The Influence of Year-end Bonuses on Colorectal Cancer Screening

, , , , ,
The American Journal of Managed Care, September 2004, Volume 10, Issue 9

Objective: To estimate the effect of physician bonus eligibilityon colorectal cancer (CRC) screening, controlling for patient andprimary care physician characteristics.

Study Design: Retrospective study using managed care planclaims data from 2000 and 2001.

Methods: Data on 50-year-old commercially insured patients ina managed care health plan were linked to enrollment andprovider files. The data included information on 6749 patients(3058 in 2000 and 3691 in 2001). Multivariate logistic regressionmodels were used to assess the association between CRC screeningreceipt and physician bonus eligibility.

P

P

Results: From 2000 to 2001, CRC screening use increased from23.4% to 26.4% ( < .01). Results from the multivariate logisticregression analysis revealed that the probability that a patientreceived a CRC screening was approximately 3 percentage pointshigher in the bonus year, 2001 ( < .01).

Conclusions: Bonuses targeted at individual physicians wereassociated with increased use of CRC screening tests. However,more research is needed to examine the effect of performance-basedincentives on resource use and the quality of medical care.Specifically, there is a need to determine whether explicit financialincentives are effective in reducing racial disparities in the qualityof patient care. This has particular relevance for CRC screeninggiven that black patients are less likely to be screened, they havehigher CRC incidence and mortality rates compared with otherracial groups, and screening has been shown to be more cost effectivein this population.

(Am J Manag Care. 2004;10:617-624)

Colorectal cancer (CRC) is the second leadingcause of cancer-related death in the UnitedStates. The American Cancer Society (ACS)estimated that there would be 147 500 newly diagnosedcases of CRC and almost 57 100 deaths in 2003.1 Inaddition, CRC is expensive to treat, with costs estimatedat $6.5 billion per year.2 Several national organizationshave recommended fecal occult blood testing(FOBT), flexible sigmoidoscopy, colonoscopy, or double-contrast barium enema as effective screeningoptions for persons aged 50 years and older.3-6 Despitean increasing body of evidence that screening of asymptomaticpersons significantly reduces mortality,7-9 thepercentage of individuals who have been screenedremains low. Estimates from the 2001 Behavioral RiskFactor Surveillance System suggest that 23.5% ofrespondents aged 50 and older reported having FOBT inthe previous year and 47.3% reported undergoing lowerendoscopy in the last 10 years.3

An extensive review of the literature has documentedvarious barriers to all types of cancer services,including screening, in different settings and populations.10 One explanation offered for the low rates ofCRC screening is the growth in managed care organizations(MCOs). A type of plan that has flourished duringthe last 2 decades is the individual practice association.Individual practice associations have grown to dominatethe health maintenance organization segment of managedcare.11 The growth in individual practice associationsmay be attributable, in part, to their willingness toaccept insurer contracts that include explicit financialincentives.12

Managed care advocates suggest that MCOs' use ofimplicit (capitation and salary) and explicit (bonuses)financial incentives are an effective means of containinghealthcare expenditures. Although there are numerousarticles examining the effect of implicit financial incentiveson physician behavior, little is known about theinfluence of explicit financial incentives on the quality ofpatient care.12 Despite the paucity of empirical evidence,the use of financial incentives is considered controversialand has been the subject of intense public scrutinyand litigation,13 because it is generally perceived thatphysicians with managed care contracts face perversefinancial incentives to limit access to services.12,14-16

To reduce public concern surrounding MCO cost-containmentmeasures, several health plans are usingexplicit physician financial incentives to improve thequality of preventive care, including cancer screening.Whether bonuses are effective in improving the qualityof preventive cancer services is an empirical question.However, it is difficult to evaluate how the structure andfinancing of MCOs affect access to and outcomes of cancercare.17,18 The reason is that managed care has rapidly evolved during the last 25 years, and little informationis available on the new types of plans and physicianincentive arrangements.12,16

