addressing social determinants of health to improve access to early breast cancer detection: results...

Addressing Social Determinants of Health to Improve Accessto Early Breast Cancer Detection: Results of the Boston

REACH 2010 Breast and Cervical Cancer Coalition Women’sHealth Demonstration Project

Cheryl R. Clark, M.D., Sc.D.,1,2,3,4 Nashira Baril, M.P.H.,4,5 Marycarmen Kunicki, M.A.,1 Natacha Johnson,1

Jane Soukup, M.Sc.,3 Kathleen Ferguson,1 Stuart Lipsitz, Sc.D.,3 and JudyAnn Bigby, M.D.,4,6

for the REACH 2010 Breast and Cervical Cancer Coalition

Abstract

Background: The Boston Racial and Ethnic Approaches to Community Health (REACH) 2010 Breast and Cer-vical Cancer Coalition developed a case management intervention for women of African descent to identify andreduce medical and social obstacles to breast cancer screening and following up abnormal results.Methods: We targeted black women at high risk for inadequate cancer screening and follow-up as evidenced bya prior pattern of missed clinic appointments and frequent urgent care use. Case managers provided referrals toaddress patient-identified social concerns (e.g., transportation, housing, language barriers), as well as navigationto prompt screening and follow-up of abnormal tests. We recruited 437 black women aged 40–75, who receivedcare at participating primary care sites. The study was conducted as a prospective cohort study rather than as acontrolled trial and evaluated intervention effects on mammography uptake and longitudinal screening rates vialogistic regression and timely follow-up of abnormal tests via Cox proportional hazards models.Results: A significant increase in screening uptake was found (OR 1.53, 95% CI 1.13-2.08). Housing concerns( p< 0.05) and lacking a regular provider ( p< 0.01) predicted poor mammography uptake. Years of participationin the intervention increased odds of obtaining recommended screening by 20% (OR 1.20, 95% CI 1.02-1.40), butthis effect was attenuated by covariates ( p¼ 0.53). Timely follow-up for abnormal results was achieved by mostwomen (85%) but could not be attributed to the intervention (HR 0.95, 95% CI 0.50-1.80).Conclusions: Case management was successful at promoting mammography screening uptake, although nochange in longitudinal patterns was found. Housing concerns and lacking a regular provider should be ad-dressed to promote mammography uptake. Future research should provide social assessment and address socialobstacles in a randomized controlled setting to confirm the efficacy of social determinant approaches to improvemammography use.

Introduction

Black women are more likely than white women to diefrom breast cancer.1 Black women have breast cancer at

younger ages, are more likely to have hormone-negative tu-mors that are less responsive to current adjuvant therapies,are more likely to present with advanced stage disease, and

have the highest breast cancer mortality rates of any racialgroup in the United States.2 Although multiple factors con-tribute to explain the higher mortality from breast cancer inblack women, lack of regular and consistent use of mam-mography screening and delayed follow-up of abnormalresults contribute to racial disparities in tumor stage at diag-nosis and breast cancer outcomes.3,4 Thus, developing

1Center for Community Health and Health Equity, Brigham and Women’s Hospital, 2Brigham and Women’s-Faulkner HospitalistProgram, and 3Division of General Medicine and Primary Care, Brigham and Women’s Hospital, Boston, Massachusetts.

4Boston REACH 2010 Breast and Cervical Cancer Coalition, Boston, Massachusetts.5Boston Public Health Commission, Boston, Massachusetts.6Executive Office of Health and Human Services, Commonwealth of Massachusetts, Boston, Massachusetts.

JOURNAL OF WOMEN’S HEALTHVolume 18, Number 5, 2009ª Mary Ann Liebert, Inc.DOI: 10.1089=jwh.2008.0972

677

culturally appropriate interventions to promote entry intoscreening, repeat mammography use, and timely follow-up ofabnormal results is a priority for black women, particularly inprimary care settings.

Recent studies of barriers and facilitators along the con-tinuum of screening and follow-up strongly implicate socialdeterminants of health as factors that place black women athigh risk for receiving poor care.5–11 Particularly high-riskgroups face challenges related to socioeconomic status (SES),language barriers, racial discrimination, and geography thathave broad-reaching consequences for their access to insur-ance coverage, breast cancer knowledge, regular providers,culturally competent doctor-patient communication, andavailability of screening facilities.8,11–16 These factors, in turn,pose logistical barriers to obtaining care (transportation,competing time demands), shape women’s preferences andhealth beliefs (trust, fear, risk perception), and interact withprovider and health systems factors (provider time pressures,fragmented care) to produce disparities in outcomes.8,11,12,17–21

As an example, qualitative work by Moy et al.8 illustrates theconfluence of these factors. In semistructured interviews withwomen of diverse backgrounds, African American partici-pants indicated that their priorities and preferences for cancerscreening were greatly affected by social problems (drugabuse, domestic violence) and poor experiences with health-care systems.

To improve mammography screening patterns amongblack women, intervention strategies must target patients,providers, healthcare systems, and community settings.4,22–28

Research syntheses and meta-analyses that analyze the im-pact of these strategies suggest that the most successfulinterventions address barriers at multiple levels.29 Amonginterventions that target patients, tailored services appear tooutperform general health education messaging.4,12,29,30 Themajority of the research in underserved populations, how-ever, relates to single-use mammography.29 Further work isrequired to identify strategies that improve use along thecontinuum of early detection among black women with highsocial and medical risks for underscreening and following upabnormal results.4

To address these issues, the Boston Public Health Com-mission, the city’s health department, formed the Racial andEthnic Approaches to Community Health (REACH) 2010Breast and Cervical Cancer Coalition to guide community-driven efforts to understand and confront social determinantsof health that may contribute to excess breast and cervicalcancer deaths among women of African descent. Funded bythe U.S. Centers for Disease Control and Prevention (CDC),REACH 2010 is a cornerstone of the CDC’s initiatives toeliminate racial and ethnic health disparities. The REACH2010 initiatives specifically adopted a community-basedparticipatory research (CBPR) approach as a strategy to createculturally targeted interventions that identify and addresscommunity concerns.31,32 Studies that address social deter-minants of health, in particular, may be strengthened by directcommunity participation so that results are tailored to theconcerns of those most affected.33 This study reports themammography screening intervention results of the REACHWomen’s Health Demonstration Project (WHDP). The WHDPwas developed to test the feasibility of performing social andmedical assessments in primary care settings and to addresswomen’s social concerns as a strategy to improve patterns of

screening mammography use and follow-up among blackwomen at risk for receiving fragmented mammography care.

Problem approach and hypotheses

In concert with the national REACH 2010 CBPR initiatives,concerned local citizens, including community activists,public health officials, and academics, convened with theBoston Public Health Commission in 2000 to form the BostonREACH 2010 Breast and Cervical Cancer Coalition (REACHCoalition). The development of the REACH Coalition and theCoalition’s prioritization of the early detection of breast andcervical cancer have been described previously.33

The underlying conceptual approach and hypothesis of theWHDP are that providing a comprehensive assessment ofwomen’s social and medical health will enable both womenand providers to develop a joint agenda to meet women’sconcerns and thereby enhance women’s (1) ability to attend tocancer screening, (2) skills to advocate for their needs, and (3)satisfaction with care.

The study aims reported here are (1) to test the feasibility ofusing a standardized health questionnaire to identify medicaland social risks among women, (2) to assess the social corre-lates of baseline mammography screening patterns in thispopulation, and (3) to determine if women with high socialand medical risks who undergo standardized risk assessmentand case management intervention can initiate and engage inregular mammography use and obtain timely follow-up forabnormal screening tests.

