RecentResearch-Risk_Factors_for_Dog_bites

Transcription

Recent Research: Risk Factors for Dog Bites
Locksley L. McV. Messam DVM., PhD
Assistant Clinical Professor (Voluntary Series)
University of California Davis: School of Medicine
Department of Public Health Sciences
llmessam@ucdavis.edu
Available online at www.sciencedirect.com
The
Veterinary Journal
The Veterinary Journal 177 (2008) 205–215
www.elsevier.com/locate/tvjl
The human–canine environment: A risk factor for non-play bites?
Locksley L. McV. Messam *, Philip H. Kass, Bruno B. Chomel, Lynette A. Hart
School of Veterinary Medicine, Department of Population Health and Reproduction, 1114 Tupper Hall,
University of California Davis, Davis, CA 95616, USA
Accepted 18 August 2007
Abstract
Few dog bite risk factor studies have been conducted. This veterinary clinic-based retrospective cohort study was aimed at identifying
human–canine environmental risk factors for non-play bites in Kingston, Jamaica (660) and San Francisco (SF), USA (452). Data were
analysed using modified Poisson regression with confounders selected using directed acyclic graphs (DAGs) and the change-in-estimate
procedure.
Dogs acquired for companionship were more likely (RR = 1.66; 95% CI 1.02–2.70) to bite than those acquired for protection. Routinely
allowing a dog into the presence of visitors was also positively associated with it biting. A dog sleeping in a family member’s bedroom was a
risk factor for biting in Kingston (RR = 2.54; 95% CI 1.43–4.54) but not in SF, while being able to leave the yard unaccompanied was a risk
factor for biting in SF (RR = 3.40; 95% CI 1.98–5.85) but not in Kingston. Overall, dogs which were less restricted in their interactions with
humans were at elevated risk for biting. An observed association with dog bites in one cultural setting might not exist in another.
Ó 2007 Elsevier Ltd. All rights reserved.
Keywords: Kingston; Jamaica; San Francisco; USA; Non-play; Dog; Bite; Risk factor
Introduction
Dog bites to humans are a worldwide problem (Chomel
and Trotignon, 1992; Bhanganada et al., 1993; Thompson,
1997; Kumar, 1999; Chen et al., 2000; Ozanne-Smith et al.,
2001; Frangakis and Petridou, 2003; Horisberger et al.,
2004; Van Eeckhout and Wylock, 2005; Morgan and Palmer, 2007). In the United States there are 300–1000 bites
per 100,000 persons per year (Beaver, 1997; Cornwell,
1997), and reports from Switzerland and Belgium have
indicated national bite rates of 180 (Horisberger et al.,
2004) and 900 (Gisle et al., 2002) per 100,000 per year,
respectively. These figures are striking given that some
studies suggest that far less than 50% of dog bites are
reported (Beck and Jones, 1985; Chomel and Trotignon,
1992; Kahn et al., 2003; De Keuster et al., 2006).
*
Corresponding author. Present address: School of Veterinary Medicine, Department of Medicine and Epidemiology, 2108 Tupper Hall,
University of California Davis, CA 95616, USA. Tel.: +1 530 752 3134;
fax: +1 530 752 0414.
E-mail address: llmessam@ucdavis.edu (L.L.McV. Messam).
1090-0233/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tvjl.2007.08.020
Research has largely focused on (1) the circumstances of
incidents (Beck et al., 1975; Beck and Jones, 1985; Szpakowski et al., 1989; Mathews and Lattal, 1994; Thompson,
1997; Guy et al., 2001a; Frangakis and Petridou, 2003;
Horisberger et al., 2004), (2) the characteristics of both biting dogs (Beck and Jones, 1985; Szpakowski et al., 1989;
Gershman et al., 1994; Mathews and Lattal, 1994; Cornwell, 1997; Thompson, 1997; Guy et al., 2001b; Horisberger
et al., 2004) and persons bitten (Beck and Jones, 1985;
Bhanganada et al., 1993; Mathews and Lattal, 1994; Cornwell, 1997; Thompson, 1997; Savino et al., 2002; Horisberger et al., 2004), (3) the estimation of public health costs
(Bhanganada et al., 1993; Weiss et al., 1998), (4) the pathological sequelae to attacks (Fishbein and Robinson, 1993;
Mendez Gallart et al., 2002; Peters et al., 2004; Van Eeckhout and Wylock, 2005), and (5) wound care for the victims
(Van Eeckhout and Wylock, 2005; Morgan and Palmer,
2007).
