GLFs are an increasing public health concern with significant economic and social consequences. It is estimated that >30% of older adults fall yearly, with >1 out of 5 incidents requiring acute care, costing more than $20 billion a year in the US (10
). Unintentional falls are currently the number one cause of injury-related death in persons aged 65 and above (16,650 in 2006), and the leading cause of nonfatal injuries treated in US emergency departments in all age groups with the exception of persons aged 15 to 24 years. Even in this group, falls are a close second (13
). Additionally, unintentional falls are estimated to contribute to 8 million emergency department visits annually. Given the sweeping demographic changes occurring in the US, it is reasonable to assume that the incidence and impact of GLFs will increase.
GLFs have appropriately drawn the attention of academic and institutional research, most of which has focused on developing strategic fall risk assessment and prevention strategies. These efforts, manifest by scoring systems such as the Morse Fall Scale or recommendations by authorities such as the ACS Subcommittee on Injury Prevention and Control, are being sporadically adopted, but with mixed results (12
). Much less has been published regarding the treatment and outcome side of this complex medical and social issue, which highlights the need for the development of evidence-based strategies in the multidisciplinary management of GLF patients (16
). A better understanding of injury patterns, predictors of morbidity and mortality, and considerations for special subgroups would clearly benefit clinicians in the triage, assessment, treatment, and disposition of these patients. While measuring the efficacy of our trauma system, our data also contribute to the identification of these factors and point to the significance of factors not yet identified.
Mortality is a multifactorial result that incorporates the variable contributions of injury, acute physiology, and the patients' physiological functional status or burden of comorbid disease. Preexisting medical conditions and older age are both well-established risk factors of mortality among trauma victims. Each becomes difficult to study independently given that the elderly are more likely to suffer from chronic medical conditions (17
). Multiple studies have implicated four disease processes that increase mortality in trauma patients: cirrhosis, cardiovascular disease, respiratory disease, and diabetes. Even after controlling for age, the effect of these diseases on mortality is significant (18
). We demonstrated that several comorbidities are associated with mortality in the present study. Malignancy, renal disorders, and cardiac disorders (arrhythmia, valve dysfunction, and congestive failure) were each independent predictors of death among GLF trauma patients in our region.
Two other reports demonstrated the relationship between comorbid conditions and mortality in patients hospitalized with injuries from GLFs. Using data from the Scottish Trauma Audit Group, Kennedy et al found that the mortality rate increased as the number of comorbid conditions increased in patients admitted following low-impact falls (20
). Similarly, Hannan et al concluded that “pre-existing conditions … are significantly related (inversely) to survival of patients with trauma from low falls.” Hannan also advocated that comorbidity and age be included in the physiologic and anatomic injury models currently used to predict survival in these patients (21
It is important to note that hypotension and metabolic acidosis on presentation to the emergency department carried the greatest odds of death in our cohort. Importantly, the strong associations with these characteristics and death are not unique to GLF patients and have been well described. The model we developed for mortality prediction included variables for age, gender, acute physiology, severity of anatomic injury, and comorbidity. The ultimate purpose of developing such a model was to arrive at a risk-adjusted odds ratio for death for each level of trauma center designation in our region, thus enabling a comparative assessment of the performance of the three levels of trauma centers in our system. Additionally, these data can be observed over time to detect temporal variation in the overall adjusted odds ratios of mortality for our region.
An effective trauma system includes proper triage of patients to a trauma center based not only on the level of injury severity but also on host factors such as age and comorbidities. Such host factors can greatly influence survival and should be integral components in the initial patient assessment. Identifying patient characteristics that adversely affect outcome should ideally prompt transport to a higher trauma center level. We observed that patients in TSA-G with the highest predicted mortality, whether based on anatomic injury severity alone or on the multivariate model we developed, were admitted and treated at the level I and II centers. We also observed that the crude mortality rate was highest at the level I center and lowest among the level III/IV centers, which seems intuitively congruent. However, after risk adjustment, the lowest odds ratios for death occurred in the level I center.
We believe that there should be only minimal differences among outcomes of level I and level II trauma centers serving a common region given the similar capabilities of levels I and II. Culica et al reviewed discharge data from the Texas Health Care Information Council to assess the outcomes of care among the regionalized trauma systems in Texas. Similar to the present study, they observed small variations in survival when the level I and II trauma centers across the state were compared (22
). The difference we observed in adjusted odds ratio for mortality between the level I center and the other trauma center levels may reflect variations in triage, transfer, referral patterns, or clinical practice.
