RE S E A R C H AR T I C L E

The Evaluation of School-Based Violence Prevention Programs: A Meta-Analysis*

HYOUN-KYOUNG PARK-HIGGERSON, PhDa

SUZANNE E. PERUMEAN-CHANEY, PhDb

ALFRED A. BARTOLUCCI, PhDc

DIANE M. GRIMLEY, PhDd

KARAN P. SINGH, PhDe

ABSTRACT

BACKGROUND: Youth violence and related aggressive behaviors have become seri-

ous public health issues with physical, economic, social, and psychological impacts

and consequences. This study identified and evaluated the characteristics of success-

ful school-based violence prevention programs.

METHODS: Twenty-six randomized controlled trial (RCT), school-based studies that

were designed to reduce externalizing, aggressive, and violent behavior between the

1st and 11th grades were analyzed for assessing the effects of 5 program characteris-

tics by comparing results of intervention groups to control groups (no intervention)

after intervention using a meta-analysis. Electronic databases and bibliographies were

systematically searched, and a standardized mean difference was used for analysis.

RESULTS: There was no significant difference between interventions, although programs

that used non–theory-based interventions, focused on at-risk and older children, and em-

ployed intervention specialists had slightly stronger effects in reducing aggression and vio-

lence. Interventions using a single approach had a mild positive effect on decreasing

aggressive and violent behavior (effect size = �0.15, 95% CI = �0.29 to �0.02, p = .03).

CONCLUSIONS: Unlike previous individual study findings, this meta-analysis did

not find any differential effects for 4 of the 5 program characteristics. In addition, the

significant effect noted was contrary to expectation, exemplifying the complexity of

identifying effective program strategies. This study adds to the current literature by

assessing the program characteristics of RCT studies in an effort to determine what

factors may affect school-based violence prevention program success.

Keywords: school-based violence prevention program; meta-analysis; randomized

controlled trials (RCTs).

Citation: Park-Higgerson HK, Perumean-Chaney SE, Bartolucci AA, Grimley DM,

Singh KP. The evaluation of school-based violence prevention programs: ameta-analysis.

J Sch Health. 2008; 78: 465-479.

aPostdoctoral Fellow, ([email protected]), The Arnold School of Public Health, University of South Carolina, 800 Sumter Street, Columbia, SC 29208. bResearch Assistant Professor, ([email protected]), Department of Biostatistics, School of Public Health, University of Alabama at Birmingham, 1665 University Blvd, Birmingham, AL 35294-0022. cProfessor, ([email protected]), Department of Biostatistics, School of Public Health, University of Alabama at Birmingham, 1665 University Blvd, Birmingham, AL 35294-0022. dAssociate Professor and Chair, ([email protected]), Department of Health Behavior, School of Public Health, University of Alabama at Birmingham, 1665 University Blvd, Birmingham, AL 35294-0022. eProfessor and Chair, ([email protected]), Department of Biostatistics, School of Public Health, University of North Texas Health Science Center, 3500 Camp Bowie Blvd, Fort Worth, TX 76107-2699.

Address correspondence to: K.P. Singh, ([email protected]), Department of Biostatistics, School of Public Health, University of North Texas Health Science Center, 3500 Camp Bowie Blvd, Fort Worth, TX 76107-2699.

*Indicates CHES and Nursing continuing education hours are available. Also available at: www.ashaweb.org/continuing_education.html

Journal of School Health d September 2008, Vol. 78, No. 9 d ª 2008, American School Health Association d 465

Youth violence and related aggressive behaviors

have become serious public health issues with

physical, economic, social, and psychological impacts

and consequences.1 According to the results from the

2005 national Youth Risk Behavior Survey, 35.9% of

the students nationwide had been in a physical fight in

the previous year,with the prevalence rates across local

surveys varying between 30.4% and 46.5%.2 Because

youth violence has drawn national attention in the

United States, numerous prevention programs have

been developed that focus on teaching children various

attitudes, knowledge, and skills in order to reduce their

involvement in violence.3 However, conflicting reports

about the evaluations of these programs exists between

the published scientific literature and the US Surgeon

General’s4 (2001) report, which concludes that little is

known about the actual effects of violence prevention

programs. Therefore, a systematic analysis and evalua-

tion of these programs are needed in order to identify

and understand the crucial factors that may affect pro-

gram success and fidelity.