Materials and Methods

Intervention: case management

Based on community feedback from focus groups andcommunity meetings,33 the REACH Coalition designed a casemanagement intervention to reduce social and health systemsbarriers to accessing care. The multipronged case manage-ment intervention provided tailored services designed to helpaddress (1) potential social, logistic, and other health statusbarriers to seeking healthcare (e.g., domestic violence, lan-guage barriers, housing concerns, food insecurity, transpor-tation, child care needs, depression, substance abuse), (2)patient-clinician communication barriers, including culturallyinadequate communication about screening recommenda-tions and abnormal results, and (3) health systems barriers,including navigation needs to help prompt and schedulescreening and track and report abnormal test results, to helpclinicians provide appropriate follow-up for abnormal results.Case managers received 12 months of training in the first yearof the study, as well as ongoing training during the inter-vention. Case management training focused on women’shealth topics, including breast and cervical cancer screeningprocedures, and the conduct of culturally appropriate socialassessments using standardized study assessment tools. Atthe initial study visit, case managers administered the studyassessment questionnaire to learn women’s medical and so-cial concerns. Case managers then provided social interven-tion through referrals to connect women to tailored medicaland social services within their health centers and local publicservice environments to help resolve these concerns (e.g.,domestic violence assistance, mental health services, transpor-tation assistance, food pantry services, transitional housing

678 CLARK ET AL.

services, ensuring availability of language interpretation). Inaddition, case managers provided navigation services forclients by tracking and contacting women who were due forscreening or follow-up for abnormal results and communi-cating steps for completing screening or follow-up to patients.Case managers accompanied clients to medical examinationsas needed to provide social support. Additionally, casemanagers communicated with providers of their clients toprompt providers to schedule screening and communicatefollow-up needed for abnormal results to clients.

Intervention sites

A request for proposal (RFP) process was used to identifyintervention sites. Selected sites were licensed communityhealth centers or primary care clinics that served large num-bers of women of African descent in Boston. Specifically,participating sites were selected that were willing to join theREACH Coalition and meet regularly with other project sitesand REACH staff; that were willing to implement projectprotocols, including study evaluation, medical record re-views, and patient satisfaction surveys; that were willing touse a common patient social assessment tool; and that werewilling to implement a tracking system for mammogram andPap smear screening and results. Six sites implemented theintervention beginning in 2001. These included an academichospital clinic with on-site mammography; a communityhealth center licensed by an academic hospital, which didnot have on-site mammography; a free-standing communityhealth center with on-site mammography services; and threefree-standing community health centers without on-sitemammography services. Sites with on-site mammographyunits had on-site radiology to interpret screening tests, andsites without on-site mammography had access to localhospital-based mammography and the Boston city mam-mography van, which makes scheduled visits to city sites thatdo not have on-site radiology services.

Study participants

Eligible participants included women aged 18–75 whowere residents of Boston, Massachusetts, who were notpregnant at the time of enrollment, and who self-identified asblack or of African descent, including women identifying asAfrican American, Haitian, African, West Indian, or from theCaribbean. Women who received their care or were interestedin initiating care at one of the six primary care sites wereeligible to receive case management intervention. Recruit-ment targeted 1000 women at risk for receiving fragmentedprimary care, as evidenced by any of the following: no med-ical appointments in 2 years, repeated use of the urgent care(walk-in) service, missed clinic appointments, and missedmammogram appointments. Women were identified throughurgent care and appointment visit logs and through referralsfrom their primary care providers, on-site social workers, andcommunity-based counselors. Between 2002 and 2006, 918women aged 18–75 were enrolled and followed longitudi-nally between January 2002 and February 2007. The Institu-tional Review Boards overseeing all sites, Brigham andWomen’s Hospital, Boston Medical Center, Beth IsraelDeaconess Medical Center, and Boston Children’s Hospital,approved this study.

The analysis of mammography use was restricted towomen aged 40–75 (n¼ 483). Women with any known canceror suspected breast cancer at enrollment were not consideredeligible for screening and were excluded from analysis(n¼ 39). Seven women who received intervention did nothave medical records at their site, leaving 437 women (90.5%)who received intervention and were included in the analysis.

Measures

Social assessment tools: Women’s Health Questionnaire,Study Questionnaire. At enrollment, women were requiredto complete the standardized social and medical assessmenttool, the Women’s Health Questionnaire (WHQ). The WHQwas originally developed by agencies of the Boston PublicHealth Commission as an assessment tool for use in primarycare settings to identify health and social risks for poor birthoutcomes among women of reproductive age.34,35

The WHQ was adapted for the present study to collectbaseline information in several domains.33 Single items col-lected information on factors thought to impede or promotemammography use, including (1) demographic characteris-tics and SES (age, country of birth, primary language, racialand ethnic identity, household income, employment status,and level of education), (2) health behaviors, family his-tory, and physical health (self-reported prior Pap smear andmammography use, immunization status, self-defined prob-lems with alcohol and drug use, family history of cancers, andself-rated health), (3) access to care (presence of and rela-tionship with a regular health provider, insurance status andtype), (4) social concerns in the past year (availability of socialsupport, presence of housing concerns including affordingrent, overcrowding, or homelessness, food security, workhistory, occupational exposures, domestic violence, neigh-borhood safety, any reported experience of discrimination,self-identified sources of stress), and (5) emotional health(depressive symptoms, anxiety symptoms), thought to berelated to mammography utilization patterns.8

Women had the opportunity to complete an optionalassessment questionnaire, the Study Questionnaire (SQ).The SQ was designed to assess women’s health beliefs andsatisfaction with processes of healthcare.33 The SQ used sin-gle items to assess issues, including reasons a woman didnot have a recent mammogram or Pap smear and women’sperceptions of the cultural competency of providers andclinic settings. Administration of the SQ took approximately1 hour. Women who completed the SQ received $15 as com-pensation for their time; 428 women (98%) completed thisquestionnaire.

Medical record review

Medical records were reviewed 3 years retrospectively andannually after enrollment to record the dates when womenobtained mammograms. Chart abstraction recorded the in-dication for the examination (screening vs. diagnostic), datesand results of mammograms, and the indicated follow-upreported by the ordering physician or radiologist, defined interms of BI-RADS classification (normal, routine intervalmammography recommended; additional views and=orultrasound recommended; recommend 3–6 month follow-up;suspicious for cancer recommend biopsy).36

SOCIAL BARRIERS TO BREAST CANCER DETECTION 679

Statistical analysis

The study sought to understand the correlates of recentsingle-use mammography at baseline (predictors of self-report of a mammogram within 2 years prior to enrollment),as well as the effect of case management in promotingmammography uptake (at least one occurrence of mammog-raphy use during the observation period, documented bymedical record review), repeat (longitudinal) use over time(percentage of recommended annual mammograms that wereactually obtained during the intervention), and timely follow-up of abnormal results.

A control group at a separate site was planned at the in-ception of the study but could not be recruited because of aloss of study funding. Rather than create a control group ofwomen who would not receive an intervention among theenrolled participants, the intervention was evaluated bycomparing screening rates among women who did not re-port having a recent mammogram at baseline (in the 2 yearsprior to study entry) to determine if rates could be increasedto match women who reported having a recent mammogramat baseline (who had a mammogram in the 2 years prior tostudy entry). Second, the analyses evaluated any increasedtrend in initial or repeat mammography screening associatedwith years of study participation to determine if womenwith longer exposure to case management intervention hada higher percentage of uptake or repeat mammographyexaminations.

