Content uploaded by Harry Prapavessis
Author content
All content in this area was uploaded by Harry Prapavessis on Feb 23, 2016
Content may be subject to copyright.
289
JOURNAL OF SPORT & EXERCISE PSYCHOLOGY, 2005,
27
, 289-310
© 2005 Human Kinetics, Inc.
Both authors were with the Dept. of Sport and Exercise Science at the Univ. of Auckland
at the time of this study. R. Maddison is now with the Clinical Trials Research Unit, School
of Population Health, Univ. of Auckland, Private Bag 92019, Auckland, New Zealand.
A Psychological Approach to the Prediction
and Prevention of Athletic Injury
Ralph Maddison and Harry Prapavessis
The University of Auckland
Two interrelated studies examined the role psychological factors play in the
prediction and prevention of sport related injury. Study 1 involved 470 rugby
players who completed measures corresponding to variables in the revised
Williams and Andersen (1998) stress and injury model at the beginning of the
2001 playing season. Prospective and objective data were obtained for both
the number of injuries and the time missed. Results showed that social support,
the type of coping, and previous injury interacted in a conjunctive fashion to
maximize the relationship between life stress and injury. Study 2 examined the
effectiveness of a cognitive behavioral stress management (CBSM) intervention
in reducing injury among athletes from Study 1 who were identi ed as having
an at-risk psychological pro le for injury. Forty-eight players were randomly
assigned to either a CBSM intervention or a no-contact control condition.
Participants completed psychological measures of coping and competitive
anxiety at the beginning and end of the 2002 rugby season. The assessment of
injury was identical to that used in Study 1. Results showed that those in the
intervention condition reported missing less time due to injury compared to
their nonintervention counterparts. The intervention group also had an increase
in coping resources and a decrease in worry following the program. Taken
together, both studies underscore the importance of (a) psychosocial factors in
identifying those athletes most vulnerable to injury and (b) cognitive behavioral
stress management programs in reducing the vulnerability to injury.
Key Words: stress-injury relationship, rugby, psychosocial factors, intervention
Athletic injury is a common occurrence and concern for those who participate
in sport and recreational activities (Uitenbroek, 1996). In the United States, high
rates of injury have been reported as a consequence of participation in recreational
activities (Booth, 1987). Similarly, in New Zealand, a country of 4 million people,
sport and recreational activities were reported to represent 16% of all new claims
to the national no-fault injury compensation scheme (Accident Rehabilitation
Compensation and Insurance Corporation, 2002). Of those claims, rugby had the
highest rate of injury and represented 4% of all new claims in 2002.
290
/
Maddison and Prapavessis
Information on the incidence and nature of Rugby Union and Rugby League
injuries suggest that injury is multifactorial, but more often than not it involves
contact due to tackling (e.g., Wilson, Quarrie, Milburn, & Chalmers, 1999). This
is consistent with the medical literature that has typically focused on physical
aspects of sport injury and largely ignored the importance of psychosocial factors.
However, during the past 15 years a substantial body of literature has developed
to con rm that psychosocial factors play a signi cant role in our understanding of
the occurrence of, reaction to, and prevention of sport injuries (Williams, 2001).
The Andersen and Williams (1988) stress and injury model has driven much of the
research conducted in this area.
The core mechanism in this model is the stress response, a bidirectional
relationship between athletes’ cognitive appraisals of demands, consequences, and
resources in the sport situation and their physiological reactions (e.g., increased
muscle tension) and attentional responses (e.g., increased distractibility, narrow-
ing of the visual eld). These variables may increase the vulnerability to injury by
disrupting one’s coordination and exibility as well as interfering with the detection
of important environmental cues.
Above the core of the model (e.g., stress response) are three major factors:
personality, history of stressors, and coping resources. These factors may operate
alone or in combination to affect the stress response, and in turn the occurrence
and severity of injury. It is suggested through the model that these psychosocial
variables in uence how athletes respond under acutely stressful situations, but
only the athletes’ response itself directly affects their susceptibility to injury. In
short, depending on the extent to which these psychosocial variables are present,
the athletes’ stress responsivity is attenuated or exacerbated (Petrie & Perna, 2004).
For example, it has been hypothesized that athletes with many life stressors, few
coping resources, and certain personality dispositions (e.g., high competitive anxi-
ety) will, when placed in a stressful situation, demonstrate a greater stress response
(e.g., generalized muscle tension and distractibility) and hence be more at risk of
injury. Athletes with this high-risk pro le will have a greater likelihood of injury
than those with the opposite pro le (Williams & Andersen, 1998).
The nal component of the model refers to interventions. It is suggested that
in order to prevent injuries caused by stress, the intervention should focus on (a)
altering the cognitive appraisal of potentially stressful events, and (b) modifying
the physiological and attentional aspects of the stress response.
The Andersen and Williams (1988) and the subsequent revised Williams
and Andersen (1998) models have provided a framework for most of the research
concerning preinjury factors. Notwithstanding this, other researchers (Perna &
McDowell, 1995; Perna, Schneiderman, & LaPerriere, 1997) have characterized
stress responses more broadly (see Petrie & Perna, 2004, for a review). Perna and
colleagues argued that to fully understand the stress response, one must consider
its effects on cognitive, emotional, behavioral, and physiological systems (Petrie
& Perna, 2004).
As to the occurrence of athletic injury, researchers have been primarily inter-
ested in determining whether certain psychosocial variables can predict vulnerability
or resiliency to injury. Although it is impossible to directly measure the proposed
stress response mechanisms during competition, much of the research has exam-
ined the relationship between life events (or life stress) and injury. A meta-analysis
provides support for the life stress/injury relationship (Kontos, 2001). Moreover,
Prediction of Athletic Injury / 291
two reviews found that athletes with high life stress were two to ve times more
likely to sustain injury than athletes with low life stress (Williams, 2001; Williams
& Roepke, 1993).
In the Williams and Andersen (1998) model is the heading of coping resources,
which incorporates a wide variety of behaviors and social networks that help indi-
viduals deal with the stresses of life. Coping resources refers both to the various
coping skills and the amount of social support an individual receives. A number of
studies have examined whether social support and/or coping resources moderate the
life stress/injury relationship. A moderator variable is a qualitative or quantitative
variable that affects the nature, direction, or strength of a relationship between an
independent or predictor variable and a dependent or criterion variable (Baron &
Kenny, 1986). For example, Petrie (1992) found that among gymnasts with low
social support (lower third distribution of social support scores), high negative
life events accounted for 6 to 12% of the injury variance outcome. No signi cant
relationship between life stress and injury outcome was found among gymnasts in
the high social support group. Petrie suggested that low social support increases
one’s vulnerability to injury whereas high social support seems to provide some
protection against injury. Similar results were reported by Patterson, Smith, Everett,
and Ptacek (1998) with ballet dancers. Furthermore, in their seminal paper, Smith,
Smoll, and Ptacek (1990a) reported that negative life events accounted for 22 to
30% of the injury time-loss variance for athletes who were low in both social sup-
port and coping skills.
Of the personality moderator variables proposed in the Williams and Andersen
model (1998), competitive anxiety has been the most widely examined. Persons with
high trait anxiety may see more situations as stressful and consequently experience
an elevated stress response, which in turn may predispose them to injury. With
respect to injury prediction, some researchers have divided their sample based on
injury status (injured vs. noninjured) and have found that greater number (or sever-
ity) of injuries was related to higher competitive trait anxiety scores (e.g., Lavelle &
Flint, 1996). Petrie (1993) showed that competitive trait anxiety had both a direct and
an indirect effect on injury. For the direct effect, Petrie found that increases in trait
anxiety were positively associated with injury rate for college football starters, but
not for nonstarters. With respect to the indirect effect, Petrie found that competitive
trait anxiety moderated the effects of positive life stress, in that greater time lost
due to injury was associated with higher levels of anxiety and stress.
As highlighted in her review, Williams (2001) stated, “except for Petrie (1993)
none of the studies reviewed employed designs that permitted testing whether their
personality variables interacted with history of stressors or with other personality
and coping variables in in uencing injury risk. Such limited designs will not elu-
cidate the potential complexity of the relationship of personality factors to injury
vulnerability and resiliency” (p. 772).
The third moderator variable category in the stress/injury model is history
of stressors, which incorporates previous injury. The reasons for including previ-
ous injury in the model are that (a) an athlete returning to sport when not fully
recovered is at greater risk of reinjury, and (b) an athlete who is physically but not
psychologically ready for return to sport risks being reinjured because of potential
negative cognitive appraisal and anxiety (Williams, 2001). The relationship between
previous injury and vulnerability to subsequent injury has received little empirical
investigation and remains to be clari ed. Hanson et al. (1992) reported that the period
292
/
Maddison and Prapavessis
of time from injury recovery was unrelated to subsequent frequency or severity
of injury, whereas others have reported a positive relationship between previous
injury and subsequent reinjury (e.g., Lysens, Van den Auweele, & Ostyn, 1986).
Van Mechelen, Twisk, Molendijk, et al. (1996) observed that previous injury was
a better predictor of injury than psychological, psychosocial, physiological, or
anthropometric factors.
As highlighted above, a shortcoming of past research is the failure to test how
other moderating variables of the model such as history of stressors and personality
might interact with coping resources (social support and coping skills) to in uence
the life stress/injury relationship. Also, notes Williams, “no studies have used the
Smith et al. (1990[a]) paper as a prototype for future injury research” (2001, p. 776),
in that similar design and statistics have not been used. Hence the aim of Study 1
was to use the revised Williams and Andersen (1998) stress and injury model as a
framework to predict sport injury in New Zealand rugby players. A second aim was
to use the Smith et al. (1990a) study as a prototype from which to base statistical
procedures and analyses to determine the extent to which individual differences
in athletes’ coping resources, history of stressors, and personality scores, both
separately and in combination, affected the magnitude of the correlation between
life stress and sport injury.
Insofar as injury prediction was concerned, the following hypotheses were
generated: A mild positive relationship would be found between life stress and sport
injury. Social support and coping would moderate relationships between life stress
and injury. Furthermore, these two variables would interact with other moderator
variables—personality, competitive anxiety, history of stressors, previous injury—in
a conjunctive fashion (multiple moderators co-occur in a speci c combination or
pattern) to maximize relationships between life stress and sport injury. Speci cally,
relationships between life stress and injury would be maximized for those athletes
low in social support, high in avoidance coping, and high in both history of previous
injury and competitive anxiety (somatic, worry, and concentration disruption).
Method — Study 1
Participants
Rugby (Union and League) players, 600 males from 37 teams, were recruited
in the present study. In all, 470 players provided complete injury and questionnaire
data and were included in the nal analysis. The participants ranged in age from
16 to 34 years (
M
= 20.69,
M = 20.69, M
SD
= 4.18) and represented various ethnic groups (NZ
European 50%; NZ Maori 17%; Paci c Islanders 28%; Other 5%). All played at a
competitive school (1st XV) or club level.
Predictor Variables
Life Stress
. The Life Events Survey for Collegiate Athletes (LESCA; Petrie,
1992) was used to assess life stress. The LESCA is a 69-item questionnaire used to
measure positive and negative life change. Participants were asked to report each
such event they had experienced in the previous 12 months. For each life event
experienced, the athlete had to indicate its impact on an 8-point Likert scale from
–4 (
extremely negative
) to +4 (
extremely positive
). In the present study we made
the following modi cations to the LESCA to increase its appropriateness to male
Prediction of Athletic Injury / 293
noncollege athletes: (a) three female items were deleted (e.g., menstrual period/
PMS), and (b) items that represented only college athletes were changed to include
noncollege athletes (e.g., “being dismissed from college residence” was changed to
“being dismissed from school or home residence”). Two life stress scores, negative
(NLE) and positive (PLE), can be derived by summing the respective life stress
values. Petrie (1992) reported adequate reliability and validity of this scale. Test-
retest reliabilities ranged from .76 to .84. In the present study only the NLE scale
was used because of its consistent relationship to injury (Kontos, 2001).
Psychological Variables
Personality-Competitive Anxiety
. The Sport Anxiety Scale (SAS: Smith,
Smoll, & Schutz, 1990b) was used to measure sport-speci c competitive anxiety.
The 21-item scale assesses three factors: somatic anxiety, worry, and concentra-
tion disruption. Recently, Dunn, Causgrove-Dunn, Wilson, and Syrotiuk (2000)
reexamined the factor structure and factor composition of the SAS using male
college and high school athletes. Overall they found that their data supported the
tenability of the original three-factor model proposed by Smith et al. (1990b).
However, two of the items (i.e., 14 and 20) originally designed to measure con-
centration disruption loaded on the worry factor. The reconstituted scale as used
in the present study, with possible scores and internal consistency values, was as
follows: somatic anxiety, 8–32 (
α
= .84), worry, 9–36 (
α
= .84), and concentration
disruption, 3–12 (
α
= .80).
Coping Resources
. A modi ed version of the Ways of Coping Scale (M-
WCS; Grove, Eklund, & Heard, 1997) was used to measure how often speci c
strategies are used in coping with the stress of competition. The M-WCS is a
26-item multidimensional scale that measures ve coping components: social
support, denial/avoidance, wishful thinking, emotional control, and effort resolve.
Cronbach’s alpha values for the separate scales were denial, .62; social support, .82;
effort, .76; wishful thinking, .78; and emotional control, .46. Emotional control was
removed from subsequent analyses. Coping consists of cognitive and behavioral
strategies used to manage stress and, according to Folkman and Lazarus (1985),
it takes two basic forms. The rst general form is termed problem-focused coping
and refers to attempting, altering, or removing the stressor. The second form is
termed emotion-focused coping and involves an attempt to regulate the emotions
or distress created by the stressor.
Endler and Parker (1990) have also proposed avoidance coping, which may
serve instrumental and person oriented bene ts by providing a break from the stressful
situation (Grove et al., 1997). Consistent with theoretical underpinnings of coping
and to reduce the number of variables for analyses, we derived a problem-focused
scale by summing values from effort and seeking social support (
α
= .84). An avoid-
ance-focused coping scale was derived by summing values from denial/avoidance
and wishful thinking (
α
= .75). The possible scores for both scales were 0–30.
Social Support
. The Social Support Questionnaire (Smith et al., 1990a) was
used to assess the amount of social support available to the individual. On separate
scales, participants indicated the extent to which each individual and group could
be counted on to provide them with (a) emotional support and caring, and (b) help
and guidance, on a scale ranging from 1 (
not all helpful
)
to 5 (
very helpful
). Scores
were summed to provide overall measures of social support and help and guidance.
294
/
Maddison and Prapavessis
Smith et al. (1990a) reported 1-week test-retest reliability of .87 for emotional sup-
port and .88 for help and guidance. In the present study the two subscales (social
support and help and guidance) were highly correlated (
r
= .90), hence only the
r = .90), hence only the r
social support subscale was used in subsequent analyses (
α
= .85). It is important
to note that the social support measure used differs from the social support subcom-
ponent of the M-WCS. The social support questionnaire is a general measure that
assesses those parties available to participants to provide support and how helpful
they are in providing emotional support, whereas the M-WCS is sport speci c and
its social support subscale assesses the seeking of social support (e.g., “I look for
help”). These two measures were mildly correlated (
r
= .26) and were considered
r = .26) and were considered r
to be different enough to be tapping separate constructs.
History of Stressors
. In the Williams and Andersen model (1998), possible
history of stressors includes previous injury, major life events, and daily hassles
(Williams, 2001). In the present study, history of stressors (previous injury history)
was assessed by asking athletes to record the number of injuries sustained during
the previous rugby season.
Dependent Variable
Injury
. Previous studies have generally classi ed injuries as those requiring
medical treatment and forcing the athlete to miss at least 1 day of play or training.
Andersen and Williams (1999) suggested there is a need to collect injury data that
re ect minor as well as major injuries, including injury that requires modi cation
to play (wearing protective head gear, strapping, etc.). They argued that collecting
this type of injury data is more in line with the prediction of their model in that
psychological factors are related to the number of injuries, regardless of severity,
that occur throughout the season.
Although we acknowledge the recommendation of Andersen and Williams
with respect to injury occurrence, their assessment of injury is not the only one
endorsed in the literature. For instance, Hodgson Phillips (2000) suggested that
time lost from participation must be recorded accurately, using data on both train-
ing and game/competition participation. She suggested that failure to do so would
result in the loss of valuable data and would not portray the true picture of injury
in that sport. Indeed, a number of studies have used time missed due to injury as
an indirect measure of injury severity (e.g., Petrie, 1992; Smith et al., 1990a).
Because injury occurrence and time missed are inexorably linked, two approaches
were taken to assess the dependant variable (injury). The rst approach assessed
time missed due to injury, which re ected total hours missed as a result of injury
during either a game or training.
Increasingly, incidence rates in all sports are being expressed as rate per
1,000 hours. Hodgson Phillips (2000) stated that this is a good approach and allows
some comparison across sports. However, a further re nement of the calculation
of incidence rates is to measure the actual exposure time at risk. In line with these
suggestions, we calculated injury time missed using the following formula.
1
Total time missed due to injury
Number of players
×
Total time played and trained
×
1,000 hrs
Prediction of Athletic Injury / 295
The second approach included an assessment of the number of injuries, which
included all injuries resulting in an athlete having to modify at least one game or
practice, or receiving treatment before or after practice, but not necessarily result-
ing in time loss. A formula identical to the one above was used to calculate total
number of injuries by substituting total time missed with total number of injuries.
As players and teams varied in the amount of time played and trained, all data
were corrected for exposure to injury and re ect injury incidence per 1,000 hours
(Hodgson Phillips, 2000).
Procedure
Approval for the study was obtained from the university ethics committee.
At the beginning of the 2001 season, Rugby Union and Rugby League clubs were
contacted to seek their willingness to participate in the present study. The principal
investigator contacted the team manager or coach to gain access to team players.
Players were given a demographic sheet (age, ethnic af liation, height, weight), a
battery of measures, and consent forms to complete prior to a team practice, which
were then collected by research assistants upon completion.
To obtain a prospective and an objective assessment of injury, the team’s
coaches or managers were paid to record injuries on a weekly basis. During the
preseason, the researchers instructed all coaches on how to complete the injury
sheets. Each week coaches or managers completed standardized injury data sheets
that indicated: (a) whether a player played a game; (b) the number of minutes
played; (c) whether an injury occurred during the game; and (d) whether the player
missed any time due to that injury. Identical data were collected for training time.
Research assistants collected the injury data sheets each week and returned them
for data entry.
Results — Study 1
Treatment of the Data
In accordance with Smith et al.’s (1990a) assessment of potential modera-
tor effects, product-moment correlations were computed between the life stress
measure (NLE) and injuries for participants in the upper and lower thirds of the
social support, coping resources, history of stressors, and personality distribution
scores. This straightforward approach is appropriate on statistical grounds, provided
that restriction in range and variance in the predictor and dependent variable are
not present in the subgroups de ned by the moderator variable(s) (Arnold, 1982;
Cohen & Cohen, 1983). In addition, regression of the injury measures on nega-
tive life events separately for each moderating condition should produce similar
Y (injury) intercepts.
Fundamental differences in intercepts could present dif culty in suggesting
vulnerability or resiliency effects, even if the correlation coef cients differ (Cohen
& Edwards, 1989). Finally, clarity in demonstrating conjunctive moderator patterns
requires that the moderator variables of interest are not highly correlated with one
another (Baron & Kenny, 1986). Preliminary checks were conducted to ensure that
none of these three statistical assumptions were violated. In addition, the distribution
of the time-loss data was positively skewed and was subjected to logarithmic trans-
formation to reduce a potential spurious in uence of extreme scores (Tabachnick &
296
/
Maddison and Prapavessis
Table 1 Descriptive and Correlation Data for the Variables of Interest (
N
= 470)
N = 470)N
Variable
M
SD
Range
1
2
3
4
5
6
7
8
9
10
1. Negative life events
9.28
10.78
0–72
_
.12
*
.23
**
.10*
.13
**
.24
**
.23
**
.24
**
.09
.11
*
2. Social support
40.65
15.60
5–83
_
.15
**
.25
**
.07
.07
–.01
.15
**
.07
.08
3. Avoidance coping
11.19
5.73
0–29
_
.34
**
.18
**
.39
**
.25
**
.11*
.01
–.01
4. Problem coping
14.82
5.75
1–30
_
.16
**
.13
**
–.03
.07
.06
.12
*
5. Somatic anxiety
16.18
4.65
8–32
_
.47
**
.21
**
.09
.04
–.07
6. Worry
17.05
4.75
9–36
_
.35
**
.11
*
.01
–.02
7. Concentration disrup.
4.73
1.76
3–12
_
.08
.07
.02
8. Previous injury
1.35
1.27
0–8
_
.12
*
.13
**
9. Injury time missed
a
3.30
7.42
0–67
_
.69
**
10. Injury number
a
0.95
1.37
0–8.9
_
Note
:
a
Represents the exposure-corrected injury time loss and injury number data per 1,000 hrs. Log-transformed injury (time missed and
number) are used in correlation matrix.
*
p
< .05;
**
p
< .01
Prediction of Athletic Injury / 297
Fidell, 2001). Finally, due to the number of correlations performed, and to reduce
the probability of Type 1 error, we set Bonferroni corrected alpha values at 0.01.
Descriptive Data
. Descriptive data for the psychosocial and injury variables
are presented in Table 1. Forty-six percent of players suffered at least one injury
during the playing season. A correlation matrix of all the variables of interest is also
presented in Table 1. As expected, mild relationships were found between negative
life events and both time loss due to injuries (
r
= .09,
r = .09, r
p =
.05) and number of injuries
(
r
= .11,
r = .11, r
p
< .05). History of stressors (previous injury) also was mildly correlated
with both time loss due to injuries (
r
= .12,
r = .12, r
p =
.01) and number of injuries (
r
=
r = r
.13,
p
< .01). Social support, coping skills (avoidance-focused), and personality
(somatic, worry, and concentration disruption) were not correlated to either injury
measure (all
p
values > .05). A mild relationship was found, however, between
problem-focused coping and injury number (
r
= .12,
r = .12, r
p
< .05). Of the moderator
variables used in the present study, only worry and avoidance coping were moder-
ately correlated (
r
= .39,
r = .39, r
p
< .01).
Testing for Moderation
. For disjunctive (single moderator) effects, nega-
tive life event/injury (time missed) correlations were generally higher for athletes
who were either low in social support, high in avoidance-focused coping, or high
in problem-focused coping. A signi cant relationship between NLE and injury
(number of injuries) was found for those in the high avoidance-focused coping
subgroup (Table 2).
In line with Smith et al. (1990a), the next logical step was to assess possible
conjunctive moderator effects of social support and coping resources by examining
subgroups of rugby players who fell in the upper and lower thirds of the distribu-
Table 2 Correlations Between Negative Life Events and Injury for Single
Moderator Variables
Moderator
Negative life events
Injury time missed
No. of injuries
variables
N
M
SD
M
SD
r
M
SD
r
Low social support
162
7.78
9.79
.30
.43
.16
.37
.42
.04
High social support
162
10.21
9.94
.37
.46
.05
.42
.43
.14
Low avoidance-
focused coping
157
7.47
9.57
.33
.44
.11
.39
.43
.08
High avoidance-
focused coping
178
11.29
12.11
.32
.44
.17
.35
.42
.23*
Low problem-
focused coping
167
7.80
9.29
.32
.46
.09
.32
.41
.09
High problem-
focused coping
178
11.29
12.12
.33
.44
.17
.42
.43
.13
Note
: High and low = upper and lower thirds of the distributions; Range restriction viola-
tions were not present in any subgroup; The
y
intercepts were similar, with the greatest
difference less than .3
SE
of the constant (intercept).
