ELDERS (age 65 years) are the fastestgrowing

Transcription

1 ACADEMIC EMERGENCY MEDICINE March 2001, Volume 8, Number Emergency Department Utilization by Noninstitutionalized Elders MANISH N. SHAH, MD, PAUL J. RATHOUZ, PHD, MARSHALL H. CHIN, MD, MPH Abstract. Objectives: To the best of the authors knowledge, no nationally representative, populationbased study has characterized the proportion of elders using the emergency department (ED) and factors associated with ED use by elders. This article describes the proportion of elder Medicare beneficiaries using the ED and identifies attributes associated with elder ED users as compared with nonusers. Methods: The 1993 Medicare Current Beneficiary Survey was used, a national, population-based, crosssectional survey of Medicare beneficiaries linked with Medicare claims data. The study population was limited to 9,784 noninstitutionalized individuals aged 66 years or older. The Andersen model of health service utilization was used, which explains variation in ED use through a combination of predisposing (demographic and social), enabling (access to care), and need (comorbidity and health status) characteristics. Results: Eighteen percent of the sample used the ED at least once during Univariate analysis showed ED users were older; were less educated and lived alone; had lower income and higher Charlson Comorbidity Index scores; and were less satisfied with their ability to access care than nonusers (p < 0.01, chi-square). Logistic regression identified older age, less education, living alone, higher comorbidity scores, worse reported health, and increased difficulties with activities of daily living as factors associated with ED use (p < 0.05). Need characteristics predicted ED use with the greatest accuracy. Conclusions: The proportion of elder ED users is slightly higher than previously reported among Medicare beneficiaries. Need (comorbidity and health status) characteristics predict ED utilization with the greatest accuracy. Key words: geriatrics; emergency medicine; utilization. ACADEMIC EMERGENCY MEDICINE 2001; 8: ELDERS (age 65 years) are the fastestgrowing segment of the population. In 1990, 31 million individuals were aged 65 years or over, and it is estimated that in 2030, 70 million individuals will be aged 65 or over. 1 Although comprising only 12% of the population in 1990, elders accounted for 15% of emergency department (ED) visits and 31% of ED expenditures. Elders made up 43% of all ED admissions and 48% of ED admissions to the intensive care unit (ICU). Finally, elder patients in the ED received 50% more radiographic and laboratory testing than younger indi- From the Section of Emergency Medicine (MNS) and the Section of General Internal Medicine (MHC), the University of Chicago, Chicago, IL (MNS); the Robert Wood Johnson Clinical Scholars Program, the University of Chicago, Chicago, IL (MNS, PJR, MHC); and the Department of Health Studies, the University of Chicago, Chicago, IL (PJR). Received September 5, 2000; revision received November 9, 2000; accepted November 20, Presented in part at the SAEM annual meeting, San Francisco, CA, May Dr. Chin is a Robert Wood Johnson Foundation Generalist Physician Faculty Scholar. Address for correspondence and reprints: Manish N. Shah, MD, Robert Wood Johnson Clinical Scholars Program, The University of Chicago, 5841 South Maryland Avenue, MC 2007 B-219, Chicago, IL Fax: ; viduals. 2 Despite the size and significance of the elder population, most existing data on ED utilization are from small, individual studies. Only the National Hospital Ambulatory Medical Care Survey (NHAMCS), through probability sampling of ED visit information, has allowed derivation of the rate of ED utilization by elder individuals. 3 However, we are unaware of any national, populationbased studies that characterize and identify factors associated with ED utilization by elders. Identifying and modifying factors responsible for ED use during an illness episode may result in better use of emergency care and better health outcomes Andersen and Newman have grouped factors generally associated with health care utilization into three domains: predisposing, enabling, and need characteristics. 12 Predisposing characteristics exist before the onset of illness and include demographic variables, such as age, sex, and marital status, and belief variables, such as trust in the health care system. Enabling, or access to care, characteristics allow an individual to obtain services. They include family resources such as income, insurance, and a usual source of health care. The third category is comprised of need characteristics, namely, comorbidity and health status. These characteristics are measured by symptom

