Embed Size (px)
Citation preview
Completion of Colorectal Cancer Screening in
Women Attending Screening Mammography1Leticia Molina, Ruth C. Carlos, MD, MS, A. Mark Fendrick, MD, Paul H. Abrahamse, MS, Qian Dong, MD,Steven J. Bernstein, MD, MPH
Rationale and Objectives. The American Cancer Society (ACS) and the United States Preventive Services Task Force (USP-STF) recommend colorectal cancer (CRC) screening to begin at age 50 in individuals at average risk for CRC. To estimate rateat which women eligible for CRC screening at the time of screening mammography attendance later completed in CRC screen-ing, we retrospectively evaluated CRC screening utilization in women who underwent screening mammography at our institu-tion.
Materials and Methods. We retrospectively identified 3357 women between the ages of 50 to 75 who received screeningmammography in 1998 at a single academic medical center using the institution’s Radiology Information System (RIS). Addi-tional information collected from the institution’s Health System Data Warehouse and the Radiology department’s mammogra-phy quality assurance data included mammography results, Breast Imaging Reporting and Database System (BI-RADS) classifi-cation of mammography findings, recommendation for screening mammography follow-up, insurance status, and CRC screeningutilization after screening mammography. After excluding women who were current with CRC screening at the time of mam-mography, we determined the proportion of eligible women who completed CRC screening after mammography. Age, insurancetype, BI-RADS code, and recommendation code were evaluated as potential predictors of CRC screening completion in eligiblewomen.
Results. Of the 3357 women between the ages of 50 and 75 who received screening mammography in 1998, only 414(12.3%) were current with CRC screening at the time of screening mammography. Of the remaining 2943 women whowere eligible for CRC screening at the time of screening mammography, 142 (4.8%) subsequently completed CRC screen-ing. Average time to completion of CRC screening after screening mammography is 35.4 months (range, 0.27–64.9).Managed care insurance was the only significant predictor of CRC screening completion after screening mammography ineligible women after adjusting for other variables (adjusted OR 1.73, 95% CI 1.21–2.47, P � .0001). Neither BI-RADSclassification nor postmammography recommendations were significantly associated with CRC screening completion.
Conclusions. Prevalence studies have demonstrated that women who were compliant with screening mammography weremore compliant with CRC screening. Our data suggest that despite this increased compliance, overall incidence of CRCscreening is low in the screening mammography population.
Key Words. Colorectal cancer; screening; mammography; teachable moment.
©
AUR, 2004In 2004, there will be 146,940 new cases of colorectalcancer (CRC) and 56,730 women will die from CRC.
Acad Radiol 2004; 11:1237–1241
1 From the Departments of Radiology (L.M., R.C.C., Q.D.) and Internal Med-icine (A.M.F., P.H.A., S.J.B.), University of Michigan, 1500 E. Medical Cen-ter Drive, Ann Arbor, MI 48109-0030. Received March 22, 2004; revisionrequested May 25; revision received June 1; revision requested June 24;revision received June 24; revision accepted July 15. Address correspon-dence to: R.C.C. e-mail: [email protected]
©
AUR, 2004doi:10.1016/j.acra.2004.07.025By comparison there will be 10,520 new cases of inva-sive cervical cancer and 3,900 women will die fromcervical cancer. Furthermore death rates have decreasedfor breast and cervical cancer between 2000 and 2004estimates, but have increased for CRC in the same timeperiod (1,2). Despite the relative risks of all three can-cers, compliance with screening guidelines varies. Inwomen who participated in the Behavioral Risk FactorsSurveillance Survey (BRFSS) administered yearly by
the Centers for Disease Control, colorectal cancer1237
MOLINA ET AL Academic Radiology, Vol 11, No 11, November 2004
screening rates are much lower than breast cancerscreening by mammography or cervical cancer screen-ing by Papanicolaou (Pap) smear (3,4). Because theBRFSS is a cross-sectional data set, the proportion ofwomen eligible for CRC screening at the time ofscreening mammography attendance who later partici-pated in CRC screening cannot be determined.
