The Influence of Insurance, Race, and Gender ... - Wiley Online Library

1260

Selassie et al.

d

INFLUENCE OF INSURANCE, RACE, AND GENDER ON ED DISPOSITION

The Influence of Insurance, Race, and Gender on Emergency Department Disposition Anbesaw Wolde Selassie, DrPH, Melissa Lee McCarthy, MS, ScD, Emily Elisabeth Pickelsimer, DA Abstract Objectives: To examine the influence of insurance, race, and gender on the likelihood of hospitalization among trauma patients. Methods: Statewide hospital discharge and emergency department (ED) visit data collected between 1996 and 2000 were merged to examine factors that influence hospitalization among patients who sustained an injury. Multivariate logistic regression was used to model the likelihood of hospitalization as a function of patient, injury, and hospital characteristics. Results: Of 1,512,611 patients who presented to an ED in South Carolina for treatment of a traumatic injury during the five-year study period, 8% were hospitalized and 92% were treated and released. One fourth (26%) of the study population was uninsured. Insurance, race, and gender were significant predictors of hospitalization despite controlling for injury severity, comorbidities, age, trauma center level, place of residence, and year of injury. Regardless of injury severity, uninsured

patients were significantly less likely to be hospitalized compared with privately insured patients (odds ratio [OR] 0.63, 99% CI ¼ 0.62 to 0.65). Among those mildly to moderately injured, patients covered by Medicare or other government insurance policies were significantly more likely to be admitted compared with those with private coverage (OR 1.46, 99% CI ¼ 1.41 to 1.52; OR 1.56, 99% CI ¼ 1.36 to 1.78). Finally, among those mildly injured, African American females were significantly less likely to be admitted compared with white females (OR 0.63, 99% CI ¼ 0.61 to 0.65). Conclusions: These results suggest that the disposition of trauma patients from the ED may be influenced by insurance and demographic characteristics in addition to the patient’s clinical condition. Key words: emergency department; disposition; race; insurance; gender. ACADEMIC EMERGENCY MEDICINE 2003; 10:1260–1270.

Numerous studies have documented that persons who are uninsured receive less medical care than those who are insured.1–3 Uninsured patients obtain fewer preventive services and do not receive as many diagnostic and therapeutic interventions for a variety of medical conditions as those who are insured.4–12 Many of these health care disparities also exist among racial minorities and women.13–23 These inequities are associated with serious, adverse outcomes such as higher mortality rates, higher hospitalization rates for ambulatory care–sensitive conditions, and higher rates of medical injury caused by negligence for

patients who are uninsured, a member of a minority, and/or female.5,7,10,11,24–30 Most studies to date have documented inequities in the receipt of specific procedures or services among different patient subgroups; relatively few have examined determinants of hospital admission.31–34 The purpose of this study was to examine the effect of insurance, race, and gender on the likelihood of hospitalization among patients who presented to any nonfederal, hospitalbased emergency department (ED) in South Carolina between 1996 and 2000 for treatment of a traumatic injury.

From the Department of Biometry and Epidemiology, Medical University of South Carolina, Charleston, SC (AWS, EEP); and the Department of Emergency Medicine, Johns Hopkins University School of Medicine, Baltimore, MD (MLM). Received August 13, 2002; revision received November 22, 2002; accepted December 2, 2002. Supported in part by cooperative agreement U17/CCU414981 from the Division of Injury and Disability Outcome Programs, National Center for Injury Prevention and Control, The Centers for Disease Control and Prevention. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the CDC. Address for correspondence and reprints: Anbesaw Selassie, DrPH, Department of Biometry and Epidemiology, Medical University of South Carolina, 135 Cannon Street, PO Box 250835, Charleston, SC 29425. e-mail: [email protected]. doi:10.1197/S1069-6563(03)00497-4

METHODS Study Design. This study relied on statewide hospital discharge and ED visit data for 1996–2000. State law mandates that all 62 nonfederal hospitals report uniform, abstracted billing data to the South Carolina State Budget and Control Board. The data files contain patient identifiers (i.e., medical record number, personal unique ID, first and last name, address, and ZIP code), demographics (date of birth, race, gender, place of residence), dates of admission/ED visit, ten International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnostic codes (one primary and nine secondary codes), external causes of injury codes, length of stay,

ACAD EMERG MED

d

November 2003, Vol. 10, No. 11

d

www.aemj.org

discharge disposition, source of admission, principal payer, hospital identification number, and trauma center level. The South Carolina Injury Surveillance System verifies the accuracy and completeness rate of these data as part of a routine evaluation protocol. The abstracted data are 99% accurate and complete.35 This study was reviewed and exempted by the institutional review board. Study Setting and Population. All patients who sustained a traumatic injury that resulted in an ED visit to any nonfederal, hospital-based ED located in South Carolina between January 1, 1996, and December 31, 2000, were eligible for the study. Trauma was defined using the ICD-9-CM diagnostic codes N800– N95936 as damage to the body resulting from an external physical force. Trauma-related ED visits/ admissions were identified by searching the primary and all secondary diagnostic code fields and selecting those records that contained at least one trauma ICD9-CM code. Injuries caused by complications (N958), late effects (N905–N909), or foreign bodies (N930– N939) were not eligible. In addition, patients who left the ED against medical advice were excluded. Study Protocol. Figure 1 summarizes the derivation of the study sample. First, the hospital discharge and ED files of patients with a traumatic injury were merged using a deterministic linkage approach. Both data sets contain a unique identification number that consists of the patient’s scrambled social security number, date of birth, race, and gender. This unique

Figure 1. Derivation of the study sample. ED = emergency department.

