Do Community Health Worker Interventions Improve Rates of ... · Science of Cancer Health...

Review

Do Community Health Worker Interventions ImproveRates of Screening Mammography in the United States?A Systematic Review

Kristen J. Wells1,2, John S. Luque2,3, Branko Miladinovic1, Natalia Vargas2, Yasmin Asvat1,2,Richard G. Roetzheim1,2, and Ambuj Kumar1,2

AbstractBackground: Community health workers (CHW) are lay individuals who are trained to serve as liaisons

between members of their communities and health care providers and services.

Methods: A systematic review was conducted to synthesize evidence from all prospective controlled

studies on effectiveness of CHW programs in improving screening mammography rates. Studies reported in

English and conducted in the United States were included if they: (i) evaluated a CHW intervention designed

to increase screening mammography rates in women 40 years of age or older without a history of breast

cancer; (ii) were a randomized controlled trial (RCT), case–controlled study, or quasi-experimental study; and

(iii) evaluated a CHW intervention outside of a hospital setting.

Results: Participation in a CHW intervention was associated with a statistically significant increase in

receipt of screening mammography [risk ratio (RR): 1.06 (favoring intervention); 95% CI: 1.02–1.11, P¼ 0.003].

The effect remainedwhen pooled data from only RCTswere included inmeta-analysis (RR: 1.07; 95%CI: 1.03–

1.12, P ¼ 0.0005) but was not present using pooled data from only quasi-experimental studies (RR: 1.03; 95%

CI: 0.89–1.18, P ¼ 0.71). In RCTs, participants recruited from medical settings (RR: 1.41; 95% CI: 1.09–1.82,

P¼ 0.008), programs conducted in urban settings (RR: 1.23; 95%CI: 1.09, 1.39, P¼ 0.001), and programswhere

CHWs were matched to intervention participants on race or ethnicity (RR: 1.58, 95% CI: 1.29–1.93, P¼ 0.0001)

showed stronger effects on increasing mammography screening rates.

Conclusions: CHW interventions are effective for increasing screening mammography in certain settings

and populations.

Impact: CHW interventions are especially associated with improvements in rate of screening mammo-

graphy in medical settings, urban settings, and in participants who are racially or ethnically concordant with

the CHW. Cancer Epidemiol Biomarkers Prev; 20(8); 1580–98. ’2011 AACR.

Introduction

In 2010, an estimated 207,090 women were diagnosedwith breast cancer (BC) in the United States (1). The U.S.Preventive Services Task Force recommends screeningmammography every 2 years for women aged 50 to 74years and recommends screening for women aged 40 to49 years be based on an individual’s risk factors (2). TheAmerican Cancer Society recommends yearly mammo-

graphy beginning at age of 40 (3). Early detection of BC isassociated with reductions in mortality and improve-ments in survival rates (4, 5).

There are significant racial and socioeconomic dispa-rities in BC mortality, survival rates, and cancer stage atdiagnosis in the United States (6–10). Women who areless likely to adhere to screening mammography guide-lines include those who belong to ethnic or racial mino-rities; lack comprehensive health insurance or a usualsource of medical care; are non-English speakers; areimmigrants; live in rural areas; or are socioeconomicallydisadvantaged (11–19).

One model of BC screening promotion that has beenimplemented and evaluated frequently is the commu-nity health worker (CHW) model. CHWs are lay indi-viduals trained to serve as liaisons between membersof their communities and health care providers andservices (20). Historically, CHWs serve low-income,medically underserved, racial/ethnic minority, andhard-to-reach populations (21).

Authors' Affiliations: 1University of South Florida; 2Moffitt Cancer Center,Tampa, Florida; and 3Georgia Southern University, Statesboro, Georgia

Note: This study was presented at the Third AACR Conference on TheScience of Cancer Health Disparities in Racial/Ethnic Minorities and theMedically Underserved on October 1, 2010.

Corresponding Author: Kristen J. Wells, University of South Florida,12901 Bruce B. Downs Blvd., MDC 27, Tampa, FL 33612. Phone: 813-396-2612; Fax: 813-905-8909; E-mail: [email protected]

doi: 10.1158/1055-9965.EPI-11-0276

’2011 American Association for Cancer Research.

CancerEpidemiology,

Biomarkers& Prevention

Cancer Epidemiol Biomarkers Prev; 20(8) August 20111580

on November 6, 2020. © 2011 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

Published OnlineFirst June 8, 2011; DOI: 10.1158/1055-9965.EPI-11-0276

http://cebp.aacrjournals.org/

Previous systematic reviews on the effectiveness ofCHWs on increasing screening mammography have sev-eral limitations. These reviews combined mammographywith other health behaviors (22), combined CHW inter-ventions with other mammography-enhancing interven-tion strategies (23–25), and combined between-groupstudies (comparing CHW interventions to comparisongroup) and within-group studies (comparing mammo-graphy rates over time in a CHW intervention group; ref.26), making it difficult to draw conclusions about effec-tiveness of CHW interventions in improving screeningmammography. To conclusively assess effect of CHWinterventions in increasing screening mammographyrates, we conducted a systematic review. The objectivewas to synthesize and critically appraise available evi-dence on effectiveness of CHW interventions in increas-ing screening mammography rates compared with acontrol group in any population.

Materials and Methods

A study protocol was developed clearly outlininga priori all stages of the systematic review process.

Search strategyWe conducted a comprehensive search of CINAHL,

Medline, PsychInfo, and Web of Science databases foryears 1980 through January 31, 2008. A broad searchstrategy was used to identify relevant articles andincluded 21 terms for CHW plus 5 BC terms to captureall studies evaluating CHW interventions to improvemammography screening. These terms included bothMeSH terms and other identified key words.Each search provided citations that were downloaded

into an Endnote database (27). After duplicate citationswere removed, all titles and abstracts were reviewed by 2study authors independently for their eligibility for inclu-sion. If a decision on inclusion was not made on initialreview, the full text article was obtained. Reference lists ofall eligible articles were also reviewed, and authors of thepresent article were asked to provide any additionalpublications not captured by the search.

Inclusion criteriaStudies published in English and conducted in the

United States were eligible for inclusion if they: (i) eval-uated a CHW intervention designed to increase screeningmammography rates in women 40 years of age or olderwithout a history of BC; (ii) were a randomized controlledtrial (RCT), case–controlled study, or quasi-experimentalstudy; and (iii) were studies in which the CHW interven-tion was delivered outside of a hospital. Studies werelimited to those conducted in the United States because ofvast differences between health systems in the UnitedStates and those in other English-speaking countries.Because CHW interventions are known by many terms,a definition of CHW was created to differentiate CHWstudies from other interventions: "any health care worker

who is involved with carrying out the intervention butwho does not necessarily have formal professional orparaprofessional education." This definition is similar toother definitions in previous reviews evaluating CHWresearch (22, 28).

Data abstractionThe primary outcome variable, receipt of mammogra-

phy, was abstracted from each article as it was reportedprior to the CHW intervention (baseline) and followingintervention (follow-up) for both participants whoreceived CHW as well as comparison groups. Data wereextracted for prespecified sensitivity analyses on samplesource (medical or community setting and urban orrural setting), description of intervention, componentsof intervention, and characteristics of CHWs. The follow-ing data were collected to assess methodologic quality ofresearch reported in each publication: generation of ran-domization sequence, matching of control to interventionparticipants in quasi-experimental studies, and use ofintent-to-treat (ITT) analysis.

Data extraction was conducted by 2 reviewers inde-pendently using a standardized data abstraction form.Any disagreements in data abstraction were resolved byconsensus in collaboration with a third author. Data wereentered into separate SPSS databases (29). Using Graph-Pad Software (30), Kappa coefficients were calculated toassess agreement between the 2 raters on 6 study vari-ables used in sensitivity analysis. Kappa coefficientscalculated for 6 variables indicated agreement rangedfrom "moderate" (0.410) to "almost perfect agreement"(0.885; ref. 31).

Statistical analysisDichotomous data (i.e., number of participants who

did and did not receive screening mammography in bothintervention and comparison groups at follow-up) wereused to calculate a risk ratio (RR), and summary results(RR) from each study were pooled under a random-effects model. A formal statistical test for heterogeneityusing an I2 test was conducted (32). The main studyanalyses were conducted using Review Manager (32).

Because cluster or group randomized trials (CRT) arefrequently used in research on efficacy of CHWs, weexplored the impact of various imputed values of intra-class correlation (ICC) on pooled estimates. ICC is the"similarity" of individuals within clusters, such asclinics or communities. A CONSORT guideline pertain-ing to CRTs explicitly recommends statistical adjust-ments for cluster randomization be used in powercalculations and analysis of primary outcomes, withreporting of ICC (33). Not accounting for clustering inanalysis of CRTs creates a "unit of analysis error" whenCRTs are combined with trials that randomized indivi-duals in a meta-analysis. Correcting for clusteringinflates variance of point estimates (RRs or ORs) inindividual CRTs, giving less weight to these studiesin random-effects meta-analysis.

