2. 274 E. Towner and T. Dowswell
countries and an increasing number of lower (WHO, 1989). Since that time, an increasing
income countries (Manciaux and Romer, 1991) number of communities around the world have
but has, until recently, been neglected on both used community-based approaches in injury
the national and international agenda [World prevention and some of these programmes have
Health Organization (WHO, 1996)]. Once over- been evaluated. Lessons learnt from such
shadowed by other causes of death and ill health, programmes would have direct application in
injuries have grown in relative importance as how best to develop, coordinate and evaluate
many diseases have been controlled. In England, injury prevention programmes at a local level.
unintentional injury is the main cause of death in Very few community-based prevention pro-
children and young people, is a major cause of grammes have been evaluated until recently.
ill health and disability, is responsible for con-
siderable financial and psychological costs and
is strongly linked with social deprivation. Its
importance as a major public health problem has METHODS
been recognized by its inclusion as a target in
‘Saving Lives: Our Healthier Nation’, England’s This paper uses as its source a systematic review
Public Health Strategy document (Secretary of of the literature, which seeks to answer the
State for Health, 1999). question ‘how effective are health promotion
There is a wide range of possible counter- interventions in preventing unintentional injuries
measures available for unintentional injury, but in childhood and young adolescents?’ (Towner
relatively few measures have been implemented et al., 2001). This source review has built on and
at a community-wide level. There has been a revised three earlier reviews published in 1993
longstanding debate within the injury field about (Towner et al., 1993) and 1996 (Nuffield Institute
the relative importance of ‘passive’ environmental for Health and NHS Centre for Reviews and
or engineering solutions (e.g. traffic calming, Dissemination, 1996; Towner et al., 1996). It
product design, playground modification) versus includes 155 studies or groups of studies published
‘active’ behavioural solutions (e.g. pedestrian between 1993 and 1996. Of these 155 studies, 10
skills training, promotion of cycle helmet were included that evaluated community-based
wearing). The community-based approach to injury prevention programmes. We are particularly
injury prevention offers the opportunity to interested in those programmes that targeted
stimulate ‘a process of cultural change which childhood injury.
allows an optimal mix of environmental and In the source systematic review, the relevant
behavioural solutions to be put into place’ literature was identified by a variety of means.
(Moller, 1992). Changes in behaviour may occur Computerized databases including MEDLINE,
at the whole community level through net- BIDS (and more recently the Web of Science)
working, mutual support and beyond this to and Excerpta Medica, and more specialized
cooperative advocacy for local policy changes. A sources such as the Transport and Road
coordinated approach by a range of agencies is Laboratory (TRL) database were searched (a
an essential ingredient: thus, injury prevention is full list of databases searched and search terms
less divided by sectoral allegiances and a common used is available on request). This electronic
culture of safety allows the adoption of comple- search was supplemented by hand searching a
mentary solutions, which should enable a number of key journals such as Accident Analysis
multiplier effect to be achieved (Moller, 1992). and Prevention and Injury Prevention, along
The philosophy behind community develop- with the reference lists of relevant published
ment set out in the Ottawa Charter for Health articles and books. In addition, key informants
Promotion incorporated a concern with reducing (researchers and specialists in the area of child
inequalities and promoting ownership of health- injury prevention) were consulted. The criteria
related issues (WHO, 1986). This broad concept for the inclusion of studies were as follows.
of safety promotion was developed in relation to
injury prevention in the mid 1970s by workers (i) They were written in English and published
at the Karolinska Institute in Sweden. The between 1975 and 2000 (the last search was
Manifesto for Safe Communities was set out at carried out in June 2000).
the First World Conference on Accident and (ii) They related to the prevention of
Injury Prevention held in Sweden in 1989 unintentional injuries (solely or in part).
3. Community childhood injury prevention: what works? 275
(iii) The target population included children A total of 15 evaluated studies were identified
15 years old and results were reported for that related to community-based studies. Five of
this group. the programmes used a simple before–after
(iv) They described either a primary interven- design, with no control group, and three of these
tion measure to prevent accidents occurring provided very few details of the intervention or
or a secondary measure to prevent or reduce evaluation. We thus excluded five studies from
the severity of injuries. the paper (Tellnes, 1985; Robertson, 1986; Sahlin
(v) They had been evaluated using some meas- and Lereim, 1990; Jeffs et al., 1993; Lindquist et
ure of outcome or impact. These included al., 1998). This paper examines the remaining
changes in injury mortality or morbidity, 10 programmes in more detail.
changes in observed or reported behaviour, In the Results, we describe the 10 programmes
environmental change or hazard removal, or identified: the features of the intervention, the
changes in knowledge or attitudes. groups or communities they target, the outcome,
and process measures used in the evaluation.
