Abstract
Background
We identified 4-year (2 years before and 2 years after the index [baseline] interview) ED use patterns in older adults and the factors associated with them.
Methods
A secondary analysis of baseline interview data from the nationally representative Survey on Assets and Health Dynamics Among the Oldest Old linked to Medicare claims data. Participants were 4310 self-respondents 70 years old or older. Current Procedural Terminology (CPT) codes 99281 and 99282 identified low-intensity use, and CPT codes 99283–99285 identified high-intensity use. Exploratory factor analysis and multivariable multinomial logistic regression were used.
Results
The majority (56.6%) of participants had no ED visits during the 4-year period. Just 5.7% had only low-intensity ED use patterns, whereas 28.9% used the ED only for high-intensity visits, and 8.7% had a mixture of low-intensity and high-intensity use. Participants with lower immediate word recall scores and those who did not live in major metropolitan areas were more likely to be low-intensity-only ED users. Older individuals, those who did not live in rural counties, had greater morbidity and functional status burdens, and lower immediate word recall scores were more likely to be high-intensity-only ED users. Participants who were older, did not live in major cities, had lower education levels, had greater morbidity and functional status burdens, and lower immediate word recall scores were more likely to have mixed ED use patterns.
Conclusions
Nearly half of these older adults used the ED at least once over a 4-year period, with a mean annual ED use percentage of 18.4. Few, however, used the ED only for visits that may have been avoidable. This finding suggests that triaging Medicare patients would not decrease ED overcrowding, although continued surveillance is necessary to detect potential changes in ED use patterns among older adults.
Keywords: Emergency department; Utilization patterns; Medicare, Claims data
Americans rely on hospital emergency departments in growing numbers because of the skilled specialists and advanced technologies they offer. At the same time, the increasing use of the emergency care system also represents failures of the larger health care system—the growing numbers of uninsured Americans, the limited alternatives available in many communities, and the inadequate preventive care and chronic care received by many. These demands can degrade the quality of emergency care and hinder its ability to provide urgent and life-saving care to seriously ill and injured patients wherever and whenever they need it. (1)
So begins the foreword to the three-volume Institute of Medicine (IOM) Report on the Future of Emergency Care (2). The lead volume in that IOM report defines the problem in a nutshell—a “national epidemic of overcrowded EDs (Emergency Departments) and trauma centers.” In simple economic terms, there is a critical imbalance between demand and supply.
From 1993 to 2003 there was a net loss of 703 hospitals (11%) in the United States, a 198,000 (17%) decline in the number of inpatient beds, and a net decline of 425 EDs (9%) (3). At the same time, the U.S. population grew by 12%, hospital admissions increased by 13%, and ED visits went up by 26% (4). As a result, national studies have found that 91% of EDs report overcrowding as a problem (5), with 40% of all EDs reporting that the problem occurs daily (6), and that academic medical center EDs were overcrowded 35% of the time (7,8).
To understand the demand side of the equation, it helps to consider the payment sources. There were four major payers for the 114 million ED visits in 2003: private health insurance (36%), Medicaid (21%), Medicare (16%), and self-pay (14%) (4). Although the factors associated with the demand for ED services likely differ by payment source, little has been done to identify these factors from a population-based perspective (2,8). This is especially surprising with respect to older adults given the potential demand that the “graying of America” may place on EDs (2). Moreover, the emergency medicine community is concerned that the role of older adults was surprisingly downplayed by the IOM Report (9), especially given the monopsonistic ability to intervene via Medicare payment policy. Accordingly, this article focuses on an initial identification of ED use patterns and the factors associated with them in a large, representative sample of older adults covered by Medicare.