Community Health Workers and Mammography

www.aacrjournals.org Cancer Epidemiol Biomarkers Prev; 20(8) August 2011 1581




A recent meta-analysis of enhanced care for depres-sion concluded that CRTs produced similar results toindividually randomized trials. However, the analysiswas based on relatively small values of ICC (ICC ¼ 0.02and 0.05; ref. 34). In contrast, a recent study reportedICC estimates in screening mammography CRTs to beas high as 0.2166 (35). A lack of adjustment for cluster-ing tends to inflate treatment effects (36). We alsoexplored the effect of cluster imbalance on pooledresults.

To conduct these analyses, similarities between indi-viduals within a cluster are measured by ICC. Given theaverage cluster size m, the design effect (DE), defined byDE ¼ 1 þ (m � 1) ICC, measures effect on variance of anestimate of treatment effect attributed to clustering. TheICC-adjusted random-effects meta-analysis was con-

ducted in the Bayesian setting due to ease of adjustingICCs for an empirical prior distribution. Using Winbugsversion 1.4.1. (37, 38), adjustment of cluster size imbal-ance was applied to DE minimum variance weightscorrected estimates, based on the Pareto principle inwhich 80% of participants belong to 20% of clusters(39). For each outcome, 5,000 burn-in simulations wereused, with the additional 45,000 simulations to obtain ORestimates. A flat uniform distribution on 0 to 1 range hasbeen recommended for priors of ICCs in case of absenceof prior knowledge (40). However, on the basis of a recentstudy of ICC estimates for cancer screening outcomes(35), unadjusted ICC estimates for mammographyscreening rates ranged from 0.0009 to 0.2166, so weadopted a conservative but informative empiricaluniform prior on (0, 0.5), for imputed values of ICC

Studies excluded Duplicate citations (n = 110)

Not relevant (n = 61)

Articles identified from database 1980–October 2008

(Total n = 265) PubMed/Medline (n = 115) PsycINFO (n = 17) CINAHL (n = 41) Web of Science (n = 92)

Potentially relevant studies identified andscreened for retrieval

(n = 128)

Articles identified in reference search or by

study authors

(Total n = 34)

Met inclusion criteria for systematic review

(n = 24)

Did not meet inclusion criteria (n = 104)

-30 did not describe CHW intervention -18 did not focus on BC screening - 8 had no comparison group - 33 did not report rates of screening mammography at follow-up - 10 were review articles - 5 reported studies outside of United States

Met inclusion criteria for meta-analysis

(n = 18)

Figure 1. Flow diagram ofliterature review process foridentifying studies evaluating useof community health workers toimprove screening mammographyrates.

Wells et al.

Cancer Epidemiol Biomarkers Prev; 20(8) August 2011 Cancer Epidemiology, Biomarkers & Prevention1582




Tab

le1.

Sum

maryof

commun

ityhe

alth

worke

rinterven

tions

forBC

screen

ingbytarget

grou

p

Firstau

thor(ref.)

Target

population

CHW

number

and

charac

teristics

Leng

th;type

oftraining

Interven

tiondes

cription

Controld

escription

African

American

Earp(46)

Rural,African

American

wom

en50

þye

ars

170vo

luntee

rlay

health

adviso

rs3–

5se

ssions

,10

–12

hours

instruction;

role

playing

,BC,an

dsc

reen

ing

practices

.

2co

mmun

ityou

trea

chac

tivities

per

mon

thUC

inco

mparison

coun

ties

4pa

idco

mmun

ityou

trea

chsp

ecialists

(indige

nous

commun

itylead

ers)

Layhe

alth

adviso

rsco

nduc

ted2wee

kly

individua

lses

sion

s

Erw

in(47)

Rural,African

American

wom

en,Mississippi

River

Delta,AR

7African

American

BC

survivors

Training

provided

;interviewing

strategies

,brea

sthe

alth

educ

ation

Chu

rch-bas

edgrou

pab

outea

rlydetec

tion

andsc

reen

ing

UC

inco

mparison

coun

ties

Free

mam

mog

ram

vouc

hers

Pas

kett(55)

Predom

inatelyAfrican

-American

wom

en40

þye

ars,

low-inc

ome

hous

ing,

Winston

–Salem

,NC

(interven

tion)

and

Green

sboro,

NC

(con

trol)

ND

ND

Educ

ationa

lses

sion

sbylayhe

alth

educ

ators

Chu

rchprogram

cond

uctedbylay

health

educ

ator

UC

inco

mparison

city

Sun

g(60)

Low-inc

ome,

inne

rcity

African

American

wom

en

Layhe

alth

worke

rsrecruitedfrom

Nationa

lBlack

Wom

en's

Hea

lthProject

10wee

ks;interviewing

andhe

alth

educ

ation

Twoho

mevisits,

1boo

ster

session.

Educ

ationab

outbreas

tan

dce

rvical

canc

er,

Pap

test

andbreas

tex

amvideo

Can

cersc

reen

ing

educ

ationa

lmaterials

Rep

roduc

tivehe

alth

educ

ation

Wes

t(63)

Rural,low-inc

ome

African

American

wom

en50

–80

Indigen

ousAfrican

American

female

health

care

worke

rs.

Training

provided

;se

mi-structured

interview's

conten

t,co

unse

lingstyle

Perso

nalized

reminder

lettersfollo

wed

byintens

ivepho

neco

unse

ling

Elig

ible

forano

-cos

tmam

mog

ram;tailo

red

letter

desc

ribingris

kforBC

Pho

neca

llinclud

edse

mi-structured

interview

eliciting

barrie

rs,facilitators

tomam

mog

raphy

Problem

solvingto

overco

meba

rriers

Provided

phon

enu

mbers

forsc

heduling

mam

mog

raphy

(Con

tinue

don

thefollo

wingpag

e)






Tab

le1.

Sum

maryof

commun

ityhe

alth

worke

rinterven

tions

forBC

screen

ingbytarget

grou

p(Con

t'd)

Firstau

thor(ref.)

Target

population

CHW

number

and

charac

teristics

Leng

th;type

oftraining

Interven

tiondes

criptio

nControld

escriptio

n

Zhu

(64)

SingleAfrican

American

wom

en65

þ,livingin

pub

licho

using

African

American

wom

enfrom

same

hous

ingco

mplex

Four

3-ho

urse

ssions

;BC,

screen

ingpractices

,pos

sible

barriers

Hom

e-bas

edBC

screen

inged

ucation

Taug

htpartic

ipan

tsto

overco

mebarrie

rsBroch

ures

Pub

licho

usingco

mplexe

s,no

interven

tion

Hispan

ic/Latina

Fernan

dez

(48)

Hispan

icfarm

worke

rsLa

yhe

alth

worke

rsLa

yhe

alth

worke

rsgive

ntraining

curriculum

includ

ing

teac

hing

guide(12lesson

s)

Individua

land

grou

pinterven

tionse

ssions

with

layhe

alth

worke

r

Com

parison

clinics,

nointerven

tion

Educ

ationse

ssions

includ

edvideo

s,flip

charts,discu

ssions

ofbrea

stan

dce

rvical

screen

ing

Nav

arro

(53)

Low-inc

omeLa

tinas

,San

Diego

,CA

36co

nsejeras

Training

provided

;co

nduc

ting

smallg

roup

sessions

onhe

alth

topics

12-w

eek,

90-m

inute

grou

pse

ssionprogram

12-w

eekCom

mun

ityLiving

Skills

program

Con

sejerastaug

htca

ncer

prev

entio

nwith

cultu

rally-appropria

tematerials.

Nav

arro

(54)

Latin

as,San

Diego

,CA

36co

nsejeras

Training

provided

;follo

wing

cons

ejeraman

ualto

cond

uctwee

klyed

ucationa

lgrou

pse

ssions

12-w

eekgrou

pse

ssions

onbreas

tan

dce

rvical

canc

erea

rlydetec

tion

12-w

eekCom

mun

ityLiving

Skills

program

Nav

arro

(51)

Latin

as,San

Diego

,CA

36co

nsejeras

Training

provided

;co

nduc

ting

smallg

roup

sessions

onbreas

tan

dce

rvical

canc

ersc

reen

ing

12-w

eekgrou

pca

ncer

screen

inginterven

tion

follo

winged

ucationa

lmaterials

12-w

eekCom

mun

ityLiving

Skills

program

Nav

arro

(52)

Low-inc

ome,

low

accu

lturated

Latin

as,San

Diego

,CA

17co

nsejeras

Five

2-ho

urse

ssions

;14

program

sess

ions

,recruitm

entstrategies

,role

playing

tolead

sessions

Twelve

90-m

inutewee

kly

grou

pse

ssions

and

2mon

thly

sessions

focu

sing

onbrea

stan

dce

rvical

canc

er

Friend

san

d/orfamily

(learning

partners)

who

rece

ived

inform

ation

from

clas

spartic

ipan

ts

Sau

aia(57)

Latin

as,CO

"Pee

rco

unse

lors,"

4Catho

licch

urch

es,Den

ver,

CO

(Promotoras

)

Training

provided

Bim

onthly

mee

tings

after

mas

san

dothe

rch

urch

even

ts

209Catho

licch

urch

esse

nt:

(Con

tinue

don

thefollo

wingpag

e)

Wells et al.





Tab

le1.

Sum

maryof

commun

ityhe

alth

worke

rinterven

tions

forBC

screen

ingbytarget

grou

p(Con

t'd)

Firstau

thor(ref.)

Target

population

CHW

number

and

charac

teristics

Leng

th;type

oftraining

Interven

tiondes

criptio

nControld

escription

Res

pected

lead

erdelivered

homilies

addressingbrea

sthe

alth

atleas

ttw

ice

atea

chch

urch

1–3ho

me-bas

edhe

alth

grou

psper

church

Letter

desc

ribingproject

Biling

ualp

rintedmaterials

displayun

itsh

ortmes

sage

sforde

livery

atpulpitan

dch

urch

bulletin

New

sletter

Sua

rez(59)

Low-inc

omeMex

ican

American

wom

en40

þ,Te

xas

Mex

ican

-American

voluntee

rs,

Span

ishsp

eake

rs

ND

Verba

land

newsletter

canc

er-scree

ning

educ

ation

Com

parison

commun

ity,

nointerven

tion

Welsh

(62)

Latin

as,CO

"Pee

rco

unse

lors",

4Catho

licch

urch

es,

Den

ver,CO

Training

provided

;stan

dardized

curriculum

Mon

thly

visits

toea

chch

urch

Promotoras

approac

hed

pee

rsafterSun

day

mas

sesan

dduring

church

fairs

,othe

rch

urch

relatedac

tivities

Promotoras

facilitated

home-bas

edplatic

as(hea

lthgrou

ps)

abou

tbreas

the

alth

209ch

urch

esmailed:

introd

uctory

letter

NCIBC

screen

ing

educ

ationa

lmaterials

displayun

itmes

sage

sto

bede

livered

atpu

lpitor

church

bulletin

s

Asian

American

/Pac

ificIsland

erBird

(43)

Vietnam

ese-American

wom

en18

þ,Califo

rnia

16indigen

ouslay

health

worke

rne

ighb

orho

odlead

ers

68ne

ighb

orho

odas

sistan

ts

Training

provided

;de

liveryof

prev

entio

ned

ucation

Provided

training

man

ual

Small-grou

pho

me-bas

eded

ucationa

lses

sion

sFo

cuse

don

gene

ral

preve

ntion,

routine

chec

kups,

breas

tan

dce

rvical

canc

ersc

reen

ing

Flip

chartan

dfacilitated

discu

ssion

Com

parison

city,no

interven

tion

Gotay

(49)

NativeHaw

aiianwom

en,

18þ,

Oah

u,HI

Native-Haw

aiianlay

health

educ

ators

Training

provided

;brea

st,ce

rvical

canc

ersc

reen

ing.

Small-grou

p,trad

ition

alHaw

aiian"talk-story"

metho

ds

Com

parison

area

,no

interven

tion

(Con

tinue

don

thefollo

wingpag

e)






Tab

le1.

Sum

maryof

commun

ityhe

alth

worke

rinterven

tions

forBC

screen

ingbytarget

grou

p(Con

t'd)

Firstau

thor(ref.)

Target

population

CHW

number

and

charac

teristics

Leng

th;type

oftraining

Interven

tiondes

cription

Controld

escription

Free

mam

mog

ram

and

Pap

test

vouc

hers

Aud

iovisu

alaidsus

edOther

Populations

Allen(41)

Wom

enwho

workin

siteswith

Service

Employe

esInternationa

lUnion

represe

ntation

80femaleem

ploye

epee

rhe

alth

adviso

rs16

hours;

canc

ersc

reen

inggu

idelines

,ep

idem

iology

,ea

rlydetec

tion

metho

ds,

commun

ityreso

urce

s

16-m

onth

prog

ram,

6sm

allg

roup

discu

ssions

,role

mod

elingindividua

lou

trea

ch,co

unse

ling,

social

supportfocu

sed

onsc

reen

ing.

Works

hopat

stud

yco

nclusion

2works

ite-w

ideca

mpaign

sAnd

erse

n(42)

Wom

en50

–80

,rural

Was

hing

ton

Wom

enfrom

partic

ipating

commun

ities

Training

provided

Teleph

oneinterven

tion

usingba

rrier-sp

ecific

coun

selingto

promote

mam

mog

raphy

Con

trol

commun

ities

,no

interven

tion

Broch

ures

Calle

(44)

African

American

andWhite

wom

en40

þ

80ACSvo

luntee

rpee

red

ucators,

Jack

sonv

ille,

Orla

ndo,

FL

Half-day

;mam

mog

raphy

andmam

mog

raphy

facilities,

breas

the

alth

guidelines

,interven

tionproce

ss,

interven

tionpractice

sess

ions

,BC

fact

shee

t,reso

urce

guidefor

mam

mog

raphy

centers

Pho

neinterven

tion

emph

asized

importanc

eof

regu

larmam

mog

rams

Set

andco

nfirm

eddate

appointmen

twou

ldbe

sche

duled

andco

mpleted.

Nointerven

tion

Dua

n(45)

African

-American

,La

tina,

andWhite

wom

en50

–80

,ch

urch

esin

Los

Ang

eles

Cou

nty,

CA

Part-tim

epee

rco

unse

lors

ND

Pee

rco

unse

lors

prov

ided

telepho

neba

rrier-focu

sed

mam

mog

raphy

coun

seling

annu

ally

for2ye

ars

Chu

rche

s,no

interven

tion

Wom

enprov

ided

inform

ationab

outris

kstatus

,BC

preva

lenc

erates,

enco

urag

edto

askphy

sician

forreferral

(Con

tinue

don

thefollo

wingpag

e)

Wells et al.





Tab

le1.

Sum

maryof

commun

ityhe

alth

worke

rinterven

tions

forBC

screen

ingbytarget

grou

p(Con

t'd)

Firstau

thor(ref.)

Target

population

CHW

number

and

charac

teristics

Leng

th;type

oftraining

Interven

tiondes

cription

Controld

escription

Margo

lis(50)

African

American

,NativeAmerican

,an

dWhite

wom

en40

þ,no

n–prim

ary-ca

reou

tpatient

clinics

Low-inc

ome,

lay

femalese

nior

aides

(layhe

alth

adviso

rs)

One

mon

th;pee

red

ucation

tech

niqu

es,breas

the

alth,

commun

icationsk

ills,

cultu

rald

iversity,prin

ciples

ofrand

omization,

adhe

renc

eto

rese

arch

protoco

l

Appointmen

tssc

hedu

led

with

femalenu

rse

practition

erforthos

edue

forsc

reen

ing

Wom

enwho

declined

screen

ingen

courag

edto

follo

wup

with

health

care

provider

Wom

enwho

were

up-to-da

tewere

enco

urag

edto

get

regu

larsc

reen

ingan

dofferedmailedreminder

UC

Pas

kett(56)

African

-American

,NativeAmerican

,an

dWhite

wom

enfrom

rural

area

san

dlow-inc

ome

bac

kgroun

d.

2NativeAmerican

andAfrican

American

lay

health

adviso

rs

1wee

kplusfollo

w-up

sess

ions

;brea

sthe

alth,

practiceon

breas

tmod

els,

reso

urce

man

ual,prac

tice

interven

tionse

ssions

3in-perso

nindividua

lvisits,follo

w-uppho

neca

lls,mailings

over

9–12

mon

ths.

Educ

ationa

lmaterials

abou

tca

ncer

risk,

overco

mingba

rriers

tomam

mog

raphy

Discu

ssionof

mam

mog

raphy

,BC,breas

tse

lf-ex

am,

sche

dulingmam

mog

raphy

2pos

tcardmailings

addressingwom

en's

read

ines

sto

chan

ge

Letter

andNCIbroch

ure

abou

tce

rvical

canc

ersc

reen

ing;

afterfollo

wup

,letter

andNCI

broch

ureab

out

mam

mog

raphy

Slater(58)

Low-inc

omewom

en,

pub

licho

using

ACSan

dresiden

tvo

luntee

rsTraining

prov

ided

;vo

luntee

rsprovided

interven

tion

scrip

tsan

dprotoco

ls.

60minute"Frie

ndto

Friend

"program

with

health

professiona

lsp

eake

r,sm

allg

roup

discu

ssions

,as

sistan

cein

obtainingmam

mog

ram,

mam

mog

ram

reminder

Delay

edinterven

tion

Free

mam

mog

rams

Ass

istanc

ewith

appointmen

tsc

hedu

ling

Free

tran

sportatio

n

(Con

tinue

don

thefollo

wingpag

e)






(ICC ¼ 0.05, 0.1, 0.2). The pooled results are reported asORs with corresponding 95% credibility intervals.

Results

Study identificationThe initial database search yielded 265 articles (Fig. 1),

of which 24 met inclusion criteria (41–64). Of the 24articles included in the systematic review (Table 1), 18(75%; refs. 41–47, 50, 52, 55–63) had extractable data.Three articles (41, 42, 62) provided data enabling calcula-tion of multiple effect sizes. One study provided data onmultiple interventions (42), one study reported results in 2different strata based on age (41), and one study reportedresults separately by race and ethnicity (62). Of the 24included studies (Table 1), 14 were (58%) RCTs (41, 42, 44,45, 50, 51, 53, 54, 56, 58, 60, 61, 63, 64) and 10 were quasi-experimental studies (43, 46–49, 52, 55, 57, 59, 62).

Outcomes of CHW intervention on receipt ofscreening mammography

All trials. Most studies (75%) collected data onscreening mammography use via self-report interviewsor surveys (41–47, 49, 51–55, 58–60, 63, 64). Other studiescollected mammography data through chart reviews (56,61), health insurance claims data (57, 62), or through acombination of self-report and chart reviews (50). Themost common time frame in which screening mammo-graphywas collectedwas 1 year (45, 49–51, 54–56, 64) and2 years (41, 46, 49, 50, 55, 57, 59, 61, 62, 64), with 2 studiesreporting time frames in between 1 and 2 years (43, 58).Several studies evaluated the CHW intervention’s effecton lifetime receipt of screeningmammography (43, 47, 49,52, 53).

Of 18 studies with sufficient data to evaluate receipt ofscreening mammography in the meta-analysis, 10 (56%)were RCTs (41, 42, 44, 45, 50, 56, 58, 60, 61, 63) and 8 werequasi-experimental (43, 46, 47, 52, 55, 57, 59, 62). Thepooled RR of obtaining screening mammography basedon 28,836 mammography events (9,342 intervention;19,494 control) was 1.06 (95% CI: 1.02–1.11, P ¼ 0.003),indicating a statistically significant effect of CHW inter-ventions on improving rate of screening mammography(Fig. 2). However, there was a statistically significantheterogeneity among included studies (I2 ¼ 80%; P <0.00001).

Quasi-experimental studies. The pooled RR ofobtaining screening mammography based on 14,677mammography events (2,235 intervention; 12,442 con-trol) in 8 quasi-experimental studies (9 comparisons; refs.43, 46, 47, 52, 55, 57, 59, 62) was 1.03 (95% CI: 0.89–1.18,P ¼ 0.71; Fig. 2), indicating no effect of CHW interven-tions on rate of screening mammography. There was astatistically significant heterogeneity among includedstudies (I2 ¼ 84%; P < 0.00001).

RCT. The pooled RR of obtaining screening mammo-graphy based on 14,159 mammography events (7,107intervention; 7,052 control) in 10 RCTs (14 comparisons;

Tab

le1.

Sum

maryof

commun

ityhe

alth

worke

rinterven

tions

forBC

screen

ingbytarget

grou

p(Con

t'd)

Firstau

thor(ref.)

Target

population

CHW

number

and

charac

teristics

Leng

th;type

oftraining

Interven

tiondes

cription

Controld

escription

Web

er(61)

African

-American

,La

tina,

andWhite

low-inc

ome,

urban

wom

en52

–77

,6prim

aryca

refacilities

Roc

hester,NY

6wom

enfrom

loca

lco

mmun

ity(4

African

American

,1Hispa

nic,

1White)

ND

1perso

nalized

mam

mog

raphy

reminder

letter

from

physician

1low

literac

ymam

mog

raphy

reminder

letter

from

commun

ityhe

alth

educ

ator

1pe

rson

alized

mam

mog

raphy

reminder

letter

from

phy

sician

plusus

ualc

are

Patient

educ

ationan

dreminders

Iden

tificationan

dremov

alof

barrie

rs

Abbreviations

:ND,no

tdes

cribed

;UC,us

ualc

are.

Wells et al.





refs. 41, 42, 44, 45, 50, 56, 58, 60, 61, 63) was 1.07 (95% CI:1.03–1.12, P ¼ 0.0005; Fig. 2) indicating a statisticallysignificant effect of CHW interventions on improving

rate of screeningmammography. There was a statisticallysignificant heterogeneity among included trials (I2¼ 78%;P < 0.00001).

Study or subgroup

RCTs

Allen, 2001 a

Allen, 2001 b

Andersen, 2000 a

Andersen, 2000 b

Andersen, 2000 c

Andersen, 2000 d

Calle, 1994

Duan, 2000

Margolis, 1998

Paskett, 2006

Slater, 1998

Sung, 1997

Weber, 1997

West, 2004

Subtotal (95% CI)

Total events

Heterogeneity: τ2 = 0.00; χ2 = 60.35, df = 13 (P < 0.00001); I 2 = 78%

Test for overall effect: Z = 3.49 (P = 0.0005)

Quasi-experimantal studies

Bird, 1998

Earp, 2002

Erwin, 1999

Navarro, 2007

Paskett, 1999

Sauaia, 2007

Suarez, 1997

Welsh, 2005 a

Welsh, 2005 b

Subtotal (95% CI)

Total events



Total (95% CI)

Total events



Events

709

532

786

813

470

446

142

321

535

184

94

71

41

18

5,162

108

224

98

56

94

359

171

25

48

1,183

6,345

Total

817

573

856

855

776

745

289

416

772

433

151

163

142

119

7,107

197

387

152

80

168

590

450

84

127

2,235

9,342

Events

628

519

799

799

462

462

104

285

447

114

85

54

17

15

4,790

54

274

89

43

54

3,338

205

574

1,756

6,387

11,177

Total

731

560

867

867

800

800

305

397

711

418

163

158

157

118

7,052

193

409

140

63

134

5,708

473

1,323

3,999

12,442

19,494

IV, Random, 95% CI

1.01 (0.97–1.05)

1.00 (0.97–1.03)

1.00 (0.97–1.02)

1.03 (1.01–1.06)

1.05 (0.97–1.14)

1.04 (0.95–1.13)

1.44 (1.19–1.75)

1.07 (0.99–1.17)

1.10 (1.02–1.19)

1.56 (1.29–1.89)

1.19 (0.98–1.45)

1.27 (0.97–1.68)

2.67 (1.59–4.48)

1.19 (0.63–2.25)

1.07 (1.03–1.12)

1.96 (1.51–2.54)

0.86 (0.77–0.96)

1.01 (0.85–1.20)

1.03 (0.82–1.28)

1.39 (1.09–1.78)

1.04 (0.97–1.11)

0.88 (0.75–1.03)

0.69 (0.49–0.96)

0.86 (0.69–1.08)

1.03 (0.89–1.18)

1.06 (1.02–1.11)

Intervention Control RR RR

IV, Random, 95% CI

0.01 0.1 1 10 100

Favors control Favors intervention

Figure 2. Forest plot of comparison: intervention versus no intervention, outcome: receipt of mammography.






Sensitivity analysesTo assess robustness of our findings and account for

observed heterogeneity among included studies, we con-ducted additional analyses. Results from quasi-experi-mental studies varied both in magnitude and direction ofeffect. Therefore, quasi-experimental studies wereexcluded from further sensitivity analysis. Sensitivityanalysis focused on the following factors that may beassociated with success of CHW interventions in the 10RCTs: methodologic quality (cluster vs. individual ran-domization, ITT analysis), choice of control intervention,method ofmeasuring study outcome (self-report vs. chartreview), setting of participant recruitment (medical vs.community and rural vs. urban), number of CHW inter-vention components, and characteristics of CHWs.

Methodologic quality. A critical appraisal of metho-dologic quality of all studies was conducted, includinginformation about study design and analyses (Table 2).

Unit of randomization. Six RCTs (60%) randomizedindividual participants to either CHW intervention orcomparison group (44, 50, 56, 60, 61, 63), whereas in theother 4, unit of randomization was a cluster or group towhich participants belonged, such as a work site orchurch (41, 42, 45, 58). None of the CRTs reported powercalculations or whether they were adjusted for clusteringeffects. Three trials (41, 42, 45) incorporated clusteringeffects in the main statistical analysis but did not reportestimated ICC values. The fourth trial (58) reported thevalue of ICC (�0.0015), which was appropriatelyassumed to indicate no design effect for the primaryoutcome measure, and data were analyzed at individualrather than cluster level.

In RCTs that randomized individual participants,CHW intervention was associated with a statisticallysignificant increase in screening mammography (RR:1.39; 95% CI: 1.13–1.70, P ¼ 0.002). In RCTs that rando-mized groups to treatment condition, CHW interventionswere not associated with increases in screening mammo-graphy (RR: 1.02; 95% CI: 1.00–1.04, P ¼ 0.05). There wasstatistically significant heterogeneity among includedtrials that used individuals as unit of randomization(I2 ¼ 80%; P ¼ 0.0002). RCTs that used groups as unitof randomization did not show heterogeneity (I2 ¼ 27%;P ¼ 0.22). Compared with unadjusted estimates, ORestimates adjusted for ICC and imbalance were statisti-cally significant, indicating that CHWs improve rates ofscreening mammography. However, adjusted ORs werecloser to the point of no effect (Fig. 3).

ITT analysis. Pooled data from 3 RCTs which con-ducted ITT analysis (50, 61, 63) indicate that a CHWintervention was not associated with increases inscreening mammography (RR: 1.48; 95% CI: 0.85–2.59,P ¼ 0.17). In 7 studies (11 comparisons) that didnot conduct ITT analysis (41, 42, 44, 45, 56, 58, 60), aCHW intervention was associated with increases inscreening mammography (RR: 1.06; 95% CI: 1.02–1.10,P ¼ 0.004). Both studies that conducted ITT analysis(I2 ¼ 82%; P ¼ 0.004) and studies that did not conduct

ITT analysis (I2 ¼ 77%; P < 0.00001) showed significantheterogeneity.

Choice of control intervention. RCTs comparingCHW interventions to routine care (7 trials involving11 comparisons due to multiple intervention groups ordifferent comparisons by age; refs. 41, 42, 44, 45, 50, 58, 60)were associated with a statistically significant increasein screening mammography (RR: 1.04; 95% CI: 1.01–1.07,P ¼ 0.007). The only RCT employing health education ascontrol (56) compared with CHW also showed a statis-tically significant benefit with use of a CHW intervention(RR: 1.56; 95% CI: 1.29–1.89; P < 0.00001). The pooled RRfrom 2 RCTs that used mammography reminders ascontrol (61, 63) showed a statistically nonsignificant dif-ference comparedwith CHW (RR: 1.82; 95%CI: 0.83–4.01,P ¼ 0.14). However, there was statistically significantheterogeneity among included trials that used routinecare (I2 ¼ 65%; P ¼ 0.002) or mammography reminders(I2 ¼ 73%; P ¼ 0.05) as control.

Method of measuring study outcome. There was sig-nificant variation in measurement of study outcome,receipt of mammography, which limited sensitivity ana-lyses that could be conducted on length of time in whichmammography was measured (Table 2). The interval ofmeasurement reported in the RCTs ranged from receiptof mammography in the past 6 months to receipt ofmammography in the past 3 years. Seven of 10 RCTsmeasured screening mammography outcome using par-ticipant self-report (41, 42, 44, 45, 58, 60, 63). Two studiesmeasured screening mammography with chart review(56, 61) and one combined self-report and a review ofpatients’medical records (50). CHW interventions wereeffective at increasing screening mammography whenmammography was measured by chart review alone(RR: 1.93; 95% CI: 1.15–3.23, P ¼ 0.01) or by chart reviewplus self-report (RR: 1.10; 95% CI: 1.02–1.19, P ¼ 0.009).The pooled RR for the 7 RCTs (11 comparisons) whichmeasured mammography using self-report indicatedthat CHW interventions were not associated withincreases in screening mammography (RR: 1.03; 95%CI: 1.00–1.06, P ¼ 0.02). There was statistically signifi-cant heterogeneity among included trials that usedchart review alone (I2 ¼ 73%; P ¼ 0.06) and self-reportalone (I2 ¼ 59%; P ¼ 0.007) to measure the mammo-graphy outcome.

Study setting. Medical versus community setting. Fivestudies recruited participants from a medical setting (50,56, 60, 61, 63) and 4 (7 comparisons) recruited participantsfrom a community setting (e.g., church, neighborhood;refs. 42, 44, 45, 58). In one study, participants wererecruited from a work setting (41) which was not codedas either a community or medical setting and was there-fore excluded from sensitivity analysis. CHWs wereeffective in increasing rates of screening mammographyin RCTs recruiting participants from a medical setting(RR: 1.41; 95% CI: 1.09–1.82, P ¼ 0.008) or communitysetting (RR: 1.05; 95% CI: 1.01–1.10, P¼ 0.02). There was astatistically significant heterogeneity among included

Wells et al.





Tab

le2.

Ran

dom

ized

controlledtrialc

haracteristic

san

dstud

yqua

lity

First

author

(ref.)

Num

ber

of

partic

ipan

ts,

bas

eline

Mea

nag

eCHW

interven

tion

partic

ipan

ts,

y

Mea

nag

eco

ntrol

participan

ts,

y

Leng

thof

interven

tion,

mo

Tim

efrom

bas

elineto

follo

w-up,

mo

Gen

erationof

rand

omization

seque

nce

Power

analys

isDes

cribed

drop

outs

Participan

tsse

lected

bec

ause

oflack

of

adhe

renc

e

Definitionan

dmea

suremen

tof

mam

mography

outco

me

ITTan

alys

isco

nduc

ted

Allen(41)

2,94

3ND

ND

1630

Worksitesblock

edon

size

ofworks

itean

dtype

ofag

ency

tocrea

te4block

s.With

inea

chblock

rand

om-num

ber

gene

ratoras

sign

edworksitesto

each

cond

ition

.

ND

Yes

No

Pas

t12

–24

mo;

self-report

No

And

erso

n(42)

8,90

7ND

ND

2436

40co

mmun

ities

assign

edto

10block

sof

4co

mmun

ities

for

block

rand

omization.

Eac

hco

mmun

itywith

inon

eof

ten

block

srand

omly

assign

edto

astud

yarm

usingon

eof

24permutation

patternsof

4for

each

block

.

ND

Yes

No

With

in24

mo

offollo

wup

interview;

self-report

No

Calle

(44)

738

ND

ND

68

American

Can

cer

Soc

iety

voluntee

rsea

chge

neratedlist

of10

wom

en.Lists

collected

andwom

enrand

omized

toeither

interven

tionor

control

grou

ps.

ND

Yes

No

Bec

ause

start

ofinterven

tion;

self-report

No

(Con

tinue

don

thefollo

wingpag

e)






Tab

le2.

Ran

dom

ized

controlledtrialc

haracteristic

san

dstud

yqua

lity

(Con

t'd)

First

author

(ref.)

Num

ber

of

partic

ipan

ts,

bas

eline

Mea

nag

eCHW

interven

tion

partic

ipan

ts,

y

Mea

nag

eco

ntrol

participan

ts,

y

Leng

thof

interven

tion,

mo

Tim

efrom

bas

elineto

follo

w-up,

mo

Gen

erationof

rand

omization

seque

nce

Power

analys

isDes

cribed

drop

outs

Participan

tsse

lected

bec

ause

oflack

of

adhe

renc

e

Definitionan

dmea

suremen

tof

mam

mography

outco

me

ITTan

alys

isco

nduc

ted

Dua

n(45)

1,11

3ND

ND

2412

30ch

urch

esmatch

edto

12block

san

drand

omized

halfthe

church

esin

each

block

topee

rco

unse

lingan

dothe

rha

lfto

control.

Match

ingva

riables

includ

edrace

/ethnicity,

mem

bersh

ipsize

,reso

urce

s,an

dden

omination.

ND

Yes

No

Pas

t12

mo;

self-report

No

Margo

lis(50)

1,65

854

.555

.91da

yplus

mailed

reminders

12þ

Patientsas

sign

edto

interven

tionor

usua

lca

reac

cordingto

whe

ther

med

ical

reco

rdnu

mber

was

oddor

even

ND

Yes

No

12or

24mo

(dep

ending

onag

e);

self-reportplus

chartreview

Yes

Pas

kett

(56)

897

54.5

55.7

12–14

12–14

Stratified

selection,

block

edrand

omization

byrace

andclinic,

cond

uctedbystud

ystatistic

ian

Yes

Yes

Yes

Pas

t12

mon

ths;

chartreview

No

Slater(58)

427

68.9

67.4

2412

41pub

licho

usinghigh

risebu

ildings

rand

omized

totrea

tmen

tor

control

grou

ps.

Buildings

stratifiedbas

edon

number

ofwom

enper

buildingwho

were

40ye

arsan

dolde

r.

ND

Yes

No

With

in15

moof

interven

tion;

self-report

No

(Con

tinue

don

thefollo

wingpag

e)

Wells et al.





Tab

le2.

Ran

dom

ized

controlledtrialc

haracteristic

san

dstud

yqua

lity

(Con

t'd)

First

author

(ref.)

Num

ber

of

partic

ipan

ts,

bas

eline

Mea

nag

eCHW

interven

tion

partic

ipan

ts,

y

Mea

nag

eco

ntrol

partic

ipan

ts,

y

Leng

thof

interven

tion,

mo

Tim

efrom

bas

elineto

follo

w-up,

mo

Gen

erationof

rand

omization

seque

nce

Power

analys

isDes

cribed

drop

outs

Participan

tsse

lected

bec

ause

oflack

of

adhe

renc

e

Definitionan

dmea

suremen

tof

mam

mography

outco

me

ITTan

alys

isco

nduc

ted

Sun

g(60)

321

ND

ND

1117

Partic

ipan

tsrand

omly

assign

edto

interven

tion

orco

ntrolg

roup

s

ND

Yes

No

Dep

ended

onag

e(w

ithin

4y

if35

–39

;with

in3y

if40

–49

;ye

arly

ifolder

than

49);

self-report

No

Web

er(61)

376

6363

47

Patientsrand

omized

toon

eof

two

grou

ps.

Ran

dom

ization

stratified

bypracticesite

Yes

Yes

Yes

During

interven

tion

period;

chart

review

Yes

Wes

t(63)

320

ND

ND

1da

y6

Env

elop

eco

ntaining

rand

omization

grou

pop

ened

byCHW

after

completin

gtelepho

neinterview

ND

Yes

Yes

Pas

t6mo;

self-report

Yes






trials for both subgroups (medical setting: I2 ¼ 81%; P ¼0.0003; community setting: I2 ¼ 70%; P ¼ 0.003).

Urban versus rural setting. Six RCTswere conducted inan urban setting (44, 45, 50, 58, 60, 61), 3 (7 comparisons)in a rural setting (42, 56, 63), and 1 (ref. 41; 2 comparisons)did not report information on setting. Studies conductedin an urban area showed (RR: 1.23; 95% CI: 1.09–1.39, P¼0.001) a statistically significant benefit associated withCHW compared with control. Pooled results for studiesconducted in a rural setting (RR: 1.05; 95% CI: 1.00–1.11,P¼ 0.06) or unknown setting (RR: 1.01; 95% CI: 0.98–1.03,P ¼ 0.69) did not show a benefit associated withCHW compared with control. There was a statisticallysignificant heterogeneity among the subgroup of studiesconducted in an urban (I2 ¼ 74%; P ¼ 0.002) or ruralsetting (I2 ¼ 78%; P ¼ 0.0003) but not the 2 includedcomparisons for the study that did not report setting (I2¼0%; P ¼ 0.75).

Number of intervention components. The CHWswere described using several different terms, such asindigenous community leaders (46), consejeras (51–54),promotoras (57), lay health advisors (46, 50, 56), lay healtheducators (49, 55, 64), community health educators (61),lay health workers (43, 60), peer health advisors, (41) peercounselors (45), community health care workers (63), andvolunteers (42, 44, 58, 59). The most commonly reportedcomponents of CHW interventions included health edu-cation (41–43, 45–49, 51–54, 56, 57, 59–64), referrals tohealth care (42, 43, 45, 47, 48, 50, 58, 59, 63), appointmentscheduling (48, 50, 56, 61), and vouchers, free mammo-grams, or lower cost mammograms (46–49, 63). Lesscommon intervention components included emotionalor social support (41, 61), financial paperwork (48, 61),communication with the health care team (61), mailedreminders (50, 61), and child care (61). Sensitivity analysis

showed a statistically significant increase in receipt ofscreening mammography associated with increase innumber of interventions. RCTs where CHW interven-tions included 1 intervention component (refs. 44, 45, 60;RR: 1.23; 95% CI: 1.00–1.52, P ¼ 0.05) or 2 interventioncomponents (refs. 41, 42, 56, 58; RR: 1.03; 95% CI: 1.00–1.07; P ¼ 0.06) were associated with a statistically non-significant increase in rates of screening mammography.Interventions that included 3 intervention components(refs. 50, 63; RR: 1.10; 95%CI: 1.03–1.19,P¼ 0.008) ormorethan 3 intervention components (ref. 61; RR: 2.67; 95% CI:1.59–4.48, P ¼ 0.0002) were associated with a statisticallysignificant increase in rates of mammography. There wasa statistically significant heterogeneity among includedRCTs in subgroups where CHW used 1 (I2 ¼ 75%; P ¼0.02) or 2 intervention components (I2¼ 74%; P¼ 0.0004).There was a statistically nonsignificant heterogeneityamong included RCTs in the subgroup with 3 CHWintervention components (I2 ¼ 0%; P ¼ 0.81).

Racial or ethnic concordance of CHW and target popu-lation. The studies reviewed targeted a number ofdifferent populations. Ninety-two percent of the studiesincluded ethnic or racial minority participants or focusedon low-incomepopulations. Pooled results fromRCTs thatreportedmatching CHWand target population by race orethnicity (refs. 44, 56, 61, 63; RR: 1.58; 95%CI: 1.29–1.93,P <0.0001) and that did not report matching CHW and targetpopulation by race or ethnicity (refs. 41, 42, 45, 50, 58, 60;RR: 1.03; 95%CI: 1.01–1.05,P¼ 0.02) showed a statisticallysignificant improvement in adherence to screening mam-mography; however, the effect was stronger for the RCTsthat matched CHWs to participants by race and ethnicity.There was not statistically significant heterogeneityamong either subgroup of studies (concordant: I2 ¼45%; P ¼ 0.14; not concordant: I2 ¼ 45%; P ¼ 0.06).

Effect of ICC adjustment on OR estimates

Method

Unadjusted 1.36 (1.18–1.57)

1.34 (1.22–1.49)

1.32 (1.21–1.43)

1.32 (1.21–1.44)

1.31 (1.21–1.42)

1.32 (1.21–1.43)

1.31 (1.21–1.44)

ICC = 0.05, balanced

ICC = 0.1, balanced

ICC = 0.2, balanced

ICC = 0.05, pareto imbalance



OR (95% Cl)

0 OR = 1 2

Favors control Favors intervention

Figure 3. Effect of ICC adjustmenton OR estimates.

Wells et al.





Discussion

The result from this meta-analysis of 18 studies enrol-ling a total of 26,660 participants indicates that CHWinterventions are associated with a statistically significantincrease in rates of screening mammography but tend tohave stronger effects in specific settings and studydesigns and when participants and CHWs were similarethnically or racially. When RCTs were compared withquasi-experimental studies, the significant increase inscreening mammography rates due to CHW interven-tionswas observed in RCTs but not in quasi-experimentalstudies. In RCTs, all studies showed either a neutral(intervention neither increased nor decreased mammo-graphy rates) or positive intervention effect (interventionincreased mammography rates), indicating that CHWswere more effective than control in improving screeningmammography rates. In quasi-experimental studies,direction of intervention effects was both positive andnegative. These findings point out to inherent biasesassociated with observational study designs. Overall,there was a significant heterogeneity in studies, withvarying populations, varying lengths of intervention,varying lengths of follow-up after intervention, and vary-ing time frames for the mammography outcome. Imput-ing different values of ICC had minimal effect on thedistribution of point estimates of ORs or credibility inter-vals. So, there was little impact of similarity of differentclusters (e.g., clinics) as a result of using a CRT design onoverall effectiveness of CHW interventions on screeningmammography.Overall, our findings are similar to results of previous

studies (21, 65). Results of the present study are alsosimilar to findings of a recent systematic review (26)which found that a subset of CHW interventions wasassociated with significantly greater screening mammo-graphy utilization rates when compared with controls orother interventions (mail, print, minimal CHW). Ourstudy improves on this recent review by providing ameta-analysis of all studies evaluating impact of CHWson screening mammography. In addition, through sensi-tivity analyses of data from RCTs, the present studyprovides information about specific situations whereCHW interventions are most likely to be beneficial (inter-ventions conducted in urban settings, recruitment ofparticipants from medical settings, measuring mammo-graphy outcome using a chart review, CHW interven-tions with 3 or more components, and interventionswhere CHW and patients are racially or ethnically con-cordant). The findings of our study differ from those ofLewin and colleagues (22), who found no beneficial effect(RR: 1.05; 95% CI: 0.99–1.12, P ¼ 0.10) of CHW interven-tions on screening mammography based on a meta-ana-lysis of 4 RCTs conducted internationally. While ouranalysis includes 3 of the 4 studies reported by Lewinand colleagues (22), our analysis also includes twice asmany RCTs as well as quasi-experimental studies. Thus,our results came from a larger pool of data that may not

be as strongly influenced by results of an individualstudy.

A lack of information reported in articles did not allowfor detailed sensitivity analysis of methodologic studyquality. Sensitivity analysis comparing unit of randomi-zation indicates CRTs of CHW interventions did not leadto increased mammography rates, whereas studies thatrandomized individuals were associated with a signifi-cant increase inmammography. In addition, there was nosignificant increase in mammography screening in 3studies that utilized ITT analysis to evaluate a CHWintervention (50, 61, 63), but there was a significantincrease in studies without ITT analysis. This lack ofintervention effect may be related to smaller pooledsample size of the 3 studies (n ¼ 2,019) or may suggestpossible biases favoring intervention effect in studies thatused per-protocol analyses, rather than ITT (66).

CHW interventions were associated with increases inrates of screening mammography in studies with routinecare or health education but not in studies with mammo-graphy reminders. Screening reminders may exert astrong effect on mammography rates, as noted by others(67, 68). Most RCTs reviewed included mammographyreferrals, reminders, or appointment scheduling as 1CHW intervention component (42, 44, 45, 50, 56, 58, 61,63). This finding points to the need to determine compo-nents of CHW interventions most strongly associatedwith increases in screening mammography to improveefficiency of CHW programs.

Both community and medical settings of recruitmentwere associated with increases in screening mammogra-phy related to a CHW intervention, but the effect wasstronger in participants recruited from a medical setting.Connecting patients to a medical setting is important indesigning a CHW program as establishing a usual sourceof care is a known predictor of mammography screening(16, 69). In addition, CHW interventions that took place inan urban settingwere associatedwith screeningmammo-graphy increases, whereas CHW interventions conductedin rural settings were not. These differences may reflectdifficulties in obtaining amammogram in a rural area (19).

RCTs that delivered at least 3 types of interventionswere associated with increases in screening mammogra-phy. These findings concur with previous findings (24)indicating that interventions with multiple interventioncomponents were associated with stronger increases inscreening mammography rates. While the studiesreviewed tended to focus or include historically or medi-cally underserved populations, the RCTs that reportedconcordance between participants and CHWs on race orethnicity indicated a stronger CHW intervention effect onscreening mammography than studies where race andethnic concordance was not described or conducted. Incontrast, a previous systematic review found concor-dance between physicians’ and patients’ race or ethnicitywas not associated with improved health outcomes orpatients’ utilization of health care (70). Sensitivity ana-lyses could not be conducted to evaluate differences in






effectiveness of CHW interventions by different popula-tion groups because the majority of the RCTs targetedmultiple underserved populations. Thus, there was noway to categorize them in a meaningful way for compar-isons.

There are several limitations to this systematic reviewand meta-analysis. The study may be biased as itincluded only published peer-reviewed articles.Although unpublished works were eligible for inclusionin the review, none met inclusion criteria. This reviewwas limited to studies published in English and imple-mented in the United States. Thus, meta-analysis resultsmay not generalize to countries providing better or worseaccess to mammography screening services. In addition,sensitivity analyses were based on a small number ofRCTs and should be interpreted with caution. The sys-tematic review is also limited by reporting of data inoriginal articles. Several sensitivity analyses could not beconducted because of significant variability in popula-tions, interventions, and study designs reported. Thisvariability is expected in the context of conductingresearch on interventions designed to meet local needsof various populations and under different breast screen-ing guidelines in place throughout the years, but varia-bility makes it difficult to combine data into meaningfulcategories.

The findings have significant implications for publichealth practice by indicating that CHW interventions areassociated with improvements in rate of 1 time screeningmammography, especially in medical settings, urban set-tings, and in participants who are racially or ethnicallyconcordantwith theCHW.However, asmammography isa behavior that must be repeated multiple times, there isstill much that is unknown about efficacy of CHW inter-ventions in increasing repeated BC screening. A recent

systematic review and meta-analysis of the effectivenessof various types of interventions designed to promoterepeat BC screening found studies that utilized screeningreminders only and studies that used more intensiveinterventions (including CHWs) were both associatedwith increases in repeat mammography (68). Futureresearch should evaluate whether CHW interventionsare associated with repeat mammography screeningand initiation of mammography screening. Futuresystematic reviews are necessary to compare CHW inter-ventions to othermammographypromoting interventionssuch as media interventions (71). In addition, future sys-tematic reviews should evaluate whether CHW interven-tions are associatedwithother cancer screeningbehaviors.Finally, future research should be conducted to defini-tively determinewhich participants and populations ben-efit most from CHW interventions and why.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Grant Support

This work was supported by a training grant from the National CancerInstitute (NCI) at the NIH (R25CA090314; Paul Jacobsen, Principal Inves-tigator) and by a grant from the NCI, through its Center to Reduce CancerHealth Disparities, NIH, Department of Health and Human Services(UO1 CA 117281-01; R.G. Roetzheim, Principal Investigator).

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received March 18, 2011; revised April 27, 2011; accepted June 1, 2011;published OnlineFirst June 8, 2011.

References1. American Cancer Society. Cancer facts and figures 2010. Atlanta, GA:

American Cancer Society, Inc; 2010.2. US Preventive Services Task Force. Screening for breast cancer: U.S.

Preventive Services Task Force recommendation statement. AnnIntern Med 2009;151:716–26.

3. American Cancer Society. American Cancer Society guidelines for theearly detection of cancer. 2010 [cited 2010 Jan 23]. Available from:http://www.cancer.org/docroot/PED/content/ped_2_3x_ACS_Cancer_Detection_Guidelines_36.asp.

4. Boyer-Chammard A, Taylor TH, Anton-Culver H. Survival differencesin breast cancer among racial/ethnic groups: a population-basedstudy. Cancer Detect Prevent 1999;23:463–73.

5. Kerlikowske K, Grady D, Rubin SM, Sandrock C, Ernster VL. Efficacyof screening mammography. A meta-analysis. JAMA 1995;273:149–54.

6. Albano JD, Ward E, Jemal A, Anderson R, Cokkinides VE, Murray T,et al. Cancer mortality in the United States by education level andrace. J Natl Can Inst 2007;99:1384–94.

7. Chu KC, Lamar CA, Freeman HP. Racial disparities in breast carci-noma survival rates: separating factors that affect diagnosis fromfactors that affect treatment. Cancer 2003;97:2853–60.

8. Clegg LX, Reichman ME, Miller BA, Hankey BF, Singh GK, Lin YD,et al. Impact of socioeconomic status on cancer incidence and stage

at diagnosis: selected findings from the surveillance, epidemiology,and end results: National Longitudinal Mortality Study. CancerCauses Control 2009;20:417–35.

9. Lantz PM, Mujahid M, Schwartz K, Janz NK, Fagerlin A, Salem B, et al.The influence of race, ethnicity, and individual socioeconomic factorson breast cancer stage at diagnosis. Am J Public Health2006;96:2173–8.

10. Vona-Davis L, Rose DP. The influence of socioeconomic disparitieson breast cancer tumor biology and prognosis: a review. J WomensHealth 2009;18:883–93.

11. Echeverria SE, Carrasquillo O. The roles of citizenship status, accul-turation, and health insurance in breast and cervical cancer screeningamong immigrant women. Med Care 2006;44:788–92.

12. Goel MS, Wee CC, McCarthy EP, Davis RB, Ngo-Metzger Q, PhillipsRS. Racial and ethnic disparities in cancer screening: the importance offoreign birth as a barrier to care. J Gen Intern Med 2003;18:1028–35.

13. Jacobs EA, Karavolos K, Rathouz PJ, Ferris TG, Powell LH. LimitedEnglish proficiency and breast and cervical cancer screening in amultiethnic population. Am J Public Health 2005;95:1410–6.

14. Kandula NR, Wen M, Jacobs EA, Lauderdale DS. Low rates of color-ectal, cervical, and breast cancer screening in Asian Americanscompared with non-Hispanic whites: cultural influences or accessto care? Cancer 2006;107:184–92.

Wells et al.





15. Kim J, Soong-Nang J. Socioeconomic disparities in breast cancerscreening among US women: trends from 2000 to 2005. J Prev MedPublic Health 2008;41:186–94.

16. Peek ME, Han JH. Disparities in screening mammography. Currentstatus,interventionsandimplications.JGenInternMed2004;19:184–94.

17. Sabatino SA, Coates RJ, Uhler RJ, Breen N, Tangka F, Shaw KM.Disparities in mammography use among US women aged 40–64years, by race, ethnicity, income, and health insurance status, 1993and 2005. Med Care 2008;46:692–700.

18. Wells KJ, Roetzheim RG. Health disparities in receipt of screeningmammography in Latinas: a critical review of recent literature. CancerControl 2007;14:369–79.

19. Doescher MP, Jackson JE. Trends in cervical and breast cancerscreening practices among women in rural and urban areas of theUnited States. J Public Health Manag Pract 2009;15:200–9.

20. Witmer A, Seifer SD, Finocchio L, Leslie J, O’Neil EH. Communityhealth workers: integral members of the health care work force. Am JPublic Health 1995;85:1055–8.

21. Swider SM. Outcome effectiveness of community health workers: anintegrative literature review. Public Health Nurs 2002;19:11–20.

22. Lewin S, Dick J, Pond P, Zwarenstein M, Aja G, van Wyk B, et al. Layhealth workers in primary and community health care. CochraneDatabase of Syst Rev 2005;CD004015.

23. Legler J, Meissner HI, Coyne C, Breen N, Chollette V, Rimer BK. Theeffectiveness of interventions to promote mammography amongwomen with historically lower rates of screening. Cancer EpidemiolBiomarkers Prev 2002;11:59–71.

24. Yabroff KR, Mandelblatt JS. Interventions targeted toward patients toincrease mammography use. Cancer Epidemiol Biomarkers Prev1999;8:749–57.

25. Spadea T, Bellini S, Kunst A, Stirbu I, Costa G. The impact ofinterventions to improve attendance in female cancer screeningamong lower socioeconomic groups: a review. Prev Med 2010;50:159–64.

26. Viswanathan M, Kraschnewski J, Nishikawa B, Morgan LC, Thieda P,Honeycutt A, et al. Outcomes of community health worker interven-tions. Rockville, MD: Agency for Healthcare Research and Quality;2009. Report No.: 09-E014.

27. Endnote X. Stamford, CT: Thompson ResearchSoft; 2006.28. Lehmann U, Sanders D. Community health workers: what do we know

about them? 2007 [cited 2010 Nov 23]. Available from: http://www.who.int/hrh/documents/community_health_workers.pdf.

29. SPSS I. SPSS statistics 17.0. 17.0 ed. Chicago, IL: SPSS Inc.; 2010.30. Graphpad Software. Quickcalcs: online calculators for sciences. La

Jolla, CA: GraphPad Software Inc.; 2010.31. Landis JR, Koch GG. The measurement of observer agreement for

categorical data. Biometrics 1977;33:159–74.32. Higgins J, Green S. Cochrane handbook for systematic reviews of

interventions. Version 5.0.0 [updated 2008 Feb]. Oxford, UK: TheCochrane Collaboration; 2008.

33. Campbell MK, Elbourne DR, Altman DG. CONSORT statement: exten-sion to cluster randomised trials. BMJ 2004;328:702–8.

34. Gilbody S, Bower P, Torgerson D, Richards D. Cluster randomizedtrials produced similar results to individually randomized trials in ameta-analysis of enhanced care for depression. J Clin Epidemiol2008;61:160–8.

35. Hade EM, Murray DM, Pennell ML, Rhoda D, Paskett ED, ChampionVL, et al. Intraclass correlation estimates for cancer screening out-comes: estimates and applications in the design of group-randomizedcancer screening studies. J Natl Cancer Inst Monogr 2010;2010:97–103.

36. Donner A, Klar N, Zou G. Methods for the statistical analysis of binarydata in split-cluster designs. Biometrics 2004;60:919–25.

37. MRC Biostatistics Unit. The BUGS project-Bayesian inference usingGibbs sampling. Cambridge, UK: Cambridge University; 2010.

38. Ntzoufras I. Bayesian modeling using WinBUGS. Hoboken, NJ: JohnWiley & Sons; 2009.

39. Guittet L, Ravaud P, Giraudeau B. Planning a cluster randomized trialwith unequal cluster sizes: practical issues involving continuous out-comes. BMC Med Res Methodol 2006;6:17.

40. Turner RM, Omar RZ, Thompson SG. Bayesian methods of analysisfor cluster randomized trials with binary outcome data. Stat Med2001;20:453–72.

41. Allen J, Stoddard A, Mays J, Sorensen G. Promoting breast andcervical cancer screening at the workplace: results from the Womanto Woman Study. Am J Public Health 2001;91:584–90.

42. Andersen MR, Yasui Y, Meischke H, Kuniyuki A, Etzioni R, Urban N.The effectiveness of mammography promotion by volunteers in ruralcommunities. Am J Prev Med 2000;18:199–208.

43. Bird JA, McPhee SJ, Ha NT, Le B, Davis T, Jenkins CN. Openingpathways to cancer screening for Vietnamese-American women: layhealth workers hold a key. Prev Med 1998;27:821–9.

44. Calle EE, Miracle-McMahill HL, Moss RE, Heath CW Jr. Personalcontact from friends to increase mammography usage. Am J PrevMed 1994;10:361–6.

45. Duan N, Fox SA, Derose KP, Carson S. Maintaining mammographyadherence through telephone counseling in a church-based trial. Am JPublic Health 2000;90:1468–71.

46. Earp J, Eng E. Increasing use of mammography among older, ruralAfrican American women: results from a community trial. Am J PublicHealth 2002;92:646–54.

47. Erwin DO, Spatz T, Scotts C, Hollenberg J. Increasing mammographypractice by African American women. Cancer Pract 1999;7:78–85.

48. Fern�andez ME, Gonzales A, Tortolero-Luna G, Partida S, Bartholo-mew LK. Using intervention mapping to develop a breast and cervicalcancer screening program for Hispanic farmworkers: Cultivando LaSalud. Health Promot Pract 2005;6:394–404.

49. Gotay C, Banner R, Matsunaga D, Hedlund N, Enos R, Issell B, et al.Impact of a culturally appropriate intervention on breast and cervicalscreening among native Hawaiian women. PrevMed 2000;31:529–37.

50. Margolis KL, Lurie N, McGovern PG, Tyrrell M, Slater JS. Increasingbreast and cervical cancer screening in low-income women. J GenIntern Med 1998;13:515–21.

51. Navarro AM, McNicholas LJ, Senn KL, Kaplan RM, Campo MC,Ropp�e B. Use of cancer screening tests among Latinas one andtwo years after participation in the Por La Vida Darnos Cuentaprogram. Women Cancer J 2000;2:23–30.

52. Navarro AM, Raman R, McNicholas LJ, Loza O. Diffusion of cancereducation information through a Latino community health advisorprogram. Prev Med 2007;45:135–8.

53. Navarro AM, Senn KL, Kaplan RM, McNicholas L, Campo MC, RoppeB. Por La Vida intervention model for cancer prevention in Latinas. JNatl Cancer Inst Monogr 1995:137–45.

54. Navarro AM, Senn KL, McNicholas LJ, Kaplan RM, Ropp�e B, CampoMC. Por La Vidamodel intervention enhances use of cancer screeningtests among Latinas. Am J Prev Med 1998;15:32–41.

55. Paskett E, TatumCM,D’Agostino Jr. R,Rushing J, VelezR,MichielutteR, et al. Community-based interventions to improvebreast andcervicalcancer screening: results of the Forsyth County Cancer Screening(FoCaS) Project. Cancer Epidemiol Biomarkers Prev 1999;8:453–59.

56. Paskett E, Tatum C, Rushing J, Michielutte R, Bell R, Foley KL, et al.Randomized trial of an intervention to improve mammography utiliza-tion among a triracial rural population of women. J Natl Cancer Inst2006;98:1226–37.

57. Sauaia A, Min SJ, Lack D, Apodaca C, Osuna D, Stowe A, et al.Church-based breast cancer screening education: impact of twoapproaches on Latinas enrolled in public and private health insuranceplans. Prev Chronic Dis 2007;4:A99.

58. Slater J, Ha CN, Malone ME, McGovern P, Madigan SD, Finnegan JR,et al. A randomized community trial to increase mammographyutilization among low-income women living in public housing. PrevMed 1998;27:862–70.

59. Suarez L, Roche R, Pulley L, Weiss N, Goldman D, Simpson D. Why apeer intervention program for Mexican-American women failed tomodify the secular trend in cancer screening. Am J Prev Med1997;13:411–7.

60. Sung JF, Blumenthal DS, Coates RJ, Williams JE, Alema-Mensah E,Liff JM. Effect of a cancer screening intervention conducted by layhealth workers among inner-city women. Am J Prev Med 1997;13:51–7.






61. Weber BE, Reilly BM. Enhancing mammography use in the inner city.A randomized trial of intensive case management. Arch Intern Med1997 10;157:2345–9.

62. Welsh AL, Sauaia A, Jacobellis J, Min SJ, Byers T. The effect of twochurch-based interventions on breast cancer screening ratesamong Medicaid-insured Latinas. Prev Chronic Dis 2005;2:A07.

63. West DS, Greene P, Pulley L, Polly K, Gore S, Weiss H, et al. Stepped-care, community clinic interventions to promote mammography useamong low-income rural African American women. Health EducBehav 2004;31:29S–44.

64. Zhu K, Hunter S, Bernard LJ, Payne-Wilks K, Roland CL, Elam LC,et al. An intervention study on screening for breast cancer amongsingle African-American women aged 65 and older. Prev Med2002;34:536–45.

65. Gibbons MC, Tyus NC. Systematic review of U.S.-based randomizedcontrolled trials using community health workers. Prog CommunityHealth Partnersh 2007;1:371–81.

66. Montori VM, Guyatt GH. Intention-to-treat principle. CMAJ 2001;165:1339–41.

67. Wagner TH. The effectiveness of mailed patient reminders onmammography screening: a meta-analysis. Am J Prev Med 1998;14:64–70.

68. Vernon SW, McQueen A, Tiro JA, del Junco DJ. Interventions topromote repeat breast cancer screening with mammography: a sys-tematic review and meta-analysis. J Natl Cancer Inst 2010;102:1023–39.

69. Swan J, Breen N, Graubard BI, McNeel TS, Blackman D, Tangka FK,et al. Data and trends in cancer screening in the United States: resultsfrom the 2005 National Health Interview Survey. Cancer 2010;116:4872–81.

70. Meghani SH, Brooks JM, Gipson-Jones T, Waite R, Whitfield-Harris L,Deatrick JA. Patient-provider race-concordance: does it matter inimproving minority patients’ health outcomes? Ethn Health 2009;14:107–30.

71. Baron RC, Rimer BK, Breslow RA, Coates RJ, Kerner J, Melillo S, et al.Client-directed interventions to increase community demand forbreast, cervical, and colorectal cancer screening a systematic review.Am J Prev Med 2008;351 Suppl:S34–55.

Wells et al.





2011;20:1580-1598. Published OnlineFirst June 8, 2011.Cancer Epidemiol Biomarkers Prev Kristen J. Wells, John S. Luque, Branko Miladinovic, et al. ReviewScreening Mammography in the United States? A Systematic Do Community Health Worker Interventions Improve Rates of

Updated version

10.1158/1055-9965.EPI-11-0276doi:

Access the most recent version of this article at:

Cited articles

http://cebp.aacrjournals.org/content/20/8/1580.full#ref-list-1

This article cites 59 articles, 5 of which you can access for free at:

Citing articles

http://cebp.aacrjournals.org/content/20/8/1580.full#related-urls

This article has been cited by 6 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cebp.aacrjournals.org/content/20/8/1580To request permission to re-use all or part of this article, use this link



http://cebp.aacrjournals.org/lookup/doi/10.1158/1055-9965.EPI-11-0276

http://cebp.aacrjournals.org/content/20/8/1580.full#ref-list-1

http://cebp.aacrjournals.org/content/20/8/1580.full#related-urls

http://cebp.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cebp.aacrjournals.org/content/20/8/1580


Do Community Health Worker Interventions Improve Rates of ... · Science of Cancer Health...

Documents

Transcript of Do Community Health Worker Interventions Improve Rates of ... · Science of Cancer Health...