1261

identification number was used to identify patients with more than one ED visit or a combination of an ED visit and inpatient admission during the study period. The mismatch error rate is \2% using this linkage method with these data sets.35 For patients admitted to the hospital through the ED, the hospital record was selected and the ED visit deleted. Interhospital transfers were identified and only the record from the final acute care facility was retained. Second, the following types of records were deleted: 1) 1,090 duplicates (i.e., matched exactly on all patient demographic and clinical characteristics); 2) 381 records that had missing covariate information (e.g., race, gender, insurance status, etc.) needed for the analysis; and 3) 1,006 deaths that occurred in the ED. Finally, 15% of patients had more than one traumarelated ED visit or hospitalization during the five-year study period. There was significant variation in the timing of these later ED visits/hospitalizations: 23% occurred within one month, 26% occurred between one and six months, 20% occurred between seven and 12 months, 20% occurred between 1 to 2 years, and 11% occurred more than two years after the initial ED visit or hospitalization. Because it was not possible to determine whether the later ED visit/hospitalization was related to the first one without reviewing over 360,000 medical records, especially among those that occurred within a short period (i.e., #6 months), the trauma-related ED visit or hospitalization associated with the most severe injury was retained for analysis. After the final study sample was derived, variables were constructed for analysis. Injury severity was determined by translating the ICD-9-CM diagnosis codes into Abbreviated Injury Scale (AIS) scores using ICDMAP-90 software.37 The AIS is an anatomic measure of injury severity that classifies over 2,000 injuries according to the body region of injury (e.g., head, chest, extremity), type of structure injured (e.g., nerve, vessel, bone), location of injury within the body region (e.g., femur, tibia, talus), and nature of injury (e.g., abrasion, burn, crush). The AIS grades each injury according to its associated threat to life on an ordinal scale from 1 (minor) to 6 (unsurvivable).38,39 AIS ratings of 3 or higher are considered significant injuries. Because the AIS characterizes only the severity of single injuries, the Injury Severity Score (ISS) was used to reflect the cumulative effect of multiple injuries.40 The ISS is the sum of the squares of the highest AIS score in each of the three most severely injured body regions. The ISS ranges from 1 to 75; the higher the score, the more severe the injury(ies). Because higher ISSs can be the result of either one severe injury or several moderate injuries to different body regions, a variable was constructed that reflected the number of different body regions injured. Analyses showed that the ISS and the numberof-body-regions-injured variable were not statistically

1262

Selassie et al.

d


correlated because of sufficient variation in the proportion with injuries to more than one body region among the different ISS categories. ICD-9-CM diagnostic codes were used to measure the number of serious pre-existing conditions each patient had. Elixhauser and colleagues developed a comorbidity scale for use with administrative data that identifies 30 conditions that are significant predictors of in-hospital mortality and resources.41 All ten ICD-9-CM diagnosis fields were searched for any of these 30 conditions, and the total number of conditions present was used to reflect each patient’s underlying pre-existing health. Finally, although race was documented in the data files as white, African American, or other, \2% of the entire study population was classified as other. Those classified as other were combined with African American and are collectively referred to as African American throughout the article. Data Analysis. The dependent variable, type of care, was defined as ‘‘hospitalized’’ for patients who were admitted to the hospital and ‘‘ED managed’’ for those treated and released from the ED. Burn patients who were transferred to a burn center (n ¼ 596) as well as patients who sustained severe trauma and were transferred to an intensive skilled nursing facility (n ¼ 744) were categorized as ‘‘hospitalized.’’ The relationship between type of care and patient, injury, and hospital characteristics was examined using a chisquare test statistic for all trauma patients who were hospitalized or treated and released from the ED.

Type of care also was modeled as a function of age, race and gender, insurance status, ISS, number of body regions injured, number of pre-existing conditions, place of residence, year of injury, and trauma center level using unconditional multivariate logistic regression techniques. The categorization of the independent variables was based on their relationship with type of care (see Table 3 for coding of variables). The effect of race and gender was examined together because the effects of each on ED disposition differed significantly depending on the category of the other. Multicollinearity among the independent variables was evaluated by assessing the deviations of the regression coefficients and their standard errors in the fitted univariate and multivariate models.42 The independent variables were entered simultaneously into the model. Because of the large sample size, an alpha level of 0.01 was chosen to reduce the likelihood of a type 1 error. Both the unadjusted (from the univariate models) and the adjusted odds ratios (ORs) (from the final multivariate model that includes all independent variables) and 99% confidence intervals (99% CIs) are reported.

RESULTS The vast majority (92.4%) of the more than 1.5 million patients who presented to a South Carolina ED between 1996 and 2000 for treatment of a traumatic injury were treated and released from the ED. More than half of the study population (53.6%) sustained minor injuries to only one body region (see Table 1).

