Factors Associated With Survival Following Blunt Chest Trauma in Older Patients: Results From a Large Regional Trauma Cooperative

Hypothesis  We hypothesized that patient factors, injury patterns, and therapeutic interventions influence outcomes among older patients incurring traumatic chest injuries.

Design  Patients older than 50 years with at least 1 rib fracture (RF) were retrospectively studied, including institutional data, patient data, clinical interventions, and complications. Univariable and multivariable analyses were performed.

Setting  Eight trauma centers.

Patients  A total of 1621 patients.

Main Outcome Measure  Survival.

Results  Patient data collected include the following: age (mean, 70.1 years), number of RFs (mean, 3.7), Abbreviated Injury Scale chest score (mean, 2.7), Injury Severity Score (mean, 11.7), and mortality (overall, 4.6%). On univariable analysis, increased mortality was associated with admission to high-volume trauma centers and level I centers, preexisting coronary artery disease or congestive heart failure, intubation or development of pneumonia, and increasing age, Injury Severity Score, and number of RFs. On multivariable analysis, strongest predictors of mortality were admission to high-volume trauma centers, preexisting congestive heart failure, intubation, and increasing age and Injury Severity Score. Using this predictive model, tracheostomy and patient-controlled analgesia had protective effects on survival.

Conclusions  In a large regional trauma cooperative, increasing age and Injury Severity Score were independent predictors of survival among older patients incurring traumatic RFs. Admission to high-volume trauma centers, preexisting congestive heart failure, and intubation added to mortality. Therapies associated with improved survival were patient-controlled analgesia and tracheostomy. Further regional cooperation should allow development of standard care practices for these challenging patients.

Outcomes from trauma are dependent on multiple factors, with age and injury severity being 2 of the strongest predictors. Injuries that may not result in poor outcomes among younger populations can cause major morbidity and possible mortality among older populations.^1,2 Osler et al³ reported that older trauma patients have higher mortality than younger trauma patients, after controlling for injury severity. Taylor et al² analyzed Maryland statewide trauma registry data and confirmed that age is an independent predictor of mortality following trauma. Indeed, the latest data from the National Trauma Data Bank indicate that trauma fatality begins to increase steadily after age 45 to 50 years.⁴ This age-related mortality disparity has also been reported for specific injury types such as traumatic brain injury, spinal injuries, and pelvic fractures.^5-7

Multiple rib fractures (RFs), associated with little morbidity and mortality in younger trauma patients, can result in major morbidity in older trauma patients. Morbidities include need for intubation (15% incidence), development of pneumonia (15%-20%), and prolonged ventilatory support (10%-20%); the incidence of mortality ranges from 2% to 20%.^3,8-11 Bulger et al¹² performed a case-control study of 277 older patients and 187 younger patients (age range, 18-64 years) with RFs. Mortality was 22% among older patients compared with 10% among younger patients. Bergeron et al⁹ similarly found that mortality associated with RF was significantly higher among older vs younger adult patients. In that study and in other studies,^9,11-14 the death rate increased with the number of RFs, with a large study¹⁴ using the National Trauma Data Bank showing a significant break point at 6 fractured ribs.

Richmond et al¹⁵ found that the likelihood of older trauma patients' experiencing a complication was increased by preexisting comorbid medical conditions. In their series, 52% of surviving older trauma patients were discharged to home, with outcomes related to injury severity, age, and premorbid functional status. In another study,¹⁶ compared with geriatric trauma patients aged 65 to 79 years, octogenarians were less frequently discharged to home (53% vs 29%) and had lower Functional Independence Measure scores in every category at the time of discharge, despite lower Injury Severity Score (ISS). Holcomb et al¹⁷ studied isolated adult patients incurring RFs, categorized by number of RFs and by age (15-44 vs ≥45 years). The mortality in their series was 3%. Older patients and those with more RFs had more ventilator days and longer intensive care unit (ICU) and hospital length of stay. These data suggest that age begins to exert a negative effect on outcomes as early as 45 years.

