admission. E-codes are present in over 90% of discharge records in the years 2008–2011 (26). E-codes describing different mechanisms of falls and motor vehicle accidents (MVA) comprised the top ten primary E-codes associated with c-spine fracture in this dataset. Therefore we classified E-codes into three categories: fall, MVA and other. The “other” category also included mechanisms for which an E-code was not listed. The categories were based on the Centers for Disease Control and Prevention (CDC) matrix of E-code groupings (21). These were used only to describe the dataset and were not included in the multivariable model as they were not direct causes of mortality.

Similar to our primary analysis, we found the top 5 primary diagnoses based on CCS codes for patients who died with cervical spine fracture. These included intracranial injury, other fractures, spinal cord injury and crushing or internal injury. We created validated ICD9-CM based codes for intracranial injury (Supplementary Table S1). The “crushing or internal injury” category was heavily weighted by patients coded for internal injury; we therefore created a validated variable “internal injury” to represent this category (20). We controlled for comorbid conditions using the CCI (24). We also evaluated elective admission, as it was likely to be protective in hospitalized patients with cervical spine fracture.

We similarly performed univariate analyses followed by a multivariate logistic regression analysis to determine which variables were significantly associated with in-hospital mortality in c-spine fracture patients and whether AS was an independently associated diagnosis.

RESULTS

Primary analysis

There were 12,484 AS hospital admissions with 2007–2011 of which 267 patients died during the hospitalization. The characteristics of the sample are presented in Table 1. The mean age of AS patients admitted was 59.2 ± 16.4 years and 71% were males. Elective admissions accounted for 24% of hospitalizations. The majority of the sample’s race was White or “Other.” Other race was weighted by admissions where race was not coded. Race did not differ significantly between patients who died and those who survived (p=0.485). The mean age of those who died was 72.8±12.9 years, 78% were males and 10% were electively admitted. The mean Charlson Comorbidity Index (CCI) of patients who survived was 0.89±1.3 whereas the mean CCI for patients who died was 1.78±2.1 (p<0.0001).

Table 1.

Characteristics and diagnoses in hospitalized ankylosing spondylitis patients comparing those that survived and died between 2007–2011.

	Survived (n=12217) n(%)	Died (n=267) n(%)
Age (years)	58.9 ± 16.4	72.8 ± 12.9
Male Gender	8630 (71%)	208 (78%)
Race
White	8543 (70%)	192 (72%)
Black	579 (5%)	13 (5%)
Hispanic	516 (4%)	12 (4%)
Other	2579 (21%)	50 (19%)
Elective admission	2950 (24%)	26 (10%)
Modified CCI1	0.89 ± 1.3	1.78 ± 2.1
Sepsis	478 (4%)	81 (30%)
Pneumonia	995 (8%)	71 (27%)
Cardiovascular disease	3851 (32%)	147 (55%)
Cervical spine fracture	364 (3%)	44 (16%)
With spinal cord injury	84 (7%)	26 (10%)
Without spinal cord injury	280 (2%)	18 (7%)
Thoracic spine fracture	345 (3%)	18 (7%)
Lumbar spine fracture	139 (1%)	8 (3%)

¹Modified Charlson Comorbidity Index per one point scale (0–38), CVD variables removed

^*Percentages will not equal 100% as patients may have more than one diagnosis.

The leading CCS-based principal diagnoses associated with AS in-hospital mortality are presented in Table 2. Sepsis and spinal cord injury were the top two primary diagnostic categories in patients who died followed by other fractures, respiratory failure and pneumonia. Although the “other fractures” category was designed to represent fractures of the spine and other bones of the trunk, all patients who died with this primary CCS code had vertebral fractures.

Table 2.

Principal diagnoses associated with AS in-hospital mortality.¹

CCS2 definitions	Died n = 267 (%)
Sepsis	37 (14%)
Spinal cord injury	26 (9%)
Other fractures3	25 (9%)
Respiratory failure	17 (6%)
Pneumonia	16 (6%)

¹Based on Clinical classification software codes

²Clinical classification software

³Other fractures include ICD9 codes for vertebral fractures and other bones of the trunk

ICD-9 code-based diagnoses associated with mortality and their frequencies are presented in Table 1. Of those who died, 30% were coded as having sepsis, 27% with pneumonia, 55% with CVD, and 16% with c-spine fracture. Of patients with c-spine fracture, 11% died, and of those with concomitant spinal cord injury 24% died. Though a diagnosis of CVD was coded in the majority of hospitalized patients in the cohort, only 4% of those coded with CVD died, thus it was a common comorbidity, but a rare cause of death in this population.