Violence prevention studies were excluded, We then discuss injury surveillance systems,
except in those cases where they were combined examine how intervention and control com-
with unintentional injury studies. munities have been chosen, and examine which
All studies were read and reviewed independ- process measures have been employed.
ently by two reviewers. Where statistical advice
or other specialized knowledge was required a
third reviewer was consulted. A standardized RESULTS
data extraction form was devised and used to
record details from each study included (avail- Intervention
able on request). Details recorded included the The 10 programmes are summarized in Table 1:
date and place of the study, the injury target (Schelp, 1987; Svanström et al., 1996) (1); (Guyer
group, and the aim, content and setting of the et al., 1989) (2); (Schwarz et al., 1993) (3);
intervention. Where interventions had been (Davidson et al., 1994; Kuhn et al., 1994) (4);
targeted at socially or economically disadvantaged (Hennessey et al., 1994; Ozanne-Smith et al.,
groups this was noted. In addition, details about 1994) (5); (Ytterstad, 1995; Ytterstad and
the evaluation were recorded. This included a Sogaard, 1995; Ytterstad and Wasmuth, 1995;
brief description of the methods used (the study Ytterstad et al., 1998) (6); (Svanström et al., 1995)
design, sample size, data collection methods, (7); (Day et al., 1997) (8); (Petridou et al., 1997)
outcome, impact and process measures). In (9); and (Coggan et al., 1998; Coggan et al., 2000)
particular, we were keen to assess how the inter- (10).
vention and control groups were selected and Six out of the 10 programmes are based on the
how comparable these groups were. A note was WHO Safe Communities Model, initially
also made of strengths and weaknesses of the developed in the community of Falköping in
evaluation. The British National Health Service’s Sweden (study 1). This model combines two
Centre for Reviews and Dissemination guide- elements: community diagnosis, which relies on a
lines on carrying out systematic literature reviews local surveillance system to provide an accurate
(Arblaster et al., 1995) were consulted for picture of the local injury problem, and a
information regarding the process of assessing reference group to coordinate activities. The six
the quality of the evidence of the various studies. programmes took place in Scandinavia, Australia
The reviewers reached a consensus decision on and New Zealand (Table 1: studies 1, 5, 6, 7, 8
the quality of the evidence. Each study was and 10). Of the remaining projects, three were
graded on a five-point scale ranging from weak conducted in the United States (2, 3 and 4) and
to good (i.e. weak, reasonable/weak, reasonable, one in Greece (9).
reasonable/good, good). Key results were recorded
and a consensus decision was made about the
effectiveness of the intervention. Details from Targeting
the data extraction forms were used to devise Five of the six programmes based on the Safe
summary tables for each study included. At this Communities Model have targeted a range of
stage, those studies where the evidence was rated ages (except study 6). The Shire of Bulla Safe
as weak were excluded. Living Program (study 5), for example, targeted all
4. Table 1: Community-based injury prevention programmes
276
Author, date and Injury target group Aims and content of Study type and Outcome impact and Key results
country and setting intervention sample size process measures
1a. (Schelp, 1987) Home and Falköping Accident Prevention Controlled trial (i) Deaths (a) Reduction of 27% in home
1b. (Svanström et al., occupational injuries Programme (a) I = Falköping (ii) Hospital admissions accidents and 28% in occupational
1996) Sweden targeted Based on community diagnosis C = Lidköping (iii) Accident and accidents
Children and older and use of reference group to (b) I = Falköping Emergency attendance Effective
people coordinate activities (pop. 32 022) (a) 1979–1982 (b) Hospital admissions increased by
Education of policy makers and C1 = Skaraborg County (b) 1983–1991 8.7% (females) and 4.9% (males)
health workers (pop. 277 397) in I. Smaller increases in C1 and C2
Range of interventions C2 = Sweden Ineffective therefore inconclusive
(pop. 8 644 125) overall
Reasonable/weak evidence
E. Towner and T. Dowswell
2. (Guyer et al., 1989) Children under Statewide Child Injury Controlled trial (i) Accident and (i) Reduction in passenger motor vehicle
USA 5 years Prevention Program (SCIPP) I = nine communities Emergency attendance injuries in I compared with C
Health promotion campaigns (pop. 139 810) (ii) Reported behaviour No evidence found in the reduction
related to burns, poisoning, falls, C = five communities (iii) Knowledge of other target injuries