Beginning in January 2001, a large managed carehealth plan operating in the southeastern United Statesimplemented a year-end bonus program that wasdesigned, in part, to improve CRC screening use amongan individual practice association's primary care physicians(PCPs). The objective of this study was to estimatethe effect of physician bonus eligibility on CRC screening,controlling for patient and PCP characteristics.From a policy standpoint, it is important to establishthat MCOs' use of explicit financial incentives contributesto improving the quality of care. This is especiallyrelevant in an area such as CRC screening, inwhich practice has been documented to be particularlydeficient.3

METHODS

Data

Census 2000: Summary File 3

Income

Educational attainment

Managed care health plan claims data for 2000 and2001 for all commercially insured persons aged 50 yearsas of January 1, 2000, and January 1, 2001, were retrospectivelylinked to enrollment and provider files toexamine the association between CRC screening ratesand year-end bonuses. The patient data includedenrollment information, demographic characteristics(age and sex), ZIP code, and CRC procedure codes.Several patient characteristics, including race, income,and educational attainment, are associated with CRCscreening receipt.19-25 Because these variables were notcollected by the health plan, we imputed this informationusing patients' 5-digit ZIP codes, which were linkedto the of the US Bureauof the Census.26 Race was defined as the percentage ofthe population in each ZIP code that is black. was defined as income per capita by ZIP code. information was used to create3 variables that categorized the patient's neighborhoodas follows: percentage with less than a high school education,percentage of high school graduates, and percentageof college graduates.

The provider data included PCP characteristics suchas sex, year of medical school graduation, medical specialty,and whether the provider was eligible for theyear-end bonus. Primary care physician experience wasmeasured in years and was calculated by subtractingthe date of a provider's medical school graduation fromthe date of his or her patient's screening. For patientsnot screened, physician experience was calculated bysubtracting the year of medical school graduation fromthe midpoint in the year (July 1) 2000 or 2001.

The selection criteria used to determine PCP bonusprogram eligibility are proprietary. Therefore, to avoidpotential bias associated with selection of PCPs into thebonus program, we excluded providers and data on thepatients of those providers who were ineligible for thebonus program (data on patients of PCPs ineligible forthe bonus are available from the author). We also limitedour sample to 50-year-old patients who were continuouslyenrolled in the health plan in calendar years2000 and 2001.

Current Procedural Terminology

The Centers for Medicare & Medicaid Services procedurecoding system and codes were used to identify CRC screeningprocedures, including the following: FOBT, flexible sigmoidoscopy,colonoscopy, and double-contrast bariumenema (a list of procedure codes used is available fromthe author). For persons reaching age 50 in 2000 or2001, the CRC screening procedure codes were used toidentify enrollees who received a CRC screening.Furthermore, if there was documentation of flexible sigmoidoscopyor colonoscopy in a prior year, the PCP wasalso given credit for a CRC screening. Screening ratesfor the bonus program were calculated 120 days afterthe close of the bonus year to minimize problems associatedwith time lags in the filing of claims data.

Statistical Analysis

t

Bivariate statistical tests (&#967;2 and tests) were used tocompare descriptive statistics across years. A multivariatelogistic regression model was used to estimate theeffect that PCP bonus eligibility had on CRC screeninguse. The dependent variable was an indicator variabledenoting whether a patient received a CRC screening.To examine the effect of the bonus, we used the year ofbonus eligibility to approximate the effects of bonuseson the likelihood of CRC screening. Independent variableswere patient sex (female),27,28 race (black),28-30 percapita income,31 and educational attainment.31-33 Inaddition to the variable denoting the year of bonus eligibility,we included the following PCP characteristics:sex (female)34 and years of experience.35,36 The independentvariable "years of experience" was squared tocapture any potential nonlinear effects of physicianexperience on CRC use. The inclusion of the squaredterm in our model was based on theory, statistical significance,goodness of fit, and biasness.37,38 We includeda term to assess the interaction between female patientand female provider. Previous research has shown thatfemale patients treated by female physicians were morelikely to receive mammograms and Pap smears.39 Thisinteraction term allowed us to determine whether thisfinding extends to CRC screening. Finally, an indicatorvariable that distinguished PCPs with an internal medicine specialty from other specialties was included tocontrol for any unobserved differences in CRC screeningrates between specialties.

All analyses were performed using SAS software, version8 (SAS Institute, Cary, NC). This study wasapproved by the Institutional Review Board of theCenters for Disease Control and Prevention.

RESULTS

Sociodemographic characteristics of the patients areshown in Tables 1 and 2. At the ZIP code level, we estimatethat 28.2% of the population was black, and percapita income averaged $24 508. Approximately 16% ofthe population had less than a high school education,54.5% were high school graduates, and 29.2% hadattended college. Approximately 53% of the patientswere female, and approximately 20% of all patients hada female PCP. Physician experienceaveraged 19.6 years.Slightly more than half (52.1%)of the patients' PCPs listedinternal medicine as their specialty.

P

P

P

P

P

Of the 6749 patients includedin our analysis, approximately25% received a CRC screening(Table 1). Overall, CRC screeninguse increased approximately3 percentage points between2000 and 2001 (23.4% vs 26.4%,<.01). Most of the increase inCRC use was attributed to a 2.8percentage points increase in theuse of FOBT (17.8% vs 20.6%,<.01). The percentage ofpatients who received a flexiblesigmoidoscopy or colonoscopyincreased 1.2 percentage pointsbetween 2000 and 2001 (8.6% vs9.8%, = .07). Total CRCscreening use differed by sex,with women more likely thanmen to have received a CRCscreening test in 2000 (27.1% vs19.0%, <.01) and 2001 (32.0%vs 20.3%, <.01).

P

Table 3 shows the odds ofreceiving a CRC screening procedurein the bonus year(2001), controlling for patientand PCP characteristics. Thecoefficient on the key variable of interest "bonus eligibility"is positive and statistically significant ( < .01),indicating that patients were more likely to havereceived a CRC screening in 2001, the year the bonusprogram took effect.

P

P

P

P

The coefficient on black race is negative and statisticallysignificant ( = .03), suggesting that blackpatients were less likely than nonblack patients tohave received a CRC screening (Table 3). The coefficienton the per capita income variable is positive butnot statistically significant ( = .54). Patients withless than a high school education were less likely tohave received a CRC screening than those with a highschool education; however, the difference was notstatistically significant ( = .74). Also, college-educatedpatients were more likely than high school—educatedpatients to have received a CRC screening,although this difference was not statistically significant( = .21).

P

P

P

P

The sign on the coefficient for the female patientvariable is positive and statistically significant ( < .01)(Table 3). The sign on the coefficient of the variableindicating whether a patient's physician was female isnegative but not statistically significant ( = .23). Theinteraction term of female patient with female physicianis positive and statistically significant ( = .02). This suggeststhat female patients treated by a female PCP weremore likely to have received a CRC screening than malepatients treated by a female PCP and male or femalepatients treated by a male PCP. Primary care providerexperience and experience squared had no statisticallysignificant effects on CRC screening. Physicians withinternal medicine listed as their specialty were significantlymore likely to provide CRC screening ( < .01).

P

P

P

Table 4 shows reestimation of the logistic regressionmodel using FOBT and, in turn, flexible sigmoidoscopyor colonoscopy as the dependent variables. Because ofthe small sample size, we combined flexible sigmoidoscopyand colonoscopy procedures into one variable.For the model that had FOBT as the dependent variable,the coefficient on bonus eligibility is positive and statisticallysignificant ( < .01). This result indicated thatthe odds that a patient received FOBT increased in thebonus year. The sign on the coefficient for the femalevariable is positive and statistically significant ( < .01),indicating that women were more likely than men tohave received FOBT. The coefficient on black race isnegative and statistically significant ( < .01), suggestingthat black patients were less likely than nonblackpatients to have received FOBT.

P

P

For the model that had flexible sigmoidoscopy orcolonoscopy as the dependent variable, the coefficienton bonus eligibility is positive ( = .08) (Table4). The coefficient on the per capita income variableis positive and statistically significant ( = .02). A$10 000 increase in income raises the probability offlexible sigmoidoscopy or colonoscopy screening byapproximately 2%.

DISCUSSION

Our analysis shows that CRC screening use increased significantly between 2000 and 2001, suggesting that year-end bonuses targeted at individualphysicians were effective in improving the delivery of CRC screeningprocedures. This finding differs from previous work that suggested that bonusestargeted at physician group practices were ineffective in improving physiciandelivery of cancer screening procedures to female Medicaid managedcare beneficiaries.41 However,our finding is consistent with previous empirical evidence thatbonuses are more effective if they are targeted at individuals as opposed to aphysician group.42

One might argue that it may be appropriate to cluster at the level of thephysician and perform the analysis using a random-effects model. Doing soyielded results that, for the key variable of interest, were similar to thoseattained at the patient level (physician-level random-effects results areavailable from the author).

Previous research suggests that sex and racialdifferences may affect CRC screening rates by type ofprocedure.28 Among Medicare beneficiaries, womenwere more likely than men to receive FOBT and lesslikely to receive invasive procedures.28 To determinewhether results from previous research were generalizableto a commercially insured population, we reestimatedour logistic regression model using FOBT and,in turn, flexible sigmoidoscopy or colonoscopy as thedependent variables. Consistent with previous work,we found that commercially insured women weremore likely than men to have received FOBT, butfound no significant gender difference for the moreinvasive screening procedures.

Our results suggest that the previously publishedfinding that black Medicare beneficiaries were less likelythan nonblacks to receive FOBT28 extends to a commerciallyinsured population. Other findings pointed toeducational attainment and income as factors accountingfor racial disparities in CRC screening amongMedicaid recipients.31-33 Because we controlled for educationalattainment and income in our model, we minimizedthese factors as possible confounders for racialdifferences in FOBT use in a commercially insuredpopulation.

Income was a statistically significant predictor offlexible sigmoidoscopy or colonoscopy use. Therefore,for the commercially insured population, our resultssuggest that costs (in particular, copays and deductibles)may be an important barrier to the use of thesemore expensive invasive procedures.

Our study had several limitations. We analyzed datafor 2 years, the year before and the year coinciding withthe implementation of the year-end bonus program.Therefore, we were unable to distinguish the effect offinancial bonuses from temporal trends. However,among health plan PCPs who were ineligible for thebonus program, CRC use remained unchanged between2000 and 2001 (26.8% vs 26.4%). The presumption thatthe lack of a temporal trend in CRC use among PCPsineligible for the bonus is applicable to PCPs eligible fora bonus provides credence to the effectiveness of thebonus program. Nevertheless, more research is neededto examine the effect of various financial incentive programsduring a longer period to better distinguish theseeffects.

Changes in health status over time and the generosityof patient health insurance benefits, includingdeductibles and copays, are factors that may influencethe use of CRC preventive screening. This informationwas unavailable. As a consequence, our estimates ofthe bonus programs' effectiveness in improving CRCscreening may be biased. Also, we analyzed data forcommercially insured patients residing in one state;therefore, our findings may not be generalizable toother areas of the country. We were unable to distinguishreliability between CRC procedures for screeningvs diagnostic purposes. Consequently, we mayhave overestimated the CRC screening rates.

Providers may sometimes forgo billing for FOBTunder more traditional reimbursement arrangementsbecause they do not consider the reimbursement to beworth the administrative burden. If so, the bonus mayinduce providers to increase their billing for FOBT.This could bias our estimates of the actual increase inscreening activities. It is beyond the capability of thepresent study to verify this possibility. However, webelieve that the improvements in screening demonstratedin this study were not merely a function ofincreased billing because we found improvements in theuse of other more expensive screening modalities, suchas flexible sigmoidoscopy, which would not have beensubject to underreporting before the bonus program.

Colorectal cancer screening is recommended forpersons aged 50 years and older. However, our analysiswas limited to persons aged 50 as of January 1, 2000,and January 1, 2001. The exclusion of older personswas designed to reduce problems associated with thetime frame surrounding current guidelines for CRCscreening. For example, the ACS43 recommends flexiblesigmoidoscopy every 3 to 5 years beginning at age50 for persons at average risk for CRC. Given that wewere limited to 2 years of data, we would have beenunable to determine with certainty whether personsaged 52 years and older had previously received one ofthe more invasive types of screening procedures withinthe ACS's recommended time frame. Therefore, byexcluding older individuals, we avoid understating CRCscreening rates. We plan future studies with at least 5years of data to determine whether our findings extendto older commercially insured persons.

Several patient characteristics, including race,income, and educational attainment, that are associatedwith CRC screening receipt were unavailable in thedata. We imputed this information using neighborhoodinformation obtained at the ZIP code level from the USBureau of the Census.26 By using neighborhood-leveldata to proxy for missing individual-level variables, ourresults are only suggestive of the true effects of race,income, and educational attainment on CRC screeninguse. Therefore, the direction of the bias of our resultscan only be determined when the individual-level databecome available. The validity of augmenting microdatawith aggregate proxies has been demonstrated44-52 (asummary is available from the author).

In general, studies that examine the influence ofexplicit financial incentives on physician behavior donot usually describe the specifics of the incentivearrangement.53 This is due, in part, to the fact thatbonus algorithms are proprietary and unavailable toresearchers. According to Kane et al,53 no study hasprovided information on the frequency and timing ofincentive payments for preventive care. The limitedamount of information that was available for thepresent study suggested that bonuses were paid annuallyand were, in part, a function of the number of aPCP's patients who were screened for CRC. We believethat the present study is one of the first to report suchinformation. However, with additional information,including the magnitude of the bonus, one would bebetter able to gauge the effectiveness of explicit financialincentives.

An absolute increase of 3 percentage points in thereceipt of CRC screening, while statistically significant,may not appear to some to be clinically significant.However, because of the low level of adoption of CRCscreening, which has been demonstrated to be effectivein reducing mortality, it is important to consider thatthe bonuses in our study were associated with a 12.8%relative increase in CRC screening. Because of the pronouncedand persistent underuse of these lifesavingscreening modalities, and in light of the fact that theNational Committee for Quality Assurance recentlyadopted CRC screening as a new Health Plan EmployerData and Information Set effectiveness-of-care measure,evidence of interventions resulting in any improvementsin screening rates should be given carefulconsideration by health policy decision makers.

CONCLUSIONS

Managed cared organizations are using explicitfinancial incentives, such as year-end bonuses, to influencephysician behavior, despite a paucity of empiricalevidence as to the effectiveness of these strategies.12The results from our study suggest that bonuses targetedat individual providers resulted in an increasein the use of CRC screening tests among a commerciallyinsured population. However, to definitivelyestablish a connection, additional information pertainingto the individual physician bonus amount, physicianimplementation expenses, and patient expenses(copays and deductibles) is required. More research isneeded to examine the effect of performance-basedincentives on resource use and the quality of medicalcare. In particular, there is a need to determine howphysicians respond to the magnitude of bonus amountsat the individual and group levels. There is also a needto investigate whether explicit financial incentives areeffective in reducing racial disparities in the quality ofpatient care. This has particular relevance for CRCscreening given that black patients are less likely to bescreened, they have higher CRC incidence and mortalityrates compared with other racial groups, andscreening has been shown to be more cost-effective inthis population.28,54,55

Acknowledgments

We thank 3 anonymous referees for helpful comments.The views expressed herein are those of the authors and donot necessarily reflect the views of the Centers for DiseaseControl and Prevention, Alliance of Community HealthPlans, Georgia Cancer Coalition, Federal Reserve Bank ofAtlanta, or the Federal Reserve System.

From the National Center for Chronic Disease Prevention and Health Promotion,Centers for Disease Control and Prevention (BSA, CF, LA); Federal Reserve Bank of Atlanta(MMP); and Kerr L. White Institute for Health Services Research (JW, JE); Atlanta, Ga.

This research was supported in part by task order 0953-016 from the Centers forDisease Control and Prevention; Alliance of Community Health Plans, Washington, DC;and Georgia Cancer Coalition, Atlanta.

Address correspondence to: Brian S. Armour, PhD, Office on Smoking and Health,Centers for Disease Control and Prevention, 4770 Buford Highway NE, MS K50 (Rhodes),Atlanta, GA 30041. E-mail: barmour@cdc.gov.

Cancer Facts and Figures, 2003

1. American Cancer Society. . Atlanta, Ga:American Cancer Society; 2003. Publication 5008.03.

Bus Health

2. Moore G. Screening is key to preventing colorectal cancer. .2001;19:40.

MMWR Morb Mortal Wkly Rep

3. Centers for Disease Control and Prevention. Colorectal cancer test use amongpersons aged &#8805;50 years, United States, 2001. .2003;52:193-196.

Gastroenterology

Gastroenterology

Gastroenterology

4. Winawer SJ, Fletcher RH, Miller L, et al. Colorectal cancer screening: clinicalguidelines and rationale [published corrections appear in .1997;112:1060 and . 1998;114:625]. .1997;112:594-642.

CA Cancer J Clin

5. Smith RA, von Eschenbach AC, Wender R, et al, ACS Prostate Cancer AdvisoryCommittee, ACS Colorectal Cancer Advisory Committee, ACS EndometrialCancer Advisory Committee. American Cancer Society guidelines for the earlydetection of cancer: update of early detection guidelines for prostate, colorectal,and endometrial cancers: also: update 2001: testing for early lung cancer detection.. 2001;51:38-75.

Guide to Clinical Preventive Services

6. US Preventive Services Task Force. . 2nded. Baltimore, Md: Williams & Wilkins; 1996.

Ann InternMed

7. Selby JV, Friedman GD, Quesenberry CP Jr, Weiss NS. Effect of fecal occultblood testing on mortality from colorectal cancer: a case-control study. . 1993;118:1-6.

N Engl J Med

8. Winawer SJ, Zauber AG, O'Brien MJ, et al, National Polyp Study Workgroup.Randomized comparison of surveillance intervals after colonoscopic removal ofnewly diagnosed adenomatous polyps. . 1993;328:901-906.

N Engl J Med

9. Mandel JS, Bond JH, Church TR, et al. Reducing mortality from colorectal cancerby screening for fecal occult blood: Minnesota Colon Cancer Control Study.. 1993;328:1365-1371.

Cancer

10. Mandelbatt JS, Yarbroff KR, Kerner JF. Equitable access to cancer services: areview of barriers to quality care. . 1999;86:2378-2390.

Handbook of HealthEconomics

11. Glied S. Managed care. In: Culyer AJ, Newhouse JP, eds. . Amsterdam, the Netherlands: Elsevier Sciences Ltd; 2000:707-753.

Dis Manag Health Outcomes

12. Armour BS, Pitts MM. Physician financial incentives in managed care: resourceuse, quality and cost implications. . 2003;11:139-147.

13. Pegram v Hedrich, 530 US 211 (2000).

N Engl JMed

14. Grumbach K, Osmond D, Vranizan K, Jaffe D, Bindman AB. Primary carephysicians' experience of financial incentives in managed-care systems. . 1998;339:1516-1521.

Med Care

15. Lee-Feldstein A, Feldstein PJ, Buchmueller T. Health care factors related tostage at diagnosis and survival among Medicare patients with colorectal cancer.. 2002;40:362-374.

Arch Intern Med

16. Armour BS, Pitts MM, Maclean R, et al. The effect of explicit financial incentiveson physician behavior. . 2001;161:1261-1266.

Am J Med

17. Shaheen NJ, Ransohoff DF. Sigmoidoscopy costs and the limits of altruism[editorial]. . 1999;107:286-287.

Ann Intern Med

18. Lewis JD, Asch DA. Barriers to office-based screening sigmoidoscopy: doesreimbursement cover costs? . 1999;130:525-530.

J Health Care Poor Underserved

19. Hardy RE, Ahmed NU, Hargreaves MK, et al. Difficulty in reaching low-incomewomen for screening mammography. .2000;11:45-57.

Am J Prev Med

20. Suarez L, Roche RA, Nichols D, Simpson DM. Knowledge, behavior, and fearsconcerning breast and cervical cancer among older low-income Mexican-Americanwomen. . 1997;13:137-142.

N Engl JMed

21. Ayanian JZ, Kohler BA, Abe T, Epstein AM. The relation between health insurancecoverage and clinical outcomes among women with breast cancer. . 1993;329:326-331.

Ethn Dis

22. Gregorio DI, Walsh SJ, Tate JP. Diminished socioeconomic and racial disparityin the detection of early-stage breast cancer, Connecticut, 1986-1995. .1999;9:396-402.

Cancer Detect Prev

23. Campbell RJ, Ferrante JM, Gonzalez EC, Roetzheim RG, Pal N, Herold A.Predictors of advanced stage colorectal cancer diagnosis: results of a population-basedstudy. . 2001;25:430-438.

J Health Care Poor Underserved

24. Polednak AP. Poverty, comorbidity, and survival of colorectal cancer patientsdiagnosed in Connecticut. . 2001;12:302-310.

Breast Cancer Res Treat

25. Kerner JF, Mandelblatt JS, Silliman RA, et al, OPTIONS Research Team.Screening mammography and breast cancer treatment patterns in older women:Outcomes and Preferences for Treatment in Older Women Nationwide Study.. 2001;69:81-91.

Census 2000: Summary File 3: Census ofPopulation and Housing

26. US Bureau of the Census. . Washington, DC: US Bureau of the Census; 2002.

Oncol Nurs Forum

27. Weinrich SP. Predictors of older adults' participation in fecal occult bloodscreening. . 1990;17:715-720.

Arch Intern Med

28. Ko CW, Kreuter W, Baldwin LM. Effect of Medicare coverage on use of invasivecolorectal cancer screening tests. . 2002;162:2581-2586.

Am J Public Health

29. Holtzman D, Bland SD, Lansky A, Mack KA. HIV-related behaviors and perceptionsamong adults in 25 states: 1997 Behavioral Risk Factor SurveillanceSystem. . 2001;91:1882-1888.

CancerEpidemiol Biomarkers Prev

30. Cooper GS, Yuan Z, Rimm AA. Racial disparity in the incidence and case-fatalityof colorectal cancer: analysis of 329 United States counties. . 1997;6:283-285.

J Gen Intern Med

31. O'Malley AS, Forrest CB, Mandelblatt J. Adherence of low-income women tocancer screening recommendations. . 2002;17:144-154.

J Natl Cancer Inst

32. Breen N, Wagener DK, Brown ML, Davis WW, Ballard-Barbash R. Progress incancer screening over a decade: results of cancer screening from the 1987, 1992,and 1998 National Health Interview Surveys. . 2001;93:1704-1713.

Am J Public Health

33. Anderson LM, May DS. Has the use of cervical, breast, and colorectal cancerscreening increased in the United States? . 1995;85:840-842.

South Med J

34. Borum ML. Cancer screening in women by internal medicine resident physicians.. 1997;90:1101-1105.

J CommunityHealth

35. Herman CJ, Hoffman RM, Altobelli KK. Variation in recommendations for cancerscreening among primary care physicians in New Mexico. . 1999;24:253-267.

Prev Med

36. Weitzman ER, Zapka J, Estabrook B, Goins KV. Risk and reluctance: understandingimpediments to colorectal cancer screening. . 2001;32:502-513.

Using Econometrics: A Practical Guide

37. Studemund AH. . New York, NY:Addison Wesley Longman Inc; 2001:167.

Am J Manag Care

38. Balkrishnan R, Hall MA, Mehrabi D, Chen GJ, Feldman SR, Fleischer AB Jr.Capitation payment, length of visit, and preventive services: evidence from anational sample of outpatient physicians. . 2002;8:332-340.

J GenIntern Med

39. Lurie N, Margolis KL, McGovern PG, Mink PJ, Slater JS. Why do patients offemale physicians have higher rates of breast and cervical cancer screening? . 1997;12:34-43.

40. Greene WH. Econometric methods. Centers for Disease Control andPrevention Workshop; August 16-17, 2002; Atlanta, Ga.

Am J Public Health

41. Hillman AL, Ripley K, Goldfarb N, Nuamah I, Weiner J, Lusk E. Physicianfinancial incentives and feedback: failure to increase cancer screening in Medicaidmanaged care. . 1998;88:1699-1701.

Q Rev Econ Finance

42. Debrock L, Arnould RJ. Utilization control in HMOs. .1992;32:31-53.

CA Cancer J Clin

43. Byers T, Levin B, Rothenberger D, Dodd GD, Smith RA, American CancerSociety Detection and Treatment Advisory Group on Colorectal Cancer.American Cancer Society guidelines for screening and surveillance for early detectionof colorectal polyps and cancer: update 1997. . 1997;47:154-160.

On the Validity of Using Census GeocodeCharacteristics to Proxy Individual Socioeconomic Characteristics

44. Geronimus A, Bound J, Neidert L. . Cambridge,Mass: National Bureau of Economic Research; 1995:1-19. NBER TechnicalWorking Paper 189. Available at: http://dsl.nber.org/papers/t0189.pdf. AccessedFebruary 21, 2004.

N Engl J Med

45. Wise PH, Kotelchuck M, Wolson ML, Mills M. Racial and socioeconomic disparitiesin childhood mortality in Boston. . 1985;313:360-366.

Pediatrics

46. Gould JB, Davey B, LeRoy S. Socioeconomic differentials and neonatal mortality:racial comparison of California singletons. . 1989;83:181-186.

Pediatrics

47. Gould JB, LeRoy S. Socioeconomic status and low birth weight: a racial comparison.. 1988;82:896-904.

Ethn Dis

48. Collins JW, David RJ. Differences in neonatal mortality by race, income, andprenatal care. . 1992;2:18-26.

Cancer Causes Control

49. Liu L, Deapen D, Bernstein L. Socioeconomic status and cancers of the femalebreast and reproductive organs: a comparison across racial/ethnic populations inLos Angeles County, California (United States). .1998;9:369-380.

JAMA

50. Etchason J, Armour B, Ofili E, et al. Racial and ethnic disparities in health care[letter]. . 2001;285:883.

J Labor Econ

51. Bound J, Brown C, Duncan G, Rodgers W. Evidence in the validity of crosssectional and labor market data. . 1992;12:345-368.

J Am Stat Assoc

52. Lee L, Sepanski JH. Consistent estimation of linear and non-linear errors-invariablesmodels and validation information. . 1995;90:130-140.

Economic Incentives: EvidenceReport/Technology Assessment: Prepared by the Minnesota Evidence-BasedPractice Center Under Contract No. 290.02.0009

53. Kane RL, Johnson PE, Town RJ, Butler M. . Rockville, Md: Agency forHealthcare Research and Quality. AHRQ publication. In press.

Cancer

54. Nelson RL, Dollear T, Freels S, Persky V. The relation of age, race, and genderto the subsite location of colorectal carcinoma. . 1997;80:193-197.

Gastroenterology

55. Theuer CP, Wagner JL, Taylor TH, et al. Racial and ethnic colorectal cancerpatterns affect the cost-effectiveness of colorectal cancer screening in the UnitedStates. . 2001;120:848-856.