Unfortunately, few investigators have employed a formal reference series in their studies (Gershman et al.,
1994; Chen et al., 2000; Guy et al., 2001c; Drobatz and
206
L.L.McV. Messam et al. / The Veterinary Journal 177 (2008) 205–215
Smith, 2003; Reisner et al., 2005), and thus research to date
has been of limited value in accurately identifying risk factors. In addition, because hospital based data formed the
basis for inferences for all except a few studies, it is questionable whether these results are applicable to the general
dog population. An analysis of a case series of 227 biting
dogs obtained from a veterinary clientele has reported that
73%, 17.9% and 21.5% animals had bitten an adult (>18
years), a teenager (13–18 years), and children (612 years),
respectively, at least once (Guy et al., 2001a). This stands in
contrast to hospital data which suggest that children are
over represented among dog bite victims (Ozanne-Smith
et al., 2001). A consequence of the limited scope of dog bite
research is the paucity of epidemiological evidence supporting the belief that a dog’s tendency to bite depends on an
interaction of genetics (including sex), early experiences,
later socialization and training, reproductive status, quality
of ownership, supervision, and the potential victim’s
behaviour (American Veterinary Medical Association,
2001).
We conducted a retrospective cohort study in San Francisco (SF), USA, and Kingston, Jamaica (JA) to identify
human–canine environmental risk factors for non-play
bites to humans. Work by a few authors has suggested that
both human–canine attitudes and interactions in the Caribbean differ considerably from those in the continental United States with some studies from Caribbean territories
reporting that 56–70% of dogs are kept entirely outdoors
(Fielding and Mather, 2001; Davis et al., 2007; OrtegaPacheco et al., 2007). In the US, this figure is 15–20%
(American Pet Products Manufacturers Association,
2005–2006). In selecting divergent cultures with respect to
attitudes to human–canine relationships, we hoped to identify, if present, heterogeneity by country.
present, their names were ranked alphabetically and the first ranked
chosen for participation.
Exposure assessment
For biters and non-biters the exposure period pertained only to the
time period up to the incident and interview respectively. Exposure
information included respondents characteristics, canine characteristics,
factors related to owner-dog habitual interactions, and factors related to
the dogs’ living environment (Fig. 1 and Table 1). Except for three agetime-related questions, all responses were categorical.
Identical data collection protocols were employed in both cities and
1120 (667 in Kingston and 453 in SF) interviews were conducted with 41
(11 in Kingston and 30 in SF) persons electing not to participate. One San
Franciscan and seven Kingstonian questionnaires were disqualified due to
participant ineligibility.
In constructing the final data set, the functional forms of ‘‘age at
acquisition’’, ‘‘current age’’ and ‘‘length of ownership’’ were determined
using fractional polynomials (Royston et al., 1999). To create the variable
‘‘Dog breed size’’ we used breed weights listed in dog breed standards
(Hart and Hart, 1988; American Kennel Club, 1997). ‘‘I don’t know’’
responses were considered missing data.
Outcome determination
Materials and methods
Outcome categories were based on answers to the following questions: (1) ‘‘Not during play, in the last two years, did the dog ever hold
onto or catch a part of any person’s body with its teeth and cause a
wound?’’, (2) ‘‘Not during play, in the last two years, did the dog ever
hold onto or catch a part of any person’s body or clothes with its teeth
but not cause a wound?’’ and (3) ‘‘During play, in the last two years, did
the dog ever hold onto or catch a part of any person’s body with its
teeth and cause a wound?’’ A dog was considered a non-play biter
(hereafter a ‘‘biter’’) if the respondent said ‘‘yes’’ to either or both of
questions 1 and 2 above, and a non-biter if the respondent said ‘‘no’’ to
all three questions. We were primarily concerned with factors motivating
a dog to attack and bite and assumed that the factors under consideration would motivate the attack but not determine whether injury
occurred. When possible, it was noted whether the victim was a family
member and/or lived in the same home as the dog though no distinction
was made in later multivariable analyses. Dogs that had bitten during
play were excluded from analysis.
Study protocol
Statistical methods
The study was approved by the Human Subjects Review Committee at
the University of California, Davis, USA.
We used directed acyclic graphs (DAGs) to create a causal diagram
(Fig. 1) defining a hypothesized causal web for dog bites. This master
DAG provided the basis for confounder selection (Greenland et al.,
1999), and a necessary set of confounders was identified for each
exposure of interest (Fig. 2 and Table 2). Relationships in the causal
diagrams were determined by subject matter considerations inclusive of
results of previous studies. A modified Poisson regression (Zou, 2004)
was used to analyze the data in SAS/STAT version 8.2. For each
exposure of interest, variables included in the relevant DAG-based
subset were used in analysis (Table 2). We employed the change-inestimate procedure (Greenland, 1989) using forward selection to select
confounders from each DAG-based subset with a P10% change in the
estimated relative risk (RR) required for retention in the model. For
each exposure of interest, we excluded all observations that had missing
values for any of the variables in the DAG-based subset of potential
confounders (Table 2). Differences in RR between cities were investigated by including an interaction term comprised of the exposure of
interest and city in the model. The term was retained in the model if
statistically significant at the 5% level. Otherwise pooled RRs were
calculated. Relative risks and their associated 95% confidence intervals
(95% CIs) were calculated using the ‘‘estimate’’ syntax in Proc Genmod
(Table 2) (Spiegelman and Hertzmark, 2005).
