Matsushima K, Schaefer EW, Won EJ, Armen SB, Indeck MC, Soybel DI. Positive and Negative Volume-Outcome Relationships in the Geriatric Trauma Population. JAMA Surg. 2014;149(4):319-326. doi:10.1001/jamasurg.2013.4834
Copyright 2014 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
In trauma populations, improvements in outcome are documented in institutions with higher case volumes. However, it is not known whether improved outcomes are attributable to the case volume within specific higher-risk groups, such as the elderly, or to the case volume among all trauma patients treated by an institution.
To test the hypothesis that outcomes of trauma care for geriatric patients are affected differently by the volume of geriatric cases and nongeriatric cases of an institution.
Design, Setting, and Participants
This retrospective cohort study using a statewide trauma registry was set in state-designated levels 1 and 2 trauma centers in Pennsylvania. It included 39 431 eligible geriatric trauma patients (aged >65 years) in the Pennsylvania Trauma Outcomes Study.
Main Outcomes and Measures
In-hospital mortality, major complications, and mortality after major complications (failure to rescue).
Between 2001 and 2010, 39 431 geriatric trauma patients and 105 046 nongeriatric patients were captured in a review of outcomes in 20 state-designated levels 1 and 2 trauma centers. Larger volumes of geriatric trauma patients were significantly associated with lower odds of in-hospital mortality, major complications, and failure to rescue. In contrast, larger nongeriatric trauma volumes were significantly associated with higher odds of major complications in geriatric patients.
Conclusions and Relevance
Higher rates of in-hospital mortality, major complications, and failure to rescue were associated with lower volumes of geriatric trauma care and paradoxically with higher volumes of trauma care for younger patients. These findings offer the possibility that outcomes might be improved with differentiated pathways of care for geriatric trauma patients.
Since 1960, the average age of the population in the United States has increased steadily and currently the proportion of the US population older than 65 years is 13.0%.1 By 2030, one-fifth of the population will belong to this age group. At the same time, the number of elderly patients being treated for trauma is also increasing, currently contributing up to 25% of acute trauma admissions.2- 4 Consensus has not been reached regarding the criteria for identifying a patient as elderly, yet advanced age is a well-recognized risk factor for mortality after major trauma.5- 8 Susceptibility to complications and death in this setting is attributed to the increased physiologic burden of preexisting medical conditions in this age group but also perhaps to nonstandard triage methods.9- 11 Along these lines, worse outcomes are observed in injured patients with do-not-resuscitate orders, for which increasing age is a significant patient-level factor.12,13 These considerations argue that suboptimal systems and pessimistic perceptions may contribute to poor outcomes in trauma. Such system and health care provider attributes may offer significant targets for improvement of care in the elderly trauma patient, different from targets that might be chosen for optimizing outcomes in younger adult patients.
Associations between high hospital volume and better mortality rate in complex surgical procedures were first reported by Luft et al14 in 1979. With few exceptions, this volume-outcome relationship has been affirmed for high-risk surgical procedures in subsequent articles.15- 17 Similarly, recent studies have provided evidence that higher volume at a trauma center is generally associated with better patient outcomes.18- 22 While most patients who undergo high-risk surgical procedures for cancer or atherosclerotic disease are elderly, such is not the case among patients treated for trauma, who are generally much younger.23,24 Inasmuch as care for geriatric trauma patients may require different kinds of specialized knowledge skill sets, we postulated that an institution’s volume of trauma care among all age groups might not be as relevant to outcomes as its trauma case volume among geriatric patients.4,25 In this study, we used a statewide registry of levels 1 and 2 trauma centers in Pennsylvania to test the hypothesis that outcomes of trauma care for geriatric patients are driven differently by the volume of geriatric trauma cases and nongeriatric trauma cases of the institution. Outcomes of interest were in-hospital mortality, major complications, and mortality after major complication (failure to rescue).
This study was approved by the institutional review board of Penn State Milton S. Hershey Medical Center; patient consent was waived. We conducted a retrospective cohort study using a statewide trauma registry in the commonwealth of Pennsylvania, also known as the Pennsylvania Trauma Outcome Study (PTOS) database.26 The PTOS database is maintained by the Pennsylvania Trauma Systems Foundation. Since 1986, more than 500 000 trauma patients have been entered in the database. All designated trauma centers are required to have an interrater reliability process to validate their data internally. The PTOS database includes trauma patients with at least 1 of the following characteristics: (1) hospital stay longer than 48 hours, (2) transfer from another institution, (3) intensive care unit admission, or (4) in-hospital mortality. Exclusions from the database included patients with (1) isolated hip fracture, (2) asphyxia, (3) drowning, and (4) hyper/hypothermia. For this study, we retrieved information from the PTOS database on geriatric patients (aged >65 years) admitted to the state-designated levels 1 and 2 trauma centers between 2001 and 2010. One institution did not submit any data for complications from 2002 to 2006 and its entries were excluded. Also excluded were patients whose systolic blood pressure or heart rate on admission was reported to be zero or who were transferred out to another trauma center.
