Risk of trauma-related hospitalization among previously injured and uninjured cohorts.
McGwin G, May AK, Melton SM, Reiff DA, Rue LW. Recurrent Trauma in Elderly Patients. Arch Surg. 2001;136(2):197-203. doi:10.1001/archsurg.136.2.197
Older patients (those aged ≥70 years) who have experienced trauma have an increased risk of recurrent trauma. Demographic, medical, and functional factors are potential contributors to the risk of subsequent trauma among injured elderly patients.
Retrospective follow-up study.
Study participants were derived from the Longitudinal Study of Aging, an extension of the 1984 National Health Interview Survey focusing on persons who were aged 70 years and older in 1984. A cohort of elderly patients participating in the Longitudinal Study of Aging and hospitalized for injury in 1985 (n
= 100) was identified using Medicare hospital discharge data. An uninjured cohort (n = 401) was also identified from the Longitudinal Study of Aging and matched for age (1 year) and sex.
Main Outcome Measures
Risk of admission for trauma among the injured cohort compared with the uninjured cohort and associations between demographic, medical, and functional characteristics and trauma recurrence.
Following adjustment for potential confounding factors, the injured cohort was 3.25 times more likely (95% confidence interval, 1.99-5.31) to be hospitalized for injury during the follow-up period compared with the uninjured cohort. Among the injured cohort, those at greatest risk of subsequent trauma included women and those with chronic medical conditions or functional impairments, the latter being the only factor independently associated with recurrence.
Elderly patients who have experienced trauma are at increased risk of subsequent injury. Interventions to reduce the likelihood of trauma recurrence should focus on those with chronic illnesses and functional impairments.
IN THE United States and many other countries, injury-related mortality rates are highest for young adults and elderly persons1; for unintentional injuries, elderly persons have the highest mortality rate.2 Among this age group, the absolute risk of death from injury is several times less than that from other causes, including heart disease, cancer, and diabetes mellitus. Nevertheless, given an injury, the risk of death is higher for elderly patients.3- 19 For nonfatal injury, hospital discharge rates are also highest among elderly persons.2 The impact of trauma among elderly patients on health care expenses and postinjury quality of life also attests to the importance of this public health issue.20- 25 Geriatric trauma is likely going to become an issue of increasing magnitude, as the number of older adults (those aged ≥65 years) increased by 2.6 million or 8% since 1990, compared with an increase of 6% for the younger than 65 years population. These increases are expected to continue through the year 2030, at which point the proportion of the US population aged 65 years and older will have increased to 21.8%. Research into the causes and prevention of injury among elderly persons has been outpaced by similar research focusing on young adults.26- 28 To effectively prevent and manage trauma in elderly patients, it is imperative that those at greatest risk for injury and poor outcomes be identified.
Trauma is rarely characterized as a chronic disease; however, research29- 38 suggests that trauma is not episodic but rather, in some populations, recurrent. Such findings can have a direct impact on injury prevention initiatives by characterizing those who are likely to have repeated incidents of trauma. Education and counseling resources can then be directed toward such individuals during the initial hospitalization in an attempt to reduce the likelihood of future events. To date, research30- 38 has identified several factors associated with trauma recurrence, including young age, male sex, penetrating trauma, and positive blood alcohol levels. Only one study29 has specifically investigated recurrent all-cause trauma in elderly patients. This retrospective study used Health Care Financing Administration data to compare the risk of injury among a cohort of persons aged 66 years and older who had experienced subsequent trauma with that of a cohort that did not. In addition to a significant association between past and subsequent trauma, the results also suggested that comorbid conditions might also be a risk factor for trauma recurrence. However, this study evaluated a limited set of variables (age, race, sex, and comorbid conditions) for their association with trauma recurrence. Assessing the role of a broader array of measures may lead to the identification of further risk factors that can then be used to create a profile of the elderly patient who has experienced trauma and who is at risk of subsequent injury.
Given that falls represent most injuries in elderly patients, research in this area is likely to be instructive for understanding injuries in general in this population. For example, the hypothesis that a history of falls is associated with future falls has been demonstrated.39 Perhaps more important, a growing body of research indicates that injuries due to falls are the result of a combination of intrinsic, situational, and environmental factors40- 45 and that multifactorial interventions focusing on such risk factors appear to reduce the risk of falling.46,47 Interventions that successfully reduce the risk of recurrent falls may benefit elderly patients who have sustained non–fall-related trauma.
The goals of this retrospective cohort study are to compare the risk of injury in a cohort of elderly patients who have experienced an injury with that in a cohort of uninjured elderly patients and to evaluate demographic, medical, and functional risk factors for recurrent trauma.
The data for this study were obtained from the Longitudinal Study of Aging (LSOA). Briefly, the LSOA is an extension of the 1984 National Health Interview Survey, following up its sample of 16 148 noninstitutionalized elderly people living in the United States, with a special focus on those who were aged 70 years and older in 1984. Because of the sampling strategy used, 7527 persons aged 70 years and older in 1984 make up the analytic cohort. Persons interviewed in 1984 were reinterviewed for the LSOA in 1986, 1988, and 1990. The response rates for the reinterviewed samples (excluding decedents) in 1986, 1988, and 1990 were 90%, 83%, and 79%, respectively. The LSOA also includes mortality data from the National Death Index and morbidity data from Medicare hospital records and other Medicare use records, including information on home health care use, hospice care, and outpatient use. Greater detail on the design and execution of the LSOA is available elsewhere.48,49
The injured and uninjured cohorts were identified using data from the Medicare Automated Data Retrieval System, part A and B tapes, which are included as part of the LSOA data sheet. The selection of the injured and uninjured cohorts began by selecting subjects with a hospital discharge diagnosis for injury (International Classification of Diseases, Ninth Revision, Clinical Modification50 codes 800-959) during 1985; these subjects were classified as the injured cohort (excluded were those with International Classification of Diseases, Ninth Revision, Clinical Modification codes 905-909 [late effects], 930-939 [foreign bodies], and 958 [complications]). Beginning in January 1, 1985, and proceeding until December 31, 1985, subjects who were hospitalized for injury were selected as members of the injured cohort. For each injured subject, 4 subjects who had not been hospitalized for injury up to and as of the date of hospitalization for the injured subject were selected from the LSOA cohort and matched to the injured subjects on age (1 year) and sex. Members of the uninjured cohort were also required to be alive at the time of the injured subjects' hospitalization. The year 1985 was selected for the identification of the cohorts instead of 1984 and 1985 because baseline interviews were conducted throughout 1984; there was concern that information would have been obtained following injury rather than before it.
The outcome of interest for this study was hospitalization for trauma subsequent to the date of the initial hospitalization. Each subject accumulated person-years from the date of the injury-related hospitalization to the date of subsequent hospitalization, death, or the end of follow-up (December 31, 1990), whichever came first. (For members of the uninjured cohort, the date of hospitalization for the injured subject to which they were matched was used for the calculation of person-time.) Information on the cause of the injury was not available.
In addition to standard demographic information (age, sex, and race), variables pertaining to health and functional status were also extracted from the database.
At the baseline (ie, 1984) interview, all subjects were asked if they had any of the following medical conditions: cataracts, glaucoma, osteoporosis, heart disease (eg, atherosclerosis or coronary artery disease), hypertension, stroke, Alzheimer disease, cancer, arthritis, or diabetes mellitus. A composite variable was also created from these data by summing the total number of these conditions that each subject reported having. Two additional measures of health status were also used: number of bed days (related to illness) in the past 12 months and number of physician visits in the past 12 months.
Three sets of functional status measures were available from the LSOA data set. The first set of measures focused on activities of daily living (ADLs) and queried subjects as to whether they had difficulty bathing, dressing, eating, transferring, walking, going outside, or using the toilet.51 The second set of measures was based on the Instrumental Activities of Daily Living (IADLs) Scale of the Older Americans Resource and Services Survey.52 Subjects were asked whether they had difficulty preparing meals, shopping for personal items, managing money, or using the telephone. Finally, functional limitations were assessed using items from the Nagi Disability Scale, including inability or difficulty in walking short distances, walking up 10 stairs, standing, sitting, stooping, reaching overhead, reaching out, using fingers, lifting light weights, and lifting heavy weights.53 For each set of measures (ADLs, IADLs, and the Nagi Disability Scale), a composite summary variable was created by summing the number of items with which subjects reported difficulty.
The functional measures are also available at the 1986 LSOA reinterview. For each set of measures (ADLs, IADLs, and the Nagi Disability Scale), a second set of composite summary variables was created by summing the number of items with which subjects reported difficulty in 1986. Using these summary variables, 3 change-in-status variables were created by subtracting the summary variables at the 1986 reinterview from the variables at the baseline interview. Therefore, positive values for these change-in-status variables represent a decline in function (ie, difficulty with more items). Finally, each change-in-status variable was dichotomized into those whose function declined and those whose function remained the same or improved.
Descriptive statistics for demographic, health, and functional status characteristics were generated and compared between the injured and uninjured cohorts using χ2 and t tests for categorical and continuous variables, respectively. Two primary sets of inferential analyses were conducted; the first set of analyses focused on the hypothesis that prior injury is a risk factor for subsequent injury. Kaplan-Meier curves were constructed to compare episodes of injury among the injured and uninjured cohorts. Relative risks (RRs) and 95% confidence intervals (CIs) comparing episodes of injury among the injured cohort with those of the uninjured cohort were calculated using Cox proportional hazards regression. The association between prior and new injury in these regression models was adjusted for age, sex, race, health (chronic medical conditions, self-reported health status, bed days, and physician visits in the past 12 months), and functional status measures.
The second set of analyses focused on the identification of demographic, medical, and functional risk factors for recurrent trauma. For these analyses, only the injured cohort was used and the change-in-status variables rather than the baseline variables were used for assessment of functional status. Kaplan-Meier curves were constructed to compare the recurrence functions according to selected characteristics (eg, presence of specific medical conditions), and the log-rank test was used to compare the functions. Cox proportional hazards regression was used to calculate RRs and 95% CIs for the association between demographic, medical, and functional characteristics and trauma recurrence. All variables found to be significant in univariate analyses were included in a multivariable regression model. Only those variables with significant, independent associations with recurrence were included in the final, multivariable regression model. All P values were 2-sided, and the significance level chosen was .05.
During 1985, 100 subjects were hospitalized for injury and, therefore, represented the injured cohort. Four hundred one uninjured subjects (4:1 ratio) were matched for age and sex to the injured cohort. For one injured subject, one of the randomly selected matched controls had missing information for certain variables. Therefore, one additional control was matched to this one injured subject. The cumulative incidence of injury-related hospitalization for the year 1985 was 20.1 per 1000 person-years. Table 1 presents the types of injuries sustained by the injured cohort.
Because of the matching, the demographic characteristics of the injured and uninjured cohorts were similar (Table 2). There were, however, fewer black subjects in the injured cohort compared with the uninjured cohort, but this difference was of borderline statistical significance. With several exceptions, the baseline health and functional characteristics of the 2 cohorts were also similar. The injured cohort was more likely to have had an increase in functional impairments compared with the uninjured cohort. The injured cohort also had more ADL limitations at baseline and was more likely to experience an increase in the number of ADL limitations.
Figure 1 presents the Kaplan-Meier plot for the risk of hospitalization for trauma among the injured and uninjured cohorts. The injured cohort was significantly more likely to be hospitalized for injury than the uninjured cohort (P<.001). The injured cohort was 3.13 times more likely to be hospitalized for injury during the subsequent 6 years, compared with the uninjured cohort (Table 3). Adjustment for demographic, medical, and functional characteristics did not materially influence this association.
Table 4 presents RRs and 95% CIs for the risk of trauma recurrence among the injured cohort. Subjects who experienced an increase in the number of ADL limitations had more than a 3-fold increased risk of subsequent trauma. Subjects with one or more IADL limitations also had a significantly increased risk of readmission for trauma. Similarly, those subjects experiencing an increase in the number of IADL limitations were more likely to experience subsequent trauma. However, elevated point estimates were observed for women, subjects with 1 to 2 or 3 to 10 chronic medical conditions, and those who experienced an increase in functional limitations. In the final multivariable model, only one variable demonstrated a significant (P = .002), independent association with the risk of trauma recurrence. Subjects experiencing an increase in the number of ADL limitations were more likely (RR, 3.16; 95% CI, 1.43-7.00) to have a subsequent injury-related hospitalization. Women had an increased risk of recurrence (RR, 3.58; 95% CI, 0.90-15.06), but the association was of marginal statistical significance (P = .06).
The topic of recurrent trauma or "trauma recidivism" has been addressed by several researchers.29- 38 With few exceptions, these studies have limited relevance to the present study in that most have focused on younger patient populations38 and have been limited to mostly urban trauma centers.31- 34,36,37 These studies have reached the homogeneous conclusion that those at risk of recurrent trauma are more likely to be black and men, to be involved in injury episodes related to alcohol, and to have penetrating injuries. Despite the select nature of these study populations, at least one population-based study of recurrent trauma has supported these general conclusions. Using data from the Nevada State Trauma Registry, Kaufman et al30 reported an increased incidence of recidivism among young men. These researchers also found that those who had injuries involving falls and cutting or piercing were at increased risk, as were those involved in intentional injury events; behavioral factors, such as positive blood alcohol levels and lack of seat belt use, were also associated with a greater likelihood of recurrent trauma. These data clearly indicate that a subgroup of the trauma patient population may benefit from interventions to reduce injury recurrence. In fact, a recent study by Gentilello et al54 screened patients who had experienced trauma for alcohol abuse (using blood alcohol levels and the short Michigan Alcoholism Screening Test), and those testing positive for blood alcohol were randomized to receive a brief intervention or not. The results indicated that the intervention group had a 47% reduction in trauma recurrence.
It is encouraging that the existing research on trauma recurrence has evolved to the point at which it is able to influence prevention efforts aimed at a large segment of the population experiencing trauma (ie, youth). However, there remains a growing segment of patients who experience trauma for whom research on injury recurrence is limited, namely, elderly patients. To date, aside from the present study, there has only been one other study in the literature that has addressed the issue of trauma recurrence in elderly patients. Gubler et al29 used data from the Health Care Financing Administration to determine the incidence and risk factors for recurrent trauma among more than 67 000 persons aged 67 years and older. These researchers reported that those with a history of trauma were more than 2 times more likely to experience future trauma, and those at greatest risk were women, the oldest old (those aged ≥82 years), and those with chronic medical conditions. The comparison of this and the present study is facilitated by the fact that both used administrative databases and are population based. In general, our results are consistent with those of Gubler et al in that we also found an increased risk, albeit of larger magnitude, of future injury among those with past injury. We also found increased risks of recurrence among women and those with one or more chronic medical conditions. We did not, however, find an increased risk of readmission for trauma with older age.
This study also has some novel aspects that are worthy of note. In addition to information on chronic medical conditions, we also had data on functional limitations (ADLs and IADLs). These data indicated that subjects with any functional impairment were not only more likely to be initially hospitalized for injury but also more likely to experience subsequent trauma. Furthermore, because of the longitudinal nature of the LSOA, we had information on changes in functional status. Using these data, it was determined that a decline in functional status (specifically in ADLs) from preinjury to postinjury was the single strongest predictor of injury recurrence. That is, injured subjects who experienced an increase in the number of ADL items (bathing, dressing, eating, transferring, walking, going outside, and using the toilet) with which they had difficulty from preinjury to postinjury were more likely to experience subsequent injury. Given that falls represent the largest category of injury-related events in elderly patients, it should not be surprising that poor functional status has been associated with first and recurrent fall events.43,45 The consistent findings are beneficial in that the fall prevention literature is extensive and can likely contribute to initiatives directed at all types of injury in elderly patients.
Given the results of this study, interventions that reduce the impact of injury on functional status are likely to have an effect on the risk of injury recurrence. Such interventions might focus on the aggressive management of the elderly patient who experienced trauma in the rehabilitation setting. Research55,56 supports the contention that more aggressive rehabilitation management planning can be effective in returning patients to a preinjury level of functioning. Fall injury prevention programs are also likely to provide insight as to effective interventions to reduce recurrent injuries in general. Several studies57- 61 have evaluated interventions focused on improving mobility and physical fitness, identifying (and modifying) environmental and personal safety risks, and modifying psychotropic drug use. The growing consensus in this literature is that multifactorial interventions are likely to be most successful for reducing the risk of initial and recurrent falls. It is suggested that patients at risk of falls based on known risk factors44 and those who have experienced a fall should be referred to programs that attempt to modify these factors in an attempt to reduce first and recurrent falls. Although lacking empirical support, it is likely that the beneficial impact of these programs will extend to elderly patients experiencing non–fall-related injuries. As the proportion of elderly persons increases and the treatment of trauma in this population continues to improve, the need for rehabilitative care for elderly patients is likely to increase. Thus, determining the most effective approach to this care represents an important area for future research, and lessons learned from the falls literature, namely, multifactorial perspectives that address medical, functional, and environmental factors, are likely to be instructive.
The results of this study should be interpreted in light of several limitations. Information on chronic medical conditions and functional status was obtained via self-report. Research62- 65 has demonstrated that there is excellent agreement between self-report and medical record diagnosis for most chronic medical conditions. Research on the validity of self-reported functional status in elderly patients is equivocal; while some studies66 have reported good agreement between actual and self-reported function, others67,68 suggest caution when using these measures. However, for any lack of agreement to affect the results of this study, the injured and uninjured cohorts must be differential in their accuracy of reporting information; we have little reason to suspect that this is the case. The use of administrative databases, such as those used in the present study, for the identification of patients who have experienced trauma has been questioned and the results have been conflicting.69,70 It is possible that the patients who experienced trauma and were identified in this study represent a subset of all "true" patients who experience trauma in the study population. However, it is unlikely that it is a differential subset and, therefore, the likely impact on the results is negligible. Finally, there is at least one competing explanation for the study's main finding. It is possible that the observed association between functional decline and trauma recurrence is confounded by injury severity. Gubler et al29 reported a positive association between injury severity and risk of injury recurrence. These researchers suggest that patients experiencing severe injury may never regain preinjury functioning, thereby being at increased risk of future injury. Without information on functional status, it was not possible for these researchers to address this hypothesis. However, the results of the present study provide supplementary evidence that this hypothesis may be correct. Unfortunately, information on injury severity was not available as part of this study, and further research is necessary to determine the exact nature of this relation. Nevertheless, it is possible that measures of functional impairment may be more useful than injury severity for the development of prevention initiatives, as severity scores may not reflect the impact of trauma on an individual's function.
The results of this study support the hypothesis that trauma is a recurrent disease in elderly persons, independent of other potential risk factors. Furthermore, among older patients who experience trauma, those who experience an increase in ADL limitations appear to be at increased risk of future injury.
This study was made possible by the Center for Research in Applied Gerontology, University of Alabama at Birmingham.
Corresponding author and reprints: Gerald McGwin, Jr, MS, PhD, Center for Injury Sciences, Epidemiology Unit, University of Alabama at Birmingham, 120 Kracke Bldg, 1922 7th Ave S, Birmingham, AL 35294-0016 (e-mail: firstname.lastname@example.org).