The intervention effect on timely follow-up of abnormalresults was determined by comparing the days to follow-upfor abnormal results found prior to case management inter-vention with those found during the intervention. Timelyfollow-up was defined by obtaining procedures within3 months or obtaining additional studies in time framesrecommended by BI-RADS classification.36

Descriptive statistical analyses were performed via theF-statistic from general linear models for mean comparisonsand chi-square tests for categorical variables; Fisher’s exacttest was used for comparisons with small numbers. Wheredistributions were skewed, median values with interquartileranges are presented and compared with Wilcoxon rank sumstatistics.

To analyze predictors of recent mammography use atbaseline, logistic regression analysis was employed to esti-mate the odds of obtaining a mammogram 2 years prior tostudy entry, measured by self-report and adjusted for cov-ariates, including baseline access to care (insurance coverage,having a regular provider), SES and social determinants ofhealth, and site of enrollment. Logistic regression was alsoemployed to estimate the adjusted odds of mammographyuptake (ever had a mammogram recorded by medical recordreview vs. never had a mammogram recorded by medicalrecord review, up to 3 years prior to enrollment or at any pointduring the study). Repeat (longitudinal) mammographypatterns were analyzed by estimating the odds of obtainingeach annual screening mammogram recommended duringthe study period. The Williams method was used to accountfor nonindependent observations of multiple mammographyevents for each woman.37 Consistent with MassachusettsDepartment of Public Health recommendations, all studyparticipants aged �40 were recommended to have screeningmammography annually. Diagnostic mammograms were not

considered screening examinations and were not counted inthe evaluation of screening patterns. In some cases, insuffi-cient clinical information was abstracted to identify the pur-pose of a mammogram coded as screening; thus, clinicaljudgment defined patients with multiple diagnostic orscreening mammograms in a year as being eligible forscreening or being under surveillance for a suspected canceror breast problem. Women not under surveillance for canceror suspected breast problems, with normal diagnostic mam-mography results, were considered eligible for furtherscreening. Women with abnormal screening or diagnosticresults were no longer considered screening candidates whenthe abnormal diagnosis was made; these results were con-sidered separately in the evaluation of abnormal results.

Cox proportional hazard models were used to model theeffect of case management exposure and the effects of cov-ariates on timely follow-up of the incident abnormal resultexperienced by a participant. Women who received late, albeitclinically indicated, follow-up were censored at the time theyreceived clinically indicated follow-up.

We report models that remove nonsignificant covariates( p> 0.05) from the analysis, except insurance status and siteof enrollment, which were included in all models. We hy-pothesized that insurance coverage and site of enrollmentmay confound the relationship between social determinantsof health and mammography utilization; thus, these covari-ates are included in all models. Although the MassachusettsDepartment of Public Health recommends annual mam-mography for women aged �40, other national guidelines donot recommend annual screening until age �50.38 To look formammography use patterns that may have related to usingdifferent guidelines among women �age 50, we reportmodels for the full cohort aged �40 older and the strata ofwomen aged �50. All analyses were conducted in SAS ver-sion 9.1 (Cary, NC).

Results

Baseline characteristics by intervention site, sitecharacteristics, and participant withdrawalcharacteristics

Study participants were similar across sites with respect totheir annual household income, experience of discrimination,and the number of social obstacles they reported (Table 1).Key differences by site of enrollment are listed in Table 1.Specifically, study participants who enrolled through the ac-ademic hospital clinic tended to have Medicaid or Medicare astheir insurance coverage, were most likely to be U.S. born, andwere most likely to have difficulties with housing (paid morethan half of annual household income in rent, homelessness).

Study participants who enrolled in the hospital-licensedcommunity health center without on-site mammographywere most likely to be non-U.S. born and had the poorest self-rated health. At the free-standing community-based healthcenter with on-site mammography, a high proportion wereuninsured, although a significant proportion had supple-mental private insurance. These participants had the highestself-rated health status. Study participants who enrolled atfree-standing community health centers without on-sitemammography were least likely to have had a regular pro-vider at enrollment. These participants were most likely toidentify themselves as having alcohol or drug problems.

680 CLARK ET AL.

Responses to the SQ showed enrollment sites were ratedsimilarly in terms of women’s satisfaction with provider’sintentions to provide their best care irrespective of race andprovider respect and courteousness. Additionally, partici-pants rated interpersonal characteristics of centers similarly(high levels of comfort receiving care and information aboutbreast problems) but thought that the free-standing centerswithout on-site mammography had the least convenient ac-cess to mammography services (Table 2).

Few women withdrew from study participation. Amongwomen due for mammography screening, 22 withdrew (6%)chiefly because of transferring care away from participatingsites (n¼ 7) or moving away from the city (n¼ 6); 7 womendied (nonbreast cancer related), 1 woman did not accept aprimary care physician at a participating site, and 1 womanno longer wished to participate. Withdrawal of two of thecommunity-based centers without on-site mammography in2005 occurred as a result of study funding cuts. The sitewithdrawals resulted in withdrawal of 51 participants (13%)whose data are analyzed here until the date of their with-drawal. Those who withdrew did not differ from those who

did not withdraw by age (49.2 years vs. 50.5 years, respec-tively, p¼ 0.75), income ( p¼ 0.99), or nativity ( p¼ 0.39).

Prediction of baseline screening rates

On average, 74% of women reported having a recentmammogram at baseline, in the 2 years prior to enrolling incase management intervention at study entry (Table 3). Thosewho enrolled at free-standing community-based health cen-ters without on-site mammography were least likely to reporthaving a recent mammogram (59%). Table 3 shows predictorsof recent baseline mammography use, stratified by age. In theadjusted model of the cohort aged�40, statistically significantbarriers to recent mammography use at baseline includedlacking a regular provider and having a low income (Table 3).An additional statistically significant social determinant ofrecent mammography use at baseline included having hous-ing concerns (Table 3). Among patients with housing con-cerns, 64% of women obtained mammograms within 2 yearsof study enrollment, compared with 79% of women withouthousing concerns (adjusted p< 0.01). Women who enrolled in

Table 1. Selected Participant Characteristics by Site of Enrollmenta

All

Academichospital

clinic, on sitemammography

Hospital-licensedcommunity

health center,no on-site

mammography

Free-standingcommunity

health center,on-site

mammography


health centers,no on-site

mammographyn¼ 437 n¼ 97 n¼ 98 n¼ 110 n¼ 132

Patient factorsMean age (SD)** 51 (8.5) 54 (8.3) 51 (8.9) 52 (8.6) 50 (7.7)Have a regular provider at enrollment?** 411 (95) 96 (99) 96 (98) 107 (98) 112 (86)

Insurance coverage**Uninsured (free care) 101 (24) 13 (14) 26 (27) 34 (33) 28 (22)Public (Medicare=Medicaid) 218 (52) 61 (64) 48 (50) 41 (39) 68 (54)Private supplemental 101 (24) 21 (22) 22 (23) 29 (28) 29 (23)

Nativity**U.S. born 288 (66) 91(94) 38 (39) 59 (54) 100 (76)Non-U.S. born 149 (34) 6 (6) 60 (61) 51 (46) 32 (24)

How would you rate your overall health?**Excellent 41 (9) 7 (7) 6 (6) 15 (14) 13 (10)Good 205 (47) 43 (44) 35 (36) 67 (61) 60 (46)Fair 156 (36) 38 (39) 45 (46) 26 (24) 47 (36)Poor 33 (8) 9 (9) 12 (12) 1 (1) 11 (8)