After adjustment for other covariates, c-spine fracture with spinal cord injury and sepsis were the diagnoses with the highest odds of death, 13.43 (95% CI 8.00–22.55; p-value <0.0001) and 7.63 (95% CI 5.62–10.36; p-value <0.0001) respectively (Table 3). C-spine fracture without spinal cord injury also remained significant (OR 2.88; 95% CI 1.67–4.95; p-value <0.0001), as did pneumonia (OR 1.94; 95% CI 1.42–2.65; p-value <0.0001). Despite the absence of CVD as a top CCS diagnostic code, it was independently associated with mortality in the adjusted model (OR 1.33; 95% CI 1.01–1.74; p-value 0.041), but associated with a lower odds ratio, as expected, compared to the diagnoses derived from the top CCS categories. Age by decade and the modified CCI were both independently associated with mortality: OR 1.61 (95% CI 1.46–1.79; p-value <0.0001) and 1.23 (95% CI 1.15–1.32; p-value <0.0001) respectively. As expected, elective admission was protective (OR 0.58; 95% CI 0.38–0.89; p-value 0.012).

Table 3.

Factors associated with in-patient mortality in hospitalized AS patients¹.

	Univariable Analysis		Multivariable Analysis
	OR (95% CI)	p-value	OR (95% CI)	p-value
Age by decade	1.83 (1.68–2.01)	<0.0001	1.61 (1.46–1.79)	<0.0001
Male Gender	1.46 (1.09–1.96)	0.010	0.95 (0.70–1.30)	0.762
Race
White	ref	ref	–	–
Black	1.00 (0.57–1.76)	0.997	–	–
Hispanic	1.03 (0.57–1.87)	0.910	–	–
Other	0.86 (0.63–1.18)	0.357	–	–
Elective admission	0.34 (0.23–0.51)	0.024	0.58 (0.38–0.89)	0.012
Modified CCI2	1.35 (1.27–1.43)	<0.0001	1.23 (1.15–1.32)	<0.0001
Sepsis	10.70 (8.11–14.10)	<0.0001	7.63 (5.62–10.36)	<0.0001
Pneumonia	4.09 (3.09–5.40)	<0.0001	1.94 (1.42–2.65)	<0.0001
Cardiovascular disease3	2.66 (2.08–3.40)	<0.0001	1.33 (1.01–1.74)	0.041
Cervical spine fracture:	2.68 (1.30–5.53)	<0.0001	–	–
With spinal cord injury	15.58 (9.86–24.63)	<0.0001	13.43 (8.00–22.55)	<0.0001
Without spinal cord injury	3.08 (1.88–5.04)	<0.0001	2.88 (1.67–4.95)	<0.0001
Thoracic spine fracture	2.49 (1.52–4.06)	<0.0001	0.95 (0.54–1.67)	0.87
Lumbar spine fracture	2.68 (1.30–5.53)	0.007	1.98 (0.90–4.31)	0.088

¹Adjusted for age and other covariates

²Modified Charlson Comorbidity Index per 1-point scale (0–38), CVD variables removed

³Includes ischemic heart disease, atherosclerosis, peripheral vascular disease, congestive heart failure and cerebrovascular disease

Secondary Analysis

Between 2007–2011 there were 53,606 admissions where c-spine fracture was listed within the discharge diagnoses, of which 408 were also coded for AS. Those with AS compared to the control population were older (mean age 67.7±15.0 vs. 57.4±22.9) and 93% were male compared to only 60% of those without AS. Fall was the predominant mechanism of injury for patients with AS and c-spine fracture (62%) whereas those without AS had nearly equal percentages of fall and MVA (35% and 37% respectively). AS patients had a lower percentage of internal and intracranial injuries when compared to non-AS controls: 7% vs. 19% (internal injury) and 9% vs. 30% (intracranial injury) respectively.