The changing demographics of the American population make study of older patients relevant to trauma practitioners. Older adults compose 12.4% of the US population, and this rate is steadily increasing. From 1900 to 2000, the older population increased more than 11-fold, while the population as a whole increased less than 4-fold. The aging population is expected to surge between 2010 and 2030 as the baby-boom generation turns 65 years and older.¹⁸ By 2030, older persons are projected to compose 20% of the US population.¹⁹

Numerous authors have looked at ways to reduce the morbidity and mortality of RFs. Among patients with blunt chest trauma studied in single-institution series, intercostal nerve blocks, thoracic epidural catheters, and noninvasive ventilatory support (ie, bilevel positive airway pressure or continuous positive airway pressure) were found to improve pulmonary function and to reduce morbidity, including number of ventilator days and development of pneumonia.^10,20-23 Evidence that these therapies are beneficial is inconclusive, and other authors show no reduction in morbidity with these therapies.^17,21,24 While epidural anesthesia, the most studied of these interventions, is associated with reduced mortality in large trauma databases, well-designed randomized trials do not show a survival benefit for this therapy.^14,24 The objectives of the present study were to examine older patients with traumatic RFs to identify patient factors and injury patterns associated with poor outcomes and to identify treatment strategies associated with superior outcomes in a well-powered multi-institutional study.

Patients older than 50 years with significant blunt chest trauma, defined as 1 or more RFs with or without pulmonary contusion, admitted to hospitals between January 1, 2002, and December 31, 2007, were retrospectively studied. To assess the isolated effect of blunt chest trauma, patients with penetrating chest trauma, an overall ISS exceeding 20, or an Abbreviated Injury Scale (AIS) score at any nonthoracic area exceeding 3 were excluded from the study.

Patient data collected include the following: age, sex, ISS, premorbid medications, injuries and AIS score for each system injured, premorbid level of functioning (independent, semi-independent, or dependent), specific characterization of thoracic trauma (number of RFs, pulmonary contusion, or flail chest), and premorbid conditions such as renal failure, cardiac disease, and chronic obstructive pulmonary disease. Preexisting conditions were scored on a scale of 0 to 3. Hospital treatments that were recorded included admission to floor, step-down unit, or ICU and use of narcotics, intercostal nerve blocks, or noninvasive ventilatory techniques. Also noted was admission to trauma service or nontrauma service, hospital trauma designation, annual trauma admissions to each hospital, annual trauma admissions with ISS exceeding 15, and in-house trauma attending presence. Events of the hospital course were recorded such as intubation, tracheostomy, ICU and hospital length of stay, development of pneumonia, death, and dis charge to home, nursing home, or rehabilitation facility. Pneumonia was defined as new or progressive infiltrate on radiograph and at least 2 of the following: fever (>101.5°F), sputum leukocytosis (>2 white blood cells per high-power field), or white blood cell count higher than 12 000/μL or less than 4000/μL in accord with the National Healthcare Safety Network guidelines²⁵ (to convert white blood cell count to ×10⁹/L, multiply by 0.001). The primary outcome measure was survival.

Individual factors (such as age, ISS, premorbid medical conditions, hospital resources, and hospital treatments) were analyzed for their association with survival. Analysis of individual variables was performed using t test and χ² analysis, as appropriate. Variables with nonnormal distributions were log transformed and reanalyzed for significance. Significance was set at 0.05. Multivariable analysis was performed for all variables identified as potentially significant on univariable analysis (P < .2) by stepwise logistic regression to determine the contribution of each variable to survival.

After obtaining institutional review board approval, each trauma center created a data file on a secure password-protected computer that included patient name, hospital medical record number, and study number. A separate file without personal health information was created that included patient study number and all study data. This latter deidentified file was sent to the sponsoring institution for analysis.

Eight trauma centers (level I and level II) participated, and 1621 patients were studied. Male sex predominated (58.7% vs 41.3%), the mean age was 70.1 years, and the mean number of RFs was 3.7. The mean AIS scores for head, face, abdomen, chest, and extremities were 0.5, 0.1, 0.4, 2.7, and 0.7, respectively. The mean ISS was 11.7. Seventy-eight percent were admitted to level I trauma centers, and the remainder were admitted to level II trauma centers. The mean annual admissions among trauma centers was 2099 patients (range, 512-2874 patients). Most patients were admitted to trauma centers with in-house attending presence (86.4%) and with ICU fellowships (74.4%).

Data on preexisting conditions were available for most patients. Common conditions included coronary artery disease (19.9%), lung disease (13.5%), congestive heart failure (CHF) (7.1%), central nervous system disease (6.3%), cirrhosis (1.1%), and end-stage renal disease or dialysis (0.8%).