Alcohol=drug**Current=past 66 (15) 15 (15) 8 (8) 9 (8) 34 (26)Never 371 (85) 82 (85) 90 (92) 101 (92) 98 (74)

Housing concerns** 146 (34) 45 (46) 25 (26) 22 (20) 54 (41)

Income<$10,000 211 (53) 46 (52) 45 (50) 60 (59) 60 (51)$10–25,000 126 (32) 26 (30) 29 (32) 25 (25) 46 (39)>25,000 61 (15) 16 (18) 16 (18) 17 (17) 12 (10)

Experience of discrimination 32 (7) 7 (7) 11 (11) 4 (4) 10 (8)Social comorbiditiesb median (interquartile range) 3 (2–4) 3 (2–4) 3 (2–4) 3 (1–4) 3 (2–4)

an (%) unless otherwise noted. Where data are missing due to item nonresponse, numbers do not sum to 100%. General linear model testsassociations between categorical and continuous variables. w2 tests associations among proportions. The Wilcoxon rank sum tests associationswhere median values are compared. The highest values are highlighted in bold.

bNumber of social comorbidities is the sum of the following social factors potentially correlated with mammography utilization: nativity,income, unemployment, housing concerns, and alcohol and drug problems.

**Statistically significant differences by enrollment site ( p< 0.05).


free-standing community centers without on-site mammog-raphy were least likely to have had a recent mammogram atbaseline after adjustment for covariates; the comparison withthe hospital clinic with on-site mammography was not sta-tistically significant. Among participants aged �50, beinguninsured was associated with a lack of recent mammogra-phy use at baseline; otherwise, site of enrollment trends didnot differ from the cohort at large (Table 3).

Trends in mammography uptake:effect of study participation

Figure 1 shows significant uptake in mammography useover the period of case management intervention, docu-mented by medical record review. Participating in the inter-vention appeared to increase the trend towardmammography uptake; an increased trend in having a singlemammogram was seen in both groups of women with( p< 0.001) and without ( p< 0.01) recent baseline use. Duringcase management intervention, uptake rates among womenwithout recent use at baseline approached single repeat ratesamong women with recent baseline use. However, an unex-plained low rate of mammography use (67%) was seen amongwomen without recent mammography use who participatedin the study for 4 years or more (Fig. 1A). Among women�age 50, an increased trend in mammography uptake was

also seen during case management intervention among wo-men without recent baseline mammography use ( p< 0.01).Uptake rates approached or exceeded rates of a single repeatmammogram among women with recent baseline mam-mography use (Fig. 1B).

Table 4 shows predictors of mammography uptake. Thenumber of years of study participation was a statisticallysignificant predictor of mammography uptake. Adjusted forcovariates, the odds of obtaining a mammogram increased by53% for each year of study participation. Lacking a regularprovider at baseline reduced the odds of mammographyuptake in the cohort aged �40 (OR 0.20, 95% CI 0.07-0.62).

Having housing concerns was a significant social barrier tomammography uptake (OR 0.40 95% CI 0.21-0.77). Free-standing community health centers without on-site mam-mography tended to have lower uptake rates than othercenters (Table 4). Predictors of mammography uptake amongwomen �age 50 were similar, although only study partici-pation and enrollment at the academic hospital clinic werestatistically significant.

Predictors of longitudinal screening

Figure 2 describes patterns of repeat (longitudinal) mam-mography use. At the end of the follow-up period, fewwomen had obtained all recommended annual screening

Table 2. Provider and Clinic Characteristics by Site of Enrollmenta

Academichospital clinic,

on-site mammography

Hospital-licensedcommunity healthcenter, no on-site

mammography

Free-standingcommunity health

center, on-sitemammography


health centers,no on-site

mammographyn¼ 97 n¼ 98 n¼ 110 n¼ 132

Provider factorsMy provider tries to deliver the

best care regardless of raceAgree 86 (93) 92 (97) 92 (97) 99 (96)No opinion 3 (3) 2 (2) 3 (3) 2 (2)Disagree 3 (3) 1 (1) 0 2 (2)

How respectful and courteous isyour provider during your visit?Completely respectful 82 (88) 89 (94) 93 (93) 95 (91)Mostly respectful 6 (6) 5 (5) 7 (7) 8 (8)Somewhat respectful 3 (3) 0 0 1 (1)Not at all respectful 2 (2) 1 (1) 0 0

Clinic factorsI feel comfortable when I come

to get my care hereAgree 88 (94) 95 (98) 102 (96) 113 (93)No opinion 3 (3) 2 (2) 4 (4) 8 (7)Disagree 3 (3) 0 0 0

The clinic has convenient accessto services for mammography**Agree 90 (95) 78 (87) 92 (92) 93 (78)No opinion 5 (5) 10 (11) 6 (6) 21 (18)Disagree 0 2 (2) 2 (2) 6 (5)

I feel comfortable getting informationabout breast problems at this clinicAgree 85 (91) 89 (93) 94 (93) 105 (86)No opinion=disagree 8 (9) 7 (7) 7 (7) 17 (14)

aFigures are n(%). Chi-square tests associations among proportions. Numbers do not sum to 100% where data are missing due to itemnonresponse. Highest values are highlighted in bold.

**Statistically significant site differences ( p< 0.05).

682 CLARK ET AL.

mammograms. Among women eligible for screening, 25%obtained all recommended annual screening mammograms,60% obtained at least half of recommended examinations(mammograms approximately every 2 years or better), and22% never obtained mammograms during the study (Fig. 2).

Study participation did not appear to reduce the differencein longitudinal patterns of mammography use among womenwith and without recent mammography use at baseline. Atthe end of follow-up, 30% of women who reported a baselinemammogram had all recommended mammography, com-pared with 12% of women without a mammogram at base-line. Whereas 70% of women with a baseline mammogramobtained at least half of recommended screening during thestudy, only 33% of women without a baseline mammogramdid so. Also among women with a baseline mammogram,17% never obtained a mammogram during the study, com-pared with 37% of those without a baseline mammogram.Longitudinal screening patterns were similar when the anal-ysis was restricted to women �age 50.

Table 5 shows the predictors of longitudinal screeningpatterns. Years of study participation in the case management

intervention increased the unadjusted odds of completing re-commended examinations by 20%, although this effect wasnot statistically significant after adjustment for covariates, in-cluding prior patterns of use. The chief statistically significantpredictors of longitudinal screening patterns were clinic fac-tors, including participants’ rating of comfort with obtaininginformation about breast health at the clinic (OR 3.16, 95% CI1.79-5.60), and the site of enrollment. Compared with womenenrolled at the free-standing community health clinics withouton-site mammography, women enrolled at the academichospital clinic with on-site mammography (OR 2.50, 95% CI1.59-3.92), the hospital-licensed community health centerwithout on-site mammography (OR 3.28, 95% CI 2.05-5.26),and the free-standing community health center with on-sitemammography (OR 1.57, CI 1.01-2.42) had a higher odds ofobtaining recommended longitudinal screening. Women whowere uninsured and women with a family history of breastcancer were less likely to obtain recommended screening ex-aminations in this cohort (Table 5). Nativity was not a statis-tically significant predictor of baseline use, but after theintervention, women who were non-U.S. born had higher

Table 3. Predictors of Baseline Mammography use. Had a Mammogram within 2 Years Prior

to Study Entrya

Age 40 and older n¼ 437 Age 50 and olderb n¼ 223

%Odds ratio adjustedfor listed covariates %

Odds ratio adjustedfor listed covariates

Had a mammogram inpast 2 years

74 � Had a mammogram in past2 years

81 �

Access to care Access to careInsurance coverage Insurance coverage

Public insurance(Medicaid=Medicare)