2 268 ELDERS Shah et al. ED UTILIZATION BY ELDERS TABLE 1. Population Description* Sample (n = 9,784) Young-old (age 66 84) (n = 8,352) Old-old (age 85) (n = 1,432) Age mean 74.7 yr 73.3 yr 88.5 yr Gender female 59% 58% 69% Race White 89% 89% 88% African American 8.0% 8.0% 8.2% Education < 12 years 42% 40% 56% Income < $25,000 74% 73% 87% ED visit 1 18% 17% 26% Visits per 100 persons Mortality (annual) 0.70% 0.57% 2.0% *Weighted analysis performed. reports, functional limitations, and comorbidity scores. Previous studies outside of emergency medicine have shown that this model can account for up to 28% of the variation in health care use The Andersen model, which has been used to evaluate primary care and inpatient settings, has been applied infrequently in the ED. 14,15 Parboosingh and Larsen used the model and found that attitude toward health care, previous experience with health care, hospital admissions, and number of sources of care predicted ED use. 14 This study was limited because it was not population-based and it included a small number of patients. A second study, performed by Wolinsky et al., sampled 401 elder individuals and found that increased ED use was associated with poor nutritional status and being widowed, and was inversely related to having a regular physician. Wolinsky and colleagues also found that need factors determined the majority of variation in use. 15 We hypothesized that a large proportion of elder individuals uses the ED for care. We further hypothesized that ED users are sicker, are of lower socioeconomic status, have no supplemental insurance, have no usual source of care, and have trouble obtaining health care as compared with nonusers. METHODS Study Design. We performed a cross-sectional study of Medicare beneficiaries using data from the 1993 Medicare Current Beneficiary Survey (MCBS), a publicly available, longitudinal panel survey of Medicare beneficiaries sponsored by the Health Care Financing Administration. 17,18 Viewed cross-sectionally at any year, the MCBS is a nationally representative, population-based survey of elder Medicare recipients linked to Medicare claims data. The MCBS sample is drawn from the Medicare enrollment file with an oversample of disabled individuals and individuals aged 85 and older. Sampling was based on 107 primary units, ZIP codes within the units, and then systematic random sampling in the subunits (ZIP codes). Interviews were performed with patient or proxy (12%) using computer-assisted patient interview methods. The response rate for the 1993 sample was 83%. 19 Cross-sectional weights were then calculated so that the data reflect the Medicare population. All individuals surveyed by the MCBS were enrolled in Medicare at the beginning of We obtained institutional review board exemption for this study. Study Population. The population sample was limited to noninstitutionalized individuals 66 years and older who were followed by the MCBS. We excluded 65-year-old persons because the process of deriving comorbidity scores required that the people be enrolled in Medicare during the previous year. These exclusion criteria limited the sample from 12,863 to 9,784 individuals. Study Protocol. The primary outcome variable was the presence or absence of at least one ED claim during 1993, identified by evaluating whether the individual had any Part B Medicare claims from an ED source. We identified predictor variables from the MCBS survey and grouped them by the characteristics of the Andersen model. We examined predisposing factors such as demographic data and whether the individual lived alone; enabling factors such as income, level of supplemental insurance coverage, the presence of a usual source of medical care, and reasons for delay to care; and need factors such as self-reported health status, deficiencies in activities of daily living (ADLs), and the Charlson Comorbidity Index score. 20,21 We used Medicare claims information in the MCBS to calculate the comorbidity score using an adaptation of the Deyo and Romano versions of the Charlson Comorbidity Index, which is validated for use with claims data to predict mortality following hospitalization. 22,23 The scoring system assigns a weight to each of a number of comorbid conditions such as renal failure, stroke, myocardial infarction, and neoplasms. The score formed by summing the weights then allows stratification into categories with varying mortality rates, with a higher score associated with a higher mortality rate. We elected to collapse the Charlson Comorbidity Index into four categories: a score of 0 (healthiest), 1 2, 3 4, and 5, and the number of