The univariate and multivariable odds of death for patients admitted with a c-spine fracture are presented in Table 5. A diagnosis of AS was associated with an increased odds of death (OR 1.61; 95% CI 1.16–2.22; p-value 0.004). Intracranial injury and internal injury were associated with an increased odds of death (OR 2.56; 95% CI 2.38–2.76; p-value <0.0001) and (OR 2.82; 95% CI 2.59–3.07; p-value <0.0001) respectively as expected. Comorbid conditions, represented by the CCI, were also associated with an increased odds of death (OR 1.17; 95% CI 1.14–1.20; p-value <0.0001), whereas, elective admission was protective (OR of 0.70; 95% CI 0.59–0.83; p-value <0.0001).

Table 5.

Diagnoses associated with in-hospital mortality in all patients admitted with cervical spine fracture between 2007–2011.

	Univariate analysis1	Multivariable analysis2
	OR (95 % CI)	OR (95 % CI)	p–value
Male	1.23 (1.14–1.31)	1.52 (1.41–1.64)	<0.0001
Age by decade	1.24 (1.22–1.26)	1.39 (1.36–1.42)	<0.0001
Elective admission	0.56 (0.47–0.66)	0.70 (0.59–0.83)	<0.0001
AS3	1.66 (1.21–2.28)	1.61 (1.16–2.22)	0.004
Intracranial injury	2.22 (2.08–2.38)	2.56 (2.38–2.76)	<0.0001
Internal injury	2.06 (1.91–2.21)	2.82 (2.59–3.07)	<0.0001
CCI4	1.23 (1.21–1.26)	1.17 (1.14–1.20)	<0.0001

¹p-values <0.0001

²Model adjusted for Elective admission, AS, intracranial injury, internal injury and Charlson index.

³Ankylosing Spondylitis

⁴Charlson Comorbidity Index

DISCUSSION

This is the first U.S. population-based study describing in-hospital mortality in AS patients. C-spine fracture with spinal cord injury and sepsis were top causes of death in this dataset and were most strongly associated with in-hospital mortality in AS patients. C-spine fracture without spinal cord injury, pneumonia, CVD and comorbidities were also significantly associated with in-hospital death. In a secondary analysis of all patients admitted with c-spine fracture, we found that an AS diagnosis was an independent risk factor for death. Patients with AS fractured their c-spines with lower velocity mechanisms of injury and sustained less poly-trauma compared to patients without AS. Falling was listed as the primary mechanism of c-spine fractures in hospitalized AS patients.

Although c-spine fracture is a well-established complication of AS, it has not previously been reported to be a top cause of death in patients with AS. This may be due to the fact that c-spine fracture is a relatively rare complication in the non-hospitalized AS population and previous cohort studies may not have been large enough to discern the signal. Additionally, as our data is gathered solely from hospitalized patients, death is driven by acute etiologies such as infections or accidents rather than sequelae of chronic diseases, which are more likely to lead to death outside of the hospital setting. Similar to our findings, retrospective studies from the surgical literature have shown that patients with AS fracture their spine at lower velocities when compared to controls (27). A small, retrospective review of patients with spinal fracture admitted to a level-I tertiary referral center found the most common location of fracture in AS patients to be located in the cervical region, a finding supported by our study (28). AS patients in that study also experienced a significant delay to spinal fracture diagnosis as well as a 21% mortality rate. Their mortality rate is higher than that reported in this study which is likely due to the high-risk trauma population studied at their center. Our study found similar c-spine fracture mortality rates in AS patients compared to those reported in a review of the spinal trauma literature, which is still nearly double that in patients without AS (11,12). The high rate of mortality in AS after c-spine fracture highlights the importance of prompt and specialized care for these patients.

Our study is the largest U.S. in-hospital population-based study evaluating mortality in AS to date, permitting the study of rare outcomes such as c-spine fracture. Prior studies have shown that CVD is a leading cause of death in patients with AS whereas our study found a high prevalence of CVD in the cohort, the diagnosis of CVD did not carry the strongest association with in-patient mortality (2–6,8,9,18). Prior studies have relied on death records (3–5) or infer cause of death based on clinic notes (4) which are weighted by deaths occurring outside of the hospital setting. CVD in particular has been shown to be over-represented on U.S. death records (29). As a result, previous studies reliant on death records may have overestimated CVD as a primary cause of death in AS patients. The study by Mok et al. was unique, showing infection to be the leading cause of death in AS patients (2). This may be explained by the inclusion of both in-patient deaths in addition to those that occurred in the outpatient setting. Increased all-cause mortality was found in AS patients in the Swedish nationwide registry. CVD was found to be the primary cause of death in their cohort and contributed to a higher proportion of deaths than in the control group (8). Additionally, they found that death from spinal trauma was more frequent in the AS group but was not one of the top causes of death. Using a retrospective population-based Ontario cohort, Haroon et al. found that vascular mortality was increased in AS patients. They did not, however, evaluate all-cause mortality, therefore it could not be determined whether cardiovascular disease was the leading cause of death in AS (9).

A limitation of our dataset was that the HCUP-NIS does not assign unique identifiers to individual patients; therefore, we could not classify multiple hospitalizations belonging to a single patient. As a result, analyses could not be adjusted for within-patient correlations, and were necessarily conducted under the assumption that each hospitalization represented an independent patient. Though patients could accrue multiple hospital admissions, our primary outcome, in-hospital mortality, represents a unique event. Additionally, deaths that may have been related to the admission but occurred after discharge could not be included in our estimate of mortality. We also relied on coding data, which is inherently subject to specificity inaccuracies. We attempted to improve specificity by excluding those with dual classification for other rheumatic diseases. It is possible that mild AS may not be coded for in our dataset, which would bias our results towards more severe disease. In order to improve the quality of our variables, we used diagnostic variables that had been previously validated or used in the published literature. Though the ICD9-CM code for AS is validated, it is done so only in a Veteran Health Administration dataset however, it has been used in many population-based studies of AS (2,8,9,18). Our dataset was also limited by the fact that no medication data was available nor were we able to ascertain AS disease severity, duration or activity. Additionally, comorbid diagnoses such as osteoporosis are unlikely to be coded for in-patient data; therefore we could not evaluate these potentially important factors.

In conclusion, we report for the first time in a population-based study that c-spine fracture is a leading cause of in-hospital mortality in AS patients. We found that c-spine fracture with and without spinal cord injury, sepsis, pneumonia, CVD and comorbid illnesses were independently associated with in-hospital mortality in AS. Additionally, we found that a diagnosis of AS independently increased the odds of death in patients admitted with c-spine fracture and patients with AS sustained the majority of their fractures after a fall. The high incidence of vertebral fractures and the associated mortality in AS warrant further study to elucidate mechanisms and associated risk factors in order to prevent this devastating event. Additional prospective studies evaluating risk factors such as disease duration, burden of radiographic damage, fall risk, osteoporosis as well as medication usage will be important to further understand the risk factors associated with the outcome of spinal fracture in AS.

Table 4.

Characteristics of all patients admitted with cervical spine fracture between 2007–2011.

	AS (n=408) n(%)	non-AS (n=53198) n(%)
Age	67.7 ± 15.0	57.4 ± 22.9
Male	378 (93%)	32035 (60%)
Died	44 (11%)	3605 (7%)
CCI1	0.9 ± 1.4	0.7 ± 1.2
Elective admission	44 (11%)	3533 (7%)
MVA2	62 (15%)	19673 (37%)
Fall	252 (62%)	18875 (35%)
Other Mechanism3	94 (23%)	14650 (28%)
Internal injury	28 (7%)	10302 (19%)
Intracranial injury	38 (9%)	15931 (30%)

¹Charlson Comorbidity Index

²Motor vehicle accident

³Other mechanism of injury

SIGNIFICANCE AND INNOVATIONS.

• Top causes for death for hospitalized AS patients were sepsis, spinal fractures, spinal cord injury and pneumonia.

• Cervical spine fracture with spinal cord injury had the strongest association with inpatient mortality in all AS admissions.

• A diagnosis of AS was associated with higher odds of death in all patients admitted with a cervical spine fracture.

• The most common mechanism of c-spine fracture in AS was fall compared to higher impact trauma in non-AS patients.