Our primary outcome of interest was in-hospital mortality of geriatric trauma patients. We calculated the rate of major complications and failure to rescue as secondary outcomes. Given the list of in-hospital complications provided by PTOS, we classified the following as major complications: acute respiratory distress syndrome, acute respiratory failure, pneumonia, aspiration/aspiration pneumonia, pulmonary embolism, fat embolus syndrome, myocardial infarction, acute renal failure, sepsis, septicemia, gastrointestinal bleeding, postoperative hemorrhage, liver failure, progression of original neurologic insult, small-bowel obstruction, empyema, acute arterial occlusion, dehiscence/evisceration, central nerve system infection, and extremity compartment syndrome.26 Patients who experienced any of these complications were classified as experiencing a major complication. Patients who died in-hospital after a major complication were classified as failure to rescue.
Average annual geriatric and nongeriatric trauma volumes were calculated for each institution from 2001 to 2010. These volumes were defined as the average number of patients admitted yearly to the trauma service from each age category (geriatric vs nongeriatric) who met our inclusion criteria. All relevant data regarding patient characteristics, including preexisting medical conditions, mechanism and severity of injury, transfer status, and do-not-resuscitate status, were extracted for the analysis. We categorized patients’ preexisting medical conditions into separate disease groups by system: heart disease, diabetes mellitus, gastrointestinal disease, hematologic/bleeding disorder, psychiatric disorder, infectious disease, liver disease, cancer, connective tissue disease, neurological disorder, obesity, pulmonary disease, and kidney disease. Diagnoses or conditions not falling into these categories were identified as other conditions.
Hierarchical logistic regression was used to investigate the relationship between institutional volume and patient outcomes. Hierarchical logistic regression is an extension of logistic regression that accounts for correlation among patients treated at the same trauma institution by including a random effect for institution in the model. For each outcome, a hierarchical logistic regression model was fitted that included average annual geriatric trauma volume and average annual nongeriatric trauma volume as fixed effects and a random intercept effect for institution. We also included adjustments for patient characteristics, injury severity, and preexisting conditions. Do-not-resuscitate status was only included in the models for in-hospital mortality and failure to rescue. Graphical methods were used to determine functional forms for each volume variable, age, Injury Severity Score (ISS), and year of admission. Linearity was reasonable for trauma volumes and year of admission. Quadratic functions were appropriate for age and ISS. All other variables were modeled as categorical with a chosen reference group. Odds ratios (ORs) and their corresponding 95% CIs were reported for all parameters. For the linearly modeled volume variables, ORs were reported for an increase of 100 patients per year. For the quadratic effects of age and ISS, ORs were reported for the third quartile vs first quartile. All patients in the data set were included in the model of in-hospital mortality. Patients who died without a major complication listed in the database were, by definition, considered not to have died as the result of failure to rescue. These patients were excluded from the models of major complications and failure to rescue. All patient records with a missing value for any variable included in the models were excluded.
Within the PTOS database, a total of 39 431 geriatric patient entries from 20 state-designated trauma centers were identified (12 level 1 and 8 level 2 centers). Table 1 shows the average annual volume of geriatric and nongeriatric trauma patients at each institution; the Figure shows a scatterplot of these data. The average annual volumes ranged from 62 to 554 for geriatric patients and from 161 to 1140 for nongeriatric patients. The correlation coefficient was 0.75.
Patient demographics, preexisting medical conditions, and severity of injury for geriatric patients are shown in Table 2. About one-quarter of geriatric patients were 85 years or older. Most blunt injuries were due to accidental falls, and very few patients (<1%) sustained a penetrating injury. Approximately 43% of patients had an ISS greater than 15 and 29% were transferred from an outside institution.
Unadjusted rates of in-hospital mortality, major complications, and failure to rescue are shown for each institution in Table 1. The final hierarchical logistic regression model for in-hospital mortality is presented in Table 3. A total of 2447 patients (6%) had missing data for covariates and were excluded from the model. The model shows that a larger institutional volume of geriatric trauma patients was significantly associated with decreased odds of mortality (OR, 0.75 for a 100-patient increase; 95% CI, 0.61-0.92; P = .005), but that a larger volume of nongeriatric trauma patients was not significantly associated with mortality (OR, 1.08 for a 100-patient increase; 95% CI, 0.98-1.19; P = .11). The interpretation of the model is that, given trauma centers with the same nongeriatric volume, the odds of in-hospital mortality for a geriatric patient was 25% lower at a center that treated 100 more geriatric patients annually.
The final hierarchical logistic regression models for major complications and failure to rescue are presented in Table 4 and Table 5, respectively. Patients with missing complications (227 patients) or a death without a listed complication (2682 patients) were excluded from these models, as were 1656 patients (4.5%) with missing covariate data. An additional 685 patients (1.5%) with missing do-not-resuscitate status were excluded from the model for failure to rescue. The fitted models indicated that a larger institutional volume of geriatric trauma cases was significantly associated with lower odds of a major complication (OR, 0.79 for a 100-patient increase; 95% CI, 0.63-0.99; P = .04), and lower odds of failure to rescue after a major complication (OR, 0.65 for a 100-patient increase; 95% CI, 0.44-0.98; P = .04). In contrast, a larger volume of nongeriatric trauma patients was significantly associated with higher odds of a major complication (OR, 1.14 for a 100-patient increase; 95% CI, 1.03-1.27; P = .02) but was not significantly associated with failure to rescue (OR, 1.18 for a 100-patient increase; 95% CI, 0.98-1.44; P = .09).
Using the Pennsylvania statewide trauma database, we found that institutions with higher case volumes for geriatric trauma have lower in-hospital mortality rates. Our data also suggest that geriatric trauma patients were less likely to develop major complications at trauma centers with higher geriatric trauma volume and that patients who developed major complications were more likely to be successfully rescued at these centers. Furthermore, our study showed that the institutional nongeriatric trauma volume adversely affects the outcomes of geriatric trauma patients, most prominently influencing the rate of major complications. These observations suggest that an institution’s interventions to improve outcomes among its trauma patients may need to consider the needs of specific subgroups such as the elderly.
National and statewide registries have been used, with conflicting results, to explore the volume-outcome relationship in trauma populations.18- 22 In the University Health System Consortium Trauma Benchmarking Study, the threshold for high volume was defined as 650 severely injured blunt and penetrating injury cases admitted annually; using this definition, Nathens et al18 found that high-volume centers were significantly associated with lower inpatient mortality rates and shorter lengths of hospital stay compared with lower-volume centers. In contrast, in a retrospective analysis of Patient Discharge Data of the State of California, London and Battistella19 found no clear correlation between outcomes and institutional volume of trauma cases. Demetriades et al20 used the National Trauma Data Bank (NTDB) to analyze outcomes of care for several high-grade injuries in levels 1 and 2 trauma centers accredited by the American College of Surgeons and were unable to demonstrate different outcomes for institutional case volumes less than 240 cases annually. Focusing on geriatric patients involved in major motor vehicle collision, Pandya et al23 analyzed results in a registry from New York state. Their analysis indicated that, for a state-designated level 1 or 2 trauma center, an incremental increase of 100 geriatric motor vehicle collision cases annually was associated with a 32.5-fold increase in the likelihood of survival to hospital discharge. In the current study, we confirmed and extended these observations by showing that lower rates of complications and failure to rescue are observed in centers with higher trauma volume specifically in the elderly.
The association between volume and mortality rate is well recognized for high-risk surgical procedures, but the specific attributes of the high-volume surgeon and institution that drive better outcomes are not fully understood.17,27 Recent work has emphasized that higher mortality rates after surgical procedures in low-volume centers are not necessarily owing to higher complication rates but to recovery from complications.28 From the perspective of quality of care provided by physicians and institutions, rescue from serious complications after surgical procedures is an emerging target for study and improvement.29,30 Systematic studies are only just beginning to emerge of the factors that contribute to the rates of complications and failure to rescue in the trauma population.
To date, there are contradictory data to indicate potential underlying factors associated with the institutional variation in the mortality rate of trauma patients. Glance et al31 used the most recent version of the NTDB to suggest that the institutional variation in the mortality of trauma patients is driven by rates of failure to rescue, not by rates of complications or mortality independently. On the other hand, using the same version of the NTDB, Haas et al32 showed that rates of complication and failure to rescue were significantly higher in trauma centers with higher mortality rates. These observations would suggest that under some circumstances, the complication rate drives the rate of failure to rescue and thus overall mortality. Interpretation of these studies is limited by the institutional variability in complication data reported in the NTDB.33 Using the PTOS database with robust data on complications, we found that 6.8% of geriatric patients died without major complications, a proportion substantially higher than that reported in elderly patients undergoing complex but still elective operations. Our analysis also suggests that mortality rates in the elderly trauma population are connected to the initial mortality rate (ie, not due to failure to rescue), higher complication rates, higher failure-to-rescue rates, and the case volume for nonelderly patients. This latter factor may be a previously unrecognized confounder in apparently contradictory analyses.