suffocations and passenger (pop. 146 866) (ii) Exposure to prevention messages
motor vehicle injuries associated with safety behaviour
(iii) 42% of households with children
in I exposed to one or more
interventions
Partially effective
Good evidence
3. (Schwarz et al., 1993) General population Safe Block Project Controlled trial (i) Observation of hazards (i) Intervention homes significantly
USA Focus on urban In poor inner city community I = census tracts in (ii) Knowledge more likely to have Ipecac and
African–American Community workers and Philadelphia, 3004 (iii) Community smoke detectors (minimal–
population community representatives homes involvement moderate effort), but fewer
involved in home inspections C = census tracts in differences for home hazards
and educational programme Philadelphia, 1060 requiring major effort
Focus on falls, fires, scald burns, homes (ii) Distinct difference between I and
poisonings and violence C houses in safety knowledge
(iii) Community representatives
recruited for 88% of blocks
Partially effective
Good/reasonable evidence
4a. (Davidson et al., Children aged Safe Kids/Healthy Controlled trial (i) Deaths (i) Significant reductions in injuries in
1994) 5–16 years Neighborhoods Injury I = Central Harlem (ii) Hospital admissions I and C areas
4b. (Kuhn et al., 1994) Disadvantaged Prevention Program Pop. of children 17 years (iii) Participation in study In I, 44% reduction in targeted
USA community Coalition of organizations = 28 457 injuries
aimed to reduce outdoor injuries C = Washington Heights In I, decline specific to targeted
in children and reduction of Pop. of children 17 years injuries
assaults to children = 66 305 (iii) 10 000 children participated in
Involved playground specific programmes
renovation, safety equipment, Partially effective/inconclusive
supervised activities and Reasonable evidence
education (26 organizations)
5. Table 1: continued
Author, date and Injury target group Aims and content of Study type and Outcome impact and Key results
country and setting intervention sample size process measures
5a. (Ozanne-Smith et al., 1994) All ages Shire of Bulla Safe Living Controlled trial (i) Mortality and (i) Little evidence of reduction of
5b. (Hennessey et al., 1994) All injury types Program I = Shire of Bulla morbidity data injury morbidity
Australia Based on Falköping model and (pop. 28 347) (ii) Observed behaviour Some evidence for telephone
injury surveillance C = Shire of Melton (iii) Area-wide survey of reduction in minor
Aimed to prevent injuries, (pop. 28 812) environmental change injuries
reduce hazards and increase (iv) Attitudes knowledge (ii) Increased use of safety devices
public awareness and equipment—helmets, safety
113 preventive programmes, seats, smoke detectors
with emphasis on training (iii) Hazard reduction ( 50% of
professionals, environmental recommendations following
modification, audit and playground safety audit enacted)
advocacy (iv) Increased community awareness
Partially effective
Good evidence
6a. (Ytterstad and Wasmuth, 1995) General population Harstad WHO Safe (a) Controlled trial (a) Mortality data (a) 27% reduction in overall traffic
6b. (Ytterstad, 1995) but specific Community Programme I = Harstad Hospital admissions injury rate
6c. (Ytterstad and Sogaard, 1995) components All ages, all injury types (pop. 22 000) A and E attendance Significant reduction for
6d. (Ytterstad et al., 1998) targeted at children programme over a period of C = Trondheim Primary care 0–9 years and 15–24 years
Norway Targeted at children 7–9 years (pop. 134 000) (b) Hospital admissions Partially effective/inconclusive
0–4 years of age (a, b) Targets included child (b) Before and after A and E attendance (b) 0–15 years—37% reduction in
pedestrians and cyclists—infant study (c) Mortality data cyclist injuries and 54%
car loan schemes, lobbying for I = Harstad Hospital admissions reduction in pedestrian
cycle paths (pop. 22 000) A and E attendance injuries—decreased exposure
(c) Burn prevention— C = Trondheim (d) Morbidity data Partially effective/inconclusive
counselling, professional (pop. 134 000) Outpatient admissions
awareness raising, safety devices (c) Controlled trial records (c) 53% reduction in burn injury
(d) Burn prevention—cooker I = Harstad rates in I, 10% increase in C1
guards and lowering tap water (pop. 22 000) and 14% decrease in C2
thermostats C1 = Trondheim Admissions in I in later period
Educational activities (pop. 134 000) less severe
Programme focused on its own C2 = six towns around Effective
sustainability Harstad (d) Decrease in burn injury rates at
(pop. 14 000) 51.5% in I1, 40.1% in I2 and
(d) Controlled trial increase of 18.1% in C
I1 = Harstad Inconclusive
(pop. 23 000) Reasonable evidence
I2 = six towns around
Harstad
(pop. 14 000)
C = Trondheim
Community childhood injury prevention: what works?