Thirty-five percent of patients were admitted to an ICU during their hospital stay, with a mean ICU length of stay of 16.5 days and a mean hospital length of stay of 27.5 days. Twelve percent of patients were intubated, and 4.3% required a tracheostomy. The mean postinjury day (PID) on which tracheostomies were performed was PID 12.3. Twenty-three percent were performed between PIDs 0 and 7, 53.5% were performed between PIDs 8 and 14, and 23.3% were performed later than PID 14. Almost 4% of patients developed pneumonia. Overall mortality was 4.6%.

On univariable analysis, increased mortality was associated with admission to high-volume trauma centers and level I centers, centers having more annual trauma admissions with ISS exceeding 15, and centers with a dedicated trauma ICU, ICU fellowships, or trauma fellowships (Table 1). On univariable analysis of patient demographics, increased mortality was associated with the following: preexisting coronary artery disease or CHF; increasing age, ISS, and number of RFs; and increasing AIS scores for head, chest, abdomen, and extremities (Table 2). Sex and preexisting obesity, central nervous system dysfunction, lung disease, cirrhosis, or renal failure showed no association with mortality. On univariable analysis of patient hospital course factors, increased mortality was associated with intubation, tracheostomy, elevated creatinine level, development of pneumonia, and number of days in a cervical collar (Table 3).

On multivariable analysis, strongest predictors of mortality were admission to high-volume trauma centers, preexisting CHF, intubation, and increasing age and ISS (Table 4). The finding that admission to high-volume trauma centers was associated with poorer survival led to post hoc analysis of differences among low-volume (<1000 trauma admissions per year), medium-volume (1000-2000), and high-volume (>2000) trauma centers. High-volume trauma centers treated patients with more RFs, greater severity of associated injuries, and increased ISS (Table 5). When the same logistic analysis was performed on only the most severely injured patients (ISS >15 [n = 421], increasing ISS, preexisting CHF, and trauma center volume no longer statistically significantly predicted mortality.

Using the predictive model on the whole data set, we evaluated effects of the following on mortality: tracheostomy, epidural catheters, noninvasive ventilation, ketorolac tromethamine, intermittent intravenous narcotics, patient-controlled analgesia, intercostal nerve blocks, and transdermal narcotics. Tracheostomy had a protective effect on survival. Patient-controlled analgesia approached statistical significance (Table 6). Other therapies had no appreciable effect on survival.

The aging of the American population and the ubiquitous nature of trauma demand that we continue to improve our therapeutic interventions among older patients. With few subjects, single-institution studies are often underpowered to detect differences in survival. Multi-institutional studies can evaluate the effect of different treatment strategies among large patient populations. National databases (such as the National Trauma Data Bank or the University HealthSystem Consortium database) are multi-institutional but do not always include critical data points and have some data inaccuracies. Our multi-institutional study included data from trauma registries, supplemented by medical record review. To date, our study among 8 trauma centers and 1621 patients is one of the largest to examine isolated blunt chest injury in older patients.

The effect of older age and increasing ISS on mortality is not surprising. While many individuals maintain organ function during the aging process, age serves as a surrogate marker for physiologic derangements and dysfunction among large study populations.²⁶ With age comes hyperplasia of the intima and media of the vasculature, myocardial ventricular hypertrophy, diminished elastic recoil of the lung and reduced functional residual capacity, loss of renal mass and lower glomerular filtration rate, and decreased absorptive capacity of the gut. Although these changes may not manifest among persons in a normal nonstressed state, increased demands of injury often result in decreased capacities of these various organ systems. Based on our study findings, a patient admitted to a high-volume trauma center with an average injury (4 RFs, 11.7 ISS, intubation, and no coronary artery disease or CHF) would have a predicted mortality of 2.3% at age 50 years and a predicted mortality of 19.8% at age 80 years. Effects of severe coronary artery disease or CHF on outcomes of blunt chest trauma are expected. A lack of physiologic reserve renders moderate acute stress of injury as potentially fatal and severely compromises injury recovery. Intubation was also associated with worse outcomes. This may represent a surrogate marker for underlying lung dysfunction or pulmonary contusion, increased injury severity, and other aspects of hospital care. The extent of involvement among these factors requires more specific data, aided by prospective data collection.

The most surprising finding on univariable analysis was increased mortality among trauma centers with high volume, in-house trauma attending presence, a dedicated trauma ICU, ICU fellowships, or trauma fellowships. On multivariable analysis, admission to a high-volume trauma center remained associated with increased mortality. There are 2 potential explanations for this finding. The first explanation is that high-volume trauma centers do not provide better care and actually have worse outcomes than lower-volume trauma centers. This contradicts the seminal work by Nathens et al,²⁷ who concluded that trauma outcomes improved when severely injured patients were cared for in a high-volume trauma center. A methodological difference is that Nathens et al found a survival benefit when the patient was very ill (shock or coma on admission to a trauma center with a high volume of severely injured patients [>650 patients per year with ISS >15]). Our study by design looked at unisystem injury among patients with less severe injury. A link between trauma center volume and improved outcomes might be irrelevant in our setting and may allow for variations in treatment to determine outcomes. The second explanation is that high-volume trauma centers deliver quality care with good outcomes but have the highest concentrations of the sickest patients, unaccounted for by our predictive equation. In our study, this increased burden of acuity in high-volume centers is reflected by number of RFs, overall ISS, and AIS scores for head, abdomen, and extremities. Furthermore, when the same logistically derived equation was used among a subset of patients with ISS exceeding 15, trauma center volume was no longer statistically significant in our study. Determining which of these explanations is correct is beyond the scope of a retrospective study. To distinguish which hypothesis is true, randomization of patients to centers of different sizes would be needed.

The second overall goal of this study was to identify treatment strategies that may be protective and provide superior outcomes in these challenging older patients with blunt chest trauma. Once the predictive model was created, several therapeutic interventions and their effects on survival were assessed. Only tracheostomy was protective of survival. Patient-controlled analgesia showed a trend toward improving survival but did not attain statistical significance. Epidural catheters, noninvasive ventilatory support, and intercostal nerve blocks did not seem to affect mortality, likely because of their infrequent use. The effect of tracheostomy on survival is understandable because as it reduces the need for sedation, decreases the work of breathing, and improves pulmonary toilet. It also facilitates temporary return to ventilatory support as opposed to an emergency reintubation. The literature is divided as to whether early or late tracheostomy should be the standard of care in modern ICUs.^28-30 The present study is underpowered for commenting on that question.

This study has limitations. Retrospective data were used, and the study is nonrandomized. Such analyses are prone to selection bias and, in general, are more suitable for developing study questions rather than answering scientific questions. This study also did not record whether the number of RFs was determined by plain radiograph or by computed tomography. An RF detected on plain radiograph has a greater negative effect on outcomes than a fracture detected on computed tomography.³¹ No data were recorded about whether epidural catheters infused narcotics only, anesthesia only, or both. The next steps for this regional trauma cooperative and for these data will be to develop a regional treatment guideline and to assess prospectively the ability of the cooperative to improve outcomes in older patients with blunt chest trauma. Evidence suggests that treatment algorithms can be stratified based on patient age, number of RFs, and admission vital capacity.³²

In conclusion, a large regional trauma cooperative confirmed that older age and increased ISS are independent predictors of survival in older patients incurring traumatic RFs. Contributing to mortality were CHF, intubation, and admission to high-volume trauma centers. Therapies associated with improved survival were patient-controlled analgesia and tracheostomy. Further regional cooperation should allow for development of standard care practices for these challenging patients.

Correspondence: David T. Harrington, MD, Rhode Island Hospital, Warren Alpert Medical School, Brown University, 593 Eddy St, Ambulatory Patient Care Bldg, Room 443, Providence, RI 02903 (dharrington@usasurg.org).

Accepted for Publication: December 4, 2009.

Author Contributions:Study concept and design: Harrington, Velmahos, and Brotman. Acquisition of data: Harrington, Phillips, Zacharias, Velmahos, Rosenblatt, Winston, Patterson, Desjardins, Winchell, Brotman, Churyla, Schulz, Maung, and Davis. Analysis and interpretation of data: Harrington, Phillips, Machan, Zacharias, Rosenblatt, Desjardins, Winchell, and Davis. Drafting of the manuscript: Harrington, Phillips, Patterson, Churyla, Schulz, and Maung. Critical revision of the manuscript for important intellectual content: Harrington, Machan, Zacharias, Velmahos, Rosenblatt, Winston, Desjardins, Winchell, Brotman, Churyla, Maung, and Davis. Statistical analysis: Harrington and Machan. Administrative, technical, and material support: Harrington, Phillips, Zacharias, Velmahos, Winston, Patterson, Brotman, and Maung. Study supervision: Harrington, Zacharias, Velmahos, and Davis.

Financial Disclosure: None reported.

Previous Presentation: This paper was presented at the 90th Annual Meeting of the New England Surgical Society; September 13, 2009; Newport, Rhode Island; and is published after peer review and revision.