The American Cancer Society (ACS) and the UnitedStates Preventive Service Task Force (USPSTF) recom-mend CRC screening beginning at age 50 in individualsat average risk for CRC (5,6). To estimate the rate atwhich women newly eligible for CRC screening at thetime of screening mammography attendance later sponta-neously participate in CRC screening, termed the conver-sion rate, we evaluated subsequent CRC screening inwomen between the ages of 50 and 75 years old who re-ceived screening mammography services at our institu-tion.
MATERIALS AND METHODS
Patient Population and Data SourcesWe retrospectively identified 3357 women between
the ages of 50 and 75 years old (average age 62) whounderwent screening mammography in 1998 at a uni-versity hospital in the Midwest. Mammograms wereidentified using the institution’s Radiology InformationSystem (RIS) using CPT code 676092. MammographyQuality Assurance (QA) data, accessed through theRIS, was used to determine BI-RADS classification andmammographer’s recommendation for screening mam-mography follow-up. The institution’s Health SystemData Warehouse, which integrates information from thehospital cost-accounting, professional billing, and clini-cal information systems, was used to identify the pa-tient’s insurance status at the time of screening mam-mography and CRC utilization after screening mam-mography. The study was approved by the InstitutionalReview Board.
Data AnalysisWe initially determined the proportion of women who
were current with CRC screening at the time of screeningmammography using the Health System Data Warehousedescribed above. Based on U.S. Preventive Services TaskForce (USPSTF) and American Cancer Society (ACS)guidelines, women were current with CRC screening if
they had completed fecal occult blood testing (FOBT)1238
within the past year, flexible sigmoidoscopy or double-contrast barium enema within the past five years, orcolonoscopy within the past ten years (5,6). These womenwere excluded from the final study group.
The final study group was composed of women whowere eligible for CRC screening at the time of screeningmammography. The primary outcome that we consideredwas CRC screening completion in women eligible forCRC screening. Completion was defined as utilization ofFOBT, flexible sigmoidoscopy, barium enema, orcolonoscopy after screening mammography. We evaluatedcompletion of CRC screening through December 2003.
Potential predictors of completion of CRC screening ineligible women that we considered were age, health carecoverage, BI-RADS code, and postmammography recom-mendation. Age was treated as a continuous variable.Health care coverage was dichotomized into managedcare and nonmanaged care. BI-RADS classification wasdichotomized into normal/benign findings or indetermi-nate/probably malignant findings. Mammographer’s rec-ommendation for follow-up after the screening mammog-raphy was dichotomized into “routine screening mam-mography in 12 months or no additional follow-uprecommended” or “additional imaging, tissue sampling, orclinical follow-up.”
Statistical AnalysisEach potential predictor variable was first screened for
its relationship with CRC screening compliance. Univari-ate analysis was evaluated using the chi-squared testwhen the predictor was a dichotomous variable, and usingthe Student t-test when the predictor was a continuousvariable. Multivariate analysis was subsequently per-formed using all 4 predictor variables. As part of ourmultivariate regression analyses, we conducted a statisti-cal diagnostic test using the variance inflation factor todetermine the presence of significant collinearity betweeneach of the predictor variables with particular emphasison correlation between BI-RADS classification and post-mammography recommendations.
All statistical analyses were done using Stata 7.0 soft-ware (Stata Corp., College Park, TX), and statistical sig-nificance was set at 0.05.
RESULTS
Colon Cancer Screening StatusA total of 3357 women between the ages of 50 and 75
(see Table 1) underwent screening mammography in
ase s
Academic Radiology, Vol 11, No 11, November 2004 CANCER SCREENING BEHAVIOR
1998. Of these women, 414 (12.3%) were current withCRC screening at the time of screening mammography.Increasing age (adjusted odds ratio [OR] 1.03, 95% confi-dence interval [CI] 1.01–1.04, P � .0001), having man-aged care insurance (adjusted OR 1.58, 95% CI 1.27–1.97, P � .0001) and having a “normal or benign” BI-RADS classification (adjusted OR 2.95, 95% CI 1.03–8.43, P � .04) were significantly associated with beingcurrent with CRC screening at the time of mammography(see Table 2). The 414 women who were current withCRC screening were subsequently excluded from furtheranalysis.