72 1.11 (0.54–2.28) Public insurance(Medicaid=Medicare)

77 0.37 (0.10–1.36)

Uninsured=free care 72 0.98 (0.44–2.18) Uninsured=free care 79 0.24 (0.06–0.98)Private insurance 78 Reference Private insurance 89 ReferenceNo regular provider 30 0.10 (0.03–0.33)

Social determinants Social determinantsIncome Had housing concerns 73 0.65 (0.28–1.50)<$10,000 73 0.32 (0.12–0.89) No housing concerns 84 Reference$10,000–$25,000 69 0.33 (0.12–0.90)>$25,000 89 Reference

Had housing concerns 64 0.55 (0.33–0.93)No housing concerns 79 Reference

Site of enrollment Site of enrollmentAcademic hospital

clinic, on-sitemammography

77 1.82 (0.93–3.54) Academic hospital clinic,on-site mammography

77 2.11 (0.85–5.24)

Hospital-licensedcommunity healthcenter, without on-sitemammography

80 2.23 (1.10–4.49) Hospital-licensed communityhealth center, without on-sitemammography

91 6.38 (1.64–24.76)

Free-standing communityhealth center on-sitemammography

84 2.90 (1.41–5.98) Free-standing community healthcenter on-site mammography

90 4.88 (1.51–15.80)

Free-standing communityhealth centers withouton-site mammography

59 Reference Free-standing communityhealth centers withouton-site mammography

66 Reference

aUnadjusted frequency of mammogram use (%) and adjusted odds ratios with 95% confidence intervals. Mammography use at baseline ismeasured by self-report. Statistically significant odds ratios (p<0.05) in bold.

bIn women aged 50 and older, too few participants were without a regular provider (3%) and too few degrees of freedom to include incomevariable in multivariable analysis.


odds of obtaining all recommended mammography in ourcohort (OR 1.68, 95% CI 1.15-2.47).

Among women �age 50, the intervention (years of studyparticipation) did not appear to increase longitudinal screen-ing rates after adjustment for covariates (OR 1.01, 95% CI 0.77-1.33). Predictors of longitudinal screening patterns amongwomen �age 50 included being uninsured, nativity, beingcomfortable with information on breast health provided bythe clinic, site of enrollment, and baseline utilization patterns.

Trends in abnormal results

Table 6 shows the percentage of women with timely follow-up for abnormal mammography results. Fifteen percent

(n¼ 66) of the cohort had abnormal findings confirmed bymedical record review. We found that 22 women hadabnormal results found prior to case management interven-tion; 44 women had abnormal results found during the casemanagement intervention. The percentage of women whoreceived timely follow-up was high; however, there was nodifference in the percentage of women whose abnormal resultwas found and resolved in a timely manner before casemanagement intervention (86%) and those found and re-solved in a timely manner during the case management in-tervention (84%). Women with self-reported health problemswere less likely to receive timely follow-up than women whodid not report health problems (HR 0.49, 95% CI 0.25-1.00).Compared with women who enrolled in free-standing com-

A. Percentage of women with a single screening mammogram

28

67

9487

81

50 72 82

0102030405060708090

100

1 year 2 years 3 years 4 years or more

Number of years women participated in study

1 year 2 years 3 years 4 years or more

Number of years women participated in study

Per

cen

t o

f w

om

en w

ith

at

leas

t o

ne

scre

enin

g m

amm

og

ram

At least one repeatmammogram obtained,among women who hada mammogram in priortwo years- time trendwithin group p < 0.0001

At least onemammogram obtained,among women who didnot have a mammogramin prior two years- timetrend within group p <0.01

B. Percentage of women with a single screening mammogram

in women over 50

94

31

8783

52

8582

100

0

10

20

30

40

50

60

70

80

90

100

Per

cen

t o

f w

om

en w

ith

at

leas

t o

ne

scre

enin

g m

amm

og

ram

At least one repeatmammogram obtained,among women over 50who had a mammogramin prior two years- timetrend within group p <0.0001

At least onemammogram obtained,among women over 50who did not have amammogram in prior twoyears- time trend withingroup p < 0.01

FIG. 1. Mammography uptake patterns.

684 CLARK ET AL.

Table 4. Effect of Years of Study Participation on Mammography Screening Uptake:

Had at Least One Mammograma

Age 40 and older n¼ 437 Age 50 and older n¼ 223

Effect of studyparticipation

Adjusted forlisted covariates


Adjusted for listedcovariates

Years of study participation 1.59 (1.23–2.06) 1.53 (1.13–2.08) 1.81 (1.21–2.72) 1.85 (1.19–2.88)

Patient factorsNo regular provider 0.20 (0.07–0.62) -Public insurance (Medicaid=Medicare) 0.98 (0.45–2.12) 1.34 (0.39–4.61)Uninsured=free care 2.08 (0.73–5.91) 2.25 (0.47–10.77)Private insurance Reference ReferenceHad housing problems at baseline 0.40 (0.21–0.77) 0.65 (0.21–2.03)

Clinic factorsAcademic hospital clinic, on-site

mammography4.12 (1.54–11.04) 4.51 (1.06–19.22)

Hospital-licensed communityhealth center, without on-sitemammography

4.44 (1.56–12.62) 3.22 (0.76–13.54)

Free standing community healthcenter on-site mammography

1.52 (0.66–3.47) 1.48 (0.41–5.37)

Free-standing community healthcenter without on-site mammography

Reference Reference

aModels predict odds of screening mammography uptake, defined as ever having at least one mammogram, based on a medial recordreview of mammography utilization from 3 years prior to study entry until the end of study participation, adjusted for listed factors. Modelsexclude women under surveillance for suspected cancer or breast problems. Statistically significant odds ratios (p<0.05) in bold.

60%

70%

30%

12%

25%33%

17%

37%

22%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

All participantsN = 390

Participants witha prior recentmammogram

N = 287

Participants withouta prior recentmammogram

N = 103

Per

cen

t o

f w

om

en w

ith

list

ed p

atte

rn

of

mam

mo

gra

ph

y u

se

Percentage of women who obtained allrecommended annual mammograms duringstudy

Percentage of women who obtained at leasthalf of recommended mammograms duringstudy (bi-annual screening or better)

Percentage of women due for screening whonever obtained mammograms during study

FIG. 2. Patterns of longitudinal mammography use: percentage of women who obtained recommended mammograms bythe end of the study period. Longitudinal screening patterns between 2002 and 2007. Percentages do not sum to 100% becausethe percentage of women with biannual screening or better includes women who screened annually. Women who obtainedmammograms but underscreened (less than biannual screening) are not pictured.


Table 5. Effect of Years of Study Participation on Longitudinal Patterns of Mammography

Use: All Eligible Participantsa

Age 40 and older n¼ 390 Age 50 and older n¼ 196





Years of study participation 1.20 (1.02–1.40) 1.06 (0.88–1.28) 1.22 (0.98–1.52) 1.01 (0.77–1.33)

Patient factorsHad recent mammogram at baseline 2.16 (1.51–3.09) 1.94 (1.07–3.52)Public insurance (Medicaid=Medicare) 0.72 (0.50–1.05) 0.85 (0.47–1.54)Uninsured=free care 0.54 (0.35–0.85) 0.42 (0.20–0.87)Had housing problems at baseline 0.85 (0.61–1.18) 0.86 (0.51–1.45)Non-U.S. born 1.68 (1.15–2.47) 2.41 (1.29–4.49)Family history of breast cancer 0.64 (0.44–0.94) 0.74 (0.42–1.32)

Clinic factorsFeels comfortable getting breast

information at this clinic3.16 (1.79–5.60) 10.8 (3.13–37.23)

Academic hospital clinic, on-sitemammography

2.50 (1.59–3.92) 2.21 (1.14–4.28)

Hospital-licensed community healthcenter, without on-site mammography

3.28 (2.05–5.26) 3.14 (1.40–7.04)

Free-standing community healthcenter on-site mammography

1.57 (1.01–2.42) 1.77 (0.88–3.54)

Free-standing community healthcenters without on-site mammography

Reference Reference

aModels estimate odds of obtaining all recommended annual screening mammography adjusted for listed factors. Includes women whowere due for screening mammography and not under surveillance for suspected or diagnosed breast cancer. Analysis examines patterns ofscreening mammography use and excludes diagnostic mammograms. Statistically significant odds ratios (p<0.05) in bold.

Table 6. Abnormal Results: Predictors of Timely Follow-Up

Abnormal results n¼ 66

% with timelyfollow-upa

Median numberof days to follow-up


(unadjusted HR)

Adjusted forlisted covariatesb

(HR)

Abnormal result found during study 84 17 days 0.83 (0.47–1.46) 0.95 (0.50–1.80)Abnormal result found before study 86 9 days Reference Reference

Patient factorsHas other health problems 70 24 days 0.59 (0.32–1.09) 0.49 (0.25–1.00)No other health problems 91 11 days Reference ReferenceUninsured=free care 91 21 days 1.21 (0.49–2.99) 1.75 (0.64–4.81)Public insurance (Medicaid=Medicare) 86 11 days 1.24 (0.60–2.58) 1.62 (0.72–3.67)Private insurance 75 19 days Reference Reference

Clinic factorsAcademic hospital clinic and

hospital-licensed community health center94 7 days 2.28 (1.22–4.27) 2.54 (1.30–4.97)

Free-standing community health centeron-site mammography

82 79 days 0.86 (0.38–1.95) 0.90 (0.38–2.15)

Free-standing community health centerswithout on-site mammography

73 38 days Reference Reference

aTimely follow-up for biopsy and repeat imaging is defined as �3 months; timely follow-up when 3–6 month follow-up indicated isdefined as �7 months. Cox proportional hazard models estimate hazard ratios for timely follow-up, where those with late or no follow-up arecensored as non-events. Statistically significant odds ratios (p<0.05) in bold.

bHazard ratios also adjust for BI-RADS category of indicated follow-up: needed 3–6 month follow-up (OR 0.20, 95% CI 0.06-0.70) or neededbiopsy (OR 0.85, 95% CI 0.19-3.83) compared with reference category, need for additional or repeat imaging.

686 CLARK ET AL.

munity health centers without on-site mammography,women in the academic hospital clinic and hospital-licensedcommunity health center had higher rates of timely follow-up(HR 2.54, 95% CI 1.30-4.97).

Discussion

The REACH Coalition WHDP employed case managers toaddress social determinants of health in primary care settingsas a novel strategy to promote complete mammographyscreening and follow-up in a cohort of women of Africandescent who had been receiving fragmented care. The studyfound case management was associated with increasedmammography uptake rates, although no increase in repeat(longitudinal) mammography use was found. By addressingparticipant-identified social concerns, we report a novelfinding that housing concerns posed an important barrier tomammography uptake in this cohort. We also found, con-sistent with prior literature, that having a regular provider isan important facilitator of single-time mammography use.21

Additionally, we found that insurance type and site of carewere the significant correlates of repeat mammography use.We found that self-rated health problems and site of care weresignificant correlates of following up abnormal results in atimely manner. Finally, a hypothesis of the WHDP related tocultural competency: we found that women in this cohortrated their providers and sites of care highly along thesemeasures, and, therefore, these interpersonal aspects of pri-mary care factored less in patterns of mammography use andfollow-up in this study.

The demonstration project has important limitations thatshould be noted. First, the small numbers of women in eachcategory, particularly small numbers of women>age 50, likelylimited the power to detect patterns that were otherwise ob-served in the analyses that included all participants aged �40.For example, low statistical power likely resulted in the non-statistical trend associated with housing as a barrier to mam-mography uptake in the subset of women >age 50. We usedFisher’s exact test where appropriate because of small cell sizesand used a priori hypotheses to remove nonstatistically sig-nificant covariates to preserve power in multivariable analyses.

Second, the study assessment tools, the WHQ and SQ,employed single items rather than scales to assess social andmedical risks. The tools were designed for simplicity of use inprimary care settings; however, they are limited in theirability to measure complex concepts, such as acculturation,experiences of discrimination, and appraisals of clinical set-tings. For example, our study does not find evidence of per-ceived discrimination as a significant determinant ofmammography use, as was suggested by qualitative work.8,33

Our negative findings are similar to those of Dailey et al.,14

who use a validated measure of perceived discrimination. Incontrast, however, Taylor et al.15 found evidence for per-ceived discrimination in breast cancer incidence. It is possiblethat our measure of discrimination reflects known difficultiesassessing perceived discrimination or that perceived dis-crimination has greater effects on somatic disease thanmammography use.14,15

Third, we note several observed processes that affected thestudy: loss of funding led to a loss of the original control groupsite prior to baseline as well as two of the intervention sitesduring the study period. The lack of a control group could

introduce important threats to validity, including an inabilityto distinguish study effects from historical trends. Surveil-lance data collected by the CDC during the study periodsuggest that local trends for recent mammography use re-mained stable during the intervention period.39,40 These datashowed that between 2001 and 2006, 2-year mammographyuse among black women �age 50 in Massachusetts was highand remained stable at 90.4%–93.5% during the study period;the lowest prevalence of use occurred in 2006. In contrast,among women aged�50 in our cohort, we observed increasesin mammography uptake within 2 years of study participa-tion, making a historical trend unlikely as an explanation forthe study trends we observed. Furthermore, black women inthe surveillance population had a considerably higher SESthan our high-risk, low-SES cohort; roughly 46% of the sur-veillance population had annual household incomes of<$25,000 compared with 85% of women in our study popu-lation. Nonetheless, by 2 years of study participation, usagerates in our cohort approached that of the surveillance group,suggesting some mitigation of the effect of socioeconomicdisadvantage experienced in our low-SES group with respectto mammography uptake.

The lack of randomized controlled design also introducespotential confounding, and we cannot fully exclude unmea-sured factors that might have affected our results. We wereable to assess the contribution of factors related to access tocare and SES and found that the effect of case managementwas robust to these known correlates of mammography up-take. However, we were unable to assess the impact of healthbeliefs on mammography use. In our cohort, too few womenendorsed concerns related to health beliefs suggested in theliterature (e.g., x-rays cause cancer, physician=medical mis-trust, embarrassment, fear=fatalism) to include these variablesin multivariable models and assess potential confoundingeffects.6,41,42 In addition to potential confounding bias, wenote that our results may not generalize to populations wherehealth beliefs represent larger factors in mammography use.

With respect to our findings on repeat mammography,without a control group, we cannot fully exclude differentialattrition (withdrawal among women without desire formammography) as an explanation for our results. Except forwomen who could not participate because of site withdrawal,the intervention was able to follow and retain 94% of womenwho enrolled, making selection bias due to differential attri-tion unlikely as an explanation for our findings. Further, inthis open cohort design, we were able to use an analyticalstrategy that estimated the odds of completing recommendedmammograms to minimize the effect of site withdrawal in thelongitudinal analysis.

