Effect of Medicare Coverage on Use of Invasive Colorectal Cancer Screening Tests FREE

COLORECTAL CANCER (CRC) is the third most common cancer diagnosed in the United States and the second leading cause of cancer death.¹ The estimated lifetime risk of CRC is 6%, and approximately 2.5% of the population will die of this disease. In recent years, screening for CRC has become more widely accepted, and several professional medical societies have adopted guidelines for CRC screening.^2- 5 The 2 most common methods of screening for CRC are fecal occult blood testing (FOBT) and flexible sigmoidoscopy (FS). The former has been demonstrated to reduce CRC mortality by 15% to 33%,^6- 9 and the latter by 50% to 80%.^10- 12 Colonoscopy and double-contrast barium enema have also been advocated as screening tests in certain patient populations.

Current guidelines from the World Health Organization, the American Gastroenterological Association, and the American Cancer Society recommend FOBT yearly and FS every 3 to 5 years beginning at age 50 years in average-risk individuals.^2- 5 The US Preventive Services Task Force recommends CRC beginning at age 50 years.⁴ The evidence was believed to be good for FOBT, FS, and colonoscopy, but the optimal screening methods were not definied. In addition, some guidelines provide the alternative option of a total colon examination, either colonoscopy or double-contrast barium enema, every 5 to 10 years in patients at average risk.¹³

Despite recently published guidelines and studies demonstrating the effectiveness of CRC screening in reducing mortality rates, few people older than 50 years are screened according to current recommendations. In the 1997 Behavioral Risk Factor Surveillance System survey, only 20% of respondents older than 50 years nationally reported having FOBT during the preceding year, and 30% reported having undergone FS or proctoscopy in the previous 5 years.¹⁴ These percentages changed little in a follow-up survey performed in 1999. At this time, only 20.6% of respondents older than 50 years reported undergoing FOBT in the previous year, and 33.6% reported having undergone FS or colonoscopy in the previous 5 years.¹⁵ Results of a recently released study¹⁶ showed that only 12% of Medicare beneficiaries in the United States had FOBT or FS in 1995 or 1996. Utilization rates varied greatly but were highest in the Northeast and Southeast.¹⁶ It is unclear whether these tests were conducted for screening or diagnostic purposes in this study, and the actual frequency of screening tests in this population may be somewhat lower.

Although these studies have documented that the rate of CRC screening in the general population is suboptimal, it is not clear whether this is due to lack of insurance coverage for screening tests or other factors. Beginning January 1, 1998, the Health Care Financing Agency (now known as the Centers for Medicare and Medicaid Services, Rockville, MD) began coverage of CRC screening procedures for Medicare beneficiaries. The purpose of this study was to determine whether the frequency of invasive CRC screening procedures, including FS, barium enema, and colonoscopy, increased with the institution of insurance coverage for Medicare beneficiaries. We also sought to determine whether the frequency of screening differed in various population subgroups.

We used the Physician/Supplier Standard Analytic File and the Denominator File, which are administrative databases covering Medicare beneficiaries that are maintained by the Centers for Medicare and Medicaid Services. The Physician/Supplier Standard Analytic File contains claims data for outpatient physician and supplier services, including the date of the visit, the diagnoses associated with the visit (coded using International Classification of Diseases, Ninth Revision [ICD-9], codes), and the procedures performed during the visit (coded using Current Procedural Terminology [CPT] codes). The Denominator File contains information about date of birth, race, sex, place of residence, vital status, Medicare Part A and Part B eligibility, and enrollment in capitated health plans. We specifically examined utilization rates in calendar years 1994, 1995, and 1998.

All Medicare beneficiaries who listed a Washington State ZIP code of residence at the time of a claimed medical service in calendar years 1994, 1995, and 1998 were eligible for inclusion. Beneficiaries who died during the study year, who were younger than 65 years, and who were enrolled in a capitated health plan for any part of the study year were excluded. We excluded patients enrolled in capitated health plans because they may have received screening tests that were not submitted as claims while enrolled in these plans. Based on ICD-9 codes in the Physician/Supplier Standard Analytic File, we also excluded patients with a personal history of colon polyps (V12.72), CRC (V10.05), or inflammatory bowel disease (555.x and 556.x), since these patients are at increased risk of CRC and are usually recommended to have more frequent surveillance. Patients without a history of colon polyps, CRC, or inflammatory bowel disease were analyzed as average risk. We did not attempt to identify or exclude patients with a family history of CRC, as we did not believe that these patients could be reliably identified from the ICD-9 diagnosis codes available.

We examined the use of FS, colonoscopy, and double-contrast barium enema. We did not analyze the use of FOBT in 1994 and 1995. Before 1998, a claim for FOBT may not have been submitted consistently, since it may have been considered part of a routine physical examination and since screening tests were unlikely to be reimbursed. Therefore, any analysis of FOBT utilization rates will probably be inaccurate before 1998. In 1998, a Health Care Financing Agency common procedures coding system (HCPCS) code for FOBT as a CRC screening test was introduced, and we were able to analyze the use of screening FOBT in this year. We analyzed only the first test performed each year, as subsequent tests may have been performed to evaluate further any abnormalities found on the initial test.

In 1994 and 1995, there were no specific CPT or HCPCS codes for screening FS, colonoscopy, or double-contrast barium enema. Because these tests may be used for either screening or diagnosis, we inferred that they were performed for screening using the following algorithm. We first identified these procedures from CPT codes (colonoscopy codes: 44388, 44389, 44392, 44393, 44394, 45378, 45380, 45383, 45384, and 45385; sigmoidoscopy codes: 45300, 45305, 45308, 45309, 45315, 45320, 45330, 45331, 45333, 45338, and 45339; and barium enema codes: 74270 and 74280). We then assumed that the procedures were performed as screening tests if there were no ICD-9 diagnosis codes of gastrointestinal tract symptoms, weight loss, or anemia (abdominal pain codes: 787.3, 789.0x, and 789.6x; altered bowel habits codes: 564.0 and 787.x; gastrointestinal bleeding code: 578.x; positive FOBT code: 792.1; weight loss code: 783.2; iron deficiency anemia code: 280.x; and anemia, unspecified, code: 285.9) associated with any physician visits within the previous 3 months. Because of this 3-month exclusion rule, we analyzed only claims submitted between April 1 and December 31 of each year so that we would have diagnosis data from the 3 months before a test was performed. Use of this screening algorithm may have underestimated the actual frequency of screening, as one of these diagnoses may have been given to justify performing the test, even if the true indication was screening. However, we have no way of accurately distinguishing the procedures coded in this way.

In 1998, HCPCS codes were assigned for FOBT (G0107), FS (G0104), colonoscopy (G0105 and G0121), and double-contrast barium enema (G0106, G0120, and G0122) as CRC screening procedures, so these codes can be used to identify any of these screening tests for this year. However, if screening FS or colonoscopy were performed and a therapeutic procedure such as a biopsy were performed, these procedures may have been coded using the appropriate CPT code instead of the HCPCS code. Because of this coding provision, use of HCPCS codes alone would have underestimated the true frequency of screening. To identify such procedures whose initial indication was CRC screening, we used the same selective algorithm described previously for the 1994 and 1995 analyses. We assumed that procedures coded with HCPCS screening codes or identified through this algorithm were intended as screening tests.

As a comparison to using this algorithm to identify screening tests, we also identified all procedures in 1994 and 1995 simply by CPT codes, regardless of any associated diagnosis codes. In 1998, we identified all procedures by HCPCS and CPT codes, again without attention to the associated diagnosis codes. This would identify all FSs, colonoscopies, and double-contrast barium enemas performed, whether they were intended to be diagnostic or screening.

We determined the proportion of all beneficiaries at average risk for CRC who underwent FS, colonoscopy, or double-contrast barium enema in 1994, 1995, and 1998. These frequencies were calculated using both tests identified by the selective algorithm described previously (hereafter identified as screening tests) and all tests identified by all CPT or HCPCS codes (hereafter identified as all tests) regardless of any associated diagnosis codes.

We also analyzed the frequency of screening test use in population subgroups defined by race, sex, age, and place of residence (urban vs rural). Race was categorized as white, black, Asian, Hispanic, or other or unknown. Age was categorized as 65 to 69 years, 70 to 74 years, 75 to 79 years, or 80 years and older. Patient residence was determined by the residential ZIP code for the last recorded medical encounter in the study year. In cases in which a residential ZIP code was missing, we used the residential ZIP code listed in Medicare's Denominator File. Patient residence was defined as rural or urban depending on the Health Service Area in which the patient lived. Rural Health Service Areas include all ZIP codes that are closest in proximity by road miles to a rural hospital, as defined by the Washington State Department of Health.

Multivariable logistic regression analysis was used to determine the relative odds of receiving any test or a screening test after adjustment for potential confounding factors. To examine the effect of insurance coverage, we used the year of service as a proxy for coverage of screening tests. Other variables examined as potential predictors included age, sex, race, and urban vs rural residence. Significance of the regression models was tested using the log-likelihood statistic, and the method of Hosmer and Lemeshow was used to assess goodness of fit of the regression models. This study was approved by the institutional review board of the University of Washington.

Demographic characteristics of the beneficiaries are given in Table 1. Because demographics and results were similar for 1994 and 1995, only data for 1994 are presented in the tables. There were more female than male beneficiaries, especially at older ages, and beneficiaries were predominantly white. The overall percentages of beneficiaries receiving any test or screening test are given in Table 2. The overall utilization rate of tests was low, with fewer than 7% of beneficiaries receiving any test in any given year. The use of screening tests was also low, with less than 4% of beneficiaries receiving screening in any year. Men, whites, and urban residents received statistically significantly more tests than women, nonwhites, and rural residents, respectively (P<.001 for all comparisons). The utilization rate for screening tests was highest in people aged 70 to 74 years but declined in older age groups. Hispanics were less likely to receive a colon test than other racial groups. Results for 1995 were similar to those for 1994, although the proportion of beneficiaries receiving any colon test or a screening test was slightly lower in 1995 (data not shown). The increase in the utilization rate of all tests between 1994 and 1998 was statistically significant (P<.001), although the absolute increase in utilization rates was small (<1%) (Table 2). The utilization rate of invasive screening tests actually decreased slightly between 1994 and 1998 (P = .02).

To assess the sensitivity of the screening algorithm, we determined whether procedures identified using HCPCS codes for screening tests in 1998 would have been identified as screening by the algorithm. Overall, 82% of procedures identified as screening from HCPCS codes would have been designated as such from the algorithm.

Although we could not assess use of screening FOBT in 1994 and 1995, we assessed use of this screening modality in 1998 using the specific HCPCS code G0107. Use of this modality was infrequent in 1998, with fewer than 7% of beneficiaries receiving this service (Table 2). Use of FOBT varied significantly by demographic group, with women using this screening test more often than men. Whites were screened more frequently than nonwhites, and younger people were screened more frequently than older people. Rural residents were more likely to use screening FOBT than urban residents. Hispanics were much less likely to receive screening FOBT than other races.

Although the choice of the optimal screening test continues to be debated, colonoscopy is generally believed to be the gold standard for diagnosis of colon polyps and cancer.^17- 18 However, it can be a difficult procedure, with risk of adverse events, and the likelihood of a patient undergoing colonoscopy may differ in various demographic subgroups. We therefore determined whether use of colonoscopy differed by patient subgroups. In this comparison, we determined the proportion of all tests performed that were colonoscopies compared with FS or barium enema. The results are similar for use of any colon test or screening colon tests (Table 3). The proportion of tests performed that were colonoscopies increased from 1994 to 1998, and men were more likely to undergo colonoscopy than women. Older patients were more likely to undergo colonoscopy than younger patients, and rural patients were more likely than urban patients. Blacks were the racial subgroup most likely to undergo colonoscopy.

Last, we developed a multivariate logistic regression model to determine the odds of receiving colon procedures in different years after adjustment for differences in patient demographics (Table 4). For colon procedures as a whole, compared with 1994, the odds of receiving a colon test were significantly lower in 1995 and higher in 1998 (P<.001 for all comparisons). The odds of receiving any colon tests were similar for men and women. Patients aged 75 to 79 years were more likely to receive any colon test compared with other age groups (P<.001), and whites were more likely to be tested than other racial groups (P<.001). Rural residence was inversely associated with undergoing any colon procedure (P<.001). For screening colon procedures, the odds of receiving a test were not significantly different between 1994 and 1998, suggesting that insurance coverage did not increase the screening rate (P = .33). Women were significantly less likely to be screened than men, and older individuals were less likely to be screened than younger individuals (P<.001 for both comparisons). All nonwhite racial groups and rural residents were significantly less likely to be screened than whites and urban residents, respectively (P<.001 for all comparisons).

The frequency of CRC screening did not increase significantly from 1994 to 1995 to 1998. Utilization rates of colon tests were low in the overall population and in all subgroups studied. Less than 6% of the population received any type of colon test, and less than 4% received screening tests in any of the 9-month periods examined. If FS was performed at the recommended interval of every 5 years, one could expect that approximately 20% of the population would undergo FS every year. Use of the noninvasive screening FOBT was similarly low, at less than 7% of the general population in 1998. This proportion is lower than that previously reported in similar populations, suggesting that use of screening FOBT is underreported.^14- 16 Nevertheless, if this test were performed annually as recommended, one could expect that a substantially greater proportion of patients would undergo this test every year. Even if use of FOBT is underreported, it is difficult to imagine that underreporting accounts for the degree of underuse seen herein. Our results show that screening procedures are underused and that utilization rates did not meaningfully change within a year after initiation of insurance coverage. Although the absolute differences between testing and screening rates in population subgroups were low, extrapolating these differences in screening rates for 3 to 5 years could result in clinically meaningful differences in the likelihood of being screened. For example, in 1998, approximately 3.6% of men and 2.9% of women had a screening test performed. If extrapolated over 5 years, this would translate to a difference of 3.5% in the percentage of people screened.

Significant differences in the utilization rates of various medical procedures have been shown by race and sex. For example, Medicare data on hospital discharges from 1986 to 1992 showed that blacks underwent coronary artery bypass surgery, total hip replacement, and percutaneous transluminal angioplasty less often than whites.¹⁹ Similar findings have also been noted in the Veterans Affairs Medical System.²⁰ Women also were less likely to undergo invasive cardiac procedures than men.²¹ Our study extends these findings by showing that nonwhites were significantly less likely to undergo diagnostic and screening colon procedures than whites. Although similar proportions of women and men had any test performed, fewer women underwent screening examinations. Women were less likely to receive the gold standard colonoscopy but more likely to receive FOBT than men. We could not determine from our data whether this difference in use of screening tests was due to physician recommendation, patient preference, or both. Blacks were the racial group most likely to undergo colonoscopy as a diagnostic or screening test, although they were less likely than whites to undergo procedures in general. These differences in use patterns persisted even after institution of universal insurance coverage for these screening procedures in January 1998.

This study has several limitations. First, we analyzed data from only 3 years—2 before and 1 after institution of insurance coverage for screening. We could not assess temporal variability or secular trends in the use of screening services before coverage of the tests. At the time this study was undertaken, only claims data from 1998 were complete and available. Because we analyzed data from 1998 only, the first year of screening coverage, it may be that utilization rates have subsequently increased to more acceptable levels. Further analyses should be performed to assess this possibility. Second, we used administrative claims databases to assess health services utilization. The accuracy of outpatient claims coding has not been established for the diagnoses and procedures we evaluated. However, coding of claims for inpatient services and certain procedures seems to be fairly reliable and accurate.^22- 23 Also, some services may be provided without a claim being submitted. In this case, we could underestimate the actual use of procedures. This may be a larger problem for more minor tests such as FOBT. However, because the claims data are used for billing and reimbursement, most major services, including FS, colonoscopy, and double-contrast barium enema, are likely to be billed.

We analyzed data from only a single state, and these results may not be directly generalizable to other parts of the country. Because most Medicare beneficiaries are older than 65 years, we could not assess use of screening or the effect of insurance coverage in people aged 50 to 65 years, in whom screening is generally recommended. We also excluded patients who enrolled in capitated health plans, whose patterns of health service use may differ from those receiving traditional fee-for-service Medicare. Patients in capitated health plans may be screened more frequently than those in indemnity plans, as has been reported for mammography and Papanicolaou smears.²⁴ As more patients decide to enroll in these plans, screening frequency should also be studied in these settings. However, the utilization data available for this study were incomplete for these patients. Studying utilization data within capitated health plans will likely become more feasible as the Centers for Medicare and Medicaid Services requires more detailed claims data from these plans.

Although the longer-term effects of Medicare policy remain to be seen, our data show that provision of insurance coverage for a screening test does not immediately increase use of such tests. Even with provision of insurance coverage, screening rates were low overall, and women and minorities were still less likely to receive these tests. Data showing that blacks are statistically significantly less likely to be screened are important in light of the fact that blacks have higher incidence and mortality rates of CRC²⁵ and that screening is most cost-effective in this group.²⁶ These findings suggest that insurance coverage is only one factor affecting the use of screening tests. Other factors affecting screening rates may be related to patient and physician knowledge and attitudes and are not likely to be easily affected by changes in insurance coverage. Previous studies have shown that educational programs for patients,²⁷ computerized reminders for providers to order screening tests,²⁸ and use of ancillary support staff to provide screening²⁹ can increase compliance with screening procedures. Because utilization rates are low overall, further efforts should be directed at developing and implementing similar interventions, targeted at either patients or providers, to increase the frequency of CRC screening in all population subgroups.

Accepted for publication May 2, 2002.

This work was supported by the Robert Wood Johnson Clinical Scholars Program at the University of Washington (Seattle) and by an Outcomes Research Grant from American Digestive Health Foundation (Bethesda, MD)–TAP Pharmaceuticals.

We thank Roger A. Rosenblatt, MD, MPH, and L. Gary Hart, PhD, for their development of the database used in this analysis.

The opinions expressed in this work are those of the authors and do not necessarily reflect those of the Robert Wood Johnson Foundation.

Corresponding author and reprints: Cynthia W. Ko, MD, MS, Division of Gastroenterology, University of Washington, Box 356424, Seattle, WA 98195 (e-mail: cwko@u.washington.edu).