Regardless of age, trauma in adult patients is managed by the same trauma team. We found that outcomes among geriatric patients might be influenced negatively by the case volume for nongeriatric trauma. With respect to mortality and failure to rescue, our results suggest a trend, although not statistically significant, toward adverse effects of higher volume for younger patients. However, for complication rates, the adverse influence of nongeriatric volume on geriatric outcome is statistically significant. These findings emphasize the importance of differentiated pathways of care for specific high-risk groups of patients. In addition, they underscore the value of analysis such as that reported here, demonstrating that it is possible to detect the influence of care for one group of patients on the care of another group of patients especially in resource-limited settings.
Recent reports have emphasized that such differentiated pathways of care for elderly patients may indeed be beneficial. Bradburn et al4 compared the mortality rate of geriatric trauma patients before and after the implementation of a high-risk geriatric protocol. Their protocol included a consultation with a geriatrician, monitoring in the intensive care unit, and laboratory tests, all of which were designed to prevent a complication after injury. The patients who received geriatric protocol had lower odds of mortality. Similarly, Lenartowicz et al25 evaluated differences in geriatric-specific clinical outcome, trauma quality indicators, and discharge destination before and after the implementation of a geriatric trauma consultation service. At their institution, the geriatric trauma consultation service was consulted for all trauma patients aged 60 years and older admitted to the trauma service to be evaluated for any geriatric issues. In addition to a significantly lower incidence of delirium, they found that severely injured geriatric patients were less likely to be transferred to a long-term acute care facility in the postgeriatric trauma consultation service period. In a separate study, a designated unit for geriatric patients with acute medical disorders was shown to be associated with a lower risk for functional decline and a higher rate of being discharged home.34 It remains to be determined whether such preferred pathways for care can offset the potentially detrimental influence of case volume for nongeriatric patients.
There were methodological limitations in our study. First, we could not adjust for patient comorbidities using a quantitative method (ie, Charlson Comorbidity Index score) because the list of preexisting conditions by the PTOS did not cover all relevant comorbidities.35 However, we adjusted for individual groups of comorbidities and the PTOS data had a low rate of patients with missing data. Second, we could not follow up the postdischarge mortality but only evaluated the mortality during the hospital stay. Third, outcomes captured in the PTOS could be influenced by variations in the availability of long-term acute care hospitals or nursing homes or variability in criteria used to determine transfer. Finally, our findings may not apply to other regions of the country or countries in which there may be different mechanisms of injury or differences in access to care in an accredited trauma center. For example, it might be postulated that better outcomes might be observed in some hospitals owing to a higher prevalence of admissions for accidental fall and such admissions might be observed more frequently in rural rather than urban hospitals. Based on the comparability of case mix in our patient population with that reported in a recent Canadian study,8 we believe our findings are likely to be generalizable to other systems in which geriatric trauma is common.
A larger hospital volume of geriatric trauma patients and nongeriatric trauma patients impacted the outcomes of geriatric trauma patients. A significantly lower major complication rate and the higher rate of the successful rescue of patients who developed major complication appear to be crucial associated factors for better outcomes at higher geriatric volume centers. These results should help focus the discussion about how to allocate geriatric trauma patients to appropriate receiving centers.
Corresponding Author: Kazuhide Matsushima, MD, Department of Surgery, University of Southern California, 1200 N State St, Inpatient Tower (C), Room C5L100, Los Angeles, CA 90033 (firstname.lastname@example.org).
Accepted for Publication: July 24, 2013.
Published Online: January 22, 2014. doi:10.1001/jamasurg.2013.4834.
Author Contributions: Dr Matsushima and Mr Schaefer had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Matsushima, Won, Indeck, Soybel.
Acquisition of data: Matsushima, Won.
Analysis and interpretation of data: Matsushima, Schaefer, Won, Armen, Soybel.
Drafting of the manuscript: Matsushima, Won, Soybel.
Critical revision of the manuscript for important intellectual content: Matsushima, Schaefer, Armen, Indeck, Soybel.
Statistical analysis: Matsushima, Schaefer.
Obtained funding: Matsushima.
Administrative, technical, and material support: Armen, Soybel.
Study supervision: Armen, Indeck, Soybel.
Conflict of Interest Disclosures: None reported.
Additional Contributions: The data used in this study were provided by the Pennsylvania Trauma Systems Foundation, Mechanicsburg, Pennsylvania. The foundation specifically disclaims responsibility for any analyses, interpretations, or conclusions. The Pennsylvania Trauma Outcome Study (PTOS) was the source of the data.