3 ACADEMIC EMERGENCY MEDICINE March 2001, Volume 8, Number deficiencies in ADL score into three categories: 0, 1 2, and 3. Data Analysis. First, we performed univariate analysis to determine associations between potential predictor variables and the presence of at least one ED visit during We used chi-square tests for binomial and ordinal variables. We evaluated continuous variables with the t-test comparing the ED-visit group with the no-ed-visit group. To identify potential patterns within the data, we performed stratified analyses by type of supplemental insurance and by the Charlson Comorbidity Index score. We chose these variables because we hypothesized that they represent major factors contributing to rates of ED utilization, and because we believed an interaction could exist between these and other factors. The predisposing, enabling, and need characteristics were then reexamined. Association between the predictor and outcome variable was determined by using logistic regression models with interaction terms. We used multivariable logistic regression to identify characteristics associated with ED utilization while controlling for other factors and to assess the variation explained by the utilization models. We entered factors into the multivariable regression model if, on univariate analysis, they were associated with ED use with a p-value less than 0.20 or if we judged a priori that it was important either as a predictor variable or as an adjustor. These decisions were based on clinical judgment and data available in the literature. We decided that age, sex, education, insurance status, income level, living arrangement, delay of care due to cost, and presence of a usual source of care would be automatically included in the model. We evaluated the accuracy of the multivariable logistic regression models via the concordance index for each model. 24 The concordance index estimates the probability of agreement between predicted and observed responses in a randomly selected pair of subjects. A value of 0.50, therefore, indicates that the model is no more accurate for predicting ED use than crude prevalence data would be and a value of 1.0 indicates perfect accuracy to predict ED use. Thus, the concordance index can be used to describe the accuracy of a logistic regression model in determining the outcome. Since the MCBS oversamples certain populations, we were required to factor in the sample weights and replicate weights that were provided as part of the MCBS data. We accomplished this by using the sample weight options within the statistical software. We used the MCBS replicate weights to derive variances, thus allowing us to correctly calculate confidence intervals. We used STATA 6.0 (College Station, TX) for all TABLE 2. Prevalence of Characteristics as Organized by the Andersen Model* No ED Visit (%) ED Visit (%) p-value Predisposing characteristics Age 85 years <0.001 Gender female Race <0.001 White African American Education < 12 years <0.001 Lives alone Enabling characteristics Income < $25, <0.001 Supplemental insurance <0.001 None Private Public Lack usual source of care <0.001 Trouble getting care <0.001 Delayed care due to cost Need characteristics General state of health <0.001 Excellent Very good Good Fair Poor Activities of daily living deficiencies < Charlson Comorbidity Index < *Some totals are less than 100% due to the exclusion of the refuse to answer/don t know category. Weighted analysis performed. data analysis and used a p-value of 0.05 for statistical significance. RESULTS Table 1 describes the population sample used for analysis, and stratifies by age group. The old-old (age 85) have a higher ED utilization rate, a lower income, and less education, and are more often female than the young-old (age 66 84). Of particular interest, we found that individuals aged 85 years or older have 43 ED visits per 100 persons per year and individuals between 66 and 84 years of age have 28 ED visits per 100 persons per year. Univariate Analysis. Table 2 shows the distribution of the predictor variables by the outcome variable, the presence or absence of any ED visits

4 270 ELDERS Shah et al. ED UTILIZATION BY ELDERS TABLE 3. Logistic Regression Model of ED Utilization* Odds Ratio 95% CI Predisposing characteristics Age 85 years , 1.43 Gender female , 1.30 Education < 12 years , 1.41 Race White Reference African American , 1.32 Other , 5.72 Lives alone , 1.29 Enabling characteristics Income < $25, , 1.37 Supplemental insurance Private Reference Public , 1.32 None , 1.30 Trouble obtaining care , 1.57 Delay care due to cost , 1.18 Lacking usual source of care , 1.28 Need characteristics Self-reported health Excellent Reference Very good , 1.60 Good , 1.73 Fair , 2.18 Poor , 3.31 Activities of daily living deficiencies 0 Reference , , 2.11 Charlson Comorbidity Index Score 0 Reference , , , 11.0 *Weighted analysis performed. during the year. The factors are grouped by the three domains (predisposing, enabling, and need) of the Andersen model. For example, of those using the ED for care, 60.9% were female. In contrast, of those not using the ED, 58.7% was female. The ED users were more likely to be older, have lower incomes, be less educated, and live alone than the nonusers (Table 2). The ED users were also more likely to have publicly-funded or no supplemental insurance and were sicker and less functional than the nonusers. Interestingly, they were also more