Study participants
Study participants were clients in the waiting rooms of eight veterinary
clinics participating from May 30th to August 9th 2003, in Kingston and
from three veterinary clinics in SF from 20th October 2003 to 10th January 2004. Both sets of clinics were located within areas 65 square miles in
their respective cities. All clinics were privately owned with caseloads of
>90% companion animal (dogs and cats). Clients were eligible to participate only if they had a dog present at the time of the interview, had owned
the dog for P24 h, lived 7 days a week in the same home as the dog, and
were P18 years of age.
Data collection
Respondents were approached, following clinic registration but
prior to being seen. The same interviewer administered the questionnaire to over 99% of respondents and dog-related information pertained only to the dog present. Whenever more than one dog was
Hours per day
locked up
Can leave premises
by itself
Hours per
day chained
207
Means of acquisition
(Dog’s origin)
Age at
acquisition
Current age
Sight/Hearing
problems
Allowed around
visitors/strangers
Length of
ownership
Removed
when fearful
Non-Play Bite
Neuter/Sex
Status
Reason for
acquisition
Sleep in bedroom
Removed
when growls
Breed size based
on breed weight
standard
Hours per
day inside
Country
Housing
(yard space)
Breed
Children (5-15
years) in home
Fig. 1. Master directed acyclic graph (DAG) showing hypothesized causal web of dog bites. Bold lines represent causal relationships between exposures
and non-play bites. Dotted lines represent causal relationships between exposures.
Results
Study population characteristics
Data for 161 biters and 951 non-biters were analysed. Of
these, 660 (59%) were from Kingston and 452 (41%) were
from SF. Most respondents were female, though more so
in SF (61%) than in Kingston (54%) and 17% of SF respondents vs. 23% in Kingston answered jointly with another
person. Respondents in Kingston were slightly older
(53% >40 years) than in SF (43% >40 years), but more frequently had a child aged 5–15 years living with them (35%
vs. 12%). Twenty-one percent of Kingstonian dogs were
born at home compared to 1% in SF and more Kingston
(99%) than SF (64%) dogs had yard space at their disposal.
Overall 70% (55% in Kingston vs. 99% in SF) of dogs with
yard space spent some portion of the day inside the house
(or apartment). In the SF sample, dogs born at home and
dogs acquired for protection and other reasons excluding
companionship both had prevalences of approximately 1%.
The 2 year incidence of non-play bites was 12.5 (Kingston) and 17.4 (SF) per 100 dogs and proportionately more
SF (91%) than Kingstonian respondents (76%) witnessed
the incident. In Kingston, 34 (41%) bites broke the victim’s
skin compared to 26 (33%) in SF. Of the victims for which
we had relevant information, 57% (49% in Kingston vs.
62% in SF) were family members and/or lived in the same
home as the dog. Overall 36% of victims were family members and/or lived with the dog, while 27% were not. Most
persons in this latter category were familiar with both
dog and owner. The relationship of the remaining 37% of
victims was not specified. In Kingston, 76% and 64% of
the bites sustained by family and non-family members,
respectively, were witnessed by the respondent. In SF these
percentages were 95 and 87, respectively.
Dogs in Kingston were acquired at younger ages than in
SF. The inter-quartile ranges (IQR) were 3–11 weeks and
8–12 weeks, respectively. Dogs in Kingston were also more
recently acquired (IQR = 1 month – 1.5 years vs. 4 months
– 6.75 years) and younger (IQR = 10 weeks – 2 years vs. 11
months – 7.5 years). Neuter status was markedly different
between cities, with intact dogs accounting for 90% of
Kingston and 22% of SF dogs.
Canine characteristics
Being born at the respondent’s home was inversely associated with biting. Compared to spayed female dogs, all
other categories of dogs had elevated risks for biting
(Table 2). Intact males were 1.68 (95% CI 1.05–2.71) times
more likely to bite than castrated males, but 0.80 (95% CI
0.55–1.14) times as likely to bite as intact females. Both
Rottweilers and Labradors had lower risks of biting compared to German Shepherds with RR = 0.38 (95% CI
0.13–1.09) and 0.24 (95% CI 0.07–0.82), respectively. Shih
Tzus had similar risks of biting to German Shepherds
(Table 2). A sight or hearing problem in the dog was inversely associated with biting.
Environmental factors
The presence of children (5–15 years) in the home had a
slight positive association with dog bites (Table 2), while
having yard space was inversely associated with biting
208
Table 1
Distribution of biting and non-biting dogs by exposure status and city of origin: Kingston (KGN), Jamaica and San Francisco (SF), USA
Exposure
Exposure categories
Non-play bites
c
By characteristics of the respondents
Respondent’s age (years)
Non-bites
c
KGN n (%)
SF n (%)
KGN n (%)c
SF n (%)c
620
21–30
31–40
41–50
51–60
61–70
P71
6 (7)
19 (23)
21 (26)
19 (23)
12 (15)
3 (4)
2 (2)
0 (0)
13 (17)
31 (40)
14 (18)
12 (15)
7 (9)
1 (1)
30 (5)
100 (17)
136 (24)
113 (20)
91 (16)
70 (12)
35 (6)
5 (1)
94 (26)
111 (30)
79 (21)
45 (12)
26 (7)
9 (2)
Total: 1104
82
78
575
369
Respondent’s gender
Male
Female
32 (39)
50 (61)
28 (35)
51 (65)
270 (47)
308 (53)
150 (40)
223 (60)
Total: 1110
82
79
578
373
Method of response
Alone
Spouse/companion helped
Child helped
Other individual helped
63 (77)
4 (5)
12 (15)
3 (4)
67 (85)
5 (6)
4 (5)
3 (4)
444 (77)
45 (8)
59 (10)
30 (5)
309 (83)
39 (11)
11 (3)
14 (4)
Total: 1112
82
79
578
373
Born at home
Acquired
12 (15)
70 (85)
1 (1)
78 (99)
125 (22)
452 (78)
2 (<1)
371 (99)
Total: 1112
82
79
577
373
Male (intact)
Male (castrated)
Female (intact)
Female (spayed)
40 (49)
4 (5)
34 (41)
4 (5)
14
33
11
20
298 (52)
19 (3)
221 (38)
36 (6)
47 (13)
150 (40)
29 (8)
145 (39)
Total: 1105
82
78
574
371
German Shepherd
Rottweiler
Labrador
Shih Tzu
Other
9 (11)
4 (5)
1 (1)
9 (11)
59 (72)
2 (3)
0 (0)
2 (3)
2 (3)
73 (92)
29 (5)
33 (6)
4 (1)
24 (4)
488 (84)
2 (<1)
4 (1)
27 (7)
7 (2)
333 (89)
Total: 1112
82
79
578
373
P9.0 kg (20 lbs)
<9.0 kg (20 lbs)
Unknown
32 (39)
35 (43)
15 (18)
50 (63)
23 (29)
6 (8)
257 (44)
110 (19)
211 (37)
232 (62)
119 (32)
22 (6)
Total: 1112
82
79
578
373
Yes
No
2 (2)
80 (98)
2 (3)
76 (97)
18 (3)
549 (97)
34 (9)
325 (91)
Total: 1086
82
78
567
359
33 (40)
49 (60)
8 (10)
71 (90)
199 (34)
379 (66)
44 (12)
329 (88)
Total: 1112
82
79
578
373
Yard space
No yard space
80 (98)
2 (2)
47 (60)
31 (40)
569 (99)
6 (1)
240 (64)
133 (36)
Total: 1108
82
78
575
373
19–24
13–18
7–12
1–6
0
34 (41)
4 (5)
6 (7)
11 (13)
27 (33)
56 (71)
13 (17)
9 (11)
0 (0)
1 (1)
114 (20)
35 (6)
26 (4)
132 (23)
269 (47)
271 (73)
59 (16)
29 (8)
10 (3)
4 (1)
Total: 1110
82
79
576
373
By characteristics of the dog
Dog’s origin
Dog’s sex and neuter status
Breed
Dog breed size (based on breed standard)
Sight/hearing problems
By characteristics of the dog’s living environment
Children (5–15 years) in home
Yes
No
Housing
Dog in house (h/day)
(18)
(42)
(14)
(26)
209
Table 1 (continued)
Exposure
Sleep in family member’s bedroom
Dog chained (h/day)
Dog locked up (h/day)
Can leave premises unaccompanied
By characteristics of human–canine interactions
Major reason for getting dog
Allowed into presence of strangers
Dog removed/allowed to retreat when fearful
Dog removed/allowed to retreat when growls
a
b
c
Exposure categories
Non-play bites
Non-bites
KGN n (%)c
SF n (%)c
KGN n (%)c
SF n (%)c
Yes
No
34 (42)
47 (58)
61 (78)
17 (22)
84 (15)
492 (85)
273 (73)
100 (27)
Total: 1108
81
78
576
373
19–24
13–18
7–12
1–6
0
1 (1)
1 (1)
2 (2)
9 (11)
69 (84)
0 (0)
0 (0)
0 (0)
0 (0)
79 (100)
20 (4)
6 (1)
24 (4)
19 (3)
508 (88)
0 (0)
1 (<1)
2 (<1)
7 (2)
363 (97)
Total: 1111
82
79
577
373
19–24
13–18
7–12
1–6
0
1 (1)
3 (4)
9 (11)
5 (6)
63 (78)
1 (1)
2 (3)
9 (12)
7 (9)
59 (76)
49 (8)
27 (5)
64 (11)
27 (5)
410 (71)
3 (<1)
5 (1)
47 (13)
19 (5)
299 (80)
Total: 1110
81
78
577
373
Yes
No
16 (20)
65 (80)
8 (10)
70 (90)
95 (16)
480 (84)
4 (1)
369 (99)
Total: 1107
81
78
575
373
Included protection (not companionship)a
Included companionship (not protection)b
All other combinations
16 (20)
38 (46)
28 (34)
1 (1)
56 (80)
22 (28)
144 (25)
194 (34)
239 (41)
5 (1)
260 (70)
108 (29)
Total: 1111
82
79
577
373
Yes
No
Sometimes
47 (62)
15 (20)
14 (18)
68 (88)
3 (4)
6 (8)
239 (48)
206 (41)
56 (11)
353 (95)
12 (3)
7 (2)
Total: 1026
76
77
501
372
Yes
No
Sometimes
Situation never occurred
7 (9)
5 (6)
0 (0)
70 (85)
8 (10)
2 (3)
5 (7)
62 (80)
67 (12)
19 (3)
1 (<1)
485 (85)
49 (13)
15 (4)
1 (<1)
302 (82)
Total: 1098
82
77
572
367
Yes
No
Sometimes
25 (31)
34 (42)
4 (5)
18 (22)
8 (10)
21 (27)
7 (9)
41 (53)
110 (20)
97 (17)
13 (2)
345 (61)
16 (4)
40 (11)
8 (2)
308 (83)
Total: 1095
81
77
565
372
Acquired for protection or for protection and other reasons excluding companionship.
Acquired for companionship or for companionship and other reasons excluding protection.
Percentages do not add to 100 due to rounding error.
(RR = 0.86; 95% CI 0.57–1.30). Dogs that spent 1–6 h per
day inside were no more likely to bite than those that were
not allowed inside, but dogs that spent P7 h per day
inside, were twice as likely to bite than those not allowed
inside (Table 2). In Kingston, dogs that slept in a family
member’s bedroom were more than twice as likely to bite,
while in SF they were no more likely to bite than those that
did not.
As no chained dogs were biters in San Francisco it was
impossible to estimate an SF specific RR for biting. When
the data from both cities were pooled, chaining was weakly
associated with biting and the pooled RR = 1.15 (95% CI
0.66–1.99) not substantially different from the Kingston
specific RR = 1.28 (95% CI 0.71–2.31). Compared to dogs
that were not confined, dogs confined for 1–6 and 7–12 h
per day had increased risks of biting, while those confined
210
Hours per day
inside house
Removed
when growls
Sight/hearing
problem
Reason for
acquisition
Removed
when fearful
Hours per
day chained
Breed size based
on breed weight
standard
Hours per day
locked up
Housing
(yard space)
Breed
Current
age
Country
Respondent’s
gender
Respondent’s age
Allowed around
visitors/strangers
Method of
completion
Non-Play Bite
Fig. 2. Directed acyclic graph (DAG) used to select a sufficient set of potential confounders for control of the effect of ‘‘Allowed around visitors/
strangers’’ on dog bites (heavy arrows). Bold arrows show confounder (ovals) relationships. Dotted lines show relationships with variables on causal
pathway (rectangles).
for 13–18 and 19–24 h per day had decreased risks of biting
(though the estimates in the latter two categories were very
imprecise) (Table 2). Leaving the owner’s premises unaccompanied was strongly associated with biting in SF but
not in Kingston.
Human–canine interactions
Dogs acquired or kept for reasons that included companionship but not protection were 1.66 (95% CI 1.02–
2.70) times as likely to bite as those acquired or kept for
reasons including protection but not companionship.
Allowing a dog into the presence of strangers or visitors
to the home was also associated with an increased risk of
biting (RR = 1.77; 95% CI 1.03–3.04).
Not routinely removing (vs. always removing) a dog
and/or allowing it to retreat was associated with an elevated risk of biting both when it was fearful (RR = 2.21;
95% CI 1.14–4.28) and when it growled (RR = 1.30; 95%
CI 0.90–1.90). These two categories of dogs also had elevated risks of biting compared to those for which the situation never occurred (Table 2).
Discussion
Epidemiological studies on dog bites have differed in
their sources of study populations. Various investigators
have used geographic location (Gershman et al., 1994),
place of occupation (Drobatz and Smith, 2003), registration status (Reisner et al., 2005) and presence at a veterinary clinic (Guy et al., 2001c). Both reported (Gershman
et al., 1994; Drobatz and Smith, 2003) and unreported
(Guy et al., 2001c; Reisner et al., 2005) bites have been used
as outcomes. This study differs from previous studies in
including bites which did not break the skin and in excluding play bites. We reasoned that from a point of view of
risk factors for biting, all dogs that attack and make contact with the teeth belong to the same source population.
Play bites were excluded on the grounds that they were
likely to be aetiologically distinct from non-play bites. In
using two questions in parallel to determine outcome status, we increased the sensitivity of detecting instances when
a dog attempted to bite and made contact. The refusal rate
of 3.7% is similar to that reported by Guy et al. (2001c) and
confirms the effectiveness of using veterinary clients as a
data source.
The inverse association between being born at home and
dog bites is consistent with a previous report that dogs bred
at home were under-represented among dogs showing
dominance aggression and social fears Serpell and Jagoe
(1995). This association might be a manifestation of the
effects of the origin of the dog and/or the age at which
the dog was acquired. Some evidence for the effect of origin
is provided by Serpell and Jagoe (1995) who reported that
among dogs found unowned and those acquired from pet
shops or breeders there was a higher prevalence of dominance aggression when compared to dogs bred at home.
In this data set approximately a quarter of all canine participants were 11 weeks or older when acquired and the
incidence of biting was higher with increased age at acquisition for dogs acquired at up to approximately 6 months
of age and remained constant thereafter (L.L.McV. Messam, unpublished Ph.D. thesis).
Previously reported associations between sex-neuter status and dog bites are inconclusive, with stronger associations reported for males (Gershman et al., 1994; Drobatz
211
Table 2
Adjusted relative risks (RR) for associations between selected variables and non-play bites, Kingston (KGN), Jamaica and San Francisco (SF), USA
Exposure
By characteristics of the dog
Dog’s origin
Exposure categories
Born at home
Acquired
RR
95% CI
Variablesa causing change in RR
P10%b
<10%c
0.71e
1
0.41–1.25
A
B, C, D,
O
2.56
1.52
3.22
1
1.51–4.34
0.94–2.46
1.86– 5.59
A, F
B, C, D,
E, L, M
N
2.27
0.86
0.54
2.09
1
1.33– 3.88
0.33–2.22
0.18–1.64
1.22–3.59
A
B, C, D,
K, M, O
0.41
1
0.15–1.09
F
A, B, C
D,
1.13g
1
0.80–1.58
A
B, C, D
0.86f
1
0.57–1.30
A
B, C, D
K
1.97e
1.90e
2.18e
1.00e
1
1.17–3.32
0.99–3.62
1.18–4.02
0.51–1.96
A, F, H
I
B, C, D
M, O
2.54h
1.11
1
1.43–4.54
0.67–1.85
F, H, P
B, C, D,
I, M,O
1.15
1
0.66–1.99
F
A, B, C
D, H, I,
0.44
0.93
1.15
1.71
1
0.07–2.76
0.35–2.46
0.72–1.83
1.02–2.86
F, U
A, B, C
D, H, I,
M
1.04
3.40i
1
0.63–1.72
1.98–5.85
F
B, C, D
M, O, Q
R
Total: 1100d
Male (intact)
Male (castrated)
Female (intact)
Female (spayed)
Total: 1026d
Breed
German Shepherd
Rottweiler
Labrador
Shih Tzu
Other
Total: 1100d
Yes
No
Total: 1025d
By characteristics of the dog’s living environment
Children (5–15 years) in home
Yes
No
Total: 1104d
Housing
Yard space
No yard space
Total: 1101d
19–24
13–18
7–12
1–6
0
Total: 1044d
Sleep in family member’s bedroom
Yes (KGN)
Yes (SF)
No
Total: 1042d
Dog chained/leashed (h/day)
1–24
0
Total: 974d
19–24
13–18
7–12
1–6
0
M, U
Total: 973d
Yes (KGN)
Yes (SF)
No
Total: 1042d
(continued on next page)
212
Table 2 (continued)
Exposure
Exposure categories
RR
95% CI
Variablesa causing change in RR
P10%b
By characteristics of human–dog interactions
Major reason for getting dog
Included protection (not companionship)g
Included companionship (not protection)h
All other combinations
<10%c
0.82e
1.36e
1
0.49–1.38
0.99–1.99
1.77
1
1.03–3.04
F, H, I
A, B, C
D, J, O, M
0.78
1.71
1
0.46–1.30
1.06–2.76
F
A, B, C,
D, H, I, K,
M, O, U
2.97
3.55
1
1.95–4.52
2.54–4.97
F, G, U
A, B, C, D
H, I, J, K,
M, O, V
A, B, C,
D, K, O,
Total: 1100d
Allowed into presence of strangers
Yes/Sometimes
No
Total: 948d
Yes
No/sometimes
Total: 961d
Dog removed/allowed to retreat when growls
Yes
No/Sometimes
Total: 892d
A = Country, B = Respondent’s age, C = Respondent’s gender, D = Method of response, F = Current age, G = Dog’s sex/neuter status, H = Breed,
I = Dog breed weight based on breed standards, J = Sight or hearing problem, K = Children (5–15 years) in home, M = Major reason for getting dog,
O = Housing, P = Dog in house, Q = Dog chained, R = Dog locked up, U = Allowed in presence of strangers, V = Dog removed/allowed to retreat when
fearful.
a
Both sets of variables together comprise hypothesized necessary set of confounders in causal web.
b
Retained in final model.
c
Not retained in final model.
d
Total number of participants (1112) minus the number of participants with missing data for at least one of the variables in the necessary set of
confounders.
e
Pooled RR heavily influenced by Kingston estimate.
f
Pooled RR heavily influenced by SF estimate.
g
h
i
Interaction with country (p = 0.002).
and Smith, 2003; Reisner et al., 2005), females (Guy et al.,
2001c), intact (Gershman et al., 1994; Guy et al., 2001c),
and neutered dogs as well (Drobatz and Smith, 2003; Reisner et al., 2005). These conflicting results are not surprising
because the relationship between sex and aggression varies
with aggression type (Borchelt and Voith, 1996) and age.
The present study found intact dogs more likely to bite
and that neuter status modified the effect of sex. These
results concur best with results from the only other study
to estimate the effect of sex and neuter status while controlling for age (Guy et al., 2001c).
Both studies with reference series (Gershman et al.,
1994) and those without (Beck et al., 1975; Szpakowski
et al., 1989; Thompson, 1997; Ozanne-Smith et al., 2001;
Mendez Gallart et al., 2002; Horisberger et al., 2004) have
reported German Shepherds as having among the highest
frequencies within samples of aggressive and biting dogs.
While our results are consistent with those findings, to
the best of our knowledge it has not been previously
reported that Shih Tzus have similar risks for biting as German Shepherds and higher risks than Rottweilers and
Labradors. We speculate that the inverse association
between a sight or hearing problem and dog bites is due
to reduced interaction with humans. Owners, as a precau-
tion might restrict the interactions of these dogs with
humans, recognising that the dog’s diminished vision
and/or hearing might render it more uncertain and thus
more likely to respond to human interaction with
aggression.
Previous studies found more than one child (Gershman
et al., 1994) and the presence of teenagers (Guy et al.,
2001c) in the home to be positively associated with dog
bites. Although not contradicting those findings, the association seen in the present study was weak. An inverse association between bites and having yard space is consistent
with the increased risk of biting among dogs allowed inside
for more than 6 h daily compared to dogs not allowed
inside. Though possibly consistent with a negative correlation between problem behaviour in dogs and the size of
yard space at their disposal (Kobelt et al., 2003) there is
no obvious explanation for the threshold effect at 6 h or
why the RRs of biting are essentially equal for dogs inside
for 7–12, 13–18 and 19–24 h per day. If this relationship is
causal, it might indicate that >6 h per day inside is necessary to facilitate the development of certain human–canine
interactions or dynamics which facilitate dog bites. A time
dependent threshold effect would be consistent with territorial aggression being at the root of many of these incidents
as the dog would need to be established in the area for it to
then become territorial (Moyer, 1968). However we did not
have information on the proportion of dog bite incidents
occurring inside and/or in the context of territorial
aggression.
Gershman et al. (1994) found that dogs that were
chained vs. not chained had increased odds of biting. After
adjusting for age, our results show a slightly increased risk
for biting though a negative association with chaining is
also compatible with the data. This RR for chaining was
heavily influenced by the data collected in Kingston as
there were no chained biters in SF. Compared to dogs that
are never confined, only for dogs locked up for 1–6 h daily
did our results show both a substantially increased risk of
biting and exclude with high probability, a protective role
of confinement. Similarly, we found proportionately more
biters among dogs chained from 1–6 h daily than for any
other time periods. If aggression is indeed caused by poor
socialisation secondary to chaining or other confinement,
as has been claimed (Lockwood, 1995), these results would
indicate that only in the case of shorter daily periods of
restraint or confinement do the aggression-promoting
influences of restraint and/or confinement counteract their
obviously beneficial effect on limiting dog bite opportunity.
Dogs that sleep in a family member’s bedroom were at
higher risk of biting only in Kingston, while dogs that are
able to leave their owner’s premises unaccompanied were
at higher risks for biting only in SF. It is noteworthy that
the owners of both categories of dogs were a minority
(18% in Kingston and 3% in SF) and in these regards, displayed ownership characteristics atypical of their environments. These results are an indication that circumspection
should be exercised in generalising results from one cultural milieu to another.
The finding that dogs kept for reasons including companionship but not protection were more likely to bite than
those kept for reasons including protection but not companionship might seem surprising. Having a dog for reasons including protection but not companionship was
positively associated with restraint (chaining, locking up)
and negatively associated with the dog being allowed inside
the house and with it being allowed around visitors or
strangers to the home (results not shown). It is likely, therefore, that the effect of reason for acquisition on a dog’s
likelihood of biting is a result of its effect on the frequency
and nature of the dog’s interactions with humans. This is
given credence by a number of Kingstonian participants
reporting that they restricted their dog’s interaction with
non-household members to enhance its capabilities as a
watchdog. Less restriction may be proffered for why dogs
allowed around visitors or strangers to the home had elevated risks of biting.
The higher risks of biting among dogs that are not
removed, left alone or allowed to retreat from a situation
after they growl or show fear can be reconciled with fear
and growling being possible warning signs of impending
aggression. Though we are unable to tell if the dog bites
213
actually occurred within the context of such events, these
results may indicate that dogs living in homes where they
are not allowed to retreat after growling or showing fear
were also likely to experience other circumstances in which
their management or lack thereof induced them to bite.
Our estimates would be biased if potential canine participants were censored prior to study enrolment but consequent to dog bites or to exposures related to dog bites.
From 2000 to 2006, 9/15 veterinarians employed at the
Kingston clinics euthanized six dogs because of aggression
to humans, while approximately 2% of the SF dog population are confiscated/relinquished yearly due to aggression
to humans (SF Animal Care and Control: personal communication, 2006). We therefore suspect that bias due to
censorship was negligible in this study. Residual confounding of RR estimates is also possible due to absence of data
on potentially confounding variables. If some exposures
occurred consequent to dog bites, temporal bias could
occur in which an apparent causal exposure-dog bite relationship may actually be a dog bite–exposure relationship.
As information was garnered only by respondent recall we
could not independently verify this. Also, if recall on the
part of respondents was imprecise this could result in
biased estimates due to misclassification of exposures.
There are restrictions on the applicability of the results
of this study to the general population of dog owners.
Although the percentages of dogs taken to veterinarians
is not known precisely for SF and Kingston, in the US
approximately 84% of dog owners report taking their dog
to the veterinarian within the previous year (The American
Veterinary Medical Association, 2002). While no such
information is available for Jamaica, this figure is likely
to be similar to the 58% reported for New Providence,
Bahamas (Fielding and Plumridge, 2005). It is likely therefore that the SF sample comprises a larger percentage of
the dog owning population in SF than the Kingston sample. For both cities some exposures had low prevalence
and thus low statistical power may have mitigated against
us detecting differences in city specific RRs. In these circumstances the pooled RR estimates were heavily influenced by the city with higher exposure prevalence (Table
2 superscripts e and f).
Nevertheless, this study contributes uniquely to the epidemiological literature on dog bites; it explicitly states its analytic assumptions regarding the causal web of dog bites; it
examines exposures not previously studied; it is the first
dog bite study to quantify associations in terms of relative
risks, and the first to compare populations from different
countries. This comparison of both cities has highlighted
two issues worth considering. Firstly, important samplebased differences between the distributions of human–canine
environments exist between cities. The low prevalence of
dogs born at home, dogs acquired for reasons which
included protection but not companionship, dogs always
kept outdoors, dogs chained on an average day (in SF)
and dogs without yard space at their disposal (in Kingston),
suggest that causal pathways of dog bites involving these
214
environments might not be important in these cities. Secondly the differences between SF and Kingston specific relative risks observed for ‘‘sleeping in a family member’s
bedroom’’ and ‘‘being able to leave the yard unaccompanied’’ suggest that an environmental risk factor may have
different effects in different countries.
Conclusions
This study suggests that dogs acquired for companionship, dogs allowed into the presence of strangers and visitors to the home, dogs with fewer restrictions placed on
their daily freedom of movement, and, possibly, interactions with humans, are at elevated risk for biting. This
study also suggests that distinct differences exist between
countries with regard to both the prevalence of certain
human–canine environmental exposures and their effect
on the risk of dog bites. For each cultural context, prevention strategies for dog bites would benefit from a consideration of which types of human–canine interaction or
environmental factors frequently place dogs in situations
in which they are more or less likely to respond by biting.
Being modifiable, these factors are amenable to public
health intervention in the form of public education.
Acknowledgements
The authors are grateful for the assistance of staff of the
following veterinary clinics in data collection; All Creatures, All Pets, Animal Care, Chang’s Veterinary Clinic,
JSPCA, Noah’s Ark, Phoenix Veterinary Clinic, Veterinary
Medical (Kingston), Pet’s Unlimited, SFSPCA, the Avenues (SF).
References
American Kennel Club, 1997. The Complete Dog Book. Howell Book
House, New York.
American Pet Products Manufacturers Association, 2005–2006. APPMA
national pet owners survey. American Pet Products Manufacturers
Association, Greenwich, CT.
Beaver, B.V., 1997. Human–canine interactions: a summary of perspectives. Journal of the American Veterinary Medical Association 210,
1148–1150.
Beck, A.M., Jones, B.A., 1985. Unreported dog bites in children. Public
Health Reports 100, 315–321.
Beck, A.M., Loring, H., Lockwood, R., 1975. The ecology of dog bite
injury in St. Louis, Missouri. Public Health Reports 90, 262–267.
Bhanganada, K., Wilde, H., Sakolsataydorn, P., Oonsombat, P., 1993.
Dog-bite injuries at a Bangkok teaching hospital. Acta Tropica 55,
249–255.
Borchelt, P.L., Voith V.L., 1996. Aggressive behavior in dogs and cats. In:
