Article, Emergency Medicine

Racial disparities in insurance reimbursement for physician professional services in the ED

a b s t r a c t

Objective: We sought to determine whether racial disparities exist in emergency physician professional services reimbursement from insurance. We hypothesized that insured adult African American emergency department (ED) visits are reimbursed at a lower level than White visits.

Methods: We conducted a retrospective, observational cohort study of insured adult White and African American ED visits (January 1, 2012, to June 30, 2013) to a tertiary center. We downloaded for each included visit age, sex, race, residential zip code, Insurance type, admission status, Current Procedural Terminology (CPT) Evaluation and Management (E/M) code charge reimbursement, and median Household income for residential zip code. We chose as our primary outcome measure visit mean total insurance reimburse- ment/work relative value unit (wRVU). We report racial variation for this outcome measure with 95% confidence intervals (CI) and present the ? coefficient related to African American race within a multivariable regression model.

Results: A total of 50 297 visits met inclusion criteria (35 574 Whites and 14 723 African Americans). Overall, mean total insurance reimbursement/wRVU for White visits was $39.99 (95% CI, 39.80-40.18), for African American visits, $34.15 (95% CI, 33.88-34.42); P b .01. At the CPT E/M code level, African American visit reimbursement was lower than for White visits, ranging from $2.18/wRVU (95% CI, 0.87-3.49) (99282) to

$7.55/wRVU (95 CI, 6.52-8.58) (99285). At the primary insurance level, African American visits showed lower reimbursement than White visits, ranging from $1.70/wRVU (95% CI, 0.75-2.65) in Commercial insurance to

$7.70/wRVU (95% CI, 5.42-9.98) in other insurance. Within the multivariable regression model, the ?

coefficient for African American race was -$1.51/wRVU (95% CI, -1.85 to -1.18); P b .001.

Conclusion: In this single-center study, professional services reimbursement was lower for African American ED visits compared with those of Whites.

(C) 2014

Introduction

Numerous studies have identified disparities in access to care, treatment, and clinical outcomes between African American and White study subjects [1-7]. The Institute of Medicine, in their landmark report Unequal Treatment: Confronting Racial and Ethnic Disparities in Healthcare, concluded that a combination of structural and attitudinal factors, including cultural, financial and policy barriers, contributes to the impediments faced by African Americans in receiving quality health care [8].

? Data from this study will be presented in abstract form at the 2014 Society for Academic Emergency Medicine Annual Meeting, Dallas, TX.

* Corresponding author at: Department of Emergency Medicine, Allegheny Health Network, 320 East North Ave, Pittsburgh, Pennsylvania 15212. Tel.: +1 412 3596 180;

fax: +1 412 359 8874.

E-mail address: [email protected] (A. Venkat).

However, one area that is largely unexplored in the literature on racial disparities in health care is whether they extend to the reimbursement by insurance sources for provider services. Although the Institute of Medicine report alluded to how the structure of insurance programs may be a source of unequal outcomes between minority and non-minority patients [8], there are limited data examining whether insurance reimbursement for health care services may be disparate between African American and white patients. If it is established that racial disparities exist in the reimbursement of physician professional services, this may be both an explanation for variations in health care outcomes between African American and White patients and a target for intervention by insurance payers and policymakers alike.

In many ways, insurance payment for professional services in the emergency department (ED) is an Ideal setting in which to examine whether racial disparities exist in the reimbursement for medical care.

http://dx.doi.org/10.1016/j.ajem.2014.06.029

0735-6757/(C) 2014

Emergency departments are legally mandated in the United States to provide care to the extent of their medical capabilities for any patient who might present regardless of ability to pay [9]. In addition, although individual centers may not have in-network contracts with the insurance source of a patient, the ED is generally viewed as a setting where a patient can expect care regardless of insurance status. We sought to examine whether disparities exist in insurance reimbursement for emergency physician professional services between African American and White patient visits. We hypothesized that insured adult African American ED visits are reimbursed at a lower level than adult White ED visits after controlling for age, sex, visit complexity distribution, primary insurance category, admission status, median

household income, and presence of supplemental insurance.

