The American Journal of Managed Care June 2011
Emergency Department Visits in Veterans Affairs Medical Facilities
Objective: To identify the frequency of, and risk factors for, repeat emergency department (ED) visits and hospitalizations following a treat-andrelease ED visit in patients from Veterans Affairs Medical Centers (VAMCs).
Study Design: Retrospective cohort study.
Methods: Subjects were veterans who visited 1 of 102 VAMC EDs between October 1, 2007, and June 30, 2008. Generalized estimating equations were used to identify factors related to repeat ED visits and hospitalizations within 30 days of the index ED visit.
Results: At their index ED visit, 80% of veterans were treated and released. Of these, 15% returned to the ED and 5% were hospitalized in the next 30 days. In adjusted= models, factors associated with increased odds of repeat ED visits included homelessness (odds ratio [OR] 1.70; 95% confidence interval [CI] 1.59, 1.82) and having a previous ED visit (OR 1.66; 95% CI 1.58, 1.74). Odds of hospitalization were higher among older (OR 1.35; 95% CI 1.26, 1.46), homeless (OR 1.61; 95% CI 1.44, 1.80), and functionally impaired (OR 1.52; 95% CI 1.35, 1.76) veterans, those with greater comorbidity (OR 1.31; 95% CI 1.27, 1.34), previous hospitalization (OR 2.48; 95% CI 2.28, 2.70), and an original ED visit related to a chronic condition (OR 1.30; 95% CI 1.23, 1.37). Among veterans who returned to the ED, 71.7% did not see another VA outpatient provider between their original and return visits.
Conclusions: A substantial proportion of veterans treated and released from VAMC EDs returned to the ED or were hospitalized within 30 days.
(Am J Manag Care. 2011;17(6):e215-e223)
This study examined emergency department (ED) visits in the Veterans Affairs (VA) system.
- The majority of VA ED visits resulted in patients being treated and released, rather than admitted to the hospital.
- In the 30 days following a treat-and-release VA ED visit, about 1 in 7 veterans returned for another unplanned visit to the ED. The risk of return was higher among homeless veterans and those with previous ED use.
- Among veterans who returned to the ED, nearly three-fourths had not seen another VA outpatient provider between their original and return ED visits.
In EDs affiliated with nonfederal US hospitals, an estimated 85% of patients evaluated are not admitted to the hospital at the conclusion of their visit; rather, they are discharged home.3 Emergency department visits that do not result in admission, commonly referred to as treat-and-release visits,4 are important, not only because of their frequency, but also because repeat ED visits and/or hospitalizations are relatively common among some patient groups (5%-19% within 30 days after an index treat-and-release ED visit).5-8 Veterans who utilize the VA health system are more likely than the general population to report poor physical and mental health and to have chronic health conditions.9,10 While these characteristics suggest that VA users may be disproportionately at risk for repeated ED visits, veterans’ access to VA primary care may mitigate against this risk. Thus, this study had 4 specific goals: (1) to describe national VAMC ED disposition rates; (2) to characterize the population of
veterans with treat-and-release ED visits; (3) to determine the frequency of repeat ED visits, hospitalizations, and deaths experienced by veterans after being treated and released from a VAMC ED; and (4) to identify factors that predict repeat ED visits and hospitalizations within 30 days of the original treat-and-release ED visit.
We conducted a retrospective cohort study using a national sample of veterans who had a VAMC ED visit over a 9-month observation period. Approval for the study was obtained from the Institutional Review Board of the Durham VAMC.
The data sets used for this study were drawn from administrative files maintained at the VA’s central data repository, the Austin Information Technology Center. Age, sex, and date of death were drawn from the Vitals Mini File.11 Dates of VA health service use (including outpatient visits, ED use, and hospitalizations) and data for all other independent variables were drawn from the VHA Medical SAS Datasets.12
The cohort consisted of VA users with an ED or urgent care clinic visit between October 1, 2007, and June 30, 2008. Patients who received care at facilities that used only urgent care clinic codes during the study period were excluded because these facilities lack medical-surgical beds and/or an intensive care unit, which affects the emergency care that can be provided there.1 For the remaining facilities, ED and urgent care clinic were considered together because of variation in coding practices (ie, some VAMCs use separate codes for ED and urgent care clinic and some do not). The sample was limited to previous VA users because patients who are new to the VA can initiate contact with the system through the ED; therefore, their visit may not be associated with an acute illness or injury. As a final step, we randomly sampled 20% of veterans from each eligible VAMC ED who met the above criteria. This sampling strategy was used to maintain the feasibility of analyses while ensuring patient representation from all eligible VAMC EDs.
Emergency Department Disposition. Possible disposition paths at the end of the index ED visit were (1) treated and released, (2) hospital admission, or (3) died in the ED. Patients were considered to have been treated and released if there was no record of a hospital admission within 1 day of their ED visit. Patients who died on the same date as their ED visit, but did not have a record of hospital admission, were considered to have died in the ED.
Dependent Variables. The main dependent variables of interest were ED visits and hospital admissions within 30 days of the index ED visit. These were considered separately because previous data demonstrated that risk factors differ for these 2 outcomes.13 To avoid doublecounting events, an ED visit followed by a hospitalization (within 1 day) was considered a hospitalization only.14 Multiple ED codes on the same date of service were considered as 1 visit.
Independent Variables. To identify predictors of our primary outcomes, we used the behavioral model of healthcare utilization which considers predisposing, enabling, and need variables.15,16 Predisposing characteristics included age (≥55 or <55 years), sex, race (African American/non–African American/ unknown or missing), and marital status (unmarried or other). Age as a continuous variable violated the linearity assumption of the analysis model; therefore, age was dichotomized at 55 years after visual inspection of the relationship between age and outcomes.17 Enabling resources included income (copay exempt due to low income or other), homeless (yes or no), and uninsured (yes or no). Homelessness was defined as receiving VA care related to homelessness in the previous year, according to clinic codes (528, 529, or 590) or diagnosis codes (V60.0 and V60.1). Need variables included comorbidity (Diagnostic Cost Group score), service-connected disability (>50% or other), treat-and-release ED visits within the previous 180 days (yes or no), hospital admissions within the previous 180 days (yes or no), 5 or more primary care provider (PCP) visits in the previous year, and use of Aid and Attendance benefits. Diagnostic Cost Group score was calculated using diagnoses from both inpatient and outpatient files in the previous fiscal year. Diagnostic Cost Group scores have been shown to predict hospitalization and mortality among veterans as well as other comorbidity scores.18,19 Aid and Attendance benefits are available to veterans with severe functional impairments that necessitate regular assistance from another person in the veteran’s home.20 Index ED visit characteristics included day of visit (weekend or weekday) and facility location (Northeast, South, Midwest, West). Using previously described methods,21 4 ED discharge diagnosis groups were identified (injury/musculoskeletal, chronic condition, infection, non-musculoskeletal symptoms). Emergency department discharge diagnosis and geographic region variables were coded using effect coding (ie, the mean of the means of the groups was used as the reference instead of 1 of the groups).22,23
Descriptive statistics were used to characterize the patient sample. Generalized estimating equations were used to examine the influence of patient and ED visit factors on outcomes. The first equation predicted repeat treat-and-release ED visits, and the second equation predicted hospital admission within 30 days of the index visit. Multivariable models included only subjects who survived through the 30-day observation period. Candidate patient variables (selected a priori according to the Andersen model) and ED visit variables were entered into the model simultaneously. A variable was considered significant if P <.05 after adjustment for variables already in the model. After assessing model fit and checking for collinearity among independent variables, a C statistic was calculated to evaluate concordance between predicted and observed outcomes (ie, how well the model predicted outcomes).24 Results from the generalized estimating equation models were expressed as odds ratios (ORs) and 95% confidence intervals (CIs). All analyses were performed at the patient level with adjustment for clustering of patients within facilities; an intraclass correlation coefficient was calculated to estimate the effects of clustering of patients within facilities. Analyses were conducted using SAS version 9.1 (SAS Institute, Inc, Cary, North Carolina).
Sample Construction and Characteristics
A total of 765,732 unique patients had 1 or more visits to a VAMC ED during the 9-month observation period. After excluding patients without VA use in the previous year (n = 90,051) and those seen at a facility that used only urgent care clinic codes (n = 26,144), 649,537 eligible veterans remained. From these patients, a random 20% sample was drawn from each of 102 VAMCs. In our final sample of 128,174 individuals, 102,516 (80%) were treated and released, 25,630 (20%) were admitted to the hospital, and 28 (0.02%) died in the ED. Subsequent analyses focused on the 102,516 veterans who were treated and released. The mean age of this predominantly male sample was 58.7 years. As shown in Table 1, more than 1 in 3 patients had low income (39.4%) and 5.6% used VA services designed for the homeless. On average, patients had seen their PCP on 4 occasions in the year prior to the index ED visit. Twenty-three percent of ED visits were related to injuries or acute musculoskeletal conditions, and 21% were related to chronic medical conditions.
Frequency of Repeat Emergency Department Visits, Hospitalizations, and Deaths
Overall, 19.3% of veterans who were treated and released at their index ED visit had 1 or more of the following events within 30 days: repeat ED visit, hospitalization, or death. A total of 15,647 veterans (15.3%) had a repeat ED visit (total of 19,791 visits), 5117 veterans (5.0%) were hospitalized, and 399 veterans (0.4%) died. The times to first repeat ED visit and hospital admission both followed a similar pattern, with a higher frequency of events occurring during the first 7 days (Figure).
Risk Factors for Repeat Emergency Department Visits
Patient- and visit-level predictors of repeat ED visits within 30 days are presented in Table 2. Being unmarried, having a higher comorbidity burden, having a service connected disability, and having 5 or more PCP visits in the previous year were associated with small but significantly increased odds of repeat ED visits. The odds of a repeat ED visit were significantly higher for patients who were homeless (OR 1.70; 95% CI 1.59, 1.82), had a previous ED visit (OR 1.66; 95% CI 1.58, 1.74), or had a previous hospital admission (OR 1.28; 95% CI 1.22, 1.35). Emergency department visits that occurred on the weekend, those related to infection, and those that occurred in VAMCs in the West were also associated with increased odds of repeat visits, although the magnitude of the association was modest.
Risk Factors for Hospitalization