Table 1Characteristics of the 3357 Women BetweenScreening Mammography in 1998
Patient characteristics
NumberAge, mean (range)Health insurance
Managed careNon-managed care
BI-RADS classificationNormal or benign findingsIndeterminate or probably malignant findingsNot specified
Classification of mammographer recommendatiRoutine screening mammography in 12 montAdditional imaging, tissue sample or clinical c
BI-RADS, breast imaging reporting and datab
Table 2Predictors of Current CRC Screening at the Time ofScreening Mammography in the 3357 Women
Predictors of current CRC screening Adjusted OR (95% CI)
Age 1.03† (1.01–1.04)Health insurance
Having managed care 1.58† (1.27–1.97)BI-RADS classification
Normal or benign findings 2.95* (1.03–8.43)Classification of mammographer
recommendationsRoutine screening mammography in
12 months 1.78 (0.71–4.47)
BI-RADS, breast imaging reporting and database system; CRC,colorectal cancer; OR, odds ratio.
*P � .05, †P � .0001
Completion Versus Non-completion of ColonCancer Screening in Eligible Women
One-hundred and forty-two of the remaining 2943women (4.8%) who were eligible for CRC screening atthe time of screening mammography completed CRCscreening (see Table 3). The average time to completionof CRC screening after screening mammography was35.4 months (range, 0.27–69.2). On multivariate analysis,the only significant predictor of CRC screening comple-tion following a screening mammogram in the eligiblewomen was having managed care insurance (adjusted OR1.73, 95% CI 1.21–2.47, P � .0001). Neither BI-RADSclassification nor the mammographer’s follow-up recom-mendation was significantly associated with CRC screen-ing completion.
We performed standard statistical evaluation of thedata designed to evaluate correlation between predictorsusing standard statistical diagnostic testing as described inthe method. Statistical diagnostic testing demonstrated nosignificant collinearity between BI-RADS classificationand postmammography recommendations, consistent withindependence of each of the predictor variables.
DISCUSSION
The American Cancer Society predicted decliningdeath rates for breast and cervical cancer between 2000
Ages of 50 and 75 Who Underwent
Not current withCRC screening
Current withCRC screening
N (%) N (%)2943 (87.7) 414 (12.3)
62 years (50–75) 60 years (50–75)
1422 (48.3) 236 (57.0)1521 (51.7) 178 (43.0)
2866 (97.4) 396 (95.7)66 (2.2) 16 (3.9)11 (0.4) 2 (0.5)
2803 (95.2) 391 (94.4)ltation 140 (4.8) 23 (5.6)
ystem; CRC, colorectal cancer.
the
onshsonsu
and 2004; however, predicted rates of CRC deaths in-
1239
ase s
MOLINA ET AL Academic Radiology, Vol 11, No 11, November 2004
creased during the same interval (1). Although CRCscreening has been shown to reduce mortality due tocolorectal cancer, its use continues to lag behind breastcancer screening (7).
Previous prevalence data from the Behavioral RiskFactors Surveillance Survey demonstrate that women whowere compliant with screening mammography were morecompliant with CRC screening (8) with a baseline preva-lence of approximately 30%. We therefore expected tofind a fairly high rate of CRC screening compliance inwomen who screen for breast cancer. However, our studyfound that only 414 women (12.3%) were compliant withCRC screening at the time of screening mammography.Furthermore, only 142 of the eligible women (4.8%) re-ceived CRC screening over the following 6 years.
Many reasons exist as to why a woman may not com-plete CRC screening. Inadequate understanding of thehealth benefits screening provides, failure of a physicianto recommend the procedure, poor insurance coverage,and psychological fear of the procedure all may reduceCRC screening (9,10). Educational efforts have beenshown to modestly improve FOBT compliance. Couplingeducational intervention with teachable moment increasesthe effect. Screening mammography represents a potential“teachable moment” (11) for providing CRC educationdesigned to improve overall knowledge of CRC screeningand to address specific psychosocial concerns.