METHODS
Data
We conducted a secondary analysis of the baseline interview data from the Survey on Assets and Health Dynamics Among the Oldest Old (AHEAD). The design and sampling approach in the AHEAD have been well-described elsewhere (10–13). Because African Americans, Hispanics, and Floridians were oversampled, all analyses are weighted to adjust for the unequal probabilities of selection due to the multistage cluster and oversampling. The AHEAD provides a nationally representative probability sample of 4310 men and women who were 70 years old or older and self-respondents at baseline (1993) and whose survey data could be linked to their Medicare claims. Medicare claims were available from January 1991 through December 1996. For each AHEAD participant, we used all Medicare claims available within a 4-year window centered on the date of their baseline in-home interview (i.e., 2 years prior to and 2 years afterward).
Measurement
To identify ED visits in the Medicare claims, we used Current Procedural Terminology (CPT) codes (14). Following the IOM report (2), we relied on CPT codes 99281–99285, which account for >80% of all Medicare expenditures for ED services. CPT codes 99281 and 99282 correspond to payment for a low intensity of physician services, which could have been provided in alternative settings (i.e., physician offices or other outpatient settings), suggesting that such ED visits might have been potentially avoidable or nonurgent (2). In contrast, CPT codes 99284 and 99285 correspond to payment for a high intensity of physician services for life-threatening situations, and for which ED visits are more clearly appropriate. CPT code 99283 corresponds to payment for a medium intensity of physician services which, although not life-threatening, may also be appropriate for ED visits.
Analytic Approach
We began with a two-stage exploratory factor analytic (EFA) approach (15–18). To determine empirical clusters, the first stage of our EFA approach began with 20 variables created from the harvest of the Medicare claims. These variables represented the annual total of ED visits corresponding to each of the five CPT codes. We expected a relatively simple factor structure to emerge from the EFA of these 20 variables (four annual totals for each of five CPT codes) in which the annual numbers of ED visits with CPT codes 99281 and 99282 would all load on one factor, and that the annual numbers of ED visits with CPT codes 99283–99285 would all load on another factor. In the second stage of our EFA approach, we constructed variables representing the annual counts of the numbers of ED visits in each of the two expected clusters (CPT codes 99281 and 99282 and CPT codes 99283–99285) and conducted EFA on those eight variables (the annual number of ED visits in each of the two CPT code groups).
We then examined the evidence of criterion validity (19) for our classification approach consistent with prior triaging studies (20–23). Specifically, we examined arrival by ambulance and hospitalization transition rates, mean length of stay and charges (if admitted), and mortality (by CPT code). We also applied an alternative International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)-based approach developed by Billings and colleagues (24,25) to our data, and cross-classified the resulting mean estimated probabilities of their ED visit classification types by CPT codes. Finally, we compared and contrasted the top ICD-9-CM codes associated with low- versus high-intensity ED visits.
Multivariable multinomial logistic regression (26) was used to compare and contrast characteristics associated with each ED use pattern, using no ED use as the overall reference group. More than 40 variables traditionally used in studying the demand for health care (27) were considered, including sociodemographics, socioeconomics, lifestyle, disease history, and functional health status. Model development and evaluation followed standard procedures (28,29), with the final model restricted to factors with significant (p < .05) independent associations with ED use patterns.
RESULTS
Descriptive
Table 1 contains the means (or proportions), standard deviations, and coding algorithms for each of the sociodemographic, socioeconomic, lifestyle, disease history, and functional health status variables considered as possible predictors of ED use patterns. Among the 4310 AHEAD participants in the analytic sample (weighted N = 4337), the mean age was 77 years, 35% were men, 9% were African American, 4% were Hispanic, and 43% were widowed. One-fourth had only been to grade school, and mean income was $25K. One-fourth had arthritis, 8% had angina, 13% had cancer, 11% had diabetes, 46% had hypertension, 4% had fractured a hip, and 7% had psychological problems. The mean number of activities of daily living (ADL) difficulties was 0.29, and the mean number of instrumental ADL (IADL) difficulties was 0.18. The mean annual percentage of participants having any ED visits was 18.4% (range = 14.8%–20.6%). The 4-year period prevalence rate of participants having any ED visits was 44.4%.
Table 1.
Coding Algorithms, Means (or Proportions), and Standard Deviations of the Potential Predictor Variables of ED Use Patterns Among the 4310 AHEAD Self-Respondents (Weighted N = 4337)
| Variable | Coding Algorithm | Mean | Standard Deviation |
|---|---|---|---|
| Sociodemographics | |||
| Age | Number of years | 77.33 | 5.67 |
| Gender | 1 = men, 0 = women | .35 | .48 |
| Race | |||
| White (reference group [R]) | 1 = yes, 0 = no | .86 | .34 |
| African American | 1 = yes, 0 = no | .09 | .28 |
| Hispanic | 1 = yes, 0 = no | .04 | .20 |
| Other | 1 = yes, 0 = no | .01 | .10 |
| Living children | Actual number | 2.68 | 2.15 |
| Household size | Actual number | 1.80 | .90 |
| Widowed | 1 = yes, 0 = no | .43 | .50 |
| Population density | |||
| Major city (≥250K; R) | 1 = yes, 0 = no | .70 | .46 |
| Small city (20–250K) | 1 = yes, 0 = no | .13 | .33 |
| Rural county (<20K) | 1 = yes, 0 = no | .18 | .38 |
| Socioeconomics | |||
| Education | 1 = grade school, 0 = beyond | .25 | .43 |
| Working for pay | 1 = yes, 0 = no | .09 | .28 |
| Poor neighborhood safety | 1 = yes, 0 = no | .03 | .17 |
| Household income | Number of dollars | 25,432.60 | 37,470.12 |
| Household wealth | Number of dollars | 180,067.84 | 292,862.02 |
| Disease history | |||
| Major condition count | Actual number | 1.47 | 1.21 |
| Angina | 1 = yes, 0 = no | .08 | .27 |
| Arthritis | 1 = yes, 0 = no | .25 | .43 |
| Cancer | 1 = yes, 0 = no | .13 | .34 |
| Diabetes | 1 = yes, 0 = no | .11 | .32 |
| Heart attack | 1 = yes, 0 = no | .06 | .25 |
| Hip fracture | 1 = yes, 0 = no | .04 | .20 |
| Hypertension | 1 = yes, 0 = no | .46 | .50 |
| Lung disease | 1 = yes, 0 = no | .09 | .28 |
| Psychological conditions | 1 = yes, 0 = no | .07 | .26 |
| Stroke | 1 = yes, 0 = no | .08 | .27 |
| Weight to height | |||
| Underweight | 1 = yes, 0 = no by NIH | .03 | .17 |
| Appropriate (R) | 1 = yes, 0 = no by NIH | .83 | .38 |
| Obese | 1 = yes, 0 = no by NIH | .14 | .35 |
| Functional health | |||
| Self-reported to be often bothered by pain | 1 = yes, 0 = no | .32 | .47 |
| Poor self-rated memory | 1 = yes, 0 = no | .04 | .21 |
| Poor self-rated hearing | 1 = yes, 0 = no | .05 | .22 |
| Poor self-rated vision | 1 = yes, 0 = no | .08 | .27 |
| Depressive symptoms | Actual number of 8 | 1.67 | 1.96 |
| ADL count | Actual number of 7 | .29 | .74 |
| IADL count | Actual number of 3 | .18 | .47 |
| CAGE (alcoholism) cut-down question | 1 = yes (need to cut-down on drinking), 0 = no | .07 | .26 |
| Driving status | |||
| Never had a license | 1 = yes, 0 = no | .11 | .32 |
| Unable to drive now (R) | 1 = yes, 0 = no | .19 | .38 |
| Able to drive now | 1 = yes, 0 = no | .70 | .46 |
| TICS cognitive battery | 0 = worst, 15 = best | 11.96 | 2.97 |
| Immediate word recall | 1 = below 50th percentile | ||
| 0 = above 50th percentile | .50 | .50 | |
| Delayed word recall | 1 = below 50th percentile | ||
| 0 = above 50th percentile | .56 | .50 | |
| Self-rated overall health | |||
| Excellent | 1 = yes, 0 = no | .11 | .32 |
| Very good/Good/Fair (R) | 1 = yes, 0 = no | .78 | .41 |
| Poor | 1 = yes, 0 = no | .11 | .32 |
Note: ED=emergency department; AHEAD=Survey on Assets and Health Dynamics Among the Oldest Old; NIH=National Institutes of Health; ADL=activities of daily living; IADL=instrumental ADL; CAGE=screening for alcohol abuse (Cut-down, Annoyed, Guilty, Eye Opener); TICS=Telephone Interview for Cognitive Screening.
Factor Analysis
We began with an EFA of the 20 variables representing the annual total of ED visits corresponding to each of the five CPT codes. A two-factor common model was initially extracted, and we expected these two factors to be modestly correlated. The initial results obtained (data not shown) were generally consistent with these expectations and revealed the anticipated principal factor loadings, no evidence of factorial complexity, and a modest correlation between the two factors. This indicated the appropriateness of combining CPT codes 99281 and 99282 into the low-intensity category and CPT codes 99283–99285 into the high-intensity category. We then performed EFA on the eight variables reflecting the annual numbers of low intensity (CPT codes 99281 and 99282) and high intensity (CPT codes 99283–99285) ED visits. As shown in Table 2, this EFA resulted in the expected simple two-factor structure.
Table 2.
Factor Loadings for the Number of Low Intensity (CPT Codes 99281 and 99282) and High Intensity (CPT Codes 99283–99285) ED Visits, by Observation Year, Among 4310 AHEAD Self-Respondents
| Annual Number of ED Visits of … | Factor 1 | Factor 2 |
|---|---|---|
| Low intensity during the 1st year | .044 | .456* |
| Low intensity during the 2nd year | .020 | .648* |
| Low intensity during the 3rd year | −.109 | .744* |
| Low intensity during the 4th year | .037 | .565* |
| High intensity during the 1st year | .801* | −.045 |
| High intensity during the 2nd year | .791* | .033 |
| High intensity during the 3rd year | .769* | −.073 |
| High intensity during the 4th year | .545* | .097 |
Notes:
Principal factor loading.
CPT = Current Procedural Terminology; ED = emergency department; AHEAD = Survey on Assets and Health Dynamics Among the Oldest Old.
Criterion Validity
We then examined the evidence of criterion validity (19) consistent with prior triaging studies (20–23). As shown in Table 3, with three exceptions there is a direct, monotonic relationship between increasing relative effort intensity levels (CPT codes) and the criteria for ED appropriateness. Table 4 contains the mean estimated probabilities of ED visit types from the ICD-9-CM based approach developed by Billings and colleagues (24) by CPT codes. With two exceptions, there is a direct, monotonic relationship between increasing relative effort intensity levels and the appropriateness of the ED visits (chi-square = 213.11; df = 12; p < .0001). Moreover, both errors involve the CPT code (99281) with the smallest ED visit share.
Table 3.
Criterion Validation Data for the CPT Code-Based Classification Approach
| Criterion Variable | CPT Code 99281 | CPT Code 99282 | CPT Code 99283 | CPT Code 99284 | CPT Code 99285 |
|---|---|---|---|---|---|
| Percentage of all ED visits, % | 4.3 | 14.9 | 31.4 | 27.4 | 21.9 |
| Arrival by ambulance, % | 10.3 | 11.5 | 18.2 | 26.7 | 35.2 |
| Admitted to hospital, % | 6.0 | 8.2 | 14.3 | 26.2 | 38.2 |
| Length of stay, mean days if admitted | 5.2 | 6.9 | 8.6 | 7.1 | 7.4 |
| Medical charges, mean $ if admitted | 7909 | 9844 | 11,070 | 11,341 | 13,175 |
| All-cause NDI mortality by 2 years after the baseline interview, % dead |
9.3 | 3.3 | 6.9 | 12.9 | 22.3 |
Note: CPT = Current Procedural Terminology; ED = emergency department; NDI = National Death Index.
Table 4.
| Billings and Colleagues ED Visit Types | CPT Code 99281 | CPT Code 99282 | CPT Code 99283 | CPT Code 99284 | CPT Code 99285 |
|---|---|---|---|---|---|
| Nonemergent visit | .214 | .195 | .159 | .118 | .080 |
| Emergent visit, primary care treatable | .294 | .330 | .298 | .247 | .207 |
| ED care needed, but the condition was preventable/avoidable |
.062 | .090 | .103 | .153 | .189 |
| ED care needed, and the condition was not preventable or avoidable |
.250 | .199 | .253 | .290 | .353 |
| Not classifiable by the Billings and Colleagues algorithm |
.180 | .186 | .187 | .192 | .171 |
| Total | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 |
Note: ED = emergency department; CPT = Current Procedural Terminology.
We also identified the five most frequent ICD-9-CM codes associated with low- versus high-intensity ED visits. For high-intensity ED visits, these were chest pain (ICD-9-CM code 786.5), congestive heart failure (428.0), respiratory distress and insufficiency (786.09), pneumonia (486.0), and other symptoms involving the abdomen or pelvis (789.0) and accounted for 18.5% of all high-intensity visits. In contrast, for low-intensity ED visits, the top five ICD-9-CM codes were open wounds of fingers (883.0), unspecified (ICD-9-CM code 401.9), other symptoms involving the abdomen or pelvis (789.0), other disorders of the urethra or urinary tract (599.0), and epistaxis (nosebleed; 784.7) and accounted for 11.1% of all low-intensity visits. These striking differences in the underlying diagnostic reasons for these ED visits provide further criterion validity for our approach.
ED Use Patterns
Based on these results, we placed each AHEAD participant into one of four ED use patterns: (i) no ED use, (ii) low-intensity-only ED use, (iii) high-intensity-only ED use, and (iv) mixed (low and high) intensity ED use. The majority of participants (56.6%) had no ED use during the 4-year period. Only 5.7% were low-intensity-only ED users (mean = 1.40 ED visits), whereas 28.9% were high-intensity-only ED users (mean = 2.02 ED visits). Mixed (low and high intensity) ED use was found for 8.7% of the AHEAD participants (mean = 4.67 ED visits).
Multivariable Model
Of the >40 variables considered as potential predictors of ED use patterns (see Table 1), only 12 had significant independent associations. Table 5 contains the adjusted odds ratios (AORs) obtained from the final model. As shown, the only variables that significantly discriminated those who had no ED use from those who had low-intensity-only ED use were population density and cognitive ability. Participants who lived in rural counties and small cities and those who scored in the lower half on the immediate word recall test were more likely to be low-intensity-only ED users (AORs = 2.923, 2.287, and 1.553, respectively; all p<.001).
Table 5.
Adjusted Odds Ratios From the Final Multivariable Multinomial Logistic Regression Model of ED Utilization Patterns Among 4135 AHEAD Self-Respondents With Complete Data (Weighted N = 4172)
| Variable | No ED Use Group (Reference) |
Low-Intensity-Only ED Use Group |
High-Intensity-Only ED Use Group |
Mixed-Intensity ED Use Group |
Overall p Value |
|---|---|---|---|---|---|
| Age | 1.000 | 1.015 | 1.050* | 1.028† | .001 |
| Rural county | 1.000 | 2.923* | 0.750† | 1.970* | .001 |
| Small city | 1.000 | 2.287* | 0.966 | 1.541‡ | .001 |
| Grade school education | 1.000 | 0.915 | 0.986 | 1.444† | .029 |
| Major comorbid condition count | 1.000 | 1.090 | 1.372* | 1.500* | .001 |
| Angina | 1.000 | 1.269 | 1.507† | 1.850* | .005 |
| Hypertension | 1.000 | 0.980 | 0.817‡ | 0.586* | .001 |
| Psychological problem | 1.000 | 0.929 | 0.747 | 0.492† | .016 |
| Depressive symptom count | 1.000 | 0.992 | 1.049† | 1.096† | .008 |
| IADL count | 1.000 | 0.736 | 1.239† | 1.1331 | .010 |
| Lower half immediate word recall scores | 1.000 | 1.553† | 1.264† | 1.404† | .001 |
| Poor self-rated health | 1.000 | 1.144 | 1.740* | 1.733† | .001 |
Notes:
p ≤ .001.
p ≤ .01.
p ≤ .05.
ED = emergency department; AHEAD = Survey on Assets and Health Dynamics Among the Oldest Old; IADL = instrumental activities of daily living. Chi-square reduction from the original null model was 532.9 (36 df), p ≤ .001.
In contrast, all but 3 of the 12 variables included in the final model discriminated between participants with no ED use and those with high-intensity ED use only. The risk (i.e., odds) of high-intensity-only ED use increased by 5% with each year of age, by 5% with each depressive symptom, by 51% for those with angina, by 37% with each major comorbid condition, by 74% for those reporting poor self-rated health, by 24% for each IADL limitation, and by 26% for those in the lower half on the immediate word recall test. Participants with hypertension were 18% less likely, and those living in rural counties were 25% less likely, to be high-intensity-only ED users.
All but 1 of the 12 variables included in the final model also discriminated between participants with no ED use and those with mixed (low and high) intensity ED use patterns. The risk of mixed intensity ED use increased by 3% with each year of age, by 97% for those living in rural counties, by 54% for those living in small cities, by 44% for those with only a grade school education, by 10% with each depressive symptom, by 85% for those with angina, by 44% with each major comorbid condition, by 73% for those reporting poor self-rated health, and by 40% for those in the lower half on the immediate word recall test. Those with hypertension were 41% less likely to be mixed-intensity ED users.
DISCUSSION
In a nationally representative sample of 4310 older adults, we used CPT codes to identify four ED use patterns across 4 years of data. Only 5.7% of the sample participants were low-intensity-only ED users, whereas 28.9% were high-intensity-only ED users. Mixed (low and high) intensity ED use was found for 8.7% of participants. This pattern of ED usage leads to our first conclusion. Although nearly half of our nationally representative sample of older adults used the ED at least once over a 4-year period, the demand that they placed on EDs would only be modestly alleviated by focusing on triaging presumably nonurgent (low intensity) and potentially avoidable visits to alternative care sources and sites, even if effective triaging procedures were widely introduced and implemented.
We also used sociodemographic, socioeconomic, lifestyle, disease history, and functional health status measures in a multivariable, multinomial logistic regression to predict ED use patterns. Compared to older adults who did not use the ED (the reference group), the only distinguishing features of participants who used the ED only for low-intensity visits were that they more likely to have resided in rural counties or smaller cities (vs major metropolitan areas) and scored in the lower half on the immediate word recall test (a measure of cognitive performance). Given the limited modifiability of these three traits (although not the processes that led to them), we conclude that it is unlikely that an effective intervention for triaging potentially avoidable ED visits to alternative sources of health care can be developed. Moreover, increased reliance on the ED for potentially avoidable visits among those living in less metropolitan areas likely reflects restricted access to alternative sources of health care in the first place.
Compared to older adults who did not use the ED, those who used the ED for high-intensity-only visits were more likely to be older, to be men, to have greater morbidity and comorbidity, to have scored on the lower half on the immediate word recall test, and to perceive their health to be poor. Thus, the principal drivers of high intensity (and presumably appropriate) use of the ED are generally medical need. The immutability of most of these factors further attests to the low likelihood that reducing the demand on and overcrowding of the ED can be readily achieved. Similarly, the major predictors of mixed (low and high) intensity ED use (compared to those of no ED use) were principally measures of medical need, except for the large effect of lower educational attainment. Participants who only completed grade school were 44% more likely to use the ED for mixed-intensity levels. Given the limited modifiability of educational attainment, this also does not bode well for potential interventions.
This initial investigation of ED use patterns is not without limitations. We focused on CPT codes for relative effort intensity levels rather than diagnoses, although we also showed that the correspondence between the two was considerable. Our approach did not consider attrition from Medicare claims due to a participant’s movement into managed care participation, or death (although all participants were alive and not participating in managed care Medicare plans for the first 2 years of the 4-year period). We have also not addressed continuity of care, the fractional share of ambulatory care provided through ED visits, or the volume of ED visits. Furthermore, the day of the week on which the ED use occurred cannot be addressed using these AHEAD data, nor can ED use that occurs when the participants are not in the vicinity of their home residence. Finally, our data are about a decade old, and there is evidence in the IOM report (2) that the volume of ED has increased since then. There is no indication, however, that the patterns of ED use among older adults have changed. Nonetheless, continued surveillance to detect potential changes in ED use patterns among older adults is recommended before these results should be considered definitive.
Those limitations notwithstanding, our findings do shed light on the current ED crisis. Unfortunately, we do not have an easy solution to offer. Simply put, the results suggest that little alleviation of the ED overcrowding phenomenon is likely to accrue from focusing on older adults.
ACKNOWLEDGMENTS
This research was supported by National Institutes of Health grants R01 AG-022913 and R03 AG027741 to Dr. Wolinsky. Dr. Wolinsky is the Associate Director of the Center for Research in the Implementation of Innovative Strategies in Practice (CRIISP) at the Iowa City VA Medical Center, Dr. Rosenthal is the Director of CRIISP, and Dr. Richardson is a CRIISP Statistician. CRIISP is funded through the Department of Veterans Affairs, Veterans Health Administration, Health Services Research and Development Service (HFP 04-149).
REFERENCES
- 1.Fineberg HV. Hospital-Based Emergency Care: At the Breaking Point. Washington, DC: National Academy Press; 2006. Foreword. [Google Scholar]
- 2.Institute of Medicine. Hospital-Based Emergency Care: At the Breaking Point. Washington, DC: National Academy Press; 2006. [Google Scholar]
- 3.American Hospital Association. [Last accessed December 18, 2006];TrendWatch Chartbook. 2005 Available at: http://www.ahapolicyforum.org/ahapolicyforum/trendwatch/chartbook2005.html.
- 4.McCaig LF, Burt CW. National Hospital Ambulatory Medical Care Survey: 2003 Emergency Department Summary. Hyattsville, MD: National Center for Health Statistics; 2005. [Google Scholar]
- 5.Richardson LD, Asplin BR, Lowe RA. Emergency department crowding as a health policy issue: past development, future directions. Ann Emerg Med. 2002;40:388–393. doi: 10.1067/mem.2002.128012. [DOI] [PubMed] [Google Scholar]
- 6.Derlet R, Richards J, Kravitz R. Frequent overcrowding in U.S. emergency departments. Acad Emerg Med. 2001;8:151–155. doi: 10.1111/j.1553-2712.2001.tb01280.x. [DOI] [PubMed] [Google Scholar]
- 7.Weiss SJ, Derlet R, Arndahl J, et al. Estimating the degree of emergency department overcrowding in academic medical centers: results of the national ED overcrowding study (NEDOCS) Acad Emerg Med. 2004;11:38–50. doi: 10.1197/j.aem.2003.07.017. [DOI] [PubMed] [Google Scholar]
- 8.Bradley VM. Placing emergency department crowding on the decision agenda. J Emerg Nurs. 2005;31:247–258. doi: 10.1016/j.jen.2005.04.007. [DOI] [PubMed] [Google Scholar]
- 9.Berger E. Geriatrics and the IOM: did report forget the elderly? Ann Emerg Med. 2006;48:723–725. doi: 10.1016/j.annemergmed.2006.10.009. [DOI] [PubMed] [Google Scholar]
- 10.Myers GC, Juster FT, Suzman RM. Asset and Health Dynamics Among the Oldest Old (AHEAD): initial results from the longitudinal study. J Gerontol Psychol Sci Soc Sci. 1997;52(Spec No:v–viii) [PubMed] [Google Scholar]
- 11.Juster FT, Suzman RM. An overview of the Health and Retirement Study. J Human Resources. 1995;30:S7–S56. [Google Scholar]
- 12.Wolinsky FD, Miller TR, Geweke JF, et al. An interpersonal continuity of care measure for Medicare Part B claims analyses. J Gerontol Psychol Sci Soc Sci. 2007;62B:S160–S168. doi: 10.1093/geronb/62.3.s160. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wolinsky FD, Miller TR, Geweke JF, et al. Hospital episodes and physician visits: the concordance between self-reports and Medicare claims. Med Care. 2007;45:300–307. doi: 10.1097/01.mlr.0000254576.26353.09. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.American Medical Association. Current Procedural Terminology 2004: Professional Edition. Chicago, IL: AMA Press; 2003. [Google Scholar]
- 15.Nunnally JC. Psychometric Theory. New York: McGraw Hill; 1967. [Google Scholar]
- 16.Kim J-O, Mueller CW. Introduction to Factor Analysis. Beverly Hills, CA: Sage; 1981. [Google Scholar]
- 17.Kim J-O, Mueller CW. Factor Analysis. Beverly Hills, CA: Sage; 1981. [Google Scholar]
- 18.Norman GR, Streiner DL. Biostatistics–The Bare Essentials. St. Louis, MO: Mosby; 1994. [Google Scholar]
- 19.Shadish WR, Cook TD, Campbell DT. Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Boston: Houghton Mifflin; 2002. [Google Scholar]
- 20.Dong SL, Bullard MJ, Meurer DP, et al. Emergency triage: comparing a novel computer triage program with standard triage. Acad Emerg Med. 2005;12:502–507. doi: 10.1197/j.aem.2005.01.005. [DOI] [PubMed] [Google Scholar]
- 21.Maningas PA, Hime DA, Parker DE, McMurray TA. The Soterion Rapid Triage System: evaluation of inter-rater reliability and validity. J Emerg Med. 2006;30:461–469. doi: 10.1016/j.jemermed.2005.05.037. [DOI] [PubMed] [Google Scholar]
- 22.Dong SL, Bullard MJ, Meurer DP, et al. Predictive validity of a computerized emergency triage tool. Acad Emerg Med. 2007;14:16–21. doi: 10.1197/j.aem.2006.08.021. [DOI] [PubMed] [Google Scholar]
- 23.Baumann MR, Strout TD. Triage of geriatric patients in the emergency department: validity and survival with the Emergency Severity Index. Ann Emerg Med. 2007 doi: 10.1016/j.annemergmed.2006.04.011. In press. [DOI] [PubMed] [Google Scholar]
- 24.Billings J, Parikh N, Mijanovich T. Emergency department use in New York City: a substitute for primary care? Commonwealth Fund Issue Brief. 2000a November; [PubMed] [Google Scholar]
- 25.Billings J, Parikh N, Mijanovich T. Instructions for Use of the SAS and SPSS Versions of the ED Classification Algorithm. New York: NYU Center for Health and Public Service Research; 2007. [Google Scholar]
- 26.Hosmer DW, Lemeshow S. Applied Logistic Regression. New York: Wiley; 1989. [Google Scholar]
- 27.Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav. 1995;36:1–10. [PubMed] [Google Scholar]
- 28.Concato J, Feinstein AR, Holford TR. The risk of determining risk with multivariable models. Ann Intern Med. 1993;118:201–210. doi: 10.7326/0003-4819-118-3-199302010-00009. [DOI] [PubMed] [Google Scholar]
- 29.Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15:361–387. doi: 10.1002/(SICI)1097-0258(19960229)15:4<361::AID-SIM168>3.0.CO;2-4. [DOI] [PubMed] [Google Scholar]