SE of the constant (intercept). SE
*
p
*p*
< .01
298
/
Maddison and Prapavessis
tions on both measures. Descriptive and correlation data are presented in Table 3
for the four respective subgroups of athletes on the life events measures and for
the transformed injury variables. As can be seen in Table 3, greater correlation
coef cients for NLE and injury (both time missed and injury number) were seen
in the low social support and high avoidance-focused coping subgroup, as well as
in the low social support and high problem-focused coping subgroup (time missed
only). Statistical tests for differences among two or more correlation coef cients
(Edwards, 1984, p. 74) were applied and revealed that the coef cients of the four
respective conditions were not statistically different (
χ
2
= 4.16,
p
= .10,
df
= 3).
df = 3). df
Next, possible conjunctive moderator effects of social support and coping
resources with the addition of personality or previous injury were assessed by select-
ing rugby players who fell in the upper and lower third distributions on all three
measures (e.g., social support, coping, and personality, or social support, coping,
and previous injury). Descriptive data for the eight subgroups for three moderating
variables are presented in Table 4. As can be seen, a signi cant NLE/injury (time
missed) relationship was found for athletes in the low social support, high avoid-
ance-focused, and high previous injury subgroup. Statistical tests for differences
among two or more correlation coef cients (Edwards, 1984, p. 74) were applied
and revealed that the coef cients of the eight respective conditions did not differ
signi cantly (
χ
2
< 14.06,
p
> .05,
df
= 7). No statistically signi cant conjunctive
df = 7). No statistically signi cant conjunctive df
relationships using three moderator variables were found when avoidance-focused
coping was replaced with problem-focused coping.
To further explore the nonsigni cant differences among the correlation coef-
cients, we followed a hierarchical regression approach using dummy variable
moderators as recommended by R.E. Smith (personal communication, July 14, 2003)
Table 3 Correlations Between Negative Life Events and Injury for Two Moderator
Variables
Moderator
Negative life events
Injury time missed
No. of injuries
variables
N
M
SD
M
SD
r
M
SD
r
Social support/ Avoidance coping
Low-low
66
6.5
9.75
.33
.43
.12
.43
.42
.02
Low-high
53
9.92
10.60
.30
.42
.39*
.32
.41
.32*
High-low
49
7.63
7.02
.35
.43
.04
.41
.43
.21
High-high
64
12.43
11.62
.38
.46
.07
.41
.42
.12
Social support/ Problem coping
Low-low
80
7.36
7.55
.30
.48
.11
.28
.41
.09
Low-high
48
9.18
13.15
.31
.38
.38*
.47
.44
.06
High-low
38
8.44
7.37
.38
.45
.02
.40
.41
.02
High-high
84
11.11
11.02
.41
.49
.06
.43
.43
.15
Note
: High and low = upper and lower thirds of the distributions; Range restriction viola-
tions were not present in any subgroup; The
y
intercepts were similar, with the greatest
difference less than .3
SE
of the constant (intercept). *
SE of the constant (intercept). *SE
p
of the constant (intercept). *p of the constant (intercept). *
< .01
Prediction of Athletic Injury / 299
Table 4 Correlations Between Negative Life Events and Injury for Three
Moderator Variables
Moderator
Negative life events Injury time missed
No. of injuries
variables
N
M
SD
M
SD
r
M
SD
r
Social support/ Avoidance coping/ Somatic
Low-low-low
27
5.67
10.27
.26
.43
.17
.37
.41
.02
Low-high-low
17
7.53
8.48
.20
.32
.49
.31
.43
.47
Low-high-high
17
11.47
14.28
.47
.50
.35
.38
.44
.28
Low-low-high
17
9.64
12.99
.35
.46
.07
.42
.39
.08
High-high-high
52
13.08
11.72
.43
.49
.08
.42
.41
.23
High-high-low
11
8.09
10.26
.16
.25
.35
.36
.51
.46
High-low-low
18
10.77
8.42
.39
.42
.32
.55
.43
.12
High-low-high
14
4.78
5.36
.32
.49
.54
.19
.35
.45
Social support/ Avoidance coping/ Worry
Low-low-low
35
3.48
3.34
.27
.41
.29
.36
.43
.22
Low-high-low
16
9.31
11.13
.49
.46
.49
.47
.46
.48
Low-high-high
26
11.34
11.26
.27
.43
.33
.26
.38
.26
Low-low-high
9
17.11
21.50
.44
.51
.05
.37
.40
.09
High-high-high
31
17.71
11.35
.32
.47
.23
.35
.44
.40
High-high-low
16
6.06
5.81
.54
.47
.25
.56
.41
.37
High-low-low
28
8.71
6.74
.32
.44
.01
.39
.39
.04
High-low-high
13
5.15
5.01
.34
.42
.44
.28
.40
.36
Social support/ Avoidance coping/ Concentration disruption
Low-low-low
31
4.22
4.91
.27
.34
.26
.45
.44
.34
Low-high-low
9
6.77
9.95
.15
.38
.27
.18
.38
.33
Low-high-high
32
10.62
11.04
.35
.44
.40
.36
.42
.38
Low-low-high
23
9.95
14.63
.42
.55
.05
.43
.41
.16
High-high-high
35
14.57
12.99
.39
.44
.13
.41
.41
.26
High-high-low
15
7.93
6.70
.54
.59
.18
.48
.46
.16
High-low-low
27
6.74
5.04
.40
.48
.29
.42
.47
.09
High-low-high
16
8.75
9.03
.26
.38
.32
.37
.42
.34
Social support/ Avoidance coping/ Previous injury
Low-low-low
47
5.89
8.74
.23
.39
.01
.37
.44
–.05
Low-high-low
30
6.23
7.23
.25
.40
.11
.29
.41
.26
Low-high-high
23
14.74
12.38
.38
.45
.56*
.37
.45
.38
Low-low-high
19
8.00
12.07
.58
.45
.21
.57
.35
–.01
High-high-high
29
15.96
12.71
.41
.49
.18
.45
.45
.26
High-high-low
35
9.51
9.89
.35
.44
–.07
.37
.38
–.11
High-low-low
28
8.64
7.31
.33
.46
–.10
.36
.39
.12
High-low-high
21
6.28
6.54
.37
.40
.30
.46
.48
.37
Note
: No moderating effects were found when avoidance coping was replaced with prob-
lem-focused coping; Range restriction violations were not present in any subgroup; No
statistically signi cant conjunctive relationships using 3 moderator variables were found
with problem-focused coping. *
p
with problem-focused coping. *pwith problem-focused coping. *
< .01
300
/
Maddison and Prapavessis
based on an amendment to their 1990a published work. To do this we rst had to
treat each of the eight social support/ coping/ personality (or history of stressors)
subgroups as dummy variables entering NLE, the dummy coded groups, and the
NLE X dummy variable product score in that order. A signi cant product-term
interaction,
R
2
change = .06,
F
change (1, 123) = 8.89,
F change (1, 123) = 8.89, F
p
< .01, was found at Step
3. For those low in social support, high in avoidance-focused coping, and high on
previous injury, negative life events explained a signi cant amount of the injury
variance (beta = .26). Similar results were seen for number of injuries sustained,
R
2
change at Step 3 = .09,
F
change (6, 119) = 2.11,
F change (6, 119) = 2.11, F
p
= .05. Negative life events
explained a signi cant amount of injury variance (beta = .20) for those low in social
support, high in avoidance-focused coping, and high on previous injury. No other
moderating effects were found.
Discussion — Study 1
In general, our results provide support for the utility of the revised Wil-
liams and Andersen (1998) stress and injury model in predicting injury among
New Zealand rugby players. Beyond this observation, a number of aspects related
to speci c results should be highlighted. First, evidence for conjunctive and not
disjunctive moderation is provided. These results mirrored those of Smith et al.
(1990a), suggesting that when social support and coping are considered separately,
the amount of injury variance accounted for by negative life events is minimal
(up to 5% was seen in the high avoidance-focused coping subgroup, see Table 2).
However, the amount of injury variance (10–15%) accounted for by negative life
events was larger for those both low in social support and high in avoidance-focused
coping, as well as for those low in social support and high in problem-focused
coping (see Table 3).
The conjunctive relationship between low social support and high avoid-
ance-focused coping is intuitively appealing when one considers that avoidance
coping was a product of the factors denial and wishful thinking. This suggests
that individuals with an insuf cient social network, who also use more disengage-
ment, passivity, and/or fantasy coping strategies (e.g., denial and wishful thinking),
may deal with the rigors of training and competition by avoiding these situations
because they cause distress. If these efforts are ineffective, or rather the situation
is unavoidable (e.g., competition), the stress response may be enhanced. The role
of problem-focused coping (a product of the factors seeking social support and
effort/resolve) is less intuitively appealing. A possible explanation for this high
subgroup being at risk of injury is that inappropriate attempts to seek support, or
trying “too hard” with respect to effort, might prove frustrating and may exacerbate
the stress response. The joint in uence of coping resources and social support in
increasing one’s vulnerability to injury highlight the importance of these variables
when trying to understand life stress/injury relationships.
Second, this study extends the ndings of Smith et al. (1990a) by adding a
third moderating variable (personality or history of stressors) with social support
and coping to maximize life stress/injury relationships. Speci cally, when partici-
pants in the low social support, high avoidance-focused coping, and high previous
injury condition were considered, negative life events accounted for 31% of the
variance in injury (Table 4). This is an impressive nding when one considers that
the amount of injury variance explained by physical, environmental, and biome-
Prediction of Athletic Injury / 301
chanical factors leaves less to be accounted for by psychological factors. Statistical
tests for differences between more than two correlations showed, however, that
the correlations between the respective conditions did not differ signi cantly from
each other. These nonsigni cant ndings are consistent with those of Smith et al.
(1990a) and can most likely be attributed to sample size. As Arnold (1982) has
suggested, the subgroup correlation analyses require large sample sizes to provide
suf cient power.
A hierarchical regression approach using the moderators as dummy variables,
however, did show that history of stressors (previous number of injuries) acts with
social support and coping in a conjunctive pattern to produce a maximum moderator
effect. These results suggest that athetes low in social support, high in avoidance-
focused coping, and high on previous injury were most likely to miss time due to
injury, as well as sustain more injuries than those with the opposite pro le. Research
in this area has generally supported the notion that previous injury is related to
subsequent risk of injury risk (e.g., Lysens, Steverlynck, Van den Auweele, et al.,
1984). Previous injury may be important for a number of reasons. First, the athlete
may not be prepared physically or psychologically to return to sport, increasing
the likelihood of negative cognitive appraisals such as anxiety and fear of reinjury.
Second, injury may alter the athlete’s ef cacy to perform the task at the level
required, or he/she may be prone to attentional distuption due to worry or concern
over his/her injury (Williams, 2001). Our results highlight previous injury as an
important variable to consider in future research on life stress.
Unfortunately, evidence for a conjunctive moderator effect for personality
(i.e., competitive anxiety) was not evident. The correlation approach offered some
support for somatic anxiety, worry, and concentration disruption interacting with
both social support and coping to improve the relationships between negative life
events and injury; however, the relationships did not meet the statistical signi cance
criteria. We concur with Smith et al.’s (1990a) sentiments that there are opportuni-
ties to study the “interactions among personality variables…. To the extent that
moderator variables are selected on the basis of sound theory and previous research,
this approach need not be a ‘ shing expedition’ that capitalizes on chance ndings”
(1990, p. 368). It is worth noting that, due to insuf cient sample size in the present
study, we were unable to examine all possible multiple moderator effects using
the correlation approach. Larger studies are warranted to examine more than three
variables from the various components of the model: personality/anxiety; history
of stressors/previous injury; coping resources/social support and coping skills.
Third, a number of methodological and analytical issues warrant comment. For
instance, the Williams and Andersen (1998) model proposes a relationship between
the stress response and occurrence of injury, and does not speci cally refer to time
missed as an injury outcome. Despite this, time missed has often been used as an
injury dependant variable in a variety of studies as an indirect measure of the severity
of injury. We advocate that both types of information (time missed and number of
injuries) be collected so as to fully understand the complex relationship between
psychological factors and injury. From an analytical perspective, it is important to
be aware that injury data will always be skewed—because a large proportion of
scores will equal zero (no injury). Appropriate transformation techniques should
always be applied to injury data before analyzing and evaluating the results.
2
In the
present study we did not assess whether subsequent injury was in fact a recurrence
of a previous injury; this should be considered in future work (Williams, 2001).
302
/
Maddison and Prapavessis
A logical extension of the present study is to examine whether a psychologi-
cal intervention would be of bene t to those athletes identi ed as being at risk of
injury. As highlighted earlier, the least researched area of the Williams and Andersen
model (1998) is the implementation and assessment of psychological interventions
that are proposed to diminish the stress response and reduce one’s vulnerability to
injury. The effectiveness of interventions in reducing the vulnerability to injury is
a fertile area for research and was the aim of Study 2.
Study 2
The Williams and Andersen (1998) stress and injury model posits that in order
to prevent injuries caused by stress, an intervention should focus on (a) altering the
cognitive appraisal of potentially stressful events and (b) modifying the physiologi-
cal and attentional aspects of the stress response. In addition, these interventions
and others may be used to directly in uence the moderator variables under coping
resources and personality factors.
Only a handful of studies have offered empirical support for the effective-
ness of psychological interventions to prevent or reduce athletic injury (Davis,
1991; Johnson, Ekengren, & Andersen, 2005; Kerr & Goss, 1996; May & Brown,
1985; Perna, Antoni, Baum, Cordon, & Schneiderman, 2003; Schomer, 1990). Of
these, only three (Johnson et al., 2005; Kerr & Goss, 1996; Perna et al., 2003) offer
experimental evidence for a reduction in injury, all using a cognitive behavioral
stress management (CBSM) approach.
Kerr and Goss (1996) examined the effect of a CBSM intervention based on
Meichenbaum’s (1985) Stress Inoculation Training in the reduction of life stress and
injury. Although a signi cant treatment effect occurred for decreased life stress, a
nonsigni cant effect was found for injury reduction. Kerr and Goss’ explanation for
the nonsigni cant ndings related to the late introduction (halfway) of relaxation
and distraction control skills into the program. However, Andersen and Stoové
(1998) argued that the small number of participants in each group and the resultant
lack of power was most likely the reason for this lack of effect. Perna et al. (2003)
have provided evidence supporting the ef cacy of a CBSM intervention in reduc-
ing injury and illness among college athletes. Moreover, the CBSM intervention
was related to decreased cortisol and negative affect (indices of exercise training
maladaption). Most recently, Johnson et al. (2005) found that an intervention in
the form of a brief prevention program was effective in reducing the number of
injuries compared to a control group.
Although results from these studies are encouraging, no studies have tested
their intervention using athletes prospectively identi ed as having an at-risk psy-
chological pro le for injury. Study 2 was therefore undertaken to examine the
effectiveness of a CSBM intervention in reducing the vulnerability to injury among
athletes from Study 1 who were identi ed as having an at-risk psychological pro-
le to injury. A second purpose was to determine what might explain a positive
result. With respect to the rst purpose, it was hypothesized that athletes at risk of
injury in the CBSM intervention would have fewer injuries and miss less time due
to injury than their control condition counterparts. For the second purpose, it was
hypothesized that athletes in the CBSM intervention would also report increased
coping resources and decreased competitive anxiety (somatic, worry, concentration
disruption) compared to athletes in the control condition.
Prediction of Athletic Injury / 303
Method — Study 2
Participants
It was calculated that a sample of approximately 38 participants in each
condition (CBSM vs. control) would be needed to provide power of 80% (
α
= .05)
and to detect a large effect (i.e.,
d
= .40) between conditions on the variables of
d = .40) between conditions on the variables of d
interest (Cohen, 1988). Sample size derivations were made on the outcome vari-
ables, namely injury number and time missed. For the injury variables, effect sizes
(ES) between those in the intervention and those in the control have been shown
to be large (ES = .67) (Kerr & Goss, 1996; Perna et al., 2003). Sixty-four rugby
players from Study 1 who were most vulnerable to injury (see Procedure) agreed
to participate in the present study. Forty-eight participants provided complete data
in Study 2. They ranged in age from 17 to 33 years (
M
= 20.98,
M = 20.98, M
SD
= 1.42) and
represented a variety of ethnic groups (NZ European, 57%; NZ Maori, 6%; Paci c
Islanders 33%; Other 4%).
Participants were randomized to either an intervention (cognitive-behavioral
stress management program) or a control condition using Statistical Packages for
Social Sciences (SPSS) 11.5 software. Comparison of baseline group means for age,
injury, and psychological variables revealed no differences, suggesting pretreatment
group equivalence on these variables.
Procedure
Study approval was obtained from the university ethics committee prior to
the start of this project. At the beginning of the 2002 season (T1), Rugby Union
and Rugby League players from Study 1 were contacted to seek their willingness
to participate in the present study. Players were classi ed as at risk of injury if
they met any of the moderating conditions found in Study 1 (e.g., low in social
support, high in avoidance coping). Those who provided consent attended the
department’s testing laboratory for a study brie ng. All participants received
verbal and written information. They were instructed to complete a demographic
sheet (age, ethnic af liation, height, weight) before completing a battery of psy-
chological measures. Repeat questionnaire assessment took place at the end of
the 2002 season (T2).
To obtain a prospective assessment of injury, the researchers trained each
participant to record injuries on a weekly basis. To encourage completion of the
injury sheet, a small nancial incentive was provided to all athletes. Research assis-
tants also contacted the players each week to encourage completion and to collect
injury data sheets for data entry. The standardized injury sheets (as used in Study
1) indicated whether the player had played a game, the number of minutes played,
whether an injury had occurred, and any time missed due to that injury; identical
data were collected for training time.
Psychological Variables
Personality-Competitive Anxiety
. The Sport Anxiety Scale (SAS; Smith
et al., 1990b) was used to measure sport-speci c competitive anxiety. Cronbach
alpha values for the subscales were as follows for Time 1 and Time 2, respectively:
somatic, .86 and .86; worry, .80 and .82; concentration disruption, .80 and .82.
304
/
Maddison and Prapavessis
Coping Resources
. The Athletic Coping Skills Inventory-28 (ACSI-28;
Smith, Schutz, Smoll, & Ptacek, 1995) was employed to assess how often speci c
strategies are used for coping in a sporting context.
3
The ACSI-28 is a 28-item
multidimensional scale that measures seven coping components: (a) coping with
adversity; (b) peaking under pressure; (c) goal setting/mental preparation; (d) con-
centration; (e) freedom from worry; (f) con dence and achievement motivation;
and (g) coachability. The scale can also be summed to represent total personal
coping resources.
4
In the present study we used the composite score (total personal
coping resources). Cronbach alpha values for the coping scale were
α
= .77 at T1
(preseason) and
α
= .82 at T2 (end of season).
Dependent Variables
Injury
. Assessment of injury in Study 2 was identical to that used in Study
1, and included time loss due to injury and number of injuries sustained throughout
the 2002 season.
CBSM Intervention
The rst author developed a structured 6-session CBSM intervention based on
Meichenbaum’s (1985) Stress Inoculation Training (SIT). The 6 sessions, each last-
ing 90–120 minutes, were delivered in a weekly format during a 4-week preseason
period. This approach differed from that taken by Kerr and Goss (1996) and ensured
that players were exposed to the relevant information early in the season. Therefore
players could apply the relevant skills throughout the season. As a control check,
participants were questioned in order to determine their previous experience with
mental skills training. The level of mental skills expertise among the players was
low at baseline, with most reporting minimal knowledge or formal experience.
Following the SIT format, the sessions were structured to include concep-
tualization, practical skills acquisition, and application. Each session consisted of
instructive information with some form of practical experiential exercise. The whole
program was supported with written information, which encouraged the completion
of a number of home-based exercises. In line with Meichenbaum’s suggestions,
the conceptualization components addressed the rationale for behavior change by
describing the physiological and behavioral sequelae of life and competitive stress,
as well as the possible implications for athletic performance and vulnerability to
injury. Participants were also informed regarding the ef cacy of cognitive behavioral
interventions to relieve psychological distress and enhance athletic performance
(Session 1).
With respect to skill acquisition, participants were trained in somatic- and
cognitive-based relaxation strategies (e.g., progressive muscle relaxation, and
autogenic techniques during Sessions 2 and 3) and other cognitive-based strate-
gies (e.g., imagery and cognitive restructuring) during Sessions 4 and 5. The nal
session addressed additional strategies (e.g., goal setting and event planning) as
well as providing a nal review to facilitate the ongoing use of CBSM techniques.
Open discussion was encouraged during all group sessions. Finally, those in the
intervention group were telephoned monthly to discuss application of the strate-
gies and to reinforce the use of the various CBSM techniques. Attendance at the
CBSM sessions in association with the telephone calls gave the only indication of
the athletes’ compliance to the program.
Prediction of Athletic Injury / 305
Control Group
Identical data collection procedures (psychological and injury) were employed
for the control group, but they did not participate in the CBSM program. At the end
of the present study the control group was offered the CBSM information.
Results — Study 2
Injury
A one-way ANCOVA (intervention vs. control) was conducted on each injury
measure. Injury scores from Study 1 (2001 season) served as the covariate. Prior to
these analyses, the assumptions underlying the use of ANCOVA (i.e., reliability of
covariates, linear relationship between dependent variable and covariates, homoge-
neity of regression slopes) were tested and satis ed (Tabachnick & Fidell, 2001).
Results showed a signi cant condition (control vs. intervention) effect for total
time missed,
F
(1, 46) = 4.58,
p
< .05, Eta squared (
η
2
) = .07, but not for number
of injuries sustained,
F
(1, 46) = 1.02,
p
= .32,
η
2
= .03. Participants in the stress
management intervention reported missing less time due to injury during 2002 (
M
= 5.19,
SD
= 5.90) compared to their nonintervention counterparts (
M
= 12.91,
SD
= 15.90). Furthermore, the intervention group only marginally increased the amount
of time missed in 2002 compared to 2001 (
M
= 4.47,
SD
= 4.63), whereas the control
group missed notably more time due to injury in 2002 compared to 2001 (
M
= 8.31,
M = 8.31, M
SD
= 8.70). Although statistically nonsigni cant, a similar pattern of results was
found for injury number. The number of injuries in 2002 vs. 2001 were more for
the control group (2001
M
= 1.57,
M = 1.57, M
SD
= 1.34; 2002
M
= 3.43,
M = 3.43, M
SD
= 1.96) than for
the intervention group (2001
M
= 1.82,
M = 1.82, M
SD
= 1.66; 2002
M
= 2.85,
M = 2.85, M
SD
= 1.60).
Psychological Variables
A one-way ANCOVA (intervention vs. control) was conducted on each psy-
chological variable at T2. Psychological variable scores at T1 (baseline) served as
the covariate. Once again, all assumptions underlying the use of ANCOVA were
tested and satis ed. Results showed a signi cant condition effect for the following
variables: total coping resources,
F
(1, 46) = 9.10,
p
< .004,
η
2
= .16; and worry,
F
(1, 46) = 4.33,
p
< .05,
η
2
= .14. A nonsigni cant trend for concentration disrup-
tion,
F
(1, 46) = 2.97,
p
= .09, was also found,
η
2
= .04. ANCOVA results showed
no signi cant effects for somatic anxiety,
F
(1, 46) = .06,
p
> .05,
η
2
= .01.
The intervention group reported a sharp increase in total coping resources
at T2 (
M
= 86.08,
M = 86.08, M
SD
= 8.22) compared to T1 (
M
= 78.16,
M = 78.16, M
SD
= 9.49), whereas
the control reported a modest increase in coping resources from T2 (
M
= 80.70,
M = 80.70, M
SD
= 7.40) compared to T1 (
M
= 78.40,
M = 78.40, M
SD
= 7.60).
The intervention group also
reported a decrease in worry at T2 (
M
= 13.95,
M = 13.95, M
SD
= 2.95) compared to T1 (
M
=
M = M
16.30,
SD
= 3.85), whereas the control group showed no change in worry at T2 (
M
= 14.09,
SD
= 3.93) compared to T1 (
M
= 14.60,
M = 14.60, M
SD
= 4.93). Although results for
concentration disruption were statistically nonsigni cant, they were in the same
direction as worry, showing an overall decrease in concentration disruption for the
intervention condition (T1,
M
= 3.65,
M = 3.65, M
SD
= 1.11; T2,
M
= 3.34,
SD
= 1.11) but an
increase in the control condition (T1,
M
= 3.80,
M = 3.80, M
SD
= 1.54; T2,
M
= 4.04,
SD
=
1.64) over time.
306
/
Maddison and Prapavessis
To further elucidate the relationships among the intervention, the psychologi-
cal variables, and injury, we conducted a path analysis (Pedhazur, 1982). Results
showed that the indirect effects of the intervention through worry, concentration
disruption, and total coping resources to time missed due to injury (path coef cients
ranged between .01 and .07) were less than the direct effect of the intervention to
time missed due to injury (.30). A similar pattern of results was found when number
of injuries was examined. Hence no support was found for mediation.
Discussion — Study 2
The primary aim of Study 2 was to examine the effectiveness of a CBSM inter-
vention in reducing injury among rugby players previously identi ed as having an
at-risk psychological pro le to injury. A secondary aim was to investigate potential
reasons should the CBSM intervention prove fruitful. With respect to the primary
aim, results supported a reduction in injury vulnerability for those who completed
a 6-session CBSM intervention. Speci cally, the amount of time missed due to
injury was reduced in the intervention group compared to the control group. A
similar pattern of results was shown for occurrence of injury, although the effect
was statistically nonsigni cant. Taken together, these results echo those of other
researchers (Johnson et al., 2005; Kerr & Goss, 1996; Perna, Antoni, Kumar, Cruess,
& Schneiderman, 1998) and support the use of CBSM interventions.
With respect to the secondary aim, the intervention was associated with an
increase in coping resources and decreased worry and concentration disruption.
Path analysis revealed, however, that the direct effect of the intervention to injury
(time missed and number of injuries) was greater than the indirect effect through the
psychological variables. These results indicate that none of the intervention-driven
changes in the psychological variables mediated the intervention effect to injury.
The change in psychological factors as a result of a CBSM program are consistent
with the ndings of Johnson et al. (2005), who reported changes in somatic anxiety,
worry, and coping (coachability, freedom from worry, con dence, and achievement
motivation). Mediation was not discussed in their study.
Our results suggest that the direct effect of the intervention to injury may
have resulted from some other effect, over and above that explained by changes
in coping resources and competitive anxiety. It is plausible that some change in
variables not assessed in the present study led to the injury effect. For example,
changes in cognitive, behavioral, or physiological responses to stress may have
resulted from the intervention. Petrie and Perna (2004) stated that an individual’s
response to stress is best understood when one considers the effects to cognitive,
emotional, behavioral, and physiological systems.
It is possible that the current intervention had effects on these systems, which
resulted in change to the injury variables. For example, Perna et al. (2003) found
that a CBSM intervention effect on injury and illness was mediated by an interven-
tion reduction in negative mood. Our intervention may also have had an effect on
physiological (i.e., muscle tension) and cognitive (i.e., narrowing of visual eld)
stress response processes. For example, Rogers and Landers (2005) have provided
evidence that peripheral narrowing during stress mediates the life stress/injury
relationship. These possibilities are still speculative and more research is needed to
explore the mechanisms that explain the effectiveness of injury-reduction interven-
tions. To date there is only limited research on the relationship between the stress
Prediction of Athletic Injury / 307
response and injury/illness (Andersen, & Williams, 1999; Perna & McDowell,
1995; Rogers & Landers, 2005), and much more work is needed to fully examine
potential stress-related mechanisms. The Williams and Andersen (1998) model and
the Petrie and Perna (2004) extension both provide useful frameworks by which to
examine these possible mechanisms.
Although the magnitude of the effects for the intervention on the injury
variables ranged between 3% and 7% and can be considered modest, they are
nevertheless clinically relevant. The minimal time and expense associated with our
intervention and that of Johnson et al. (2005) should encourage coaches, athletes, and
sports organizations to fund and participate in similar injury prevention programs.
Perhaps if our intervention had focused on enhancing the mental skills of players
for a longer time period, the magnitude of the effect may have been greater (cf.
Kerr & Goss, 1996; Perna et al., 2003).
As with all empirical investigations, the present study is not without limita-
tions. The rst author developed and implemented the intervention, which may have
in uenced the results through experimenter bias (expectancy effect). However, it
is more likely that these changes were due to the intervention rather than to some
Pygmalion response. Because an attention control condition was not used in this
study, it is possible that change in the injury and psychological variables was due
to contact with the researchers rather than to the intervention. Future researchers
should consider including an attention control group to eliminate the possibility of
a Hawthorne effect.
With respect to injury, the present study relied on the athletes themselves,
rather than the coaches as in Study 1, to provide injury data. Although research
assistants collected the data on a weekly basis, this approach has limitations and
future studies should continue to provide more objective assessments of injury.
Finally, this study incorporated a multicomponent approach as recommended in
stress management training, thus we were unable to clarify the individual contri-
bution of the treatment components of this study (e.g., relaxation training and/or
cognitive restructuring). Other researchers may wish to clearly delineate what
speci c subcomponents of a CBSM intervention contribute the most to reduction
in injury and illness (Perna et al., 2003).
Overall Summary and Conclusions
Study 1 showed that social support, the type of coping, previous injury, and
competitive anxiety interacted together to maximize life stress and injury relation-
ships. Study 2 found that a CBSM intervention was effective in reducing injury
among athletes from Study 1 who were identi ed as having an at-risk psychological
pro le. Taken together, both studies underscore the importance of (a) psychosocial
factors in identifying those athletes most vulnerable to injury and (b) cognitive
behavioral stress management programs in reducing their vulnerability to injury.
References
Accident Rehabilitation Compensation and Insurance Corporation. (2002).
Accident Reha-
bilitation Compensation and Insurance Corporation Report 2002
. Wellington, New
Zealand.
Andersen, M.B., & Stoové, M.A. (1998). The sanctity of
p
< .05 obfuscates good stuff: A
comment on Kerr and Goss.
Journal of Applied Sport Psychology
,
10
, 168-173.
308
/
Maddison and Prapavessis
Andersen, M.B., & Williams, J.M. (1988). A model of stress and athletic injury: Prediction
and prevention.
Journal of Sport & Exercise Psychology
,
10
, 294-306.
Andersen, M.B., & Williams, J.M. (1999). Athletic injury, psychosocial factors, and percep-
tual changes during stress.
Journal of Sport Sciences
,
17
, 735-741.
Arnold, H.J. (1982). Moderator variables: A clari cation of conceptual, analytic, and psycho-
metric issues.
Organizational Behavior and Human Performance
,
29
, 143-174.
Baron, R.M., & Kenny, D.A. (1986). The moderator-mediator variable distinction in social
psychological research: Conceptual, strategic, and statistical considerations.
Journal
of Personality and Social Psychology
,
251
, 1173-1182.
Booth, W. (1987). Arthritis Institute tackles sports.
Science
,
237
, 846-847.
Cohen, J. (1988).
Statistical power analysis for the behavioral sciences
. Hillsdale, NJ:
Erlbaum.
Cohen, J., & Cohen, P. (1983).
Applied multiple regression/correlation analysis for the
behavioral sciences
(2nd ed.). Hillsdale, NJ: Erlbaum.
Cohen, S., & Edwards, J.R. (1989). Personality characteristics as moderators of the relation-
ship between stress and disorder. In R.W.J. Neufeld (Ed.),
Advances in the investigation
of psychological stress
(pp. 235-283). New York: Wiley.
Davis, J.O. (1991). Sports injuries and stress management: An opportunity for research.
The
Sport Psychologist
,
5
, 175-182.
Dunn, J.G.H., Causgrove-Dunn, J.C., Wilson, P., & Syrotiuk, D.G. (2000). Reexamining the
factorial structure of the Sport Anxiety Scale.
Journal of Sport & Exercise Psychol-
ogy
,
22
, 183-193.
Edwards, A.L. (1984).
An introduction to linear regression and correlation
(2nd ed). New
York: Freeman.
Endler, N.S., & Parker, J.D.A. (1990). Multidimensional assessment of coping: A critical
evaluation.
Journal of Personality and Social Psychology
,
58
, 844-854.
Folkman, S., & Lazarus, R.S. (1985). If it changes it must be a process: Study of emotion
and coping during three stages of a college examination.
Journal of Personality and
Social Psychology
,
48
, 150-170.
Grove, R.J., Eklund, R.C., & Heard, P.N. (1997). Coping with performance slumps: Factor
analysis of the Ways of Coping in Sport Scale.
The Australian Journal of Science and
Medicine in Sport
,
29
(4), 99-105.
Hanson, S.J., McCullagh, P., & Tonymon, P. (1992). The relationship of personality
characteristics, life-stress, and coping resources to athletic injury.
Journal of Sport
& Exercise Psychology
,
14
, 262-272.
Hodgson Phillips, L. (2000). Sports injury incidence.
British Journal of Sports Medicine
,
34
, 133-136.
Johnson, U., Ekengren, J., & Andersen, M.B. (2005). Injury prevention in Sweden: Helping
soccer players at risk.
Journal of Sport & Exercise Psychology
,
27
, 32-38.
Kerr, G., & Goss, J. (1996). The effects of a stress management program on injuries and
stress levels.
Journal of Applied Sport Psychology
,
8
, 109-117.
Kontos, A.P. (2001, October).
A meta-analytical review of the life stress and injury rela-
tionship in sport
. Paper presented at the Association for the Advancement of Applied
Sport Psychology Conference, Orlando, FL.
Lavelle, L., & Flint, F.A. (1996). The relationship of stress, competitive anxiety, mood state,
and social support to athletic injury.
Journal of Athletic Training
,
31
, 296-299.
Lysens, R., Steverlynck, A., Van den Auweele, Y., LeFevre, J., Renson, L., Classens, A., &
Ostyn, M. (1984). The predictability of sports injuries.
Sports Medicine
,
1
, 6-10.
Prediction of Athletic Injury / 309
Lysens, R., Van den Auweele, Y., & Ostyn, M. (1986). The relationship between psycho-
social factors and sports injuries.
Journal of Sports Medicine and Physical Fitness
,
26
, 77-84.
May, J.R., & Brown, L. (1985). Delivery of psychological services to the U.S. Alpine ski team
prior to and during the Olympics in Calgary.
The Sport Psychologist
,
3
, 320-329.
Meichenbaum, D. (1985).
Stress inoculation training
. New York: Permagon Press.
Patterson, E.L., Smith, R.E., Everett, J.J., & Ptacek, J.T. (1998). Interactive effects of life
stress as predictors of ballet injuries.
Journal of Sport Behavior
,
21
, 101-112.
Pedhazur, E.J. (1982).
Multiple regression in behavioral research
(2nd ed., pp. 577-635).
New York: Holt.
Perna, F.M., Antoni, M.H., Baum, A., Cordon, P., & Schneiderman, N. (2003). Cognitive
behavioral stress management effects on injury and illness among competitive athletes:
A randomized clinical trial.
Annals of Behavioral Medicine
,
25
, 66-73.
Perna, F.M., Antoni, M.H., Kumar, M., Cruess, D.H., & Schneiderman, N. (1998). Cogni-
tive-behavioral intervention effects on mood and cortisol during exercise training.
Annals of Behavioral Medicine
,
20
, 92-98.
Perna, F.M., & McDowell, S.L. (1995). Role of psychological stress in cortisol recovery
from exhaustive exercise among elite athletes.
International Journal of Behavioral
Medicine
,
3
, 13-26.
Perna, F.M., Schneiderman, S.L., & LaPerriere, A. (1997). Psychological stress, exercise,
and immunity.
International Journal of Sports Medicine
,
18
, S78-S83.
Petrie, T. (1992). Psychosocial antecedents of athletic injury: The effects of life stress and
social support on female collegiate gymnasts.
Behavioral Medicine
,
18
, 127-138.
Petrie, T.A. (1993). Coping skills, competitive anxiety, and playing status: Moderating effects
on the life stress–injury relationship.
Journal of Sport & Exercise Psychology
,
15
,
261-274.
Petrie, T., & Perna, F. (2004). Psychology of injury: Theory, research, and practice. In T.
Morris & J. Summers (Eds.),
Sport psychology: Theory, applications, and issues
(pp.
547-571). Australia: Wiley.
Rogers, T.J., & Landers, D.M. (2005). Mediating effects of peripheral vision in the life
event stress/athletic injury relationship.
Journal of Sport & Exercise Psychology
,
27
, 271-288.
Schomer, H.H. (1990). A cognitive strategy training program for marathon runners: Ten
case studies.
South African Journal for Research in Sport, Physical Education and
Recreation
,
13
, 47-48.
Smith, R.E., Schutz, R.W., Smoll, F.L., & Ptacek, J.T. (1995). Development and validation
of a multi-dimensional measure of sport-speci c psychological skills: The Athletic
Coping Skills Inventory-28.
Journal of Sport & Exercise Psychology
,
17
, 379-398.
Smith, R.E., Smoll, F.L., & Ptacek, J.T. (1990a). Conjunctive moderator variables in vulner-
ability and resiliency research: Life stress, social support and coping skills and ado-
lescent sport injuries.
Journal of Personality and Social Psychology
,
58
, 360-370.
Smith, R.E., Smoll, F., & Schutz, R. (1990b). Measurement and correlates of sport-speci c
cognitive and somatic anxiety: The Sport Anxiety Scale.
Anxiety Research
,
2
, 263-
380.
Tabachnick, B.G., & Fidell, L.S. (2001).
Using multivariate statistics
(3rd ed). New York:
Harper & Row.
Uitenbroek, D. (1996). Sports, exercise, and other causes of injuries: Results of a population
survey.
Research Quarterly for Exercise and Sport
,
67
, 380-385.
310
/
Maddison and Prapavessis
Van Mechelen, W., Twisk, J., Molendijk, A., Blom, B., Snel, J., & Kemper, H.C.G. (1996).
Subject-related risk factors for sport injuries: A 1-yr perspective study in young adults.
Medicine and Science in Sports and Exercise
,
28
, 1171-1179.
Williams, J.M. (2001). Psychology of injury risk and prevention. In R.N. Singer, HA. Haus-
enblas, & C.M. Janelle (Eds.),
Handbook of sport psychology
(2nd ed., pp. 766-786).
New York: Wiley & Sons.
Williams, J.M., & Andersen, M.B. (1998). Psychosocial antecedents of sport injury: Review
and critique of the stress and injury model.
Journal of Applied Sport Psychology
,
10
, 5-25.
Williams, J.M., & Roepke, N. (1993). Psychology of injury and injury rehabilitation. In
R.N. Singer, M. Murphey, & L.K. Tennant (Eds.),
Handbook of research in sport
psychology
(pp. 815-839). New York: Macmillan.
Wilson, B.D., Quarrie, K.L., Milburn, P.D., & Chalmers, D.J. (1999). The nature and cir-
cumstances of tackle injuries in Rugby Union.
Journal of Science and Medicine in
Sport
,
2
, 153-162.
Notes
1
Number of players was included in the injury calculation for two reasons: (a) Hodgson
Phillips (2000) recommended this approach to evaluate the potential risk of injury to athletes
in a team sport. A re nement of the calculation of incidence rates is to measure the actual
exposure time at risk. Thus injuries are re ected as player exposure hour. (b) The sample
used in the present study included Rugby Union and Rugby League codes. Team number
for Rugby Union is 15 and for Rugby League it is 13. Hence it was necessary to ensure that
the data were representative for both codes.
2
As highlighted in this paper, the injury data were skewed and subjected to appropriate
transformation techniques. However, to address concerns that these transformed data still vio-
lated normality assumptions associated with parametric tests, we performed identical analyses
on data from all injured athletes, including a random selection (15%) of noninjured athletes.
Results were almost identical to the transformation log injury data presented in this paper.
Thus there is a high degree of con dence that our data are both valid and interpretable.
3
A different coping scale (ACSI-28) was used in Study 2 from that in Study 1 because
we wanted to know whether the intervention would have an effect on the total personal coping
resources an individual resorted to in athletic situations, rather than getting an assessment of
the type of coping strategies an individual resorted to in dealing with the stress of competi-
tion (i.e., avoidance vs. problem focused).
4
In the present study, condition effects were found for the following subscales of the
ACSI-28: concentration,
F
(1, 46) = 8.49,
p
< .006,
η
2
= .16; and peaking under pressure
F
(1, 46) = 8.33,
p
< .01,
η
2
= .15.
Acknowledgment
These two studies were supported by a grant from the New Zealand Accident Com-
pensation Corporation (ACC). We wish to thank all the Auckland-based rugby team players,
coaches, and trainers who participated in this study.
Manuscript submitted
: June 17, 2004
Revision accepted
: May 1, 2005