(pop. 134 000)
277
6. 278
Table 1: continued
Author, date and Injury target group Aims and content of Study type and Outcome impact and Key results
country and setting intervention sample size process measures
7. (Svanström et al., 1995) Children 0–14 years Lidköping Accident Prevention Controlled trial (i) Hospital discharge (i) From 1983 to 1991 a reported
Sweden Programme I1 = Lidköping register data annual decrease in hospitalized
Community-wide injury (pop. 35 949) (ii) Process data: notes injuries of 2.4% (boys) and
prevention programme C1 = four surrounding and reports of health 2.1% (girls) in I1
(a) Surveillance of injuries municipalities (pop. planners In C1, increase in hospitalized
(b) Provision of information 42 078) injuries of 0.6% (boys) and
(c) Training C2 = Skarabourg county 2.2% (girls)
E. Towner and T. Dowswell
(d) Supervision (pop. 278 162) In C2, decrease of 1.0% (boys)
(e) Environmental measures and 0.3% (girls)
Specific activities—bicycle Inconclusive
helmet campaigns, first aid Reasonable/weak evidence
training for parents, loan schemes,
removal of local hazards
8. (Day et al., 1997) General population Latrobe Valley Better Health Before and after study (i) Emergency (i) Overall decline in rate of
Australia Project (non-targeted injuries Department attendance from 6594 to
All ages, community-based used for comparison presentations 4821/100 000 for targeted injuries,
approach to prevent injuries, data) (ii) Self-reported injury compared with a small decrease
reduce hazards and increase (a) Injury surveillance Knowledge in non-targeted injuries
public awareness system (Victorian (iii) Playground hazards Significant decrease in
(a) Home Injury Surveillance playground injuries among
(b) Sports System) 5- to 14-year olds
(c) Playground injuries, and (b) Telephone survey Estimated 908 injuries prevented
(d) Alcohol misuse among youth 375 households pre- (ii) Telephone survey
test, 400 households Non-significant decrease in rate
Exhibition, home safety training, post-test of self-reported injuries from
education for new mothers 62.7 to 48.2/1000. Non-significant
Protective sports equipment increase in injuries requiring
promoted medical attention from 24.5%
Playground safety— to 31.9%.
environmental measures Modest increases in knowledge
Mass media and community (iii) Evidence of hazard removal in
event playgrounds
Process: 46 000 educational contacts
with community on home injury
prevention
Evidence of institutionalization of
programme
Partially effective/inconclusive
Reasonable/weak evidence
7. Table 1: continued
Author, date and Injury target group Aims and content of Study type and Outcome impact and Key results
country and setting intervention sample size process measures
9. (Petridou et al., 1997) Young people Greek Island Community Injury Controlled trial (i) Self-reported injuries (i) No difference in accidents
Greece 0–18 years Prevention Project I = Island of Naxos (ii) Observed hazards reported in I and C
Older adults 65 years Multi-faceted intervention (172 households) Attitudes (ii) For I, improvements on 11 out
involving local community C = Island of Spetses Knowledge of 28 hazard variables
leaders and activities for parents, (177 households) (iii) Improvements in 1/28 hazard
teachers and children variables (improvement related
Home visits, counselling on to changes that could easily or
home hazards cheaply be implemented)
Partially effective
Reasonable/weak evidence
10a. (Coggan et al., 1998) General population Waitakere Community Injury Controlled trial (i) Injury rates (hospital (i) No significant reductions in
10b. (Coggan et al., 2000) with specific Prevention Project (WCIPP) I = Waitakere admissions and census admissions overall in I, C1
New Zealand components targeted All ages, all injuries, prevention (pop. 155 000) data) in I, C1 and C2 and C2
at children 0–14 years programme based on WHO C1 = comparison (ii) Data from Land In children 0–14 years, decrease
of age model. community Transport Safety in admissions in I, no decrease
Multicultural urban Seven priority areas—Maori, (pop. = 147 000) Authority and Fire in C1 or C2 (sig)
community Pacific, children, young people, C2 = rest of Auckland Service (ii) Land Transport data annual
older people, alcohol and roads. (iii) Self-reported injury increase of 7% in adults
Three approaches and appropriately restrained in front
(a) Promotion (iv) Self-reported seats, 7% increase in children in
(b) Education and awareness behaviour (telephone I (C1, not clear)
(c) Advocacy and environmental survey n = 4000 in I (iii) No reduction in self-reported
change and C1) injury in I and C1, but fewer
(v) Reach/awareness in injured people required medical
Range of activities including total population and treatment
promotion of car restraints, cycle organizations (iv) Significant increases in
helmets, smoke alarms, burn and ownership of child restraints,
scald education pool fencing, stair gates and
protective sports equipment in
I compared with C1
(v) 85% of organizations in I aware
of intervention compared with
25% in C1
Partially effective
Good/reasonable evidence
Pop., population; I, intervention; C, control.
Community childhood injury prevention: what works?
279
8. 280 E. Towner and T. Dowswell
age groups and injuries occurring in home, school evaluation. For example, the recruitment of
and leisure environments. The three US studies representatives of neighbourhood housing
targeted children, and the Greek island study blocks was regarded as a measure of community
targeted young people and older adults. The involvement in the Safe Block Project (study 3).