Using the setting of screening for one type of cancer(such as screening mammography) to encourage compli-
Table 3Characteristics of 2943 Eligible Women BetwScreening After Screening Mammography VeCRC Screening
Patient characteristics
NumberAge, mean (range)Health insurance
Managed careBI-RADS classification
Normal or benign findingsNot specified
Classification of mammographer recommendatiRoutine screening mammography in 12 montAdditional imaging, tissue sample or clinical c
BI-RADS, breast imaging reporting and datab
ance with screening for another type of cancer (such as
1240
colorectal cancer) represents a novel use of the diagnosticimaging experience.
Other investigators have evaluated the use of diagnos-tic testing as a method to improve health behaviors, mostnotably for smoking cessation. Hepper et al. demonstratedthat within 2 to 3 years following their first spirometrytest, patients were more likely to quit smoking if the testwas abnormal compared to patients with a normal test(12). Women participating in a comprehensive lung can-cer screening protocol that included spiral computerizedtomography had a 16% quit rate after the procedure (13).Schnoll et al. further demonstrated that 59% of current orformer smokers queried about lung cancer screeningwould be interested in receiving smoking cessation inter-vention with screening. In the same population, 52%noted that they would quit if the scan was positive, com-pared to 19% if the scan was negative (14).
We posit that there may be psychosocial factors thataccount for a high level of compliance with screeningmammography, estimated at approximately 70% to 80%(8,15). Understanding and applying these psychosocialfactors may increase compliance with CRC screening,using the screening mammography experience as a deliv-ery setting for targeted education (16,17).
Our study does have several limitations. First, the preva-lence of CRC screening in our study was lower than in thenationally representative BRFSS database. This may be dueto differences in the data collection methods for the twostudies. The BRFSS relies on self-reported CRC screening
50–75 Years Old Who Completed CRCEligible Women Who Did Not Complete
Did not completeCRC screening
Completed CRCscreening
N (%) N (%)2801 (95.2) 142 (4.8)
60 years (50–75) 59.4 years (50–75)
1335 (47.7) 87 (62.3)
2726 (97.3) 140 (98.6)10 (0.4) 1 (0.7)
2666 (95.2) 137 (96.5)ltation 135 (4.8) 5 (3.5)
ystem; CRC, colorectal cancer.
eenrsus
onshsonsu
while our study was based on administrative data. Second,
e_
D.
Academic Radiology, Vol 11, No 11, November 2004 CANCER SCREENING BEHAVIOR
none of the patients who received CRC screening underwentFOBT. In a previous analysis of the BRFSS 2001 data, wedemonstrated that 22% of respondents who were currentwith CRC screening had received FOBT (8). Institutionalquality assurance performed subsequent to the current analy-sis demonstrated that FOBT is usually performed in the cli-nician’s office and although the results are recorded in thepatient’s medical chart, they are not consistently captured byadministrative databases. Thus our estimate of CRC utiliza-tion represents the lower bounds of institutional compliancewith CRC screening. Even assuming that similar utilizationof FOBT was achieved in our current population, the preva-lence of CRC screening would be less than 35%, markedlylower compared to screening mammography rates. Finally,patients often come to our institution by referral from theirphysician. If a woman routinely receives her health care atanother facility, we may underestimate the true rate of CRCscreening after mammography because we used institutionalutilization data.
CONCLUSION
Prevalence studies have demonstrated that women whowere compliant with screening mammography were morecompliant with CRC screening. Our data suggest that de-
Table 4Predictors of CRC Screening Completion in the 2943 EligibleWomen
Predictors of CRC screeningcompletion Adjusted OR (95% CI)
Age 1.00 (0.98–1.02)Health insurance
Having managed care 1.73† (1.21–2.47)BI-RADS classification
Normal or benign findings 1.72* (0.28–10.87)Classification of mammographer
recommendationsRoutine screening mammography in
12 months 1.07 (0.33–3.45)
BI-RADS, breast imaging reporting and database system; CI,confidence interval; CRC, colorectal cancer; OR, odds ratio.
*P � .05; †P � .0001
spite this increased compliance, overall incidence of CRC
screening is low in the screening mammography popula-tion. Screening mammography may represent a “teachablemoment” for improvement of CRC screening adherence.Table 4.
REFERENCES
1. American Cancer Society. Cancer deaths to decline in 2004. January2004. Available at: http://www.cancer.org/docroot/NWS/content/NWS_1_1x_Cancer_Deaths_to_Decline_in_2004.asp. Accessed January2004.
2. American Cancer Society. How many women get cancer of the cervix?Revised November 2003. Available at: http://www.cancer.org/docroot/CRI/content/CRI_2_2_1X_How_many_women_get_cancer_of_thcervix_8.asp?sitearea. Accessed January 2004.
3. Seeff LC, Shapiro JA, Nadel MR. Are we doing enough to screen forcolorectal cancer? Findings from the 1999 Behavioral Risk Factor Sur-veillance System. J Fam Pract 2002; 51:761–6.
4. Shapiro JA, Seeff LC, Nadel MR. Colorectal cancer-screening tests andassociated health behaviors. Am J Prev Med 2001; 21:132–7.
5. American Cancer Society. ACS cancer detection guidelines. RevisedJanuary 2004. Available at: http://www.cancer.org/docroot/PED/con-tent/PED_2_3X_ACS_Cancer_Detection_Guidelines_36.asp?sitearea�PEAccessed January 2004.
6. U.S. Preventive Services Task Force. Screening for colorectal cancer:recommendations and rationale. July 2002. Agency for Healthcare Re-search and Quality, Rockville, MD. Available:http://www.ahrq.gov/clinic/3rduspstf/colorectal/colorr.htm
7. Selby JV, Friedman GD, Quesenberry CPJr, Weiss NS. Effect of fecaloccult blood testing on mortality from colorectal cancer. A case-controlstudy. Ann Intern Med 1993; 118: 1–6.
8. Carlos RC, Fendrick AM, Bernstein SJ. Can breast and cervical cancerscreening visits be used to improve colorectal cancer screening? J GenIntern Med 2004; 19s1:127.
9. Harewood GC, Wiersema MJ, Melton LJ 3rd . A prospective, controlledassessment of factors influencing acceptance of screening colonos-copy. Am J Gastroenterol. 2002; 97:3186–94.
10. Walsh JM, Posner SF, Perez-Stable EJ. Colon cancer screening in theambulatory setting. Prev Med 2002; 35: 209–18.
11. McBride CM et al. Understanding the potential of teachable mo-ments: the case of smoking cessation. Health Educ Res 2003; 18:156 –170.
12. Hepper NG, Drage CW, Daview SF, Rupp WM, LaMothe J, Schoen-felder PF, Munson P. Chronic obstructive pulmonary disease: a com-munity-oriented program including professional education and screen-ing by a voluntary health agency. Am Rev Respir Dis 1980; 121:97–104.
13. Schnoll RA, Miller SM, Unger M, McAleer C, Halbherr T, Bradley P.Characteristics of female smokers attending a lung cancer screeningprogram: a pilot study with implications for program development.Lung Cancer 2002; 37:257–265.
14. Schnoll RA, Bradley P, Miller SM, Unger M, Babb J, Cornfeld M. Psy-chological issues related to the use of spiral CT for lung cancer earlydetection. Lung Cancer 2003;39(3):315–25.
15. Shapiro JA. Colorectal cancer-screening test and associated healthbehaviors. Am J Prev Med 2001;21:132–137.
16. Carlos RC, Fendrick AM. Improving cancer screening adherence: Usingthe “Teachable Moment” as a delivery setting for educational interven-tions. Am J Managed Care 2004; 10:247–48.
17. Carlos RC. The added value of screening mammography in improved
screening for other cancers. J Am Coll Radiology 2004;1:591-596.1241