Two other meta-analyses have examined school-

based violence prevention programs. According to the

result of Mytton et al5 (2002), school-based violence

prevention programs modestly reduced both student

aggressive behaviors and school or agency actions in

response to this aggressive behavior. However, the con-

tent of the prevention program was not provided,

which makes it difficult to assess either the scope or

the programmatic characteristics that had led to this

reported success. The program content may have ac-

counted for the considerable variation in the reviewed

programs’ effectiveness.

Derzon and Wilson6 (1999) also reported that

school-based interventions were effective in prevent-

ing and reducing violence and other antisocial behav-

iors through influencing various mediating conditions

and behaviors. However, the authors also did not

account for the content of the program and pooled

the data by type of outcome. As a result, it is difficult

to assess how the programs were conducted or how

rigorously each program was evaluated. Because they

conducted their analysis at the group level rather than

at the individual level, themediators identified through

the meta-analysis cannot be interpreted as those that

actually contributed to the success of each program.

Finally, the meta-analysis included studies that were

not randomized controlled trials (RCTs), which might

have reduced the effect of other confounding variables.

Consequently, both meta-analyses did not provide

crucial information nor clarify the process on how

school-based violence prevention programs could be

effective in reducing aggressive or violent behaviors.

Therefore, this study adds to the current literature by

assessing the program characteristics of RCT studies in

an effort to determine what factors may affect school-

based violence prevention program success.

The purposes of this study were to provide

a descriptive overview of the current research on

school-based violence prevention programs, to evalu-

ate those programs in order to identify the factors that

lead to a program’s success, and to explore the eff-

ective strategies needed to develop programs that

prevent or modify the aggression and violence of

children and adolescents (hereafter referred to as

‘‘violence’’).

While most violence prevention program evalua-

tions focused on the characteristics of intervention

(ie, ‘‘increasing knowledge’’ or ‘‘improving social

skills’’), this study focused on the program character-

istics related to program delivery (ie, program target

or program type). From the literature review, the fol-

lowing 5 program characteristics were found most

often in the trials and differentially related to program

success: (1) the application of theory,6,7 (2) the type

of program such as universal or selective,5,8-10 (3) the

number of programs such as single- or multiple-

approach interventions,6,9,11-13 (4) the characteristics

of the target population,8,14,15 and (5) the type of

instructor, such as the use of specialists.16-18 Using

these characteristics, the following 5 hypotheses based

on the effect sizes were proposed:

1. Theory-based programs may have a stronger effect on

reducing violence than non–theory-based programs. Pro-

grams need to be evaluated from their theoretical

perspectives accorded with the program’s design.

Expected outcome should be derived from these

perspectives about the underlying processes that

contribute to the development of violent behavior

and how these developmental processes and paths

can be altered.7 For testing the first hypothesis, a

theory-based program was operationalized as

a study that supported its hypothesized outcomes

with logical or theoretical reasoning. For example,

DuRant et al (1996) and Ngwe et al (2004) both

applied Social Cognitive Theory to their program

for reducing violent behaviors. Although both stud-

ies reported that their program had significant

effects in decreasing violence behavior, DuRant

et al (1996) did not explain how their program could

make significant decreases in violence use. Thus,

this program was classified as nontheory based. On

the other hand, Ngwe et al (2004) clarified the

logical flow on how their program worked by mea-

suring mediators/determinants such as behavioral

intentions, attitudes, and so forth. Therefore, this

program was classified as theory based.

2. Selective programs may have a stronger effect on reducing

violence than universal prevention programs. Universal

programs were defined as primary preventions that

targeted everyone in the school, while the selective

program was a secondary prevention program that

targeted only the at-risk groups in the school.

466 d Journal of School Health d September 2008, Vol. 78, No. 9 d ª 2008, American School Health Association

3. Programs that applied multiple-approach interventions

may have a stronger effect on reducing violence than sin-

gle-approach programs.Amultiple-approach program

involved the parents, peers, and/or the community

in addition to the school’s curriculum. Those inter-

ventions that used the curriculum only in the class

were defined as a single-approach program.

4. Programs focused on younger students (third grade or

lower)mayhave a stronger effect on reducing violence than

programs focused on older students. Steinberg19 (2002)

defined adolescence as the second decade of life (10

through 19). Based on this definition, we compared

the first grade (7 years old) through third grade

(9 years old) to adolescence.

5. Programs conducted by specialists (personnel from outside

of the school) may have a stronger effect on reducing vio-

lence than programs conducted by schoolteachers.

Although the role of teachers in program success is

critical,16 it is not easy to get the support from teach-

ers in reality. They did not deliver all lessons as

planned,20 showed lack of participation,21 or even

felt resentment due to the additional burden of

teaching a new curriculum that was different from

the school’s educational mission.12 Therefore, we

assumed that programs conducted by specialists

may have been more effective for reducing violence

than programs conducted by schoolteachers.

METHODS

To evaluate the school-based violence prevention

programs, a meta-analysis was conducted. The follow-

ing 7 inclusion criteria were established: (1) the pro-

gram was designed to prevent aggression or violence

in children and adolescents; (2) the study population

was students enrolled in the 1st through the 11th

grades; (3) the experimental design randomly as-

signed the participants to either an intervention or

a control group; (4) the study outcome was defined as

externalizing, aggressive, or violent behavior (ie,

scores of aggression, use of violence/violent or exter-

nalizing behavior); (5) the type of intervention was

designed to reduce aggressive or violent behavior by

providing education, improving social skills, or chang-

ing the environment; (6) the program was conducted

in the school; and (7) the study had published the

necessary statistical information in order to determine

the direction of the effect and to conduct the

meta-analysis.

Procedures To identify relevant prevention programs, searches

through electronic databases and by hand were

conducted. For the electronic searches, ‘adolescen*’,

‘violen*’, ‘aggress*’, ‘externalizing’, ‘school’, ‘preven-

tion’, and ‘intervent*’ were the key terms used in the

following databases: Pubmed, Medline, PsychInfo,

Criminal Justice Info, National Institute of Justice,

ArticleFirst, BasicBIOSIS, Educational Resource Infor-

mation Centre (1970-2004), and the National Crimi-

nal Justice Reference Service (1970-2004). The

searches were restricted to studies conducted in the

United States and published in English but not

restricted by ‘‘date.’’

Dissertation abstracts were excluded from the study

because of the problem of accessibility. The hand

searches involved examining the references of the

published articles identified through the electronic

searches. We could contact only 1 author of the eligi-

ble trials to obtain missing data.

While the first author searched the relevant data,

titles, abstracts, and keywords were screened to

exclude ineligible trials. Next, full texts of remaining

reports were reviewed. Therefore, additional ineligible

trials were excluded.

Coding of the Studies Eligible studies were coded directly into an Excel

database by the first author. The coding was reviewed

by the third author, and questions were resolved

through discussion among authors. While coding the

data, several studies were excluded due to the lack of

measurable outcomes. A summary table (Table 1) was

created to record each study’s information pertaining

to the students’ characteristics, research methodology

(research design and assignment), intervention com-

ponents (duration and instructors), time to measure,

applied theory, outcome, measures of violence or

aggressive behaviors, and evaluation. A statistical out-

come table was created to extract detailed data on

outcome measures using statistical analysis but is not

shown because of limited space.

Statistical Analysis The effects of 5 program characteristics were as-

sessed by comparing results of intervention group to

control group after intervention. Because several

studies included in this meta-analysis did not report

pretest results, the effect sizes were calculated from

the posttest results. The unit of randomization (indi-

vidual, class, or school) was not taken into consider-

ation due to the small number of studies available for

the comparisons.

The overall effect size was calculated for the full

sample, and each hypothesis was tested using RevMan

4.2 (Cochrane Centre) program44,45 and SAS 9.1 for

Windows. The study-specific difference between inter-

vention and control groups for each comparison re-

ported in the study was pooled to produce an overall

estimate of the effect. Pooled results were expressed as

standardized mean differences (SMDs) with 95% confi-

dence intervals. Because each study in themeta-analysis

Journal of School Health d September 2008, Vol. 78, No. 9 d ª 2008, American School Health Association d 467

Ta b le

1. Su m m ar y o f Sc h o o l- B as ed

V io le n ce

P re ve n ti o n P ro g ra m s (T o ta l2 6 St u d ie s) *

So u rc e/ P ro g ra m

Ty p e

R an

d o m iz at io n

[U n it ] I, C ,S

In te rv en

ti o n

Ti m e to

M ea su re

P ro g ra m

O u tc o m e

M ea su re m en

t Ev al u at io n

Le e et al 2 2 (1 97 9) :9 th

gr ad e: th eo ry b as ed ,

se le ct iv e, si ng le ,4 th [ ,

sp ec ia lis t p ro gr am

Ra nd om

ly as si gn ed ; N = 30

(m = 24 , f = 6, ag gr es si ve

st ud en ts ) [I]

d 3 co nd iti on s: (1 ) as se rt io n tr ai ni ng :

1 p er w ee k, ei gh t 50 -m

in ut e

as se rt io n tr ai ni ng

se ss io ns ;( 2)

p la ce b o: ‘‘h ow

to m ak e a de ci si on ’’;

(3 ) no

tr ea tm en t

Pr et es t; p os tt es t

A ss er tio n tr ai ni ng

A gg re ss iv en es s;

as se rt iv en es s; as se rt iv e

sk ill

Pe er R, SR ;S R, p ee rR ;

ob se rv at io n

A ss er tio n sc al e in cr ea se

si gn ifi ca nt ly ; sl ig ht

de cr ea se

in se lf- ra te d ag gr es si on ; no

si gn ifi ca nt

ef fe ct

re ga rd in g as

p ee r ju dg m en t

H ue y an d Ra nk

2 3 (1 98 4) :

8t h, 9t h gr ad es :t he or y

b as ed ,s el ec tiv e, si ng le ,

4t h[

,s p ec ia lis t

p ro gr am

Ra nd om

ly as si gn ed ; n = 48 ;

ag gr es si ve

b la ck

m al es

[I]

d 3 co nd iti on s: (1 ) pr of es si on al or

pe er co un se lo r as se rt iv e tr ai ni ng

gr ou ps :t ot al 8 ho ur s (2 p er w ee k

fo r 4 w ee ks ); (2 ) p ro fe ss io na lo r

p ee r co un se lo r di sc us si on

gr ou p s;

(3 ) no -t re at m en t co nt ro lg ro up s

Pr et es t; p os tt es t

A ss er tio n tr ai ni ng

A ss er tiv e sk ill ; an ge r

le ve l; ag gr es si ve

be ha vi or ;c la ss ro om

ag gr es si ve

b eh av io r

A ut ho r; au th or , SR ;

p sy ch ol og is ts

ob se rv at io n; TR

(u si ng

BR PT )

A ss er tiv e tr ai ni ng

gr ou p s sc or ed

si gn ifi ca nt ly lo w er

in cl as sr oo m

ag gr es si on , re sp on de d m or e

as se rt iv el y th an

th e no –

at te nt io n co nt ro l gr ou p s

H er rm

an n an d

M cW

hi rt er 2 4 (2 00 3) :

7t h- 9t h gr ad es :t he or y

b as ed ,s el ec tiv e, si ng le ,

4t h[

,s p ec ia lis t

p ro gr am

Ra nd om

ly as si gn ed

to ea ch

co nd iti on

2 al te rn at iv e

m id dl e sc ho ol s, n = 20 7;

fin al ly , on ly 89

st ud en ts

w er e le ft [S ]

d 2 co nd iti on s: (1 ) SC A RE

pr og ra m :

8- w ee k p er io d 15

se ss io ns ,2

p er

w ee k fo ra p p ro xi m at el y 1 ho ur p er

da y (r ec og ni zi ng

an ge r an d

vi ol en ce

in th e co m m un ity ,

m an ag in g an d re du ci ng

an ge r in

th e se lf, de fu si ng

an ge r an d

vi ol en ce

in ot he rs ) co nd uc te d b y

gr ad ua te st ud en ts ;( 2) co nt ro l

gr ou p: th e En te r H er e cu rr ic ul um

Pr et es t; p os tt es t;

1- ye ar

fo llo w -u p

SC A RE

p ro gr am

(a tt rib ut io n

th eo ry )

A ng er ; ag gr es si on ;

at tit ud e

SR ; SR ; PR ;S R;

di sc ip lin ar y re co rd s

C as e sh ow

ed si gn ifi ca nt ly lo w er

le ve ls of

an ge r an d ag gr es si on

an d a sl ig ht ly hi gh er

le ve l of

an ge r co nt ro l. In

1- ye ar

fo llo w –

up , SC A RE

co nt in ue d to

re fle ct

si gn ifi ca nt ly lo w er

sc or es

on a m ea su re

of ag gr es si ve

an d

vi ol en t at tit ud es

Fe in dl er et al 2 5 (1 98 4) :

p ub lic

ju ni or

hi gh

sc ho ol s (1 2. 5- 15 .7

ye ar s) :t he or y b as ed ,

se le ct iv e, si ng le ,4 th [ ,

sp ec ia lis t p ro gr am

Ra nd om

ly as si gn ed

to ea ch

co nd iti on

ch os en

fr om

a b eh av io r m od ifi ca tio n

p ro gr am

fo r

m ul tis us p en de d

de lin qu en ts [I]

d 2 co nd iti on s: (1 ) 3 tr ea tm en tg ro up

(e ac h n = 6) :1 0 b iw ee kl y 50 –

m in ut e tr ai ni ng

se ss io ns

co nd uc te d b y tr ai ne d th er ap is t.

C on te nt s: b eh av io ra la nd

co gn iti ve

co nt ro ls ;( 2) co nt ro lg ro up

(n = 18 )

Pr et es t; p os tt es t

A ng er

co nt ro l

tr ai ni ng

p ro gr am

(c og ni tiv e

b eh av io ra l th eo ry )

Pr ob le m

so lv in g; lo cu s of

co nt ro l; m at ch in g

fa m ili ar

fig ur es

te st ;

se lf- co nt ro lr at in g

(a gg re ss io n)

TR Re su lts

sh ow

ed th e p ro gr am

p ro vi de s m od es t su p p or t fo r

th e ef fic ac y of

th e co gn iti ve

b eh av io ra l tr ea tm

en t

p ro ce du re s. Si gn ifi ca nt

ch an ge

sc or es

fo r tr ea tm

en t su b je ct s

on p ro b le m -s ol vi ng

ab ili ty

an d

se lf- co nt ro l

D uR an t et al 1 8 (1 99 6) :6 th –

8t h gr ad es :n on th eo ry

b as ed ,u ni ve rs al ,s in gl e,

4t h[

,s p ec ia lis t

p ro gr am

Ra nd om

ly as si gn ed

to on e of

th e cu rr ic ul a, N = 20 9 [S ]

d 2 co nd iti on s: (1 ) vi ol en ce

pr ev en tio n cu rr ic ul um

fo r

ad ol es ce nt s: te n 50 -m

in ut e

se ss io ns

he ld tw ic e a w ee k ov er 5

w ee ks ,d es ig ne d to

b e us ed

in a cl as sr oo m

fo rm

at (k no w le dg e

b as ed )—

sc ho ol 1 (n

= 14 6) ;

(2 ) co nf lic t re so lu tio n: a cu rr ic ul um

fo ry ou th p ro vi de rs am

on g m id dl e

sc ho ol st ud en ts :t en

50 -m

in ut e

se ss io ns .E ac h se ss io n co nt ai ns

at le as t 1 sk ill -b ui ld in g ex er ci se :

m ai nl y ro le -p la yi ng

(S C T b as ed

on b eh av io r ch an ge )—

sc ho ol 2

(n = 63 ); in te rv en tio n co nd uc te d

b y th e sa m e in st ru ct or

Pr et es t; p os tt es t

Th e vi ol en ce

p re ve nt io n

cu rr ic ul um

/c on fli ct

re so lu tio n:

a cu rr ic ul um

fo r

yo ut h p ro vi de rs

(S C T)

H yp ot he tic al us e of

vi ol en ce ; av oi da nc e of

vi ol en ce ; fr eq ue nc y of

us e of vi ol en ce

sc al e;

fr eq ue nc y of fig ht in g;

fr eq ue nc y of in ju ry