TABLE 1. Percent Distribution of Type of Injury by Body Region and Type of Care Type of Care System Involved

N

% Hospitalized

% ED Managed

293,988 9,006 9,953 20,881 973 230,052 23,123

27 83 60 57 45 22 13

73 17 40 43 55 78 87

59,595 315 1,298 50,032 1,721 6,229

9 76 23 6 10 28

91 24 77 94 90 72

Minor single system injuries Minor head injuries (AIS ¼ 1) Minor spine injuries (AIS ¼ 1) Minor thoracic/abdominal injuries (AIS ¼ 1) Minor upper/lower-extremity injuries (AIS ¼ 1) Minor face/neck injuries (AIS ¼ 1)

810,246 38,216 178,847 56,194 536,905 84

2 4 1 6 2 18

98 96 99 94 98 82

Minor multiple system injuries Multiple minor injuries (AIS ¼ 1) without head

348,782 348,782

4 4

96 96

Major single-system injuries Head injury only (AIS $ 3) Spine injury only (AIS $ 2) Thorax or abdominal injury only (AIS $ 2) Face/neck injury only (AIS $ 2) Extremity injury, upper/lower (AIS $ 2) Moderate head injury (AIS ¼ 2) Major/moderate multiple system injuries Major head (AIS $ 3) and other injuries (AIS $ 1) Moderate head (AIS ¼ 2) and other injuries (AIS $ 1) Other multiple injuries (AIS $ 2) without head Other injuries (AIS $ 2) with minor head (AIS ¼ 1) Other injuries (AIS $ 2) with any minor injury (AIS ¼ 1)

N ¼ 1,512,611.

ACAD EMERG MED

d


d

1263

www.aemj.org

Minor injuries were most likely to occur to the upperor lower-extremity body regions. Nearly all (98%) patients sustaining minor trauma were treated and released from the ED. Whereas major traumas (AIS $ 3) that involved single body regions represented approximately one fifth (19.4%) of the total trauma in the study, they accounted for 68.5% of the hospital admissions. Overall, persons who sustained a major head injury (i.e., AIS $ 3) with or without an associated injury to another body region had the highest hospitalization rates (76% and 83%, respectively). Higher admission rates (i.e., [40%) also were associated with major injuries (i.e., AIS $ 2) to the

spine, thoracic/abdominal, or face/neck body regions. Table 2 presents the frequency distribution of patient, injury, and hospital characteristics of the study sample by type of insurance. Approximately one fourth (26%) of the study population was uninsured. Persons who were uninsured had the lowest hospitalization rate (3.7%), whereas patients with Medicare or other governmental insurance had the highest (23.7%). Uninsured patients were more likely to be young adults (73.3%), male (57.6%), and residing in an urban area (64.1%) at the time of injury. On average, patients with Medicare or other

TABLE 2. Percent Distribution of Patient, Injury, and Hospital Characteristics by Insurance Status Insurance Status

Characteristics

N (1,512,611)

Uninsured (n ¼ 397,487)

Medicaid (n ¼ 195,509)

Medicare and Other Government (n ¼ 208,547)

Private (n ¼ 711,068)

Column Percent Type of care Hospitalized ED managed Race and gender African American male African American female White male White female Age (yr) 65 and over 45–64 25–44 15–24 0–14 Pre-existing conditions Three or more Two One None ISS Severe (15–75) Moderate (9–14) Mild (1–8) Number of body regions injured $3 body regions 2 body regions 1 body region Trauma level status Level I Level II Level III Undesignated Year of injury 2000 1999 1998 1997 1996 Residence Urban Out of state Rural

115,261 1,397,350

3.7 96.3

4.7 95.3

23.7 76.3

5.9 94.1

303,310 265,833 500,420 443,048

26.6 19.6 31.0 22.8

28.0 30.2 19.8 22.0

11.5 14.1 27.7 46.7

16.7 14.0 39.5 29.8

158,559 233,491 495,893 302,606 322,062

1.1 12.4 45.5 27.8 13.2

0.6 7.0 15.8 18.8 57.8

66.5 14.6 11.3 4.2 3.4

2.0 19.7 36.7 20.6 21.0

13,921 28,521 81,361 1,388,808

0.1 0.6 3.4 94.9

0.4 1.0 3.7 94.9

5.2 8.5 15.6 70.7

0.3 0.9 4.0 94.8

7,747 42,429 1,474,389

0.4 1.2 98.4

0.6 1.6 97.8

1.5 11.6 86.9

0.6 1.6 97.8

28,470 146,489 1,352,513

2.2 11.6 86.2

1.0 6.0 93.0

2.1 11.3 86.6

1.9 9.0 89.1

234,171 115,785 617,659 559,857

17.6 6.8 41.6 34.0

15.0 8.0 39.1 37.9

15.7 7.3 34.8 42.3

13.9 8.0 41.8 36.3

324,983 324,903 292,173 288,621 296,792

19.6 20.5 19.8 20.2 19.9

23.4 23.1 18.5 16.8 18.2

22.4 21.6 18.6 17.7 19.7

21.4 21.3 19.3 19.2 18.8

902,385 76,085 549,002

64.1 5.3 30.6

55.9 1.2 42.9

57.8 4.4 37.8

57.8 6.1 36.1

1264

Selassie et al.

d


governmental insurance were older, had one or more comorbidities, and sustained more severe trauma compared with patients in the other three insurance groups (p \ 0.01). Patients who were uninsured were not statistically different from patients with Medicaid or privately insured in terms of injury severity, number of body regions injured, and the presence of comorbidities (chi-square [ 0.01). Table 3 displays the frequency distribution of all of the independent variables by type of care. Table 4 summarizes the unadjusted and adjusted ORs of these factors on the likelihood of hospital admission. Because there was a consistent relationship between the bivariate and multivariate results, only the

adjusted ORs from the multivariate regression are reviewed below. After adjusting for all variables listed in Table 4, the most important predictor of hospital admission was injury severity. Furthermore, there was a strong dose– response relationship noted. As the severity of the injury worsened, an increasing proportion of patients were hospitalized. In addition, patients who sustained injuries to more than one body region or had preexisting conditions documented were significantly more likely to be hospitalized compared with patients with a single body region injured or without any comorbidities. Among those admitted, the most frequently documented pre-existing conditions were

TABLE 3. Percent Distribution of Patient, Injury, and Hospital Characteristics by Type of Care Type of Care Characteristics Insurance status* Uninsured Medicaid Medicare & other government Commercial Race and gender* African American female African American male White female White male Age* (yr) 65 and over 45–64 25–44 15–24 0–14 Pre-existing conditions* Three or more Two One None ISS* Severe (15–75) Moderate (9–14) Mild (1–8) Number of body regions injured* $3 body regions 2 body regions 1 body region Trauma level status* Level I Level II Level III Undesignated Year of injury* 2000 1999 1998 1997 1996 Residence* Out of state Urban Rural *p \ 0.0001.

Hospitalized (n ¼ 115, 261)

ED Managed (n ¼ 1,397,350)

397,487 195,509 208,547 711,068

3.7 4.7 23.7 5.9

96.3 94.3 76.3 94.1

265,833 303,310 443,048 500,420

4.8 6.7 9.7 7.9

95.2 93.3 90.3 92.1

158,559 233,491 495,893 302,606 322,062

27.0 9.4 5.4 4.5 3.1

73.0 90.6 94.6 95.5 96.9

13,921 28,521 81,361 1,388,808

88.0 57.4 30.1 4.5

12.0 42.6 69.9 95.5

10,191 39,893 1,462,527

91.2 83.9 5.0

8.8 16.1 95.0

28,000 144,761 1,339,850

32.3 11.8 6.7

67.7 88.2 93.3

230,407 114,915 612,497 554,792

13.6 11.0 6.2 6.0

86.4 89.0 93.8 94.0

319,280 320,481 288,161 288,223 296,412

7.2 7.2 7.7 7.7 8.5

92.8 92.8 92.3 92.3 91.5

79,187 893,290 540,134

5.2 7.0 8.7

94.8 93.0 91.3

N (1,512,611)

ACAD EMERG MED

d


d

1265

www.aemj.org

TABLE 4. Unadjusted and Adjusted Odds Ratio (OR) of Hospital Admission among Trauma Patients Characteristics Insurance status (vs. commercial) Uninsured Medicaid Medicare and other government Race and gender (vs. white female) African American female African American male White male Age (vs. 0–14) 65 and over 45–64 25–44 15–24 No. of pre-existing conditions (vs. none) Three or more Two One ISS (vs. mild ISS 1–8) Severe (ISS 15–75) Moderate (ISS 9–14) No. of body regions injured (vs. single) $3 body regions 2 body regions Trauma level status (vs. undesignated) Level I Level II Level III Year of injury (vs. 1996) 2000 1999 1998 1997 Residence (vs. rural) Out of state Urban

Unadjusted OR (99% CI)

Adjusted OR (99% CI)

0.62 (0.61, 0.64) 0.79 (0.77, 0.82) 4.99 (4.90, 5.08)

0.63 (0.62, 0.65) 0.97 (0.94, 1.01) 1.48 (1.42, 1.53)

0.47 (0.46, 0.48) 0.67 (0.66, 0.69) 0.80 (0.79, 0.82)

0.63 (0.61, 0.65) 1.20 (1.17, 1.24) 1.26 (1.23, 1.30)

11.40 3.19 1.78 1.45

(11.06, 11.74) (3.09, 3.29) (1.72, 1.83) (1.40, 1.50)

1.79 1.45 1.36 1.20

(1.70, (1.39, (1.31, (1.15,

1.88) 1.51) 1.41) 1.25)

155.98 (145.74, 166.94) 28.71 (27.79, 29.67) 9.21 (9.01, 9.42)

83.68 (77.63, 90.20) 16.70 (16.02, 17.40) 5.96 (5.78, 6.13)

199.71 (182.38, 218.68) 100.28 (96.69, 104.00)

107.08 (97.17, 118.00) 50.95 (48.90, 53.10)

6.71 (6.49, 6.94) 1.88 (1.84, 1.92)

2.82 (2.67, 2.98) 1.48 (1.43, 1.52)

2.45 (2.40, 2.51) 1.92 (1.87, 1.98) 1.04 (1.02, 1.06)

2.92 (2.83, 3.01) 1.91 (1.83, 1.98) 1.20 (1.17, 1.24)

0.83 0.84 0.90 0.90

0.65 0.71 0.84 0.88

(0.81, (0.82, (0.88, (0.88,

0.86) 0.86) 0.92) 0.92)

0.78 (0.77, 0.80) 0.86 (0.83, 0.89)

(0.63, (0.69, (0.82, (0.85,

0.67) 0.74) 0.87) 0.91)

0.58 (0.57, 0.60) 0.71 (0.67, 0.74)

(N ¼ 1,512,611).

hypertension (19%), diabetes (9%), chronic pulmonary disease (8%), fluid and electrolyte disorders (8%), cardiac arrhythmias (6%), alcohol or drug abuse (6%) and congestive heart failure (5%). The likelihood of inpatient admission also increased significantly with age, despite controlling for other factors. Patients who lived out of state or in an urban area were significantly less likely to be hospitalized (42% and 29%, respectively) compared with state residents from rural areas. Finally, patients who presented to hospitals that were designated as Level 1 trauma centers were almost three times more likely to be admitted compared with hospitals in which the trauma center level was undesignated (see Table 4). Whereas the magnitude of the difference in admission rates by trauma center level varied somewhat by injury severity, the effect was consistent across all injury-severity groups. After controlling for all of the patient, injury, and hospital characteristics listed in Table 4, significant differences in hospitalization rates by type of insurance remained. Overall, trauma patients who were uninsured were 37% less likely to be hospitalized than those with private insurance (OR 0.63; 99% CI ¼ 0.62

to 0.65). Furthermore, this insurance-related difference among the uninsured and those privately insured remained across all levels of injury severity (see Figure 2). Conversely, for patients covered by Medicare or Medicaid, the insurance-related differences in ED disposition were only significant among those who sustained less severe trauma. Patients covered by Medicare (or another government insurance plan) and Medicaid were significantly more likely to be admitted among those moderately injured (OR 1.56, 99% CI ¼ 1.36 to 1.78; OR 1.23, 99% CI ¼ 1.03 to 1.47). Among those mildly injured, patients covered by Medicare or another government insurance were 46% more likely to be hospitalized compared with those privately insured (OR 1.46; 99% CI ¼ 1.41 to 1.52). Hospitalization rates also varied significantly by race and gender among those who sustained a mild injury (see Figure 3). After adjusting for all covariates listed in Table 4, there were no significant differences in admission rates by race and gender among patients who sustained moderate to severe trauma (ISS 9þ). However, among patients who sustained a mild injury (ISS 1–8), African American females were 37%

1266

Selassie et al.

d


Figure 2. Odds of hospital admission by insurance status and trauma severity adjusted for age, place of residence, pre-existing health condition, trauma level status, year of discharge, multiple trauma, race, and gender. ISS ¼ Injury Severity Score.

less likely to be admitted compared with their white female counterparts (OR 0.63, 99% CI ¼ 0.61 to 0.65). Furthermore, among the mildly injured, both African American and white males were 21% and 29% more likely to be admitted compared with white females (OR 1.21, 99% CI ¼ 1.17 to 1.25; OR 1.29, 99% CI ¼ 1.26 to 1.32, respectively) (see Figure 3).

DISCUSSION This study aimed to examine the influence of insurance, race, and gender on a major process of care, namely, the decision to hospitalize an injured patient. Errors in process of care are usually one of omission (i.e., failure to perform a necessary function) or commission (i.e., performance of an unnecessary function). Both types of errors can be associated with access to care. In errors of omission, access to care often is impeded; in errors of commission, access is typically unproblematic or induced.43 In the case of trauma, errors of omission can have devastating consequences for the patient, whereas errors of commission can be costly to society. Whereas differences in a specific process of care may be justified depending on the severity of the illness or injury,

generally speaking, major variations in a process of care by demographic and/or socioeconomic characteristics suggest that health care services are not being equitably distributed. This study found that after controlling for a patient’s clinical condition (i.e., injury severity, pre-existing health, and age), patients who were uninsured were consistently less likely to be admitted, regardless of the severity of the injury. Other studies also have found lower rates of hospitalization among the uninsured; to the best of our knowledge, this is the first study to report a consistent effect across all injuryseverity groups. McCarthy et al. found a similar trend among children who sustained a traumatic brain injury, but the effect was not statistically significant within each injury-severity group. However, the small number of severely injured children in their study sample may have limited their ability to detect an insurance-related difference in hospital admission rates within this patient subgroup.31 Svenson and Spurlock found that among those with a less severe head injury, uninsured patients were significantly less likely to be admitted compared with those who had private insurance. They did not examine the influence of insurance on those with severe head injuries (i.e.,

ACAD EMERG MED

d


d

www.aemj.org

1267

Figure 3. Odds of hospital admission by race, gender, and trauma severity adjusted for age, place of residence, pre-existing health condition, trauma level status, year of discharge, multiple trauma, and insurance status. ISS ¼ Injury Severity Score.

intracranial bleeds and fractures), because the vast majority of them (95% to 98%) were hospitalized.32 The insurance-related difference in ED disposition found in this study between the uninsured and the privately insured is disturbing for two reasons. First, as evidenced by this study population, trauma patients are more likely to be uninsured than persons from the general population. During the study period, the proportion of South Carolina residents who were uninsured ranged from 12% to 17%.44 In contrast, during the same five-year period, 26% of all study patients, on average, were uninsured. Other studies also have found that injury occurs more frequently among the economically disadvantaged and the uninsured.45–48 Although there may be considerable controversy over whether to admit a person who sustains a mild to moderate injury, most clinicians would agree that severe trauma requires hospitalization. In this study, approximately 9% of patients who sustained either at least one severe injury or several moderately severe injuries were not admitted, regardless of insurance status. The uninsured were disproportionately represented in this subgroup compared with those privately insured. It is possible to conclude that insurance-related differences in admission rates among those with a less severe injury may represent

overadmissions among the insured; however, it is difficult to interpret these findings as other than underadmission rates among those who sustained a severe injury. These findings suggest that among those who sustain severe trauma, uninsured patients are less likely than patients with private insurance to receive adequate care. Patients covered by Medicare or another governmental insurance plan were significantly more likely to be hospitalized than those privately injured, among those who sustained mild to moderate trauma. The higher rates of admission among less severely injured Medicare patients compared with those privately insured may occur because physicians recognize that elder patients are physically more fragile and not able to recover as quickly from trauma as younger persons.47 Likewise, among those moderately injured, low-income patients covered by Medical Assistance also were significantly more likely to be admitted compared with those who have private health insurance. Higher admission rates among Medicaid patients could be related to limited access to timely follow-up care. Because Medicaid patients face more barriers to health care compared with those privately insured, ED providers may not feel as comfortable sending home these patients who have sustained moderate trauma.49

1268

Selassie et al.

d


In addition to payer status, race and gender were important determinants of ED disposition, but only among those mildly injured. There were no significant race- and gender-related differences in ED disposition among those with moderate to severe trauma. Numerous other studies have reported significantly higher rates of receiving diagnostic tests and therapeutic interventions for males and whites compared with females and African Americans.13,15–21,23,50 Furthermore, other investigators previously reported the race and gender interaction effect found in this study.51–53 For example, Schulman and colleagues reported that African American females were significantly less likely to be recommended for cardiac catheterization compared with white males, but there were no significant differences for African American males or white females.52 Given the gender and minority disparities noted by previous studies, it is not surprising that African American females were the least likely to be admitted among those mildly injured. Although insurance, race, and gender were significantly associated with ED disposition, the most important predictors of hospital admission were injury severity and preinjury health as measured by the presence of pre-existing conditions and age. The more severe the injury (as reflected by the ISS and number of body regions injured), the higher the likelihood of being hospitalized. In addition, the presence of one or more pre-existing conditions significantly increased the odds of hospital admission. Previous studies have shown that comorbidities are associated with an increased likelihood of in-hospital mortality and poorer functional outcome following trauma.54–56 Age, also a surrogate for health status, was significantly related to inpatient admission. The older the patient, the more likely he or she was to be admitted. Older age is associated with slower healing rates and more frequent complications.47,57 The results of our study also indicate that out-ofstate and urban residents were significantly less likely to be admitted than patients who resided in rural counties. Out-of-state patients may opt to be transferred to their home state rather than be admitted to a hospital in another state. In addition, the portability of some insurance policies across state jurisdictions may be limited, necessitating transfer to an in-state hospital following medical stabilization. It is not clear why urban residents were less likely to be admitted compared with rural residents. One possible explanation for this is that the inpatient bed capacity in relation to patient demand may be scarcer in urban as compared with rural areas. Finally, we also found that patients who presented to a trauma center (Levels 1–3) were significantly more likely to be admitted compared with patients who presented to a hospital that was not designated as a trauma center. This effect was consistent across all levels of injury severity. These data suggest that trauma centers manage patients

more conservatively than hospitals that are not designated trauma centers.

LIMITATIONS While administrative data are readily available, are inexpensive to acquire, and encompass large populations, using them to evaluate quality of care often is severely limited or not possible because they lack the detail needed to evaluate many aspects of care such as the interpersonal quality of care, the technical quality of care, or the appropriateness of the care received. These inherent limitations of administrative data were present in this study. Because the hospital discharge and ED visit data did not include any shortor long-term follow-up information on patients’ health outcomes, it is not clear whether insurance or race- and gender-related differences in ED disposition represent over- or underadmission rates for different patient subgroups. For example, among those who sustained severe trauma, were uninsured patients being underadmitted compared with those with private insurance? Without knowing patients’ recovery trajectory, the appropriateness of the acute-care services provided to different subgroups of this study population cannot be adequately evaluated. In addition, to assess the severity of the patients’ injuries, this study relied on ICDMAP to convert ICD-9CM diagnostic codes into AIS and ISS scores because there was no other clinical information available on the nature or severity of the injury(ies) sustained. ICDMAP has been validated for injuries severe enough to warrant hospital admission; it has not been validated for less severe injuries, such as those treated in the ED.37,58 Finally, the administrative data did not allow us to examine other factors that may influence ED disposition such as patient preference, amount of social support from family, timing of the ED visit (i.e., day vs. night), usual source of care, secondary payer sources, and experience and training of the provider.

CONCLUSIONS The results of this study suggest that ED disposition following a traumatic injury may be influenced by insurance and demographic factors, in addition to injury severity and preinjury health status. Whereas this study relied on administrative data, its limitations should affect all patient subgroups equally and not bias one subgroup more than another. Therefore, the fact that African American females who sustained mild injuries were significantly less likely to be admitted than white females who had similar characteristics is unsettling. If outpatient management of mild injuries is appropriate, then why do not all women, irrespective of race, receive the same treatment? Furthermore, among patients who sustained severe trauma, why were the uninsured significantly

ACAD EMERG MED

d


d

www.aemj.org

less likely to be admitted than those with private insurance? Although it is not possible to determine whether some patient groups were over- or underadmitted from these data, the inequitable distribution of resources that these results suggest warrants further investigation on a prospective basis in which all important clinical and social parameters can be carefully measured and examined. The authors are grateful to Mr. Leroy Frazier, Jr., and Dr. Ernest McCutcheon for their useful contribution in the conduct of the study.

References 1. Baker DW, Shapiro MF, Schur CL. Health insurance and access to care for symptomatic conditions. Arch Intern Med. 2000; 160:1269–74. 2. Committee on Monitoring Access to Personal Health Care Services. Access to Health Care in America (ed 1). Washington, DC: National Academy Press, 1993. 3. Schoen C, Lyons B, Rowland D, Davis K, Puleo E. Insurance matters for low-income adults: results from a five-state survey. Health Aff. 1997; 16:163–71. 4. Ayanian JZ, Weissman JS, Schneider EC, Ginsburg JA, Zaslavsky AM. Unmet health needs of uninsured adults in the United States. JAMA. 2000; 284:2061–9. 5. Canto JG, Rogers WJ, French WJ, Gore JM, Chandra NC, Barron HV. Payer status and the utilization of hospital resources in acute myocardial infarction: a report from the National Registry of Myocardial Infarction 2. Arch Intern Med. 2000; 160:817–23. 6. Faulkner LA, Schauffler HH. The effect of health insurance coverage on the appropriate use of recommended clinical preventive services. Am J Prev Med. 1997; 13:453–8. 7. Haas JS, Goldman L. Acutely injured patients with trauma in Massachusetts: differences in care and mortality, by insurance status. Am J Public Health. 1994; 84:1605–8. 8. Hafner-Eaton C. Physician utilization disparities between the uninsured and insured. Comparisons of the chronically ill, acutely ill, and well nonelderly populations. JAMA. 1993; 269:787–92. 9. Braveman PA, Egerter S, Bennett T, Showstack J. Differences in hospital resource allocation among sick newborns according to insurance coverage. JAMA. 1991; 266:3300–8. 10. Young GJ, Cohen BB. Inequities in hospital care, the Massachusetts experience. Inquiry. 1991; 28:255–62. 11. Hadley J, Steinberg EP, Feder J. Comparison of uninsured and privately insured hospital patients. JAMA. 1991; 265:374–9. 12. Woolhandler S, Himmelstein DU. Reverse targeting of preventive care due to lack of health insurance. JAMA. 1988; 259:2872–4. 13. Kressin NR, Petersen LA. Racial differences in the use of invasive cardiovascular procedures: review of the literature and prescription for future research. Ann Intern Med. 2001; 135:352–66. 14. Weinick RM, Zuvekas SH, Cohen JW. Racial and ethnic differences in access to and use of health care services, 1977 to 1996. Med Care Res Rev. 2000; 57(suppl 1):36–54. 15. Mayberry RM, Mili F, Ofili E. Racial and ethnic differences in access to medical care. Med Care Res Rev. 2000; 57(suppl 1): 108–45. 16. Chandra NC, Ziegelstein RC, Rogers WJ. Observations of the treatment of women in the United States with myocardial infarction. Arch Intern Med. 1998; 158:981–8.

1269

17. Lee AJ, Gelbach S, Hosmer DW, Reti M, Baker CS. Medicare treatment differences for blacks and whites. Med Care. 1997; 35:1173–89. 18. Harris DR, Andrews R, Elixhauser A. Racial and gender differences in use of procedures for black and white hospitalized adults. Ethnicity Dis. 1997; 7:91–105. 19. Giacomini MK. Gender and ethnic differences in hospital-based procedure utilization in California. Arch Intern Med. 1996; 156:1217–24. 20. Ford ES, Cooper RS. Racial/ethnic differences in health care utilization of cardiovascular procedures: a review of the evidence. Health Serv Res. 1995; 30:237–52. 21. Giles WH, Anda R, Casper ML, Escobedo LG, Taylor HA. Race and sex differences in rates of invasive cardiac procedures in US hospitals, data from the national hospital discharge survey. Arch Intern Med. 1995; 155:318–24. 22. McBean AM, Gornick M. Differences by race in the rates of procedures performed in hospitals for Medicare beneficiaries. Health Care Financ Rev. 1994; 15(4):77–90. 23. Ayanian JZ, Udvarhelyi IS, Gatsonis CA, Pashos CL, Epstein AM. Racial differences in the use of revascularization procedures after coronary angiography. JAMA. 1993; 269:2642–6. 24. Shi LY, Lu N. Individual sociodemographic characteristics associated with hospitalization for pediatric ambulatory care sensitive conditions. J Health Care Poor Underserved. 2000; 11:373–84. 25. Braveman P, Oliva G, Egerter S, Reiter R, Egerter S. Adverse outcomes and lack of health insurance among newborns in an eight-county area of California, 1982 to 1986. N Engl J Med. 1989; 321:508–13. 26. Franks P, Clancy CM, Gold MR. Health insurance and mortality, evidence from a national cohort. JAMA. 1993; 270:737–41. 27. Weissman JS, Gatsonis C, Epstein AM. Rates of avoidable hospitalization by insurance status in Massachusetts and Maryland. JAMA. 1992; 268:2388–94. 28. Billings J, Anderson GM, Newman LS. Recent findings on preventable hospitalizations. Health Aff. 1996; 15:239–49. 29. Shi LY, Samuels ME, Pease M, Bailey WP, Corley EH. Patient characteristics associated with hospitalizations for ambulatory care sensitive conditions in South Carolina. South Med J. 1999; 92:989–98. 30. Burstin HR, Lipsitz SR, Brennan TA. Socioeconomic status and risk for substandard medical care. JAMA. 1992; 268:2383–7. 31. McCarthy ML, Serpi T, Kufera JA, Demeter LA, Paidas C. Factors influencing admission among children with a traumatic brain injury. Acad Emerg Med. 2002; 9:684–93. 32. Svenson JE, Spurlock CW. Insurance status and admission to hospital for head injuries: are we part of a two-tiered medical system? Am J Emerg Med. 2001; 19:19–24. 33. Sox CM, Burstin HR, Edwards RA, O’Neil AC, Brennan TA. Hospital admission through the emergency department: does insurance status matter? Am J Med. 1998; 105:506–12. 34. Cornelius LJ. Access to medical care for black Americans with an episode of illness. J Natl Med Assoc. 1991; 83:617–26. 35. Division of Injury and Disability Prevention. Traumatic brain injury surveillance system data evaluation report. Charleston, SC: South Carolina Department of Health and Environmental Control, 1999. 36. Smith GS, Langlois JA, Buechner JS. Methodological issues in using hospital discharge data to determine the incidence of hospitalized injuries. Am J Epidemiol. 1991; 134:1146–58. 37. MacKenzie EJ, Steinwachs DM, Shankar B. Classifying trauma severity based on hospital discharge diagnosis. Med Care. 1989; 27:412–22. 38. Committee on Injury Scaling. The Abbreviated Injury Scale 1990 Revision. Des Plaines, IA: Association for the Advancement of Automotive Medicine. 1990.

1270

Selassie et al.

d


39. Petrucelli E, States JD, Hames LN. The Abbreviated Injury Scale: evolution, usage and future adaptability. Accid Anal Prev. 1981; 13(1):25–35. 40. Baker SP, O’Neill B, Haddon W, Long WB. The Injury Severity Score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma. 1974; 14:187–96. 41. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998; 36(1):8–27. 42. Darlington GA. Encyclopedia of Biostatistics. Chichester, UK: John Wiley & Sons, 1998. 43. Iezzoni LI. Assessing quality using administrative data. Ann Intern Med. 2002; 127:666–74. 44. Mills RJ. Current population reports. Census Bureau, U.S. Department of Commerce. 2001, pp 60–215. 45. Cubbin C, LeClere FB, Smith GS. Socioeconomic status and the occurrence of fatal and nonfatal injury in the United States. Am J Public Health. 2000; 90(1):70–7. 46. Cubbin C, LeClere FB, Smith GS. Socioeconomic status and injury mortality: individual and neighborhood determinants. J Epidemiol Community Health. 2000; 54:517–24. 47. Baker SP, O’Neill B, Ginsburg MJ, Li G. The Injury Fact Book (ed 2). New York: Oxford University Press, 1992. 48. The LEAP Study Group, MacKenzie EJ, Bosse MJ, Kellam JF, et al. Characterization of patients with high-energy lower extremity trauma. J Orthop Trauma. 2000; 14:455–66. 49. Medical Access Study Group. Nowhere to go: Medicaid patient access to primary care. N Engl J Med. 1994; 330:1426–30.

50. Joslyn SA. Racial differences in treatment and survival from early-stage breast carcinoma. Cancer. 2002; 95:1759–66. 51. Schneider EC, Leape LL, Weissman JS, Piana RN, Gatsonis C, Epstein AM. Racial differences in cardiac revascularization rates: does ‘‘overuse’’ explain higher rates among white patients? Ann Intern Med. 2001; 135:328–37. 52. Schulman KA, Berlin JA, Harless W, et al. The effect of race and sex on physicians’ recommendations for cardiac catheterization. N Engl J Med. 1999; 340:618–26. 53. Giles WH, Anda R, Casper ML, Escobedo LG, Taylor HA. Race and sex differences in rates of invasive cardiac procedures in US hospitals, data from the national hospital discharge survey. Arch Intern Med. 1995; 155:318–24. 54. Morris JA, MacKenzie EJ, Edelstein SL. The effect of preexisting conditions on mortality in trauma patients. JAMA. 1990; 263:1942–6. 55. McCarthy ML, MacKenzie EJ, Bosse MJ, Copeland CE, Hash CS, Burgess AR. Functional status following orthopedic trauma in young women. J Trauma. 1995; 39:828–37. 56. McCarthy ML, MacKenzie EJ. Predicting ambulatory function following lower extremity trauma using the Functional Capacity Index. Accid Anal Prev. 2001; 33:821–31. 57. Baker SP, O’Neill B, Haddon W, Long WB. The Injury Severity Score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma. 1974; 14:187–96. 58. Durbin DR, Localio AR, MacKenzie EJ. Validation of the ICD/ AIS MAP for pediatric use. Inj Prev. 2001; 7(2):96–9.