Statewide Child Injury Prevention Program in The Waitakere Community Injury Prevention
Massachusetts (study 2) selected the main injury Project was placed within local government: this
types affecting pre-school children, for which provided the council with an avenue to interact
a proven countermeasure was available, and with the voluntary sectors of the community and
thus developed programmes aimed at the preven- ‘thereby contribute to the social structure of
tion of burns, poisoning, falls, suffocations and Waitakere’ (study 10). The Safe Communities
passenger road traffic accidents. Two programmes Model advocates the need for a reference group
specifically targeted deprived communities: the to coordinate the activities of the agencies
Safe Block Project (study 3) in a poor African– involved in delivering the intervention. The New
American inner city community in Philadelphia, Zealand project stressed the pivotal role of
and the Safe Kids/Healthy Neighborhoods Pro- project coordinators.
gram (study 4), in a mainly non-Hispanic, black
community in Harlem. One programme was
based in a multi-cultural urban community Nature of the intervention
in New Zealand (study 10) and had specific An innovative feature of many programmes was
community components for Maori and Pacific the attempt to deliver a range of diverse activities
people. at the same time. Unlike some health problems, the
range of possible preventive activities is vast,
and no intervention alone is likely to result in
Length of intervention observable differences in the injury mortality or
Some programmes had been in progress for morbidity experienced by a single community.
many years. The Harstad programme (study 6) The Safe Kids/Healthy Neighborhoods Program
from Norway developed over a period of 7–9 (study 4) aimed to reduce outdoor injuries in
years. The evaluation of the Shire of Bulla Safe children. Specific interventions included the
Living Program and the Waitakere Community renovation of playgrounds, the involvement of
Injury Prevention Project (studies 5 and 10, children and adolescents in safe supervised
respectively) related to the first 3 years of longer activities, which taught them useful skills, the
projects. In contrast, the Greek island (study 9) provision of injury and violence prevention
and the Safe Block Project (study 3) inter- education and the supply of safety equipment at
ventions were both of short duration. a reasonable cost. Several of the programmes
included elements that resulted in environmental
change, or lobbied for environmental change.
Multi-agency approaches For example, lobbying for the provision of
A feature of most of the community-based cycle paths was a feature of the Harstad pro-
programmes has been the involvement of a range gramme (study 6), and a parent pressure group in
of organizations drawn from health, Local the Latrobe Valley Project (study 8) was active in
Authority, voluntary and commercial agencies. changing Council priorities with respect to the
Interventions have taken place in a variety of refurbishment of existing playgrounds and the
settings: home, school, roads and neighbour- creation of new ones. The scale of many
hoods. In the Falköping programme (study 1), for programmes meant that educational, environ-
example, importance was placed on raising mental and policy approaches were all feasible
public awareness and local journalists were and these approaches were often combined. The
members of the multi-agency group. The owner New Zealand project aimed to cover all ages and
of a local shop selling child safety products was all injury types, but in practice the focus was on
also a key member of the group. child safety.
The involvement of local people and the
development of local ownership were important
features of several programmes, and the number Evaluation
of local people participating in local programmes Of the evaluation designs employed in the 10
was sometimes used as a process measure in the programmes summarized in Table 1, none have
9. Community childhood injury prevention: what works? 281
used a randomized controlled design. Only one reporting, key informant interviews with coord-
evaluation used several intervention and control inators and with management group members,
communities (study 2), eight used one main and detailed case studies of different project
control community, and one (study 8) measured components.
success by a comparison of targeted and non-
targeted injuries.
The Statewide Child Injury Prevention Overall effectiveness
Programme in the USA (study 2) selected nine Eight of the studies were considered partially
intervention and five control communities from effective and two inconclusive (1 and 7).
351 potential cities and towns in Massachusetts,
matched for a number of relevant variables. In
the Safe Living Program (study 5), a demo- DISCUSSION
graphically matched Shire, the Shire of Melton,
was selected as a control community. In the In systematic reviews of effective injury
Greek island project, the islands of Naxos and prevention, most evaluated studies described
Spetses were selected as intervention and control relate to single countermeasures, such as the
communities. promotion of bicycle helmets or child safety seats
Two of the evaluation designs were considered (Towner et al., 2001). Community-based studies
‘good’ (studies 2 and 5), two ‘good/reasonable’ such as those described in this paper, offer the
(3 and 10), two ‘reasonable’ (4 and 6) and four opportunity to examine whether using a multi-
‘reasonable/weak’ (1, 7, 8 and 9). agency coordinated approach provides the
opportunity to change the whole culture of safety
within a community and to assess the result in
Outcome measures terms of health gain.