Similarities have been observed between outcomes when case mix adjustment was applied in other studies. MacKenzie et al compared mortality rates between level I trauma centers and non-trauma centers using data from 19 states as part of the National Study on the Costs and Outcomes of Trauma. They found that the unadjusted mortality rate was higher among patients treated in trauma centers compared to non-trauma centers. After adjustment for case mix, however, the risk of death within 1 year of injury was significantly lower for patients receiving care at a trauma center (23
). Khuri et al also found that risk adjustment had a significant impact on the rank ordering of Veterans Affairs hospitals after implementing a system for the prospective collection and comparative reporting of postoperative mortality rates after major noncardiac operations (24
). O'Connor et al sought to improve the mortality rates associated with coronary artery bypass graft surgery in Northern New England. Adjusted mortality rates were used to compare the efficacy of a coordinated intervention among the 23 cardiothoracic surgeons practicing in Maine, New Hampshire, and Vermont.
Risk-adjusted outcomes have become a popular tool for comparing the quality of care between hospitals as part of the so-called quality report cards. Perhaps the best use of risk-adjusted outcomes, however, may be the identification of opportunities for improvement, such as the refinement of our patterns of referral to higher levels of care for groups of patients at higher-than-average risk for death. As this is our first regionwide comparative assessment endeavor, additional refinement over time of our data acquisition and modeling may yield other significant covariates not included in the present model. Nonetheless, the present study provides a useful benchmark for our region's performance in triaging and treating patients injured in GLFs.
When our region is considered in total, the overall mortality rate from GLF was 2.1%. This rate compares favorably to other reports focusing on this mechanism of injury. Kennedy et al reported a 2.8% rate in patients with a mean age of 61.6 years (20
). Other authors have published GLF mortality rates ranging from 4.2% to as high as 8.9% among adult trauma patients (25
). Bergeron et al reported a 13.4% mortality rate among patients admitted to a regional trauma center in Quebec, Canada (26
). Our cohort and the Canadian study group had similar median injury severity scores of 9. These figures represent unadjusted mortality data, though at present, no omnibus metric exists to enable meaningful case mix–adjusted comparative assessment of outcomes between entire regional trauma systems.
In 2006 the ACS embarked on the Trauma Quality Improvement Program (TQIP). A key component of TQIP will be the comparative assessment of observed-to-expected mortality rates based on risk adjustment for institutional case mix (27
). A pilot study by Hemmila et al assessed the feasibility of utilizing the infrastructure of the National Trauma Data Bank to provide risk-adjusted benchmarking of trauma centers. They observed differences in the observed-to-expected mortality ratios across similarly verified trauma centers (28
). It is possible that efforts such as our study may be supplanted in the future by TQIP as a means of comparing trauma center performance across our region.
In our study, the adjusted odds of mortality remained fairly consistent from year to year. With the exception of 2001, the yearly adjusted odds of mortality did not vary significantly, as the 95% CIs spanned the 1.0 reference line. The variation in unadjusted mortality rates for the years 2002 to 2009 ranged from 1.2% in 2004 to 2.7% in 2002, indicating that the observed mortality was decreased by more than half over 2 years. Yet in the context of the risk adjustment regression model, we can see that the difference in mortality between these outlier years was less dramatic. This is valuable to us because appraisal of our performance from year to year based only on mortality incidence could lead to unfounded concern when unadjusted mortality is trending up, as in 2002. Similarly, when unadjusted mortality incidence is trending down, as in 2004, the lack of appropriate risk adjustment may allow for the misinterpretation that our outcomes had improved.
The most significant limitations of these data are that they were collected over 9 years and from 20 hospitals. As such, it is likely that some variation in data collection and management practices may have occurred between centers and over time. Nonetheless, this analysis provides a basis for future comparison, both in terms of methodology and outcomes. A second limitation to this study is the potential confounding effect attributable to interfacility transfer of GLF patients within our region. No attempt was made to determine correlations between outcome and transfer patterns. The effect of interfacility transfer, if significant, may potentially contribute to the observed difference in adjusted odds ratios for mortality between the level I and level II centers. However, we advocate that patients who require definitive management of injuries be treated in facilities equipped to accommodate the need at hand. Given that there are one level I and two level II trauma centers in TSA-G, appropriate interfacility transfer is an important component of trauma care in our region.
From these observations, we can conclude that the risk-adjusted hospital mortality for patients hospitalized for injuries sustained from GLF in TSA-G has remained consistent since the year 2003. Additionally, these patients are triaged across our region to appropriate levels of care, as demonstrated by the modest variation in adjusted odds ratios for mortality between levels of trauma center verification. More study is warranted to better define the nature of the lower adjusted odds ratio for mortality observed in the level I center compared to the other centers in the region. Understanding the relative contributions of case mix, referral patterns, and clinical practices may allow for an overall improvement in outcomes in our region for this growing population of trauma patients.