Racial disparities in the use of and indications for colorectal procedures in Medicare beneficiaries
Abstract
BACKGROUND
African Americans are diagnosed more frequently with colorectal carcinoma at a later stage compared with Caucasians. One potential reason for the disparity is a lower rate of screening examinations.
METHODS
Using Outpatient and Physician-Supplier claims for all Medicare beneficiaries age ≥ 65 years in 1999, indications for fecal occult blood testing (FOBT), sigmoidoscopy, colonoscopy, and barium enema were divided into diagnostic, surveillance, or screening categories. Annualized rates were calculated based on the number of eligible fee-for-service months.
RESULTS
Rates of FOBT (18.24% vs. 11.86%; P < 0.001) and sigmoidoscopy (3.07% vs. 2.17%; P < 0.001) were higher in Caucasians compared with African Americans, whereas rates of barium enema were higher in African Americans (2.26% vs. 1.88%; P < 0.001). Colonoscopy use was more frequent among men only in Caucasians compared with African-Americans (8.00% vs. 6.97%; P < 0.001). For FOBT, sigmoidoscopy, and colonoscopy, the racial differences in procedures performed for diagnostic purposes were of smaller magnitude than for screening; and, for colonoscopy, the use of diagnostic procedures actually was higher for African Americans.
CONCLUSIONS
Racial disparities exist not only in the use of colorectal procedures but also in the indications for such testing, with African Americans less likely to undergo screening tests. The differences are consistent with delay in ;100:418–24. © 2003 American Cancer Society.
Colorectal screening with fecal occult blood testing (FOBT), sigmoidoscopy, colonoscopy, and/or barium enema currently is recommended by all relevant professional societies.1-4 In addition, most guidelines recommend surveillance or the use of routine examinations for the follow-up of patients with a past history of neoplastic polyps or resected malignancy or to screen for dysplasia in patients with inflammatory bowel disease.1-4. However, data from population-based surveys5 suggest that the current use of screening is below targets (unpublished data),6 with screening rates somewhat lower among African Americans compared with other racial groups.7 A lower use of screening and surveillance is one potential reason for the higher colorectal carcinoma incidence and mortality observed in African Americans compared with Caucasians.7, 8 However, because previous studies have not considered screening and diagnostic procedures separately and/or have not used national-level data, this hypothesis has not been evaluated rigorously.
Therefore, we conducted the current study to investigate the use of FOBT, sigmoidoscopy, colonoscopy, and barium enema in a large cohort of Medicare beneficiaries. In addition to studying the overall procedure use, potential differences among subgroups in the indication for testing (i.e., diagnostic vs. screening) also were determined. To ascertain whether any observed disparities may be due to differences in access to endoscopic or radiographic facilities, we also included two other procedures (upper gastrointestinal endoscopy and upper gastrointestinal series) in our analyses.
MATERIALS AND METHODS
Data Sources and Procedures
All data were obtained from the Centers for Medicare and Medicaid Services, and the protocol was approved by the Institutional Review Board at the University Hospitals of Cleveland. The data were obtained from the 1999 Medicare outpatient files and Physician-Supplier or National Claims History (NCH) files. The outpatient files include claims for all outpatient hospital encounters in the Medicare population, whereas the NCH files include physician and supplier claims for Medicare beneficiaries. Data elements in both files include demographic characteristics (age, gender, race), up to 10 diagnosis codes according to the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM), 1 procedure code according to the ICD-9-CM or the Current Procedural Terminology, 4th Edition (CPT-4), and billing data. A unique identifier known as the beneficiary identification code allows longitudinal tracking of patients as well as linkage with other files. In January 1998, Medicare reimbursement for screening FOBT, sigmoidoscopy, and barium enema, as well as screening colonoscopy in patients at increased risk for colorectal carcinoma, was instituted.9 Previously, procedures were reimbursed only for evaluation of signs or symptoms.
The study was conducted at the procedure level. All claims for four different colorectal procedures and for two reference procedures (one endoscopic and one radiographic) were identified by the following procedure codes: 1) FOBT (NCH files only) was identified by CPT-4 codes 82270, 82273, and G0107 (these codes typically reflect home collection of specimens with subsequent submission to a clinical laboratory); 2) flexible sigmoidoscopy was identified by CPT-4 codes 45330, 45331, 45333, 45338, 45339, and G0104 and by ICD-9-CM codes 45.22, 45.24, 48.22, and 48.24; 3) colonoscopy was identified by CPT-4 codes 44388, 44389, 45378, 45380, 45382, 45383, 45384, 45385, and G0105 and by ICD-9-CM codes 45.23, 45.25, 45.41, 45.42, 45.43, and 48.36; 4) barium enema was identified by CPT-4 codes 74270, 74280, G0106, and G0120 and by ICD-9-CM code 87.64; 5) upper endoscopy was identified by CPT-4 codes 43234, 43235, 43239, and 43255 and by ICD-9-CM codes 45.10–45.19; and 6) upper gastrointestinal series was identified by CPT-4 codes 74240, 74241, 74245, 74246, 74247, 74249, 74250, and 74251 and by ICD-9-CM codes 87.61, 87.62, and 87.63.
The Medicare denominator file, which contains beneficiary-level enrollment data for all Medicare beneficiaries in a given year, also was used. The study was limited to individuals age ≥ 65 years who were receiving Medicare benefits through Part A and Part B. Monthly enrollment indicators were used to calculate the total number of months for which an individual was enrolled in Medicare as well as the total number of months during which a beneficiary participated in a managed care program. Fee-for-service (FFS) months, which served as the denominator in this study, were obtained by subtracting managed care months from total months. The utilization of services through managed care programs is not represented in claims files, and this strategy made it possible to combine the utilization data for individuals receiving their care through the FFS system on a partial-year basis with utilization data for individuals receiving their care exclusively through the FFS system. In addition, it enabled us to account for the utilization data of individuals who participated in Medicare for only part of the year, either because of late enrollment in the program in a given calendar year or because of change in vital status during the study period. Individuals were identified as Caucasian, African American, or all other. Because the study focused on comparing the use of services between African Americans and Caucasians, the stratified analyses encompassed only these two racial groups.
Measures
The diagnoses associated with each procedure were identified and divided into diagnostic (symptom or sign assessment), surveillance (for colonoscopy), and screening categories as follows: the diagnostic category included abdominal distension (ICD-9-CM 787.3), abdominal pain (ICD-9-CM 789.0), abdominal swelling or mass (ICD-9-CM 789.3), anemia (ICD-9-CM 285.1 and 285.9), anorexia (ICD-9-CM 783.0), bowel obstruction (ICD-9-CM 560.9), change in bowel habits (ICD-9-CM 787.9), constipation (ICD-9-CM 564.0), diarrhea (ICD-9-CM 558.9 and 564.5), fecal incontinence (ICD-9-CM 787.6), gastrointestinal bleeding (ICD-9-CM 578), heme-positive stool (ICD-9-CM 792.1), hemorrhage of rectum and anus (ICD-9-CM 569.3), iron-deficiency anemia (ICD-9-CM 280), nausea and emesis (ICD-9-CM 787.0), and weight loss (ICD-9-CM 783.2); the surveillance category included Crohn disease (ICD-9-CM 555), personal history of colorectal carcinoma (ICD-9-CM V10.05 and V10.06), personal history of colon polyps (ICD-9-CM V12.72), ulcerative colitis (ICD-9-CM 556), and none of the ICD-9-CM codes described in symptom or sign assessment; and the screening category included all procedures performed in the absence of codes associated with diagnostic indications or surveillance. In addition, a hierarchy was incorporated into the algorithm to account for procedures in mutually exclusive categories. For example, if a claim included codes for both diagnostic indications and surveillance, then the procedure was categorized as diagnostic. Although the recommended screening frequency may have varied depending on the presence or absence of a family history of colorectal neoplasia, this risk factor did not change the indication for screening or surveillance.
Analysis
Annual procedure rates were determined for FOBT, sigmoidoscopy, barium enema, colonoscopy, upper endoscopy, and upper gastrointestinal series. For each procedure listed above, the Physician/Supplier and Outpatient files were used to identify claims in eligible patients according to ICD-9-CM or CPT-4 codes. Note that although reimbursement for screening colonoscopy was provided only for patients who were at increased risk for the development of colorectal neoplasia (prior history of colorectal polyps or malignancy, family history of colorectal carcinoma, chronic inflammatory bowel disease), appropriateness or approval or denial of services for individual procedures was not evaluated. The two files were combined and records unduplicated by the beneficiary identification code, procedure, and date of service, such that a procedure documented in both files for the same individual and date of service, were counted only once.
The denominator for calculating procedure rates consisted of FFS months, as obtained from the 1999 Medicare Denominator files, which then were annualized. Rates were age adjusted and calculated according to the procedural indication (i.e., diagnostic, screening, or surveillance). The relative differences in rates between Caucasian patients and African-American patients, both for overall procedure use and procedures for specific indications, were determined. Statistical comparisons were performed using the two-sample test for binomial proportions.
RESULTS
We identified approximately 33.8 million Medicare beneficiaries, who accounted for approximately 298 million annual FFS months (Table 1). The cohort was 59% female and 87% Caucasian, and more than half of the beneficiaries were ages 65–74 years.
| Characteristic | No. (millions) | Fee for service months (millions) |
|---|---|---|
| Gender | ||
| Female | 20.00 | 179.38 |
| Male | 13.84 | 118.86 |
| Age (yrs) | ||
| 65–69 | 9.18 | 76.49 |
| 70–74 | 8.71 | 77.29 |
| 75–79 | 7.09 | 64.22 |
| 80–84 | 4.71 | 43.43 |
| ≥ 85 | 4.14 | 36.82 |
| Race | ||
| Caucasian | 29.25 | 262.31 |
| African American | 2.64 | 22.24 |
| Other | 1.95 | 13.68 |
| Total | 33.84 | 298.24 |
The race-specific and gender-specific age-adjusted annual rates of FOBT, sigmoidoscopy, colonoscopy, and barium enema, as well as the two reference procedures (upper endoscopy and upper gastrointestinal series), are shown in Table 2. Both FOBT and sigmoidoscopy were performed more commonly in Caucasians than in African Americans (FOBT: 18.24% vs. 11.86%; P < 0.001; sigmoidoscopy: 3.07% vs. 2.17%; P < 0.001), whereas barium enema was performed more commonly in African Americans (2.26% vs. 1.88% in Caucasians; P < 0.001). Colonoscopy use also was somewhat more common for Caucasians (7.36% vs. 6.88% in African Americans; P < 0.001).
| Procedure | Men | Women | ||||
|---|---|---|---|---|---|---|
| Annualized % | % differencea | Annualized % | % differencea | |||
| Caucasian | African American | Caucasian | African American | |||
| FOBT | 16.52 | 10.11 | +63.4 | 19.40 | 12.85 | +51.0 |
| Sigmoidoscopy | 3.37 | 2.09 | +61.2 | 2.88 | 2.19 | +31.5 |
| Colonoscopy | 8.00 | 6.97 | +14.8 | 6.95 | 6.85 | +1.5 |
| Barium enema | 1.68 | 2.09 | −19.6 | 2.01 | 2.36 | −14.8 |
| Upper endoscopy | 5.30 | 5.87 | −9.7 | 5.02 | 6.02 | −16.6 |
| Upper GI series | 2.28 | 2.69 | −15.2 | 2.70 | 3.18 | −15.1 |
- FOBT: fecal occult blood testing; GI: gastrointestinal.
- a P < 0.001 for all comparisons between racial groups.
Although the use of FOBT and barium enema was more common in women (FOBT: 18.78% vs. 15.97% in men; P < 0.001; barium enema: 2.04% vs. 1.71%; P < 0.001), rates of sigmoidoscopy and colonoscopy use were higher in men (sigmoidoscopy: 3.23% vs. 2.79% in women; P < 0.001; colonoscopy: 7.86% vs. 6.91%; P < 0.001). Sigmoidoscopy was performed more frequently in Caucasian patients than in African American patients regardless of gender, although the rate in African-American women (2.19%) was slightly higher compared with African-American men (2.09%; P < 0.001). Racial differences in colonoscopy use were observed mainly for men (8.00% and 6.97% in Caucasian men and African American men, respectively; P < 0.001), whereas the differences in colonoscopy rates between Caucasian women and African American women were of much smaller magnitude (6.95% in Caucasian women vs. 6.85% in African American women; P < 0.001). Both upper endoscopy and upper gastrointestinal series were performed more often in African-American patients (5.93% and 2.99%, respectively) than in Caucasian patients (5.14% and 2.54%, respectively; P < 0.001 for both comparisons), whereas rates of upper endoscopy were higher in men (5.39% vs. 5.17% in women; P < 0.001), and rates of upper gastrointestinal series were highest in women (2.77% vs. 2.33% in men; P < 0.001).
The procedure rates then were stratified by apparent indication—diagnostic, screening, and (for colonoscopy) surveillance (Table 3). For FOBT, sigmoidoscopy, and colonoscopy, the racial differences in use of procedures performed for diagnostic purposes were of smaller magnitude compared with procedures that were performed for screening or surveillance. For colonoscopy, procedure rates for diagnostic indications were higher for African-American men compared with Caucasian men (3.28% vs. 2.64%, respectively; P < 0.001) for African-American women compared with Caucasian women (3.36% vs. 2.86%, respectively; P < 0.001). In contrast, although rates of both diagnostic and screening barium enemas were higher in women and in African Americans, the differences were greater for diagnostic procedures.
| Procedure | Men | Women | ||||
|---|---|---|---|---|---|---|
| Annualized % | % differencea | Annualized % | % differencea | |||
| Caucasian | African American | Caucasian | African American | |||
| FOBT | ||||||
| Diagnostic | 4.66 | 3.89 | +19.8 | 5.32 | 4.74 | +12.2 |
| Screening | 11.72 | 6.17 | +90.0 | 13.97 | 8.06 | +73.3 |
| Sigmoidoscopy | ||||||
| Diagnostic | 1.06 | 0.81 | +30.9 | 1.12 | 0.94 | +19.1 |
| Screening | 2.15 | 1.21 | +77.7 | 1.65 | 1.19 | +38.7 |
| Colonoscopy | ||||||
| Diagnostic | 2.69 | 3.28 | −18.0 | 2.86 | 3.36 | −14.9 |
| Surveillance | 1.28 | 0.62 | +106.5 | 0.88 | 0.57 | +54.4 |
| Screening | 4.03 | 3.07 | +31.3 | 3.21 | 2.92 | +9.9 |
| Barium enema | ||||||
| Diagnostic | 1.04 | 1.40 | −25.7 | 1.28 | 1.56 | −17.9 |
| Screening | 0.61 | 0.68 | −10.3 | 0.70 | 0.78 | −10.3 |
- FOBT: fecal occult blood testing.
- a P < 0.001 for all comparisons between racial groups.
DISCUSSION
Colorectal carcinoma is currently the second most common fatal malignancy in the United States, with more than 148,000 cases and 57,000 deaths estimated annually.8 Compared with other common malignancies for which screening either is not recommended or is designed to detect established malignancy at an early stage, the goals of colorectal carcinoma screening are to detect premalignant adenomatous polyps as well as presymptomatic malignant disease. However, as recently as 2001, only 38.7% of respondents age ≥ 50 years in a population-based survey reported that they had undergone a sigmoidoscopy or a colonoscopy within the previous 5 years, and only 23.5% of respondents had received FOBT during the past year.5 Because sociodemographic characteristics are associated with the receipt of many preventive health services,7, 10, 11 there would be expected differences in the use of colorectal carcinoma screening and surveillance procedures. However, this hypothesis has not been studied well at a population level.
Using national-level population-based Medicare data, we observed that although there were racial differences in procedure use, the differences were most apparent for procedures performed for screening or surveillance purposes. For diagnostic procedures, racial differences were smaller quantitatively, and the use of diagnostic colonoscopy actually was higher in African Americans. Similarly, the rates of upper endoscopy and upper gastrointestinal series, two procedures that rarely are used for screening, were higher in African Americans and also suggest that the observed racial differences with respect to colorectal procedures are not due merely to an inability to access endoscopic or radiologic facilities. One interpretation of these findings is that African American patients and/or their healthcare providers may be more likely to defer procedures until symptoms or signs develop and are less likely to perform screening or surveillance per se.
The observed differences in screening use and indications may be one factor contributing to the higher colorectal carcinoma incidence (due to less removal of precursor lesions through polypectomy) and mortality (due to fewer diagnoses of early-stage, presymptomatic malignancies) observed in African American patients.7 Indeed, at least 50% of the observed racial disparity in colorectal carcinoma mortality is due to differences in stage at presentation.12 However, other factors, such as differences in dietary and other risk factors, the use of potentially curative treatment (such as surgical resection),13 and differences in the use of adjuvant therapy14 and postsurgical follow-up,15, 16 also are likely contributors to racial disparity. Sociodemographic factors, including income and educational level, as well as the presence of Medicare supplemental insurance also are likely contributing factors. In addition, much of the stage-specific and treatment-specific mortality differences probably are due to a higher prevalence of major comorbid illnesses in African Americans.17
Other groups also have studied the differential use of colorectal procedures in Medicare beneficiaries. Using Medicare claims in the state of Washington, Ko et al. examined rates of flexible sigmoidoscopy, colonoscopy, and barium enema for 1994, 1995, and 1998 and also evaluated FOBT use for 1998.18 Overall, less than 4% of the patients in their study received a screening procedure in a given year, with rates generally lower in women, nonwhite patients, and patients age > 80 years. However, because only 1.5% of their sample was African American, the power to discern racial differences was limited. Richards and Reker used a national 5% random sample of physician claims from 1999 to study racial differences in colorectal procedure use.19 As in the current study, rates of endoscopic procedures were lower among African Americans, rates of barium enema use were higher, and the differences persisted when the analyses were adjusted for county-level measures of socioeconomic status and physician supply. However, those analyses did not consider other procedures (such as FOBT), use Outpatient files, or stratify according to apparent indication.
The use of colorectal procedures by Medicare beneficiaries in the era before routine reimbursement also has been studied at a national level. Using a 5% sample of Part B claims for Medicare beneficiaries, Lurie and Welch20 studied the use of FOBT during 1995 (i.e., prior to Medicare reimbursement for screening FOBT). After the FOBT, 9.3% of patients had subsequent evaluations, with incomplete evaluation of the colon or unrelated testing as common as complete evaluation. In a study that used Medicare Part B claims for the state of Michigan, McMahon et al found lower use of colonoscopy and a higher rate of barium enema among African-American, female, and older patients.21 However, the procedural indications were not assessed. In contrast, in other studies of Medicare beneficiaries with an established diagnosis of colorectal carcinoma, the use of endoscopic and radiographic procedures did not differ by race.22, 23
Because of their national scope, large sample sizes, and relatively small incremental research costs, Medicare data are used commonly for epidemiological and health services research. However, there are several relevant limitations associated with Medicare claims. Because the Medicare data are collected for billing purposes and not research, there are potential concerns about both random and systematic bias in procedure coding accuracy.24 In contrast to mammography, there are no studies to date that have measured the accuracy of coding directly for colorectal screening procedures.25 In studies of inpatient claims, the accuracy of endoscopic procedure coding has been high,26, 27 though analogous data for outpatient procedures or nonendoscopic tests are not available. However, many of the procedural codes used in the current analysis are found commonly in studies of patients with colorectal carcinoma (unpublished data),23 and the accuracy of procedure codes in general is high.28, 29 There also are potential inaccuracies in diagnosis coding associated with the procedures, particularly for procedures that ordinarily would not be reimbursed for pure screening purposes (i.e., colonoscopy in average-risk beneficiaries). Whereas the specific diagnoses associated with the procedure may account for shifts in apparent procedural indication over time, it is unlikely that they would account for the observed racial differences by indication. The codes used for FOBT typically include home collection, and FOBT claims also may be missing if FOBT procedures were performed as part of digital rectal examinations and were not submitted to a clinical laboratory. However, the capture of in-office FOBT in claims data is unknown. The validity of the specific diagnostic codes to ascertain procedural indication also is not known. The analyses also were performed at the procedure level, rather than the patient level, and repeat tests or combinations of tests in individual patients were not determined. Thus, individual procedures were included separately, even if they were performed in a single patient. The findings also are limited to patients age ≥ 65 years and may not be generalizable to younger populations. However, this age group has the highest incidence of colorectal carcinoma8, 30 and, presumably, the highest yield from screening.
Although the current study did not capture patients enrolled in managed care plans, the proportion of African Americans in FFS and managed care under Medicare is similar (10% and 9%, respectively, in 2000).31 Thus, differential enrollment would not explain the findings. In addition, the analysis did not capture procedures performed during periods of managed care enrollment for patients who crossed over to FFS reimbursement. However, there is no a priori reason to suspect that racial differences in procedure use exist in only FFS settings. Finally, we could not ascertain the use of procedures prior to Medicare enrollment. If African Americans were more likely to receive screening before age 65 years, then a lower rate of subsequent procedures under Medicare would be observed. However, previous population-based studies7 do not support this observation, and if African Americans were less likely to receive screening before Medicare enrollment, then higher procedure rates would be expected in African Americans under Medicare.
The data from the current study suggest that disparities exist not only in the use of colorectal procedures but also in the indications for such testing. The observed differences in procedural indication may contribute to disparities in colorectal carcinoma incidence and mortality. Further studies should evaluate the impact of provision of reimbursement for screening colonoscopy in average-risk patients as of July 2001 and also determine whether the observed racial disparities persisted in later years.
Acknowledgements
The authors thank Ms. Fang Xu for her analytic support.