Importantly, there were processes we did not measure thatmay have helped to provide insights into the efficacy of theconceptual approach to intervene to mitigate social issues. Arepeat questionnaire was not administered to determinechange in any social barriers. Also, we could not systemati-cally capture processes of communication between womenand case managers that may have influenced outcomesamong women and between sites. Along these lines, Boboet al.43 suggest intensive contact is needed to reach womenwith abnormal tests. In the National Breast and CervicalCancer Early Detection Program, study staff attempted tolocate and contact women with abnormal results; 93% of thecohort was reached after 7.2 contact attempts per woman,


consisting of letters and phone calls.43 Future qualitative workshould describe interpersonal dynamics and successful tech-niques from the perspectives of case managers to inform bestpractices in navigator or case management interventions.

These limitations considered, the study has severalstrengths. The demonstration project was embedded in aninfrastructure provided by an active community coalition. Inthis context, the study used a conceptual approach developedthrough community participation as well as academic litera-ture, which led to the identification of a novel correlate ofmammography uptake, namely, having trouble with afford-ing housing. Second, the evaluation was able to observe theimpact of case management on different stages of thescreening process and find separate predictors of each of thesesteps. We find case management is helpful, particularly ininitiating or repeating a single mammogram in our cohort.However, insurance type and factors related to clinic site ofenrollment were important for regular repeated longitudinaluse and for following up abnormal results. The effect of in-surance coverage on mammography utilization has been welldocumented in the literature.44 Studies are underway tounderstand the impact of healthcare reform on preventiveservices in Massachusetts, where healthcare financing forlow-income women, previously supported through free carepools at clinics and community health centers, will now besupported primarily through individually purchased insur-ance plans, with subsidies provided for those with incomesbelow the federal poverty level. As of September 2008, statefunding to pay for screening mammograms, previously pro-vided through Massachusetts state-administered Women’sHealth Network funds, will in part be used to fund patientnavigators to help support women continue screening. Ourstudy suggests that in addition to navigation, this populationof black women was helped by tailored case managementthat addressed participant-identified social concerns, guidedby a social assessment tool. Our findings recommend socialassessment as an important step for navigation processesthat attempt to initiate screening practices among women ofAfrican descent. Moreover, to increase repeat mammographyuse and follow-up of abnormal results, we find clinic levelinfrastructure and unidentified site level factors need to beaddressed to ensure that women at community sites who getcare from mobile vans and other off-site venues are wellscreened. Further quality assurance work and investigationwill be necessary to discover infrastructure needs and assets atthese sites that can be building points for improving utiliza-tion. To date, few data assess the impact of case managers orlay health educators along the continuum of mammographyscreening and follow-up in quasiexperimental or randomizedsettings.23,45–47 Future research should determine the effi-ciency and cost of addressing social determinants of health asa case management strategy in a controlled, randomizedsetting.27 In our intervention, in order to serve up to 200women per site, one case manager and one on-site adminis-trator were employed. A formal study of the cost efficiency ofcase management as a strategy for increasing mammographyuptake was beyond the scope of this project.

Conclusions and Action Steps

Our study finds that a social determinants of health ap-proach to case management appears to improve uptake of

mammography, whereas site level factors and insurancecoverage were stronger correlates of repeat mammographyuse and follow-up of abnormal results. Policy implications ofour results suggest further investigation into intervening atthe level of social determinants of health, such as housingconcerns, to promote mammography use. Additionally, de-veloping process measurement methods for tracking socialdeterminants of health interventions may improve medicaland public health practice. Ultimately, the demonstrationproject showed that strategies to identify and mitigate theimpact of social determinants of breast cancer screening canbe successfully employed in primary care settings. Im-portantly, the REACH Coalition succeeded in mobilizingcommunity assets to provide solutions to this priority issuefor women of African descent.

Acknowledgments

We are grateful to Dr. Barbara Ferrer, Executive Director ofthe Boston Public Health Commission, and Ms. WandaMcClain, Executive Director of the Center for CommunityHealth and Health Equity at the Brigham and Women’sHospital, for their kind support in critiquing the presenta-tions of the study findings. We are especially grateful to theREACH case managers, health center directors, and theREACH site administrators for their tireless work providingdirect service to study participants and their feedback duringthe dissemination of the work. We thank Ms. Angela Hall-Jones, co-chair of the REACH Coalition, and Ms. LorraineHector-Watkins, REACH steering committee member, fortheir thoughtful input during the dissemination and writingof this paper. We thank the members of the REACH Coalitionfor providing the impetus and direction for this line ofinquiry. We thank Ms. Amanda Mitchell at the Center forCommunity Health and Health Equity for her administrativeassistance in the preparation of this publication.

The study was supported by a REACH 2010 grant from theCDC. Additionally, the Center for Community Health andHealth Equity and The 2006 Miles and Eleanor Shore MinorityFaculty Development Award supported C.R.C. during thewriting of this work. The financial sponsors of the researchhave not contributed to the design, interpretation, or manu-script preparation for this study. Excerpts of this paper werepresented at the biannual meeting of the Minority Women’sHealth Summit, Washington, DC, August 2007.

Disclosure Statement

The authors have no conflicts of interest to report.

References

1. Ries LAG, Melbert D, Krapcho M, et al. SEER Cancer Sta-tistics Review, 1975–2004. Bethesda, MD: National CancerInstitute, 2007. Available at seer.cancer.gov=csr=1975_2004=Accessed May 7, 2008.

2. Newman LA. Breast cancer in African-American women.Oncologist 2005;10:1–14.

3. Curtis E, Quale C, Haggstrom D, Smith-Bindman R. Racialand ethnic differences in breast cancer survival: How much isexplained by screening, tumor severity, biology, treatment,comorbidities, and demographics? Cancer 2008;112:171–180.

4. Masi CM, Blackman DJ, Peek ME. Interventions to enhancebreast cancer screening, diagnosis, and treatment among

688 CLARK ET AL.

racial and ethnic minority women. Med Care Res Rev2007;64(Suppl 5):195S–242S.

5. Use of mammograms among women aged greater orequal to 40 years: United States, 2000–2005. MMWR 2007;56:49–51.

6. Paskett ED, Tatum C, Rushing J, et al. Racial differences inknowledge, attitudes, and cancer screening practices amonga triracial rural population. Cancer 2004;101:2650–2659.

7. Yabroff KR, Breen N, Vernon SW, Meissner HI, FreedmanAN, Ballard-Barbash R. What factors are associated withdiagnostic follow-up after abnormal mammograms? Find-ings from a U.S. National Survey. Cancer Epidemiol Bio-markers Prev 2004;13:723–732.

8. Moy B, Park ER, Feibelmann S, Chiang S, Weissman JS.Barriers to repeat mammography: Cultural perspectives ofAfrican-American, Asian, and Hispanic women. Psy-chooncology 2006;15:623–634.

9. Blanchard K, Colbert JA, Puri D, et al. Mammographicscreening: Patterns of use and estimated impact on breastcarcinoma survival. Cancer 2004;101:495–507.

10. Davis C, Emerson JS, Husaini BA. Breast cancer screeningamong African American women: Adherence to currentrecommendations. J Health Care Poor Underserved 2005;16:308–314.

11. Blackman DJ, Masi CM. Racial and ethnic disparities inbreast cancer mortality: Are we doing enough to address theroot causes? J Clin Oncol 2006;24:2170–2178.