likely to identify a usual source of care. Multivariable Analysis. Table 3 describes the results from the multivariable regression model predicting at least one ED visit during the year. In separate models, the concordance index for only predisposing characteristics was 0.59, for only enabling characteristics was 0.58, and for only need characteristics was Thus, for predisposing and enabling characteristics the models had minimal accuracy for predicting ED use, while it was considerably better for the need characteristics. The concordance index for the complete model with all characteristics was 0.75, essentially the same as that for need characteristics. Stratified Analysis. Stratification by supplemental insurance plan and Charlson Comorbidity Index score provided additional details not discernible with the univariate or regression analysis. The data from insurance stratification showed that among people without Medicare supplemental insurance, a greater trend to ED use was associated with problems obtaining health care (p = 0.09 for interaction term) and delayed care due to cost (p = 0.05 for interaction term) as compared with individuals with private supplemental insurance. Also, among those individuals with Medicaid supplemental insurance as compared with private supplemental insurance, ED users tended to be more strongly associated with not having a usual place for care (p < for interaction term). Stratification by Charlson Comorbidity Index showed that only sicker individuals (Charlson Index 5) were more likely to use the ED if they reported increased difficulty getting to their physicians (p = 0.04 for interaction term). DISCUSSION Epidemiology of ED Use. This nationwide, population-based study shows that the proportion of elder Medicare beneficiaries using the ED is slightly greater than previously reported by other studies of Medicare data. Other studies, using selfreports or Medicare Part A (institutional) claims data, have shown that between 11.9% and 16.1% of elders have at least one ED visit per year. 25,26 A more recent study using Part B (physician) claims data from Washington State identified a rate of 18.1%, similar to our rate of 18%. 27 Whether the increased rate of ED utilization in claims data, as compared with self-reports, represents more valid data collection or temporal changes in utilization rates is unknown. Although we found a greater proportion of elders using the ED, we found a slightly lower rate of ED use than the NHAMCS. The 1992 NHAMCS survey found that individuals aged had 31 visits per 100 persons per year and individuals aged 75 and over had 56 visits per person per year. 28 The most likely explanation for this difference is the failure of physicians to submit claims to Medicare, resulting in no record of the ED visit. Another explanation includes potential differences between the Medicare population and the elder population at large.

5 ACADEMIC EMERGENCY MEDICINE March 2001, Volume 8, Number Predictors of ED Use. Multivariable analysis eliminated most factors as significant covariates of ED utilization (Table 3). We found that age greater than 85 years, living alone, less education, increased Charlson Comorbidity Index score, worse reported state of health, and increased ADL deficiencies were associated with ED use. Our findings contradict some findings in the smaller study by Wolinsky et al. In the multivariable analysis, we did not find any relationship of ED use to the presence or absence of a usual source of care and we did find a significant association with ADL deficiencies and comorbidity scores (Table 3). 15 Although we cannot identify the causes for these findings because this is a cross-sectional analysis of observational data, we are able to identify two major themes from the data. The first theme is that the comorbidity and health status factors, or need characteristics, predict the majority of variation in ED utilization, far exceeding the contribution of predisposing or enabling factors. However, because of the limitations of the study design, it is impossible to definitely know the order of causation, particularly in relation to the other associated factors. For example, it is possible that improved access to care might prevent the worsened health status associated with increased ED utilization. Explanations will not be available until prospective or longitudinal studies are completed. A second theme is that enabling characteristics of the Andersen model, which represent access to care factors, are not significantly associated with ED utilization in the multivariable models. In particular, financial barriers, as measured by a lack of supplemental insurance, lower income, or a reported delay to care due to cost, are not associated with ED use after adjusting for other factors. It is possible that the presence of Medicare insurance for the entire population has removed some of the barriers. However, the stratification analyses suggest that not having private supplemental insurance is associated with potential barriers to accessing health care. More detailed studies on access to care for elders are needed, particularly regarding issues such as prescription benefits. 