Skip to content
Publicly Available Published by De Gruyter August 31, 2016

Admissions after discharge from an emergency department for chest symptoms

  • Brian J. Moore EMAIL logo , Rosanna M. Coffey , Kevin C. Heslin and Ernest Moy
From the journal Diagnosis

Abstract

Background:

Often patients who present to the emergency department (ED) with chest symptoms return to the hospital within 30 days with the same or closely related symptoms and are admitted, raising questions about quality of care, timeliness of diagnosis, and patient safety. This study examined the frequency of and patient characteristics associated with subsequent inpatient admissions for related symptoms after discharge from an ED for chest symptoms.

Methods:

We used data from the 2012 and 2013 Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SID) and State Emergency Department Databases (SEDD) from eight states to identify over 1.8 million ED discharges for chest symptoms.

Results:

Approximately 3% of ED discharges experienced potentially related subsequent admissions within 30 days – 0.2% for acute myocardial infarction (AMI), 1.7% for other cardiovascular conditions, 0.5% for respiratory conditions, and 0.6% for mental disorders. Logistic regression results showed higher odds of subsequent admission for older patients and those residing in low-income areas, and lower odds for females and non White racial/ethnic groups. Privately insured patients had lower odds of subsequent admission than did those who were uninsured or covered by other programs.

Conclusions:

Because we included multiple diagnostic categories of subsequent admissions, our results show a more complete picture of patients presenting to the ED with chest symptoms compared with previous studies. In particular, we show a lower rate of subsequent admission for AMI versus other diagnoses. ED physicians and administrators can use the results to identify characteristics associated with increased odds of subsequent admission to target at-risk populations.

Introduction

In 2012, more than 5 million patients in the United States presented to the emergency department (ED) with nonspecific chest pain or coronary atherosclerosis, referred to hereafter as chest symptoms [1]. More than 85% of these visits were treat and release – that is, they did not result in a hospital admission at the time of the ED visit [1]. Cases of chest symptoms often involve multiple diagnostic tests and extensive evaluation in the ED before patients are admitted to the hospital or released to outpatient follow-up. The mean length of stay for treat-and-release ED visits for nonspecific chest pain is 7.4 h, nearly 3 h longer than the mean length of stay for visits resulting in an admission [2]. Questions about diagnostic performance, quality of care, and patient safety arise when treat-and-release patients return to the hospital within 30 days with symptoms that are the same or closely related and are admitted. Considerable attention has been paid to identifying, quantifying, and learning from ED cases that result in subsequent admission for a related condition.

In addition, diagnostic performance problems often result in high litigation settlement amounts [3]. Improved and multiple diagnostic tests that are currently available [4] bring with them increased scrutiny and expanded expectations for rapid and accurate diagnosis in the ED [5]. Previous researchers have attempted to quantify the extent to which acute myocardial infarction (AMI) diagnoses are missed in the ED and patients are discharged home [6], [7], [8]. More recent estimates of missed diagnosis of AMI in the ED have ranged from 0.5% to 2% [3], [9], [10].

These studies, while informative, have been limited by data obtained from a small number of facilities [3], [6], [7], [8] or a single payer [3], [9]. Studies that first look for AMI admissions and then track patient records backwards to identify treat-and-release ED visits [10] ignore the uncertainty involved in the initial diagnosis and the alternative conditions that chest symptoms may reflect. The current study is the first study we are aware of to address this concern by investigating the issue in the opposite direction. We identify treat-and-release ED visits for chest symptoms and then search for those patients in the community hospital system for the next 30 days using all-payer data from eight geographically dispersed states.

This approach provides a more complete profile of patients discharged home after an ED visit for chest symptoms and an estimate of the percentage of cases that could be designated as a subsequent admission for AMI. We also provide results on subsequent admissions for common principal diagnoses other than AMI that present with chest symptoms.

Methods

Data sources

The data used for this study came from the Healthcare Cost and Utilization Project (HCUP) State Emergency Department Databases (SEDD) [11] and State Inpatient Databases (SID) [12] for eight states for 2012 and 2013. The SEDD represent the universe of ED visits that do not end in a hospital admission for all community hospitals (defined as all non federal, short-term, general, and other specialty hospitals, excluding hospital units of institutions). The SID represent the universe of inpatient admissions for community hospitals, including visits that started in the ED. Data are available for all payer types, including private insurance, Medicare, Medicaid, other program, and uninsured populations. The states included in our analysis were California, Florida, Missouri, Nebraska, New York, South Carolina, Tennessee, and Vermont. We chose these eight states because their SID and SEDD include patient linkage numbers to facilitate tracking patients across multiple visits and hospital settings.

Analytic sample

Our study sample consisted of patients treated and released from the ED with a primary diagnosis of nonspecific chest pain or coronary atherosclerosis and other heart disease. Diagnoses were coded according to the International Classification of Disease, 9th Edition, Clinical Modification (ICD-9-CM). We classified visits into clinical groups using the Agency for Healthcare Research and Quality (AHRQ) Clinical Classifications Software (CCS) [13], which organizes ICD-9-CM diagnosis codes into clinically homogeneous groups. The specific CCS categories used to define index visits and admissions are available as Supplemental Data, Appendix Table 1.

The sample included 2,180,910 treat-and-release ED discharges for chest symptoms among adults (age≥18 years) from January 1, 2012 through November 30, 2013. From this population, we sequentially applied exclusion criteria to produce the final study sample (Figure 1). We excluded rehabilitation and non-community hospitals as defined by the American Hospital Association (AHA) Annual Survey of Hospitals (n=9740 discharges; 0.4%) and hospitals without data in both 2012 and 2013 (n=4569 discharges from 17 hospitals; 0.2%). We also excluded discharges missing the HCUP variables required for tracking patients over time (n=120,593 discharges; 5.5%), which represent either patients without an identification number or discharges lacking enough information to verify the identification number. Next, we excluded discharges missing length of stay (n=6635 discharges; 0.3%), missing in-hospital death indicators (n=2350 discharges; 0.1%), with erroneous records in which stays overlap or patients return after being discharged dead (n=579 discharges; 0.03%), with patients discharged against medical advice (n=132,876 discharges; 6.1%), with patients in the top 0.01% of hospital visits, which translated into 69 or more visits in 2 years (n=5484 discharges; 0.3%), or missing the primary expected payer (n=1922 discharges; 0.1%). After all restrictions had been applied, 1,896,162 index ED visits from 1029 hospitals remained in the final study sample.

Figure 1: Sample selection criteria.Source: Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project, State Emergency Department Databases for eight states (2012 and 2013).
Figure 1:

Sample selection criteria.

Source: Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project, State Emergency Department Databases for eight states (2012 and 2013).

Dependent variable: subsequent admissions

Following the index treat-and-release ED visit, we tracked patients for inpatient admissions between 2 and 30 days later to create four binary dependent variables (representing four medical conditions) that were equal to 1 if the patient had an inpatient admission and 0 otherwise. HCUP data do not capture mortality outside the hospital setting. Therefore, patients who died within 30 days of their ED discharge would have been included as patients who were eligible for subsequent admission and coded as 0 in our dependent variables. We excluded next day admissions (n=10,257; 0.5% of the analytic sample) to reduce the risk of misclassification bias due to transfers being counted as subsequent admissions. Related subsequent admissions involving several common causes of chest symptoms – cardiovascular, respiratory, and mental conditions – were categorized according to the principal diagnosis at discharge on the subsequent inpatient hospital record. For each record, we first searched for inpatient admissions for AMI within 2–30 days. If no AMI admission was identified, we searched for other cardiovascular admissions. If no other cardiovascular events were identified, we searched for admissions due to respiratory conditions, and then mental disorders. We added respiratory admissions as an outcome to capture cases in which chest symptoms identified in the index ED visit eventually were identified as a respiratory issue. We included mental disorder admissions to capture cases in which the index ED visit was attributable to mental disorders that may involve chest symptoms such as pain, discomfort, shortness of breath, or smothering sensations.

Independent variables

The independent variables were obtained from the index ED visit. These variables included patient age, sex, income quartile of the patient’s ZIP Code, primary expected payer (Medicare, Medicaid, private, uninsured, or other), race/ethnicity (Hispanic, White non-Hispanic, Black non-Hispanic, other non-Hispanic, or missing). Payer types were derived from the expected primary payment source on the discharge record. We used Elixhauser comorbidities [14] entered as a comorbidity index [15] to represent patients’ coexisting conditions. The comorbidity index condenses the impact of 29 individual comorbid conditions that could confound our results (e.g. hypertension, congestive heart failure, or chronic pulmonary disease) into a single measure [16]. Hospital characteristics, determined from the AHA Annual Survey of Hospitals, included an indicator of teaching status, bed size (<100, 101–200, 201–349, ≥350), ownership structure (public nonfederal, private nonprofit, or private for-profit), census region (Northeast, Midwest, South, and West), location (large metropolitan, small metropolitan, micropolitan, or noncore), and whether a cardiac catheterization lab was available at the site where the index ED visit occurred.

We also included hospital volume characteristics to control for differences in capacity, crowding, and weekend effects on ED diagnostic performance. As a proxy for hospital capacity, we created a daily metric capturing the percentage of ED visits resulting in admission. Specifically, we calculated the percentage of ED visits resulting in admission for each ED for each day of the calendar year and categorized the resulting daily values into quintiles. For this measure, a lower admission rate may indicate decreased availability of inpatient beds. To capture differences in ED crowding on the day of the index visit, we created a measure of daily ED volume. Specifically, we calculated the number of visits for each ED for each day of the calendar year. Next, we ordered this daily volume from lowest to highest for each ED in this sample and categorized the visits into quintiles. For this measure, cases in the top quintile may indicate unusually high volume and crowding in the ED. We also included a binary indicator for ED visits occurring on the weekend.

Statistical analyses

We conducted univariate analyses to describe the sample. We used the PROC GENMOD command in SAS software version 9.4 (Cary, NC, USA) to estimate multivariable logistic models with standard errors clustered at the hospital level. We ran four separate regressions for AMI, other cardiovascular conditions, respiratory conditions, or mental disorders.

Results

Table 1 contains summary statistics for each variable included in the multivariable regressions. Of the 1,896,162 treat-and-release ED visits for chest symptoms in our sample, we identified 3709 (0.2%) with a subsequent inpatient admission for AMI, 32,710 (1.7%) with a subsequent inpatient admission for other cardiovascular conditions, 8577 (0.5%) with a subsequent inpatient admission for a respiratory condition, and 11,785 (0.6%) with a subsequent inpatient admission for a mental disorder within 2–30 days of discharge.

Table 1:

Summary statistics for index ED discharges for chest symptoms, January 2012 through November 2013.a,b

Variablen%
Index ED discharges (chest symptoms)1,896,162100
Subsequent admissions identified
 Acute myocardial infarction37090.2
 Other cardiovascular conditions32,7101.7
 Respiratory conditions85770.5
 Mental disorders11,7850.6
Mean, %SD
Patient characteristic
 Age in years, mean51.017.6
 Age in years, %
  18–4437.0
  45–6440.1
  65–8419.2
  85+3.7
 Sex, %
  Female56.2
  Male43.8
 Primary expected payer, %
  Medicare28.8
  Medicaid16.5
  Private insurance34.9
  Uninsured (self-pay or no charge)15.7
  Other4.1
 Patient income, %
  Quartile 1 (lowest)31.6
  Quartile 224.8
  Quartile 322.7
  Quartile 4 (highest)19.1
 Comorbidity index, mean1.11.3
 Race/ethnicity, %
  Hispanic15.2
  White (non-Hispanic)56.1
  Black (non-Hispanic)19.6
  Other (non-Hispanic)6.9
  Missing2.2
Hospital facility characteristic
 Teaching hospital, %
  Yes42.5
  No57.5
 Hospital region, %
  Northeast20.0
  Midwest8.1
  South40.7
  West31.2
 Hospital location, %
  Large metropolitan area (residents≥1 million)57.9
  Small metropolitan area (50,000≤residents<1 million)30.9
  Micropolitan area (10,000 ≤residents <50,000)8.0
  Noncore-based area (residents <10,000)3.2
 Ownership status, %
  Government16.6
  Private, for-profit16.3
  Private, nonprofit67.1
 Cardiac catheterization lab available, %
  Yes65.6
  No18.5
  Unknown15.9
VariableMean, %SD
Hospital volume characteristic
 Admits from ED by percentile, %
  0–20th4.5
  21st–40th13.9
  41st–60th21.9
  61st–80th27.7
  81st–100th29.2
 ED volume by percentile, %
  0–20th9.1
  21st–40th9.6
  41st–60th22.0
  61st–80th27.4
  81st–100th29.3
 Weekend ED visit, %
  Yes24.0
  No76.0

ED, emergency department. aThe sample consists of adults (age ≥18 years) with an ED visit for either nonspecific chest pain or coronary atherosclerosis and other heart disease in January 2012 through November 2013. Admissions 1 day after discharge from an ED were excluded. Admissions were assigned to a subsequent admission category based on the Clinical Classifications Software category of the principal diagnosis. bData sources were the 2012–2013 State Inpatient Databases and State Emergency Department Databases.

Table 2 shows the results of the multivariable logistic regressions for cardiovascular-related admissions following discharge from an ED for chest symptoms, presented as odds ratios (ORs) with 95% confidence intervals (CIs). In Model 1 for subsequent AMI admissions after discharge from an ED for chest symptoms, one additional year of age was associated with higher odds of subsequent admission (OR 1.04; 95% CI 1.04–1.04). Females had substantially lower odds of subsequent AMI admissions (OR 0.49; 95% CI 0.46–0.53). Income of the patient’s ZIP Code did not have a statistically significant impact on subsequent AMI admissions. Compared with those with private insurance, the odds of subsequent AMI admission was highest for Medicaid (OR 1.45; 95% CI 1.28–1.65), then uninsured (OR 1.17; 95% CI 1.03–1.34), and then Medicare (OR 1.12; 95% CI 1.00–1.26) patients. The only race/ethnicity group that differed from non-Hispanic Whites was Hispanics, which had lower odds of subsequent AMI admissions (OR 0.89; 95% CI 0.79–0.99). Odds of a subsequent AMI admission increased for each additional point on the comorbidity index scale (OR 1.21; 95% CI 1.18–1.24).

Table 2:

Multivariable logistic regression results for cardiovascular admissions within 30 days of discharge from an ED for chest symptoms.a,b

VariableModel 1 – AMI

Admission after ED discharge
Model 2 – Other cardiovascular conditions

Admission after ED discharge
OR95% CIOR95% CI
Patient characteristic
 Age1.041.04, 1.041.021.02, 1.02
 Female0.490.46, 0.530.530.52, 0.55
 Patient income
  Quartile 1 (lowest)1.080.96, 1.221.361.29, 1.42
  Quartile 21.020.91, 1.141.251.20, 1.32
  Quartile 31.010.91, 1.121.121.08, 1.18
  Quartile 4 (highest)c
  Missing1.130.87, 1.461.461.32, 1.61
 Primary payer
  Medicare1.121.00, 1.261.711.63, 1.79
  Medicaid1.451.28, 1.652.262.15, 2.37
  Private insurancec
  Uninsured1.171.03, 1.341.311.24, 1.39
  Other1.100.88, 1.381.591.46, 1.72
 Patient race/ethnicity
  White (non-Hispanic)c
  Black (non-Hispanic)0.920.82, 1.041.010.97, 1.05
  Other (non-Hispanic)0.950.82, 1.110.820.77, 0.87
  Hispanic0.890.79, 0.990.760.73, 0.80
  Missing0.860.69, 1.070.840.73, 0.95
 Comorbidity index1.211.18, 1.241.261.25, 1.27
Hospital facility characteristic
 Teaching hospital0.920.83, 1.030.990.91, 1.08
 Hospital size, no. of beds
  <100c
  101–2000.860.75, 0.990.930.85, 1.01
  201–3490.820.70, 0.960.900.82, 1.00
  ≥3500.780.65, 0.930.990.88, 1.12
 Ownership status
  Government1.030.91, 1.161.131.05, 1.22
  Private nonprofitc
  Private for-profit1.141.01, 1.301.251.15, 1.34
 Hospital region
  Northeast0.960.83, 1.091.161.07, 1.26
  Midwest1.211.03, 1.411.291.17, 1.42
  Southc
  West1.271.15, 1.411.211.13, 1.31
 Hospital location
  Large metropolitanc
  Small metropolitan1.060.96, 1.180.790.74, 0.84
  Micropolitan1.080.92, 1.270.750.68, 0.83
  Noncore1.411.12, 1.770.840.75, 0.94
 Cardiac catheterization lab
  Available0.880.78, 0.990.980.90, 1.07
  Unknown0.910.79, 1.051.050.96, 1.14
Hospital volume characteristic
 Admits from ED by percentile
  Missing1.060.44, 2.590.930.67, 1.29
  0–20th1.191.01, 1.390.990.91, 1.06
  21st–40th1.050.94, 1.191.000.96, 1.04
  41st–60thc
  61st–80th0.960.87, 1.061.010.98, 1.04
  81st–100th0.910.82, 1.021.071.03, 1.12
 ED volume by percentile
  Missing1.140.46, 2.841.080.77, 1.51
  0–20th0.960.85, 1.081.041.00, 1.08
  21st–40th1.020.90, 1.151.061.02, 1.10
  41st–60thc
  61st–80th0.980.89, 1.080.990.96, 1.01
  81st–100th0.930.85, 1.020.980.95, 1.01
 Weekend ED visit1.071.00, 1.151.051.03, 1.08

AMI, acute myocardial infarction; CI, confidence interval; ED, emergency department; OR, odds ratio. aAll regressions were estimated with robust standard errors clustered at the hospital level with an exchangeable correlation structure. The sample consists of adults (age≥18 years) with an ED visit for either nonspecific chest pain or coronary atherosclerosis and other heart disease in the January 2012 through November 2013 time period. Admissions 1 day after discharge from an ED were excluded. Admissions were assigned to a subsequent admission category based on the Clinical Classifications Software category of the principal diagnosis. bData sources were the 2012–2013 State Inpatient Databases and State Emergency Department Databases. cRepresents the reference group.

Table 2 also contains results for facility characteristics. Lower odds of subsequent AMI admission after discharge from the ED were found for patients seen in EDs at the largest hospitals (OR 0.78; 95% CI 0.65–0.93 for ≥350 beds relative to <100 beds). Compared with visits to EDs in private nonprofit hospitals, patients with index ED visits at private for-profit hospitals experienced higher odds of subsequent AMI admission (OR 1.14; 95% CI 1.01–1.30). Compared with patients with index ED visits in hospitals in the South, patients in the Midwest (OR 1.21; 95% CI 1.03–1.41) and West (OR 1.27; 95% CI 1.15–1.41) had higher odds of subsequent AMI admission. ED patients in the least dense (noncore) locations experienced higher odds of subsequent AMI admission (OR 1.41; 95% CI 1.12–1.77) than patients in large metropolitan areas. Having a cardiac catheterization lab available at a hospital decreased the odds of subsequent AMI admission (OR 0.88; 95% CI 0.78–0.99).

Regarding hospital capacity impacts, using the midrange quintile (41st to 60th) of daily admission rates directly from the ED as the comparison group, only the lowest (0–20th) quintile of daily direct ED admissions had higher odds of subsequent AMI admission (OR 1.19; 95% CI 1.01–1.39). There were no effects of relative ED volume or weekend ED visits on subsequent AMI admissions.

The results of Model 2, which expanded the possible reasons for subsequent admission for other cardiovascular events, were similar to those of Model 1 with the following exceptions. In Model 2, patients in the three lowest income quartiles had substantially higher odds of subsequent cardiovascular admissions after discharge from an ED for chest symptoms, compared with patients who had subsequent AMI admissions. The odds of subsequent admissions after discharge from an ED for chest symptoms by insurance were substantially higher in Model 2 than in Model 1. In Model 2, patients with index ED visits in small metropolitan, micropolitan, and noncore areas had lower odds of subsequent admission than those who used EDs in large metropolitan locations, whereas only the noncore areas differed significantly in Model 1. The availability of a cardiac catheterization lab had no effect in Model 2. Although there were few significant results based on admissions from the ED or ED volume, odds of a subsequent admission were higher for weekend ED visits.

Table 3 contains results of the multivariable logistic regressions for noncardiovascular-related admissions following an ED visit for chest symptoms. Model 3 looked at subsequent admissions for respiratory conditions following discharge from an ED for chest symptoms. These results closely resemble the cardiovascular results in both significance and magnitude for most factors. However, although lower odds of subsequent admissions were seen for Hispanic patients (OR 0.68; 95% CI 0.63–0.75) in all four models, other non-Hispanic patients (OR 0.61; 95% CI 0.54–0.69) and those with missing race/ethnicity data (OR 0.78; 95% CI 0.61–0.98) also had lower odds compared with White non-Hispanic patients in Model 3. Unlike the results for Models 1 and 2, non-Hispanic Black patients had significantly lower odds of subsequent admissions (OR 0.85; 95% CI 0.79–0.91) for respiratory conditions.

Table 3:

Multivariable logistic regression results for respiratory and mental disorder admissions within 30 days of discharge from an ED for chest symptoms.a,b

VariableModel 3 – Respiratory conditions

Admission after ED discharge
Model 4 – Mental disorders

Admission after ED discharge
OR95% CIOR95% CI
Patient characteristic
 Age1.021.02, 1.020.970.97, 0.98
 Female0.850.81, 0.890.500.47, 0.53
 Patient income
  Quartile 1 (lowest)1.681.52, 1.851.341.21, 1.48
  Quartile 21.431.30, 1.571.141.04, 1.26
  Quartile 31.181.08, 1.291.020.93, 1.13
  Quartile 4 (highest)c
  Missing1.571.32, 1.862.692.30, 3.15
 Primary payer
  Medicare2.792.56, 3.043.723.42, 4.05
  Medicaid3.473.19, 3.774.394.03, 4.80
  Private insurancec
  Uninsured1.331.20, 1.491.961.80, 2.15
  Other1.691.41, 2.022.642.30, 3.04
 Patient race/ethnicity
  White (non-Hispanic)c
  Black (non-Hispanic)0.850.79, 0.910.600.56, 0.65
  Other (non-Hispanic)0.610.54, 0.690.440.39, 0.50
  Hispanic0.680.63, 0.750.500.45, 0.56
  Missing0.780.61, 0.980.820.66, 1.03
  Comorbidity index1.291.27, 1.321.301.28, 1.32
Hospital facility characteristic
 Teaching hospital0.910.81, 1.011.090.95, 1.25
 Hospital size, no. of beds
  <100c
  101–2000.970.84, 1.121.140.97, 1.34
  201–3490.900.78, 1.051.321.09, 1.59
  ≥3501.000.84, 1.191.641.32, 2.03
 Ownership status
  Government1.080.97, 1.201.251.06, 1.48
  Private nonprofitc
  Private for-profit1.231.11, 1.361.351.17, 1.56
  Hospital region
  Northeast1.100.99, 1.231.951.65, 2.29
  Midwest1.241.08, 1.411.991.70, 2.32
  Southc
  West1.030.93, 1.141.100.93, 1.29
 Hospital location
  Large metropolitanc
  Small metropolitan0.940.86, 1.020.610.53, 0.69
  Micropolitan0.810.71, 0.930.540.46, 0.64
  Noncore0.910.76, 1.100.470.38, 0.59
 Cardiac catheterization lab
  Available0.940.83, 1.051.040.89, 1.22
  Unknown0.960.84, 1.091.181.02, 1.38
Hospital volume characteristics
 Admits from ED by percentile
  Missing1.180.86, 1.611.230.86, 1.77
  0–20th1.030.91, 1.170.860.75, 0.99
  21st–40th0.980.90, 1.070.930.86, 1.01
  41st–60thc
  61st–80th1.030.96, 1.101.030.97, 1.09
  81st–100th1.080.99, 1.171.091.01, 1.17
 ED volume by percentile
  Missing1.200.86, 1.670.740.52, 1.06
  0–20th0.960.88, 1.050.980.91, 1.05
  21st–40th0.940.87, 1.021.000.94, 1.06
  41st–60thc
  61st–80th0.980.93, 1.041.000.95, 1.05
  81st–100th0.970.92, 1.030.980.93, 1.03
 Weekend ED visit1.131.08, 1.181.151.10, 1.19

CI, confidence interval; ED, emergency department; OR, odds ratio. aAll regressions were estimated with robust standard errors clustered at the hospital level with an exchangeable correlation structure. The sample consists of adults (age≥18 years) with an ED visit for either nonspecific chest pain or coronary atherosclerosis and other heart disease in the January 2012 through November 2013 time period. Admissions 1 day after discharge from an ED were excluded. Admissions were assigned to a subsequent admission category based on the Clinical Classifications Software category of the principal diagnosis. bData sources were the 2012–2013 State Inpatient Databases and State Emergency Department Databases. cRepresents the reference group.

Model 4 examined the association of discharge from an ED for chest symptoms with subsequent admissions for mental disorders. The results were similar to previous models, although the direction of the ORs for age, hospital size, and ED activity were different, and magnitudes of some other associations were stronger. In Model 4, increasing age decreased the odds of subsequent admission for mental disorders (OR 0.97; 95% CI 0.97–0.98). In Model 4, insurance effects were in the same direction as in other models but were stronger for subsequent admissions for mental disorders (e.g., for Medicaid, OR 4.39; 95% CI 4.03–4.80). Unlike the other models, patients with index ED visits in hospitals with more beds had higher odds for subsequent admissions for mental disorders. Patients with index ED visits in private for-profit hospitals had higher odds of subsequent admission for mental disorders (as was seen in the other models), as did patients receiving ED services in government hospitals (OR 1.25; 95% CI 1.06–1.48). Patients receiving ED care on days when the admission rate was the lowest quintile (0–20th percentile) had lower odds of subsequent mental disorder admissions (OR 0.89; 95% CI 0.75–0.99), whereas those with index ED visits on days in the highest quintile (81st–100th percentile) had higher odds (OR 1.09; 95% CI 1.01–1.17) than did patients in the midrange of ED activity (41st–60th percentile).

Discussion

Our findings have potential implications for evaluation of diagnostic performance. These relate to (1) the effect of insurance in the context of the Affordable Care Act and (2) variation in results for different types of patients across diagnostic categories.

The odds of subsequent admission after discharge from an ED for chest symptoms were often 2–3 times higher for uninsured patients and those covered by Medicaid, Medicare, or other programs than for privately insured patients, especially for those with respiratory or mental conditions. This finding implies that among those discharged, (1) privately insured patients are more likely than other patients to receive accurate diagnoses at the time of an ED visit for chest symptoms and (2) respiratory and mental disorder diagnoses related to chest symptoms are more likely to be missed when patients without private insurance visit the ED with chest symptoms.

As the Affordable Care Act has increased the number of patients who have private insurance via state exchanges, our findings suggest the accuracy of diagnoses may improve for these privately insured patients presenting to the ED with chest symptoms. Whether this is true depends on the underlying reasons for fewer subsequent admissions for privately insured patients. Does the relative generosity of private reimbursement influence diagnosis? Or does the clinical profile of patients with private insurance somehow drive the cognitive process of diagnosis? Or both? Considering Medicaid expansion under the Affordable Care Act might lead us to conclude the opposite – that new Medicaid patients are more likely to have subsequent admissions when they present with chest symptoms. This may be because Medicaid reimbursement rates are lower than private rates or because Medicaid patients have a younger profile, unlike the typical chest symptom patient. Furthermore, payer groups may reflect disease severity that is observed by ED clinicians but not captured by our other covariates. If our results reflect subtle differences in provider behavior rather than bias due to unmeasured patient severity, hospital administrators may be able to improve patient safety and reduce subsequent admissions by raising awareness among ED clinicians of the potential for missing diagnoses.

Compared with other diagnoses, we found fewer eventual AMI diagnoses via subsequent admission and also that patient attributes were less important in predicting subsequent AMI admissions. For example, income of the patient’s ZIP Code did not influence the odds of AMI admissions after ED release, but it was strongly associated with other diagnosis outcomes. The effects of payer and comorbidity were smaller for AMI than for other diagnoses. The acute and life-threatening nature of AMI relative to the other chronic conditions studied, the common approach of ruling out AMI first among reasons for chest pain, and the effectiveness of patient education about chest pain and heart attacks all may intensify focus on AMI as a potential diagnosis when chest symptoms are presented.

At the other end of the spectrum is the mental disorder category. Almost all the patient and hospital factors we studied had a higher association with subsequent admissions for mental disorders after discharge from an ED for chest symptoms. Public payment or uninsured patients were 2–4 times more likely than privately insured patients to return for a subsequent mental disorder admission after discharge from an ED for chest symptoms. It is notable that patients in larger hospitals were more likely to return for a subsequent mental disorder admission after discharge from an ED for chest symptoms because the opposite was the case for all other delayed diagnosis categories. Also patients in EDs in larger metropolitan areas were more likely to return for subsequent admission for mental disorders than in smaller geographic areas. These findings suggest that hospitals that deal with mental disorders on an ongoing basis (large, metropolitan, public hospitals) may be more prone to overlooking a mental disorder when patients present with chest symptoms. Or these institutions may have capacity constraints leading them to discharge these patients knowing their conditions are not life threatening.

This study has four limitations. First, because it used data from eight states, the results cannot be generalized nationwide. Second, because the data came from administrative discharge records, we could not distinguish the complexity of illness and difficulty of diagnosis of patients based on laboratory test results, physical history, mental tests, symptom details, or emergency interventions, all of which have implications for diagnostic complexity and performance. However, the Elixhauser Comorbidity system has demonstrated superior performance for measuring the effect of comorbid conditions on patient outcomes when using administrative data [17]. Third, the study may underestimate the percentage of treat-and-release ED visits for chest symptoms resulting in a subsequent inpatient admission. We excluded admissions identified 1 day after discharge from the ED to avoid counting them as subsequent admissions when they may have been direct admissions from the ED. This exclusion may have resulted in the omission of true ED discharges that subsequently were admitted within one calendar day; however, this affected <1% of the analytic sample. Fourth, as mentioned earlier, HCUP data do not capture mortality outside the hospital setting. Therefore, patients who died within 30 days of their ED discharge would have been included in the denominator of patients who were eligible for subsequent admission, which could bias the results in unknown ways.

Conclusions

Our study clarifies how often patients with ED visits for chest symptoms experience subsequent inpatient admissions for related symptoms, providing new information on patient and hospital characteristics associated with subsequent admissions for a diverse set of related clinical conditions estimated in a large, all-payer database. The clinical conditions potentially related to chest symptoms that we studied include a broad definition of cardiovascular conditions, respiratory conditions, and mental disorders, extending the generalizability of our results beyond AMI cases. Our findings could inform the work of medical specialty organizations that are developing guidelines and tools for triage of patients with chest pain. And, ED administrators might use these results to identify populations with increased odds of returning to the hospital after discharge from the ED for chest symptoms.


Corresponding author: Brian J. Moore, PhD, Truven Health Analytics, 100 Phoenix Dr Ann Arbor, Ann Arbor, MI, 48108, USA

Acknowledgments

The authors wish to acknowledge Marguerite L. Barrett for contributions to the methodological design and selection of the analytic sample, as well as Minya Sheng for programming support. The authors would like to acknowledge the HCUP Partner organizations that participated in the HCUP State Inpatient Databases (SID) and State Emergency Department Databases (SEDD). A list of all HCUP Partners is available at the HCUP User Support Website (https://www.hcup-us.ahrq.gov/partners.jsp). The HCUP Partner organizations that contributed to the data used in this study were the California Office of Statewide Health Planning and Development, the Florida Agency for Health Care Administration, the Missouri Hospital Industry Data Institute, the Nebraska Hospital Association, the New York State Department of Health, the South Carolina Revenue and Fiscal Affairs Office, the Tennessee Hospital Association, and the Vermont Association of Hospitals and Health Systems.

  1. Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission.

  2. Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect those of the Agency for Healthcare Research and Quality, the Centers for Disease Control and Prevention, the National Center for Health Statistics, or the U.S. Department of Health and Human Services.

  3. Research funding: This research was supported by the Agency for Healthcare Research and Quality, Center for Delivery, Organization, and Markets, Healthcare Cost and Utilization Project (HCUP).

  4. Employment or leadership: None declared.

  5. Honorarium: None declared.

  6. Competing interests: The funding organization(s) played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.

References

1. Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project (HCUP). HCUPnet. 2012. Available at: http://hcupnet.ahrq.gov/. Accessed: 29 Sept 2015.Search in Google Scholar

2. Moy E, Coffey RM, Moore BJ, Barrett ML, Hall KK. Length of stay in EDs: variation across classifications of clinical condition and patient discharge disposition. Am J Emerg Med 2016;34:83–7.10.1016/j.ajem.2015.09.031Search in Google Scholar

3. Schull MJ, Vermeulen MJ, Stukel TA. The risk of missed diagnosis of acute myocardial infarction associated with emergency department volume. Ann Emerg Med 2006;48:647–55.10.1016/j.annemergmed.2006.03.025Search in Google Scholar

4. Foy AJ, Guodong L, Davidson WR, Sciamanna C, Leslie DL. Comparative effectiveness of diagnostic testing strategies in emergency department patients with chest pain: an analysis of downstream testing, interventions, and outcomes. JAMA Intern Med 2015;175:428–6.10.1001/jamainternmed.2014.7657Search in Google Scholar

5. Schuur JD, Venkatesh AK. The growing role of emergency departments in hospital admissions. N Engl J Med 2012;367:391–3.10.1056/NEJMp1204431Search in Google Scholar

6. Lee TH, Rouan GW, Weisberg MC, Brand DA, Acampora D, Stasiulewicz C, et al. Clinical characteristics and natural history of patients with acute myocardial infarction sent home from the emergency department. Am J Cardiol 1987;60:219–24.10.1016/0002-9149(87)90217-7Search in Google Scholar

7. McCarthy BD, Beshansky JR, D’Agostino RB, Selker HP. Missed diagnoses of acute myocardial infarction in the emergency department: results from a multicenter study. Ann Emerg Med 1993;22:579–82.10.1016/S0196-0644(05)81945-6Search in Google Scholar

8. Pope HJ, Aufderheide TP, Ruthazer R, Woolard RH, Feldman JA, Beshansky JR, et al. Missed diagnoses of acute cardiac ischemia in the emergency department. N Engl J Med 2000;342:1163–70.10.1056/NEJM200004203421603Search in Google Scholar PubMed

9. Wilson M, Welch J, Schuur J, O’Laughlin K, Cutler D. Hospital and emergency department factors associated with variations in missed diagnosis and costs for patients age 65 years and older with acute myocardial infarction who present to emergency departments. Acad Emerg Med 2014;21:1101–8.10.1111/acem.12486Search in Google Scholar PubMed PubMed Central

10. Moy E, Barrett M, Coffey R, Hines AL, Newman-Toker DE. Missed diagnoses of acute myocardial infarction in the emergency department: variation by patient and facility characteristics. Diagn 2015;2:29–40.10.1515/dx-2014-0053Search in Google Scholar PubMed

11. Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project (HCUP). State Emergency Department Databases (SEDD) [database online]. Updated June 27, 2014. Available at: www.hcup-us.ahrq.gov/seddoverview.jsp. Accessed: 11 Feb 2015.Search in Google Scholar

12. Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project (HCUP). Overview of the State Inpatient Databases (SID) [database online]. Updated June 27, 2014. Available at: http://www.hcup-us.ahrq.gov/sidoverview.jsp. Accessed: 11 Feb 2015.Search in Google Scholar

13. Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project (HCUP). Clinical Classifications Software (CCS) for ICD-9-CM [database online]. Updated July 29, 2014. Available at: www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp. Accessed: 11 Feb 2015.Search in Google Scholar

14. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care 1998;36:8–27.10.1097/00005650-199801000-00004Search in Google Scholar PubMed

15. Van Walraven C, Austin PC, Jennings A, Quan H, Forster AJ. A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care 2009;47:626–33.10.1097/MLR.0b013e31819432e5Search in Google Scholar PubMed

16. Thomson NR, Fan Y, Dalton JE, Jehi L, Rosenbaum BP, Vadera S, et al. A new Elixhauser-based comorbidity summary measure to predict in-hospital mortality. Med Care 2015;53:374–9.10.1097/MLR.0000000000000326Search in Google Scholar PubMed PubMed Central

17. Stukenborg GJ, Wagner DP, Connors AF Jr. Comparison of the performance of two comorbidity measures, with and without information from prior hospitalizations. Med Care 2001;39: 727–39.10.1097/00005650-200107000-00009Search in Google Scholar PubMed


Supplemental Material:

The online version of this article (DOI: 10.1515/dx-2016-0014) offers supplementary material, available to authorized users.


Received: 2016-4-12
Accepted: 2016-7-18
Published Online: 2016-8-31
Published in Print: 2016-9-1

©2016 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 28.3.2024 from https://www.degruyter.com/document/doi/10.1515/dx-2016-0014/html
Scroll to top button