12. Peek ME, Han JH. Disparities in screening mammography.Current status, interventions and implications. J Gen InternMed 2004;19:184–194.

13. Litaker D, Tomolo A. Association of contextual factors andbreast cancer screening: Finding new targets to promoteearly detection. J Womens Health 2007;16:36–45.

14. Dailey AB, Kasl SV, Holford TR, Jones BA. Perceived racialdiscrimination and nonadherence to screening mammogra-phy guidelines: Results from the Race Differences in theScreening Mammography Process study. Am J Epidemiol2007;165:1287–1295.

15. Taylor TR, Williams CD, Makambi KH, et al. Racial dis-crimination and breast cancer incidence in U.S. blackwomen: The Black Women’s Health Study. Am J Epidemiol2007;166:46–54.

16. Poon EG, Haas JS, Louise PA, et al. Communication factorsin the follow-up of abnormal mammograms. J Gen InternMed 2004;19:316–323.

17. Crane LA, Kaplan CP, Bastani R, Scrimshaw SC. Determi-nants of adherence among health department patients re-ferred for a mammogram. Women Health 1996;24:43–64.

18. Oluwole SF, Ali AO, Adu A, et al. Impact of a cancerscreening program on breast cancer stage at diagnosis in amedically underserved urban community. J Am Coll Surg2003;196:180–188.

19. Purc-Stephenson RJ, Gorey KM. Lower adherence toscreening mammography guidelines among ethnic minoritywomen in America: A meta-analytic review. Prev Med 2008;46:479–488.

20. McAlearney AS, Reeves KW, Tatum C, Paskett ED. Cost as abarrier to screening mammography among underservedwomen. Ethn Health 2007;12:189–203.

21. Barrett K, Legg J. Demographic and health factors associatedwith mammography utilization. Am J Health Promot2005;19:401–405.

22. Increasing Cancer Screening. Guide to Community Pre-ventive Services Website. Centers for Disease Control and

Prevention 2007. Available at www.thecommunityguide.org=cancer=screening=default.htm Last Accessed May 7, 2008.

23. Battaglia TA, Roloff K, Posner MA, Freund KM. Improvingfollow-up to abnormal breast cancer screening in an urbanpopulation. A patient navigation intervention. Cancer2007;109(Suppl 2):359–367.

24. Tangka FK, Dalaker J, Chattopadhyay SK, et al. Meeting themammography screening needs of underserved women:The performance of the National Breast and Cervical CancerEarly Detection Program in 2002–2003 (United States).Cancer Causes Control 2006;17:1145–1154.

25. Taplin SH, Barlow WE, Ludman E, et al. Testing reminderand motivational telephone calls to increase screeningmammography: A randomized study. J Natl Cancer Inst2000;92:233–242.

26. Earp JA, Eng E, O’Malley MS, et al. Increasing use ofmammography among older, rural African Americanwomen: Results from a community trial. Am J Public Health2002;92:646–654.

27. Bastani R, Yabroff KR, Myers RE, Glenn B. Interventions toimprove follow-up of abnormal findings in cancer screening.Cancer 2004;101(Suppl 5):1188–1200.

28. Ell K, Vourlekis B, Nissly J, et al. Integrating mental healthscreening and abnormal cancer screening follow-up: An in-tervention to reach low-income women. Community MentHealth J 2002;38:311–325.

29. Legler J, Meissner HI, Coyne C, Breen N, Chollette V, RimerBK. The effectiveness of interventions to promote mam-mography among women with historically lower ratesof screening. Cancer Epidemiol Biomarkers Prev 2002;11:59–71.

30. Yabroff KR, Mandelblatt JS. Interventions targeted towardpatients to increase mammography use. Cancer EpidemiolBiomarkers Prev 1999;8:749–757.

31. Tucker P, Liao Y, Giles W, Liburd L. The REACH 2010 logicmodel: An illustration of expected performance. PrevChronic Dis, 2006. Available at www.cdc.gov=pcd=issues=2006=jan=05_0131.htm

32. Israel BA, Schulz AJ, Parker EA, Becker AB. Community-based participatory research: Policy recommendations forpromoting a partnership approach in health research. EducHealth (Abingdon) 2001;14:182–197.

33. Bigby J, Ko LK, Johnson N, David MM, Ferrer B. A com-munity approach to addressing excess breast and cervicalcancer mortality among women of African descent in Bos-ton. Public Health Rep 2003;118:338–347.

34. Ferrer B. National Association of County & City HealthOfficials Model Health Practices Database. Maternal childhealth: Women’s health questionnaire. www.archive.naccho.org=modelpractices=Result.asp?PracticeID¼43 AccessedMay 7, 2008.

35. Ferrer B, Bigby J, Gottlieb B, et al. Risk assessment tool toimprove birth outcomes in Boston [Abstract]. 128th AnnualMeeting of the APHA, 2000.

36. Eberl MM, Fox CH, Edge SB, Carter CA, Mahoney MC. BI-RADS classification for management of abnormal mammo-grams. J Am Board Fam Med 2006;19:161–164.

37. Williams DA. Extra-binomial variation in logistic linearmodels. Appl Statistics 1982;31:144–148.

38. Smith RA, Cokkinides V, Eyre HJ. Cancer screening in theUnited States, 2007: A review of current guidelines, prac-tices, and prospects. CA Cancer J Clin 2007;57:90–104.

39. Liao Y, Tucker P, Okoro CA, Giles WH, Mokdad AH, HarrisVB. REACH 2010 surveillance for health status in minority


communities—United States, 2001–2002. MMWR SurveillSumm 2004;53:1–36.

40. Abt Associates Inc. REACH 2010 risk factor survey. Year 5sampling and weighting report: Boston Public HealthCommission (Site 2), 2006. Available from National Centerfor Chronic Disease Prevention and Health Promotion,Centers for Disease Control & Prevention.

41. Fernandez ME, Palmer RC, Leong-Wu CA. Repeat mam-mography screening among low-income and minority women:A qualitative study. Cancer Control 2005;12(Suppl 2):77–83.

42. O’Malley AS, Sheppard VB, Schwartz M, Mandelblatt J. Therole of trust in use of preventive services among low-incomeAfrican-American women. Prev Med 2004;38:777–785.

43. Bobo JK, Shapiro JA, Brustrom J. Efforts to locate low-income women for a study on mammography rescreening:Implications for public health practice. J Community Health2006;31:249–261.

44. Trivedi AN, Rakowski W, Ayanian JZ. Effect of cost sharingon screening mammography in Medicare health plans.N Engl J Med 2008;358:375–383.

45. Powell ME, Carter V, Bonsi E, et al. Increasing mammog-raphy screening among African American women in ruralareas. J Health Care Poor Underserved 2005;16(4 SupplA):11–21.

46. Lewin SA, Dick J, Pond P, et al. Lay health workers in pri-mary and community health care. Cochrane Database SystRev 2005;(1):CD004015.

47. Ell K, Vourlekis B, Lee PJ, Xie B. Patient navigation and casemanagement following an abnormal mammogram: A ran-domized clinical trial. Prev Med 2007;44:26–33.

Address reprint requests to:Cheryl R. Clark, M.D., Sc.D.

Brigham and Women’s HospitalDivision of General Medicine and Primary Care

Center for Community Health and Health Equity1620 Tremont Street

Boston, MA 02120

E-mail: [email protected]

690 CLARK ET AL.

addressing social determinants of health to improve access to early breast cancer detection: results...

Documents