29 Nonfinancial barriers were also not associated with ED use in this population. Lacking a usual source of care was not associated with ED use, as suggested in other literature. 15,30 Future Interventions. As the elder population grows in the coming years, the high rates of ED use among this population that we and others have found indicate that EDs will experience an increase in volume of elder patients. This may lead to a need for additional resources, including patient care facilities and staff specially trained to help elder patients, such as social workers. More importantly, emergency physicians will need additional training in geriatric issues, which currently have limited coverage in the emergency medicine residency curriculum. 31 If comorbidity and health status factors explain the majority of ED use by elders, it may be very difficult to design and implement ED-based interventions to reduce ED use by elders. However, if some ED use is considered a failure of outpatient care, then more broad-based system interventions may be warranted to improve the health of this population. Of all of the predisposing and enabling characteristics, only living arrangement was associated with ED utilization by the elders, and this association was weak. Although further research is needed to identify why living alone is associated with increased ED use, the lack of social supports may lead to a state where the individual depends upon the ED, the only source of medical care always open, for support and medical advice. 32 A system could be developed to intervene at this point. However, because the predisposing characteristics explain so little of the variance in ED use, this intervention may not have notable impact on overall ED use. Our data suggest that improving comorbidities and health status in the outpatient setting might be the intervention that would most significantly impact ED use by the elders. This will require aggressive disease prevention, such as public health programs to stop smoking and promote exercise, and improved access to primary care essential to such efforts. In addition, improved chronic disease management of conditions such as heart failure may reduce ED utilization as it has reduced hospital utilization, and it may also improve care and quality of life. 4,33 LIMITATIONS AND FUTURE QUESTIONS This study has a number of limitations. First, as we performed a secondary analysis on the MCBS data set, we could analyze only those factors evaluated by the MCBS. As a result, we were unable to analyze factors such as medical knowledge, locus of control, health beliefs, ED satisfaction, and prescription drug coverage. Second, as this study was an analysis of survey data, it was susceptible to bias from sampling errors. However, the MCBS is a representative sample of the Medicare population. 18 This study was also susceptible to information bias from individuals being placed in incorrect categories. Moreover, the misclassification error is most likely nondifferential, thereby minimizing any association between the predictor and outcome variables. Fourth, we were dependent on the Medicare claims data for evaluating health

6 272 ELDERS Shah et al. ED UTILIZATION BY ELDERS care utilization. Medicare claims data may underestimate the frequency of health care use, particularly if claims are not always filed and paid for each episode of patient care. 34 Fifth, because this was a cross-sectional study, we are limited in our conclusions. We cannot identify temporality. Moreover, we cannot determine to what extent predisposing and enabling factors were associated with the comorbidity and severity of illness so crucial to ED utilization. A nationwide, population-based study that asks elder individuals focused questions that we were unable to assess with this data set, including health beliefs and medical knowledge, is needed. This will help identify a more complete model and will help identify modifiable factors. Following the populations longitudinally is also important. By doing so we can identify how utilization changes as the characteristics of the individual change. Finally, interventions based on modifiable factors need to be designed so as to evaluate whether the level of ED use can be altered. CONCLUSIONS The proportion of elder Medicare beneficiaries using the ED is higher than previously reported, but the rate of ED use is slightly less than reported by the NHAMCS. In the multivariable analysis, older age, less education, living alone, higher comorbidity scores, worse reported health, and increased difficulties with ADLs were associated with ED use. Comorbidity and health status characteristics predicted ED utilization