Local injury surveillance systems were not only What is apparent from the results section of
used as a means of identifying local problems and this paper is the great variety in the content of
targets for interventions, but also as a source of the intervention in the 10 programmes investi-
outcome data in programme evaluation. Such gated. Only in a few cases is the full extent of the
outcome data related primarily to Accident and intervention documented, e.g. the Safe Living
Emergency attendance and hospital admissions. Program, where details of the 113 programme
In the Harstad programme (study 6), length of components have been described. One element
hospital stay was used as a proxy measure of common to nine of the 10 programmes (the
injury severity. In a few studies (5, 8, 9 and 10), exception is programme 10) is the importance of
sample population questionnaire surveys were injury surveillance systems, not just in evaluating
used to elicit self or proxy reports of injuries as the impact of the programme, but in contributing
an outcome measure. Area-wide environmental to the intervention itself. Data collected in these
changes were measured in the Shire of Bulla systems can be utilized in generating local
Program (study 5), numbers of home hazards in interest and mobilizing community involvement,
the Safe Block Project (3) and the Greek Island attracting media and political interest, obtaining
Programme (9), and sales of safety equipment in resources and for targeting specific local
the Falköping study (1). Reported behaviour problems.
(e.g. use of safety equipment) and knowledge For injury surveillance systems to be useful for
were used as measures of programme impact in evaluation purposes (and to make comparisons
several programmes. between, or to summarize findings from similar
studies) it is necessary to have meaningful and
consistent outcome measures. Death as an out-
Process measures come is too rare an event to provide information
The Shire of Bulla Safe Living Program, the on what to target or to be used to evaluate local
Latrobe Valley Better Health Project and the campaigns. Most of the programmes have used
Waitakere Community Injury Prevention Project hospital admission or Accident and Emergency
(studies 5, 8 and 10, respectively) provide more attendance as measures of non-fatal injury in a
detailed documentation of the process of the community. There are flaws in using such
intervention. Process measures included pro- measures because they may reflect changes in the
gramme reach, community participation, media use of, and access to, health services rather than
10. 282 E. Towner and T. Dowswell
true injury rates. For instance, in the Latrobe of randomized controlled trials for use in com-
Valley Project (study 7), changes in the hospital plex interventions has been questioned (Speller
resourcing mechanism led to large-scale variations et al., 1997) and there is considerable debate on
in admission rates. One of the programmes this issue within health promotion. We agree
attempted to utilize a proxy measure of injury with this argument, but feel that the strength of
severity, which in this case was hospital bed days. the evidence is enhanced by the selection of
Injury surveillance systems are potentially appropriate control communities or comparison
expensive to establish as part of community- groups. The provision of detail about the nature
based programmes. Several programmes relied of the intervention also enhances the interpreta-
on existing (usually health care) databases for tion of results, as well as providing necessary
local injury data. Under these circumstances, information for implementation elsewhere.
data collection considerations would be likely to Evaluating the effectiveness of health promo-
have a direct effect on both the selection of tion activities in the field of childhood injury is
outcomes and the selection of controls in constrained by the wide range of injury types
programme evaluations. If existing health infor- and variety of possible interventions. As death and
mation systems are used, only a limited amount serious injury are relatively rare events, attribut-
of information is collected and outcomes tend to ing health gain to a single health promotion
relate to the uptake of health services. The intervention may not be appropriate. Under these
existence of similar data collection systems in circumstances, the collection of process data,
other areas may govern the selection of control such as information on programme reach, may
communities. This may be a very arbitrary means improve our understanding of the impact of
of choosing controls and lead to the selection of community-based approaches.
control areas that appear to be very different The range of process measures employed in
from intervention communities. the different studies was diverse. Detailed case
Of the 10 programmes reviewed in this paper, studies used in the Waitakere study, for example,
only one has included multiple intervention and documented the importance of different models
control communities: the Statewide Child Injury of programme delivery to be tailored to the
Prevention Program from the USA, which needs of different cultural groups (study 10). The
selected these communities from 351 potential Maori component of the project was based in a
sites in the state of Massachusetts. Its inter- Marae (Maori community grouping) and allowed
vention, however, only took place over a 22 month distinctive Maori perspectives of ‘a holistic view
period, far shorter than in many of the other of health and well-being’ to be incorporated into
programmes. In the other programmes only one the programme. Other process measures included
control community was selected, sometimes with the degree of community involved, as reported
comparisons with national statistics or a broader in the ‘Shire of Bulla Program’ (study 5), media
area. In the Harstad Programme, the intervention reporting (Falköping study), and indicators of
community of Harstad with a population of a shift in the culture of safety within an organ-
23 000 was compared with the city of Trondheim, ization (the local council requiring all projects
a much larger city, 1000 km away. The Safe and programmes to state how their project meets
Kids/Healthy Neighborhoods Program in Harlem or furthers safety) as in the Waitakere study
had one intervention and one control area, and (study 10).
although both were disadvantaged communities,
the demographic characteristics of the two areas
were different. Even when the intervention and CONCLUSION
control areas were of similar size and socio-
demographic mix, as in the case of the Falköping There is increasing evidence emerging regarding
programme (study 1), there was considerable the effectiveness of community-based injury
under-reporting in the control area, which prevention programmes. The use of multiple
resulted in difficulties in interpreting the results. interventions implemented over a period of
Demonstrating the effectiveness of complex time can allow injury prevention messages to be
interventions is not straightforward. Community- repeated in different forms and contexts and can
based, multi-faceted interventions that target a begin to develop a culture of safety within a
range of injury types do not lend themselves to community. Important elements of community-
experimental evaluation approaches. The value based programmes are a long-term strategy,
11. Community childhood injury prevention: what works? 283
effective and focused leadership, multi-agency Davidson, L., Durkin, M., Kuhn, L., O’Connor, P., Barlow,
collaboration, the use of local surveillance to B. and Heagarty, M. (1994) The impact of the Safe
Kids/Healthy Neighborhoods Injury Prevention program
develop locally appropriate interventions and in Harlem, 1988 through 1991. American Journal of
tailoring interventions to the needs of the com- Public Health, 84, 580–586.
munity. Time is also needed to coordinate existing Day, L., Ozanne-Smith, J., Cassell, E. and McGrath, A.
networks, and to develop new ones. However, a (1997) Latrobe Valley Better Health Project. Evaluation
of Injury Prevention Program 1992–1996, Report No.
positive and sustained impact of community- 114. Monash University Accident Research Centre/
based programmes on injury rates has not yet Victorian Health Promotion Foundation, Melbourne,
been demonstrated conclusively. There is a need Australia.
to develop valid and reliable indicators of impact Finney, J. W., Christophersen, E. R., Friman, P. C., Kalnins,
and outcome appropriate to community studies. I. V., Maddux, J. E., Peterson, L. et al. (1993) Society of
Pediatric Psychology Task Force report: pediatric
Where proxy measures are used for injury out- psychology and injury control. Journal of Pediatric
comes, it is important that there is clear evidence Psychology, 18, 499–526.
of the association between the proxy (e.g. hazard Guyer, B., Gallagher, S., Chang, B., Azzara, C., Cupples, L.
removal, knowledge gain or behaviour change) and Colton, T. (1989) Prevention of childhood injuries:
evaluation of the Statewide Childhood Injury Prevention
and injury risk (Towner et al., 1996). There is also Program (SCIPP). American Journal of Public Health, 79,
an urgent need to develop and monitor indicators 1521–1527.
to assess and monitor a culture of safety, pro- Hennessey, M., Arnold, R. and Harvey, P. (1994) The First
gramme sustainability and long-term community Three Years: Final Report of the First Three Years of the
involvement. Community-based injury prevention Shire of Bulla’s Safe Living Program (1991–1993). Shire
of Bulla, Victoria, Australia.
programmes have been hampered by the lack of Jeffs, D., Booth, D. and Calvert, D. (1993) Local injury infor-
resources allocated to both their programme mation, community participation and injury reduction.
development, and appropriate and rigorous Australian Journal of Public Health, 17, 365–372.
evaluation. Kuhn, L., Davidson, L. L. and Durkin, M. S. (1994) Use of
Poisson regression and time series analysis for detecting
changes over time in rates of child injury following a
prevention program. American Journal of Epidemiology,
ACKNOWLEDGEMENTS 140, 943–955.
Lindquist, K., Timpka, T., Schelp, L. and Ahlgren, M. (1998)
This project was funded by England’s NHS The WHO safe community program for injury
prevention: evaluation of the impact on injury severity.
Executive National R&D Programme in ‘Mother Public Health, 112, 385–391.
and Child Health’ (MCH 10-21). Manciaux, M. and Romer, C. (eds) (1991) Accidents in
Childhood and Adolescence. The Role of Research. World
Address for correspondence: Health Organization, Geneva, Switzerland.
E. Towner Moller, J. (1992) Community Based Injury Prevention. A
Community Child Health Practical Guide. National Safety Council of Australia,
University of Newcastle upon Tyne South Australia.
Donald Court House Nuffield Institute for Health and NHS Centre for Reviews
13 Walker Terrace and Dissemination (1996) Preventing unintentional
Gateshead NE8 1EB injuries in children and young adolescents. Effective
UK Health Care, 2, 1–16.
E-mail: e.l.m.towner@ncl.ac.uk Ozanne-Smith, J., Sherrard, J., Brumen, I. and Vulcan, P.
(1994) Community Based Injury Prevention Evaluation
Report: Shire of Bulla Safe Living Program. Monash
University Accident Research Centre (MUARC),
REFERENCES Victoria, Australia.
Petridou, E., Tolma, E., Dessypris, N. and Trichopoulis, D.
Arblaster, L., Entwistle, V., Lambert, M., Forster, M., (1997) A controlled evaluation of a community injury
Sheldon, T. and Watt, I. (1995) Review of the Research on prevention project in two Greek islands. International
the Effectiveness of Health Service Interventions to Reduce Journal of Epidemiology, 26, 173–179.
Variations, Report Number CRD Report 3. NHS Centre Puska, P., Toumiletito, J., Nissinen, A. et al. (1989) The
for Reviews and Dissemination, University of York, UK. North Karelia Project: 15 years of community-based
Coggan, C., Patterson, P., Brewin, M., Douthett, M. and prevention of coronary heart disease. Annals of
Norton, R. (1998) Process Evaluation Report of the Medicine, 21, 169–173.
Waitakere Community Injury Prevention Project. Injury Robertson, L. S. (1986) Community injury control programs
Research Centre, University of Auckland, New Zealand. of the Indian Health Service: an early assessment. Public
Coggan, C., Patterson, P., Brewin, M., Hooper, R. and Health Reports, 101, 632–637.
Robinson, E. (2000) Evaluation of the Waitakere Sahlin, Y. and Lereim, I. (1990) Accidents among children
Community Injury Prevention Project. Injury Prevention, below school age. Changes of incidence after intervention.
6, 130–134. Acta Paediatrica Scandanavica, 79, 691–697.
12. 284 E. Towner and T. Dowswell
Schelp, L. (1987) Community intervention and changes in for the Prevention of Unintentional Injuries. Health
accident pattern in a rural Swedish municipality. Health Education Authority, London, UK.
Promotion, 2, 109–125. Towner, E., Dowswell, T., Mackereth, C. and Jarvis, S.
Schwarz, D., Grisso, J., Miles, C., Holmes, J. and Sutton, R. (2001) What Works in Preventing Unintentional Injuries
(1993) An injury prevention program in an urban in Children and Young Adolescents? An Updated
African–American community. American Journal of Systematic Review. Health Development Agency,
Public Health, 83, 675–680. London, UK.
Secretary of State for Health (1999) Saving Lives: Our WHO (1986) Ottawa Charter for Health Promotion. WHO,
Healthier Nation. The Stationery Office, London, UK. Geneva, Switzerland.
Speller, V., Learmouth, A. and Harrison, D. (1997) The WHO (1989) Karolinska Institutet, Stockholm. Manifesto for
search for evidence of effective health promotion. British Safe Communities. Adopted at First World Conference on
Medical Journal, 315, 361–363. Accident and Injury Prevention. Stockholm, 1989.
Svanström, L., Ekman, R., Schelp, L. and Lindstrom, A. WHO (1996) Investing in Health Research and
(1995) The Lidköping Accident Prevention Development: Report of the Ad Hoc Committee on Health
Programme—a Community Approach to Preventing Research Relating to Future Intervention Options, Report
Childhood Injuries in Sweden. Injury Prevention, 1, Number TDR/GEN 96.1. WHO, Geneva, Switzerland.
169–172. Ytterstad, B. (1995) The Harstad Injury Prevention Study:
Svanström, L., Schelp, L., Ekman, R. and Lindstrom, A. hospital-based injury recording used for outcome
(1996) Falköping, Sweden, ten years after: still a safe evaluation of community-based prevention of bicyclist
community? International Journal for Consumer Safety, and pedestrian injury. Scandinavian Journal of Primary
1, 1–7. Health Care, 13, 141–149.
Tellnes, G. (1985) An evaluation of an injury prevention Ytterstad, B. and Sogaard, A. (1995) The Harstad Injury
campaign in general practice in Norway. Family Practice, Prevention Study: prevention of burns in small children
2, 91–93. by a community-based intervention. Burns, 21, 259–266.
Tones, K. and Tilford, S. (1994) Health Education. Ytterstad, B. and Wasmuth, H. H. (1995) The Harstad
Effectiveness, Efficiency and Equity. Chapman and Hall, Injury Prevention Study: evaluation of hospital-based
London, UK. injury recording and community-based intervention for
Towner, E., Dowswell, T. and Jarvis, S. (1993) Reducing traffic injury prevention. Accident Analysis and
Childhood Accidents. The Effectiveness of Health Prevention, 27, 111–123.
Promotion Interventions: a Literature Review. Health Ytterstad, B., Smith, G. and Coggan, C. (1998) Harstad
Education Authority, London, UK. injury prevention study: prevention of burns in young
Towner, E., Dowswell, T., Simpson, G. and Jarvis, S. (1996) children by community based intervention. Injury
Health Promotion in Childhood and Young Adolescence Prevention, 4, 176–180.