Methods

Study design, setting, and population

We conducted a retrospective, observational cohort analysis of all insured White and African American adult ED visits to a single center for which professional Current Procedural Terminology (CPT) Evalua- tion and Management (E/M) code charges were submitted between January 1, 2012, and June 30, 2013. This study took place at a level 1 trauma, tertiary referral center with an ED staffed by board-certified emergency physicians, emergency medicine residents, and physician assistants with an annual visit census of 48 000. Patients at this center are predominantly White or African American. Professional evalua- tion and management coding and billing are based on abstraction from the medical record by hospital-employed staff of initial patient chief complaint, written documentation by the treating emergency medicine resident or physician assistant, dictated documentation by the attending emergency physician, and a determination of the complexity of medical decision making based on ordered laboratory and imaging diagnostics and their interpretation as documented by the treating physicians as well as therapies initiated in the ED. Coders use the Lynx system to enter this information, which does not include Patient race, and generate a professional CPT E/M code that is transferred to this center’s billing database. The codes used are those specified by the American Medical Association and Medicare and adopted across payers for emergency physician professional services (99281-99285, 99291-99292) [10].

At this center, billing for professional E/M by emergency

physicians is segregated from facilities charges, regardless of whether the patient is admitted to or discharged from the hospital. Each professional service coding level has a standardized charge that is submitted to the payer source(s) of the patient with unpaid balances forwarded to the patient as guarantor after insurance payment and contracted discounts. The institutional review board of this center approved this study.

Study protocol

From this center’s billing database, we downloaded for each adult visit during the study period with a submitted CPT E/M charge the patient age, sex, race (White, African American, other minority, and unknown race sourced from patient self-identification at triage and entered into the electronic medical record system used in this study), primary and any secondary Insurance types (self-pay, Medicare, Medicaid, commercial, other insurance [eg, Workman’s Compensa- tion, auto], Medicare contracted [eg, Medicare Advantage HMO or PPO], and Medicaid contracted [eg, Medicaid managed care]), residential zip code, ED disposition (admission or discharge), CPT E/ M code, CPT E/M code charge, and CPT E/M code reimbursement (broken down by payment from payer and guarantor sources). Using patient residential zip code data, we downloaded from 2008-2012 US Census Data the Median household income for that location [11].

Reimbursement data were downloaded in December 2013 to ensure that the billing cycle for included visits would be complete.

Patient visits with no listed insurance (self-pay) were excluded from analysis. We also excluded non-African American and non-White patient visits given their small percentage in this center’s population, making it difficult to draw conclusions on reimbursement for their care. We finally excludED patient visits where the race was unknown after evaluating that the primary outcome measure below did not significantly change after the removal of these visits from the final analyzed cohort.

Outcome measure

We defined our primary outcome measure, classified by primary insurance category, for each visit as:

Mean total reimbursement for CPT E/M code minus guarantor payment (hereafter referred to as mean total insurance reimburse- ment)/work relative value units (wRVUs) generated.

This outcome measure was chosen to control for the confounders of varying CPT E/M levels between the race groups and differing guarantor payment of deductibles, copayments, and unreimbursed professional service charges.

For each visit, the Medicare-assigned wRVU value for the relevant CPT E/M code was used, as this is both a widely accepted marker for encounter complexity and physician effort that is accepted by other payer sources and is used to determine visit charges and compensa- tion for emergency physicians at this and other centers [10]. We excluded guarantor payments to focus on whether disparities exist specifically in insurance reimbursement without consideration of what contribution a patient individually might have made to compensation for professional services.

We did not evaluate reimbursement for facilities charges, as there is no factor comparable to wRVU that would allow control of physician effort or visit complexity. In addition, facility charges can vary considerably from visit to visit based on diagnostic testing, treatments initiated, and decision on patient admission.

Data analysis

Fig. 1 shows the analytic strategy adopted to determine if African American race was associated with reduced emergency physician professional service reimbursement after controlling for age, sex, CPT E/M code distribution, admission status, primary insurance category, presence of secondary insurance, and median household income. The first step was to evaluate whether the primary outcome measure differed significantly between White and African American visits using the Z test. Once that was established, it was necessary to separately evaluate whether African American race was consistently associated with reduced reimbursement using the primary outcome variable at both the ED presentation level (CPT E/M code pathway) and at the insurance source level (primary insurance category pathway). This was done to address the variability that might be present at the time of ED presentation either through differential physician evaluation and CPT E/M coding or through the differing individual insurance contract provisions that patients might have.

Using the Z test with 95% confidence intervals (CIs), we compared

the outcome measure between White and African American visits at the CPT code level. Similarly, we compared mean total insurance reimbursement/wRVU across primary insurance categories between White and African American patient visits, again using the Z test. Once established that these analyses still showed reduced reimbursement in African American visits, we used multivariable regression in separate models to establish the ? coefficient for the marginal effect of patient race to the reimbursement outcome measure controlling for patient age; sex; median household income; ED disposition; and, separately, CPT E/M code distribution or primary insurance category and presence of secondary insurance. Finally, after establishing that

Comparison of Primary Outcome Measure between White and African- American Visits within CPT E/M Level Categories (Table 2)

Multivariable Regression Model of Age, Sex, Race, Admission Status, Median Household Income and CPT E/M Code Distribution Association with Primary

Outcome Variable (Table 4)

Comparison of Primary Outcome Measure between White and African- American Visits within Primary Insurance Categories (Table 3)

Multivariable Regression Model of Age, Sex, Race, Admission Status, Median Household Income, Primary Insurance Category and Presence of Secondary

Insurance Association with Primary Outcome Variable (Table 5)

Multivariable Regression Model of Age, Sex, Race, Admission Status, Median Household Income, CPT E/M Code

Distribution, Primary Insurance Category and Presence of Secondary Insurance Association with Primary Outcome Variable (Table 6)

Analysis of Distribution of Secondary Insurance Within Primary Insurance Categories as Potential Marker of Uncaptured Socioeconomic Variables as Possible Explanations for Racial Variation in Primary Outcome Variable (Table 7)

Fig. 1. Hierarchical analytic strategy used to evaluate whether Racial disparity exists in primary outcome variable.

the ? coefficient for African American race still showed a negative marginal effect in these separate models, we performed a summary multivariable regression model that combined all of the above independent variables to establish the marginal effect of African American race on reduced reimbursement for emergency physician professional services in the study cohort.

Gross Comparison of Primary Outcome Measure between White and African- American Visits

After this last multivariable regression model established a negative marginal reimbursement effect associated with African American race, we wished to evaluate possible explanations for racial disparities in reimbursement for emergency physician professional services. Toward that end, we examined, using the Pearson ?2 test, whether the distribution of secondary insurance sources varied between White and African American patients in the study cohort. This last analysis was performed, as it was anticipated that secondary insurance variability might serve as a marker for uncaptured socioeconomic variables that might explain the disparities elucidated (see Discussion). All data are presented as analyzed at the patient visit level as, for a given encounter, CPT E/M code, ED disposition, and insurance source might vary for the same patient seen at different times. Data were downloaded directly from the billing database of this center into a Microsoft Excel Spreadsheet and analyzed using SPSS version 21.0.

Results

Characteristics of study subjects

Table 1 and the CONSORT Figure (Fig. 2) show the descriptive characteristics of all adult ED visits during the study period for which

charges were submitted to either insurance or guarantor sources for

CPT E/M codes for emergency physician professional services. Overall,

61 216 visits met this initial screening criteria. The mean total insurance reimbursement per wRVU in the entire initial screening cohort was $33.18 (Table 1). After excluding subject visits in which there was no insurance source (14.8%) or with a race category of non- African American/non-White or unknown (1.9% and 1.8%, respective- ly), 50 297 visits remained for analysis of the study hypothesis (Tables 2-7). In the cohort, there were 550 different privately administered insurance plans, 49 traditional Medicare plans (eg, Medicare Part B, Federal Black Lung, etc) and 22 Medicaid plans (eg, individual state Medicaid plans, prison inpatient, spina bifida, etc). The 550 privately administered plans were categorized as 242 plans within the commercial category (5 major insurance companies predominantly represented), 232 plans within the other category (74 Workman’s Compensation, 10 auto insurance, the remainder from life insurance, home health plans, and other types of insurance with health coverage), 45 plans within the Medicare contracted category (4 major insurance companies majority represented), and 31 plans within the Medicaid contracted category (3 major insurance compa- nies represented).

Main results

For the 35 574 White visits, the mean total insurance reimburse- ment/wRVU was $39.99 (95% CI, 39.80-40.2) in comparison to $34.15/ wRVU (95% CI, 33.88-34.42) for the 14 723 African American visits; P b .001 by the Z test. Table 2 shows that at the CPT level, with the

Table 1 Characteristics of adult ED visits with CPT code E/M charges submitted during the study period (January 1, 2012 to June 30, 2013) (n = 61 216)

Independent variable Value

Age in years (mean, interquartile range) 50.4, 33.4-63.9 Sex (%)

Male 47.7

Female 52.3

Race (%)

White 65.6

African American 30.7

Non-African American, non-White 1.9

Unknown 1.8

Admission to the hospital (%)

Yes 24.2

No 75.8

CPT code for evaluation and management (%)

99281 0.5

99282 6.4

99283 41.3

99284 38.5

99285

10.1

(Tables 4 and 5). Within the summary multivariable regression model

99291

3.1

(Table 6), the ? coefficient for African American race, controlling for age,

99292

0.1

sex, median household income, admission status, primary insurance

exception of level 99281 and 99292 (both of which had relatively few cases), African American visits had a lower mean total insurance reimbursement/wRVU in comparison to White visits, ranging from

$2.18/wRVU (95% CI, 0.87-3.49) (99282) to $7.55/wRVU (95% CI,

6.52-8.58) (99285). Table 3 shows the results of within primary insurance category reimbursement between White and African American subjects. In each primary insurance category except Medicaid contracted, the outcome of visit mean total insurance reimbursement divided by wRVU was significantly lower for African American visits. In terms of absolute difference, the range varied based on primary insurance category from $1.70/wRVU (95% CI, 0.75-2.65) in commercial insurance to $7.70/wRVU (95% CI, 5.42-9.98) in other insurance. Table 3 also shows the expected increase in professional service reimbursement if insurance payment of African American visits were raised to that of White visits in this cohort. This disparity in the primary study outcome was maintained after controlling for age; sex; median household income; ED disposition; and, separately, CPT E/M code distribution or primary insurance category and the presence of secondary insurance

Primary insurance status (%)

Self-pay 14.8

Medicare 12.6

Medicaid 3.4

Commercial 21.5

Other (eg, Workman’s Compensation, auto) 6.7

Medicare contracted (eg, Medicare Advantage HMO or PPO) 17.8

Medicaid contracted (eg, Medicaid managed care) 23.2

Visit mean total insurance reimbursement ($)/wRVU (Mean, SD) $33.18, $21.27 Median household income based on US census

data for residential zip code

category, presence of secondary insurance, and CPT E/M code distribution was -$1.51/wRVU (95% CI, -1.85 to -1.18), P b .001.

The analysis in Table 7 shows that the distribution of secondary payer sources between White and African American visits differed significantly within each primary insurance category. In the primary insurance categories with the highest disparity in reimbursement as shown in Table 3 (Medicare, other insurance, and Medicare con- tracted), there was an identifiable pattern of significant variation between African American and White subjects. In Medicare, White

Overall cohort White

African American

$44 440

$47 530

$37 320

subjects more often had commercial secondary (Medigap) payer sources, whereas African American visits had Medicaid secondary

Other minority group

$48 620

payer sources. In other insurance, White visits more likely had a

Unknown race

$48 670

secondary payer source, which was most often commercial. In Medicare

61,216 Adult Emergency Department Visits During the Study Period With Submitted Charges for CPT E/M

Codes

14.8% of Visits by Patients with No Insurance (Self Pay)

1.9% of Visits by Patients who are Non-African-American, Non-White

1.8% of Visits by Patients with Unknown Race

50,297 Visits by Insured African- American and White Patients During the Study Period with Submitted Charges for CPT E/M Codes

Fig. 2. CONSORT diagram of adult ED visits during study period with submitted charges for CPT E/M codes (January 1, 2012 to June 30, 2013).

Table 2

Visit mean total insurance reimbursement per wRVU shown by CPT E/M code level (total insurance reimbursement is from combination of primary and any secondary insurance sources)

CPT E/M code level

Patient race

No. of visits

Mean total insurance reimbursement/wRVU

95% CI

P value for comparison

99281

White

127

$41.46

37.20, 45.80

.32

African American

69

$44.49

40.30, 48.70

99282

White

1579

$38.46

37.50, 39.40

.001

African American

1268

$36.28

35.40, 37.20

99283

White

12877

$41.31

41.00, 41.70

b.001

African American

6714

$35.63

35.20, 36.10

99284

White

14904

$38.83

38.60, 39.10

b.001

African American

5195

$31.71

31.30, 32.10

99285

White

4576

$38.90

38.50, 39.30

b.001

African American

1177

$31.35

30.40, 32.30

99291

White

1472

$44.97

44.20, 45.80

.03

African American

294

$42.95

41.40, 44.50

99292

White

39

$44.33

38.80, 49.90

.46

African American

6

$38.21

18.20, 58.20

Table 3

Visit mean total insurance reimbursement per wRVU shown by primary insurance category (total insurance reimbursement is from combination of primary and any secondary insurance sources)

Primary insurance with subject race

No. of cases in category

Visit mean total insurance reimbursement ($)/wRVU

95% CI

P value for comparison

Expected revenue increase if African American reimbursement/wRVU equals White visits ($)

95% CI

Medicare White

6034

$36.10

35.80, 36.40

b.001

$5705.70

4599, 6813

African American Medicaid

White

1463

1249

$32.20

$16.60

31.50, 32.80

15.90, 17.30

b.001

$1674.40

793, 2556

African American

Commercial

728

$14.30

13.30, 15.20

$4227.90

1858, 6598

White

10104

$48.70

48.30, 49.10

b.001

African American Other insurance

White

2487

3138

$47.00

$42.30

46.10, 47.80

41.40, 43.30

b.001

$4943.40

3426, 6460

African American

Medicare contracted White

642

8185

$34.60

$42.60

32.50, 36.80

42.40, 42.80

b.001

$9279.60

7999, 10 560

African American

2442

$38.80

38.30, 39.30

Medicaid contracted

White

6864

$30.70

30.40, 30.90

.14

$2088.30

-666, 4842

African American

6961

$30.40

30.10, 30.70

Table 4

Multivariable regression model of relative contribution of independent variables to disparities in visit mean total insurance reimbursement for CPT E/M code per wRVU–CPT E/M code pathway (model R2 = 0.06)

Independent variable

Reference variable

? Coefficient (relative increase or decrease in visit mean total insurance reimbursement ($)/wRVU for independent variable)

95% CI for ?

coefficient

t-statistic for independent variable

P value for independent variable

Age Sex

Female

Male

0.11

0.38

0.10, 0.12

0.07, 0.69

25.18

2.43

b.001

.02

Income

Median household income ($’000)

0.12

0.11, 0.13

20.88

b.001

CPT E/M code

99281

99283

3.88

1.44, 6.32

3.12

.002

99282

-0.60

-1.28, 0.09

-1.72

.09

99284

-3.75

-4.11, -3.40

-20.79

b.001

99285

-5.12

-5.68, -4.56

-17.94

b.001

99291

1.65

0.74, 2.56

3.56

b.001

99292

Race

0.46

-4.63, 5.55

0.18

.86

African American

White

-3.86

-4.22, -3.49

-20.77

b.001

ED disposition Admitted

Discharged

0.78

0.37, 1.19

3.71

b.001

Table 5

Multivariable regression model of relative contribution of independent variables to disparities in visit mean total insurance reimbursement for CPT E/M Code per wRVU–primary insurance pathway (model R2 = 0.22)

Independent variable

Reference variable

? Coefficient (relative increase or decrease in visit mean total insurance reimbursement ($)/wRVU for independent variable)

95% CI for ?

coefficient

P value for independent variable

Age Sex

0.09

0.08, 0.10

b.001

Female Income

Median household income ($’000)

Male

0.44

0.04

0.16, 0.72

0.03, 0.05

b.001

b.001

Primary insurance type

Medicare

Commercial

-15.41

-16.00, -14.90

b.001

Medicaid Other

Medicare Contracted

-31.87

-7.54

-8.87

-32.60, -31.10

-8.10, -7.00

-9.30, -8.40

b.001 b.001

b.001

Medicaid Contracted

Presence of secondary insurance Yes

No

-16.38

0.53

-16.80, -16.00

0.15, 0.92

b.001

.01

Race

African American

White

-1.41

-1.74, -1.07

b.001

ED disposition Admitted

Discharged

0.69

0.35, 1.03

b.001

contracted, African American visits more likely had a secondary payer source, which was most often Medicaid. These variations may suggest the presence of uncaptured socioeconomic variables that might explain the disparities elucidated (see Discussion).

Discussion

The data presented in this study show that insurance reimburse- ment for professional physician services is lower in African American ED patient visits than in Whites after controlling for age, sex, median household income, ED disposition, primary insurance category, variation in guarantor payment (through exclusion), the presence of secondary insurance, CPT E/M code distribution, and a quantifiable measure of physician effort and visit complexity (wRVU). The reduced reimbursement extended to all identified primary insurance catego- ries, except Medicaid contracted, and those CPT E/M code levels with

the highest prevalence in the study cohort, although varied in their degree. To our knowledge, this is a novel finding and shows that disparities in health care between African American and White subjects extend beyond medical treatment and clinical outcomes to financial compensation for physician services from insurance sources. If the results we present are validated across other centers and, potentially, physician specialties, there may be broad implications for the health care system, insurance providers, and policy makers. First, our data suggest that beyond issues of insurance source, there may be a further until now unidentified Financial impact in centers that care for a disproportionate number of African American patients. As shown in Table 3, expected professional physician services reimbursement would overall be higher if insured African American visits were compensated at the same level as White visits. This decreased reimbursement, if replicated across centers and, potentially, other specialties, may contribute to the impediments discussed in the

Table 6

Multivariable regression model of relative contribution of independent variables to disparities in visit mean total insurance reimbursement for CPT E/M code per wRVU–summary (model R2 = 0.24)

Independent variable

Reference variable

? Coefficient (relative increase or decrease in visit mean total insurance reimbursement ($)/wRVU for independent variable)

95% CI for ?

coefficient

t-statistic for independent variable

P value for independent variable

Age Sex

Female

Male

0.10

0.56

0.09, 0.11

0.29, 0.84

20.66

3.99

b.001

b.001

Income

Median household income ($’000)

0.04

0.03, 0.05

8.32

b.001

Primary insurance type

Medicare

Commercial

-15.61

-16.16, -15.06

-55.57

b.001

Medicaid Other

Medicare contracted Medicaid contracted

Presence of secondary insurance

Yes

No

-31.93

-8.02

-9.02

-16.74

0.60

-32.68, -31.18

-8.60, -7.44

-9.49, -8.55

-17.13, -16.34

0.23, 0.98

-83.63

-27.16

-37.59

-82.19

3.13

b.001 b.001 b.001 b.001

.002

Race

African American

White

-1.51

-1.85, -1.18

-8.92

b.001

ED disposition

Admitted

Discharged

1.87

1.50, 2.24

9.84

b.001

CPT E/M code

99281

99283

4.98

2.78, 7.17

4.44

b.001

99282

-0.12

-0.74, 0.50

-0.39

.70

99284

-4.25

-4.57, -3.93

-26.11

b.001

99285

-5.61

-6.11, -5.10

-21.79

b.001

99291

1.43

0.61, 2.25

3.43

.01

99292

0.58

-4.00, 5.15

0.25

.81

Table 7

Distribution of secondary payor sources for primary insurance and race categories

Primary insurance category

Secondary insurance category

white race (percentage of subjects with type of secondary insurance)

African American race (percentage of subjects

with type of secondary insurance)

Pearson ?2, P value for comparison

of distribution of secondary insurance between race categories

Medicare

None

24.5%

32.7%

409.528, P b .001

Medicare

6.0%

4.4%

Medicaid

25.4%

44.6%

Commercial

42.1%

16.1%

Other

1.2%

1.2%

Medicare contracted

0.4%

0.1%

Medicaid contracted

0.4%

0.9%

Medicaid

None

94.0%

91.1%

23.752, P = .001

Medicare

1.4%

0.7%

Medicaid

0.2%

0.3%

Commercial

0.7%

1.1%

Other

1.0%

0.4%

Medicare contracted

0.0%

0.5%

Medicaid contracted

2.7%

5.9%

Commercial

None

90.7%

90.0%

114.388, P b .001

Medicare

4.5%

2.0%

Medicaid

0.8%

2.0%

Commercial

1.8%

1.6%

Other

0.4%

0.2%

Medicare contracted

0.3%

0.2%

Medicaid contracted

1.6%

4.0%

Other

None

62.9%

74.0%

153.162, P b .001

Medicare

4.9%

1.7%

Medicaid

2.6%

2.6%

Commercial

20.9%

8.7%

Other

1.0%

1.9%

Medicare contracted

5.3%

1.9%

Medicaid contracted

2.3%

9.2%

Medicare contracted

None

75.3%

46.4%

760.256, P b .001

Medicare

0.9%

1.8%

Medicaid

23.1%

51.2%

Commercial

0.5%

0.2%

Other

0.1%

0.1%

Medicare contracted

0.0%

0.1%

Medicaid contracted

0.1%

0.3%

Medicaid contracted

None

93.8%

93.8%

7.908, P = .245

Medicare

0.2%

0.1%

Medicaid

5.3%

5.4%

Commercial

0.3%

0.3%

Other

0.1%

0.0%

Medicare contracted

0.0%

0.0%

Medicaid contracted

0.3%

0.4%

Bold face indicates P b .05 for comparison between race categories for particular secondary insurance type.

Institute of Medicine report that show that access to and receipt of quality health care is more difficult for African Americans [8].

Second, with the advent of the Affordable Care Act and the rapid change in the types of insurance products that are marketed to potential subscribers (employers and individuals), our data suggest that there needs to be an evaluation as to how the structure of insurance products and programs can aid in the reduction of financial disparities in the care of African American patients. Our results do not show that there is an intentional discriminatory effect in the types of insurance marketed or available to African Americans. In fact, we purposefully have not compared exact insurance-secondary insur- ance combinations across racial categories, as we do not believe that our findings are explained by explicit discriminatory intent among insurance payers. Rather, the study results raise the question as to whether there is an insidious effect in terms of how the provisions of insurance products may be leading to the racial disparities identified in this investigation. Some possibilities include how health care insurance products are made available to individuals and employers and how their cost and coverage structures may lead individuals and employers to choose particular coverage options. In the case of other insurance (eg, auto or Workman’s Compensation), there may be additional factors such as cost to the individual, product availability based on employer, and coverage limits for health-related expenses

that may contribute to the disparities seen in this study. This stands in contrast to Medicaid contracted insurance, which at this site is represented by predominantly 3 insurance companies whose terms are closely mandated by the state Department of Public Welfare. As such, it is likely that the above uncaptured socioeconomic factors cannot manifest as with other insurance categories, leading to our finding that insurance reimbursement in this category is largely equivalent between White and African American visits.

Finally, for policymakers, the data in this study, if replicated, would suggest that their decisions on the structure and regulation of both public and commercial insurance may have an impact in reducing racial disparities in financial reimbursement for physician professional services. In the case of Medicare, which is not comprehensive insurance, the need for Medigap or other secondary insurance source may be unintentionally contributing to racial disparities in the finances of health care reimbursement. For commercial insurance, there may be a need for regulatory scrutiny in collaboration with insurance companies to ensure that reimbursement contracts with providers take into account the potential for disparities based on patient race.

It is worth considering whether the findings elucidated in this study more indicate a difference in insurance provisions, with race as a surrogate marker for this underlying variation, rather than race as the primary associated independent variable with the chosen outcome

measure. We would contend that a strength of our study is that we have attempted to evaluate through a variety of perspectives (CPT E/M code, primary insurance category, and multivariable regression model of captured independent variables separately and in summary) to isolate African American race as the most clearly associated indepen- dent variable with reduced reimbursement compared to White race. In addition, it is perhaps irrelevant if race is a surrogate marker for poor insurance provisions and subsequent reimbursement or vice versa as the ultimate outcome is the same–visits by African American patients appear to be reimbursed at a lower level with implications as described above for EDs, insurance payers, and policymakers. Given the previously cited worse clinical outcomes and access to Health care resources among African Americans in a variety of health care settings, insurance coverage that results in reduced reimbursement in this population deserves heightened scrutiny as a further potential hindrance to quality care in this underserved minority.

Assuming that the results presented here are validated, the question arises as to what might explain the disparities here elucidated, given that controlling for age, sex, primary insurance category, presence of secondary insurance, CPT E/M code distribution, median household income, and ED disposition did not remove this potential for racial differences in payer reimbursement for emergency physician professional services. Here, the data in Table 7 may be informative. In the case of Medicare subjects in our study, it is clear that most patients viewed a necessity of having secondary payer sources for comprehensive coverage. White patients were more likely to have commercial Medigap secondary payer sources, whereas African American patients were more likely to rely on Medicaid. As shown in Tables 3 to 6 as well as what is generally accepted in the realm of health care economics [12], it is clear that commercial insurance reimburses better than Public insurance. A similar association is seen in the other insurance category; White patients who were initially billed to payer sources such as auto or Workman’s Compensation insurance were more likely to have secondary commercial health care insurance in comparison to African Americans. Finally, for Medicare-contracted insurance (eg, Medicare Advantage), where secondary payer sources are generally illegal except for Medicaid [13], African American patients were far more likely to have secondary insurance in this form. Because Medicaid is a marker of lower income, it may be that the types of Medicare-contracted plans that are marketed or affordable to African Americans are less comprehensive and, therefore, require higher guarantor payment than those of Whites. Overall, types of primary insurance that were most associated with subscription of secondary insurance, a marker of coverage fragmentation, showed the highest levels of payer source payment disparity (Table 3) between African American and White emergency visits. This may suggest that regulation that limits insurance fragmentation has the potential to reduce the racial disparities described in this investigation.

Limitations

Our data are limited in its generalizability at this point due to its derivation from a single center and in our single medical discipline of emergency medicine where we have direct access to relevant data. This is perhaps inevitable given the common charge and reimbursement opacity that exists in the US health care system. As has been widely reported in the lay press [14], how charges are determined and what reimbursement is expected does not have the same transparency that exists in other industries, and there is little consistency in how final determinations are made in insurance reimbursement for physician services. Similarly, in the Institute of Medicine report, citations on the impact of insurance differences between African Americans and Whites focused more on the potential impact on access to care and urging the avoiding of payment source fragmentation; specific data on racial disparities in reimbursement for health care services were largely absent [8,15-17]. Our intention in publishing these results that show that racial disparities in health care extend to areas of physician reimbursement is to urge other centers to

evaluate their data in a similar manner and, potentially, across medical and surgical disciplines to validate or dispute our findings.

We are additionally limited by the largely dichotomized nature of our patient population (White vs African American). Clearly, in a diverse demographic nation such as the United States, additional data from centers with other predominant minority patient populations would add to the evaluation of whether non-White ED visits as a whole show decreased reimbursement for professional physician services.

Finally, 1.8% of visits in the original screening sample showed a missing value for patient race. We have attempted to address this by evaluating whether the primary outcome measure in the analyzed cohort of known White and African American visits would significantly differ with the exclusion of these visits and found that not to be the case. However, we cannot definitively rule out that knowledge of the actual patient race in these visits might marginally affect the results presented.

Conclusion

In this single-center study, insured African American patient visits to the ED resulted in less mean total insurance reimbursement per wRVU for physician professional services than White patient visits, controlling for age, sex, median household income, admission status, primary insurance category, presence of secondary insurance, and CPT E/M code distribution. These results should be validated in other centers and, potentially, disciplines to help guide policy interventions designed to address this potential contributor to racial disparities in health care.

Acknowledgment

We thank Professor Carrie Leana, Katz Graduate School of Business, University of Pittsburgh, for her insightful comments on an earlier draft of the paper.

References

  1. Barocas D, Alvarez J, Koyama T, Anderson C, Gray D, Fowke J, et al. Racial variation in the quality of surgical care for bladder cancer. Cancer 2013 [Epub Dec 11, 2013].
  2. Venkat A, Hoekstra J, Lindsell C, Prall D, Hollander J, Pollack Jr C, et al. The impact of race on the acute management of chest pain. Acad Emerg Med 2003;10(11):1199-208.
  3. Burgess D, Nelson D, Gravely A, Bair M, Kerns R, Higgins D, et al. Racial differences in prescription of opioid analgesics fro chronic non-cancer pain in a national sample of veterans. J Pain 2014;15(4):447-55.
  4. Cassie K, Cassie W. Racial disparities in the use of physical restraints in US nursing homes. Health Soc Work 2013;38(4):207-13.
  5. Lewis P, Fagnano M, Koehler A, Halterman J. Racial disparities at point of care for urban children with persistent asthma. J Community Health 2014;39(4):706-11.
  6. Samnaliev M, McGovern M, Clark R. Racial/ethnic disparities in mental health treatment in six Medicaid programs. J Health Care Poor Underserved 2009;20(1):165-76.
  7. Sineshaw H, Robbins A, Jemal A. Disparities in survival improvement for metastatic Colorectal cancer by race/ethnicity and age in the United States. Cancer Causes Control 2014;25(4):419-23.
  8. Institute of Medicine. Unequal treatment: confronting racial and ethnic disparities in health care; 2003 [Washington, DC].
  9. Center for Medicare and Medicaid Services. Emergency medical treatment & Labor Act; 2014 [Available from: http://www.cms.gov/Regulations-and-Guid- ance/Legislation/EMTALA/index.html?redirect=/EMTALA/].
  10. Center for Medicare and Medicaid Services. Physician fee schedule; 2012 and 2013 [[cited 2014 February 12]; Available from: http://www.cms.gov/apps/physician- fee-schedule/overview.aspx].
  11. US Census Bureau. 2008-2012 American Community Survey 5-Year Estimates; 2013 [Washington, DC].
  12. Fox W, Pickering J. Hospital & physician cost shift: payment level comparison of Medicare, Medicaid and commercial payers. Milliman; 2008.
  13. Center for Medicare and Medicaid Services. Medicare advantage plans; 2014 [[cited 2014 February 12]; Available from: http://www.medicare.gov/sign-up- change-plans/medicare-health-plans/medicare-advantage-plans/how-medicare- advantage-plans-work.html].
  14. Brill S. Bitter pill: why medical bills are killing us. Time 2013;181(8):16-55.
  15. Bloche M. Race and discretion in American medicine. Yale J Health Policy Law Ethics 2001;1:95-131.
  16. Institute of Medicine. Crossing the quality chasm. Washington, DC: National Academy of Sciences; 2001.
  17. Tai-Seale M, Freund D, LoSasso A. Racial disparities in service use among Medicaid beneficiaries after mandatory enrollment in managed care: a difference-in- differences approach. Inquiry 2001;38(1):49-59.

Leave a Reply

Your email address will not be published. Required fields are marked *