with the greatest accuracy, while demographic and enabling characteristics, such as supplemental insurance status or usual source of care, had minimal predictive accuracy. The authors acknowledge the assistance of James Zhang, PhD, and John Lantos, MD, in performing this study and reviewing the manuscript. References 1. Hobbs FB, Damon BL. 65 in the United States, P Current Population Reports: Special Studies. Washington, DC: U.S. Census Bureau, Strange GR, Chen EH, Sanders AB. Use of emergency departments by elderly patients: projects from a multicenter database. Ann Emerg Med. 1992; 21: Wofford JL, Schwartz E, Timerding BL, et al. Emergency department utilization by the elderly: analysis of the National Hospital Ambulatory Medical Care Survey. Acad Emerg Med. 1996; 3: Bindman AB, Grumbach K, Osmond D, et al. Preventable hospitalizations and access to health care. JAMA. 1995; 274: Bazargan M, Bazargan S, Baker RS. Emergency department utilization, hospital admissions, and physician visits among elderly African American persons. Gerontologist. 1998; 38: Baker DW, Stevens CD, Brook RH. Regular source of ambulatory care and medical care utilization by patients presenting to a public hospital emergency department. JAMA. 1994; 271: Rask KJ, Williams MV, Parker RM, McNagny SE. Obstacles predicting lack of a regular provider and delays in seeking care for patients at an urban public hospital. JAMA. 1994; 271: Safer MA, Tharps QJ, Jackson TC, Leventhal H. Determinants of three stages of delay in seeking care at a medical clinic. Med Care. 1979; 17: Ell K, Haywood LJ, Sobel E, deguzman M, Blumfield D, Ning J. Acute chest pain in African Americans: factors in the delay in seeking emergency care. Am J Public Health. 1994; 84: Gurwitz JH, McLaughlin TJ, Willison DJ, et al. Delayed hospital presentation in patients who have had acute myocardial infarction. Ann Intern Med. 1997; 126: Wester P, Radberg J, Lundgren B, Peltonen M. Factors associated with delayed admission to hospital and in-hospital delays in acute stroke and TIA: a prospective, multi-center study. Stroke. 1999; 20: Andersen R, Newman JF. Societal and individual determinants of medical care utilization in the United States. Milbank Q. 1973; 51: Evashwick C, Rowe G, Diehr P, et al. Factors explaining the use of health care services by the elderly. Health Serv Res. 1984; 19: Parboosingh EJ, Larsen DE. Factors influencing frequency and appropriateness of utilization of the emergency room by the elderly. Med Care. 1987; 25: Wolinsky FD, Coe RM, Miller DK, et al. Health services utilization among the noninstitutionalized elderly. J Health Soc Behav. 1983; 24: Kronenfeld JJ. Provider variables and the utilization of ambulatory care services. J Health Soc Behav. 1978; 19: Olin GL, Liu H. Health & health care of the medicare population. Rockwell, MD: Westat, Nov Adler GS. A profile of the Medicare Current Beneficiary Survey. Health Care Financing Rev. 1994; 15: Medicare Current Beneficiary Survey CY 1993, Section 6, pp Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. Studies of illness in the aged. The Index of ADL: a standardized measure of biological and psychosocial function. JAMA. 1963; 185: Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis. 1987; 40: Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992; 45: Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol. 1993; 46: Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996; 15: Cunningham PJ, Clancy C, Cohen JW, et al. The use of hospital emergency departments for nonurgent health problems: a national perspective. Med Care Res Rev. 1995; 52: Weiss LJ, Blustein J. Faithful patients: the effect of longterm physician patient relationships on the costs and use of health care by older Americans. Am J Public Health. 1996; 86: Rosenblatt RA, Wright GE, Baldwin L, et al. The effect of the doctor patient relationship on emergency department use among the elderly. Am J Public Health. 2000; 90: National Hospital Ambulatory Medical Care Survey Emergency Department Summary. Vital and Health Statistics. 1997; Series 13, No Blustein J. Drug coverage and drug purchases by Medicare beneficiaries with hypertension. Health Aff. 2000; 19: Medicaid Access Study Group. Access of Medicaid recipients to outpatient care. N Engl J Med. 1994; 330:

7 ACADEMIC EMERGENCY MEDICINE March 2001, Volume 8, Number Jones JS, Rousseau WE, Schropp MA, Sanders AB. Geriatric training in emergency medicine residency programs. Ann Emerg Med. 1992; 21: Andren KG. A study of the relationship between social network, perceived ill health, and utilization of emergency care. Scand J Soc Med. 1988; 16: Culler SD, Parchman ML, Przybylski M. Factors related to potentially preventable hospitalizations among the elderly. Med Care. 1998; 36: Fisher ES, Baron JA, Malenka DJ, Barrett J, Bubolz TA. Overcoming potential pitfalls in the use of Medicare data for epidemiologic research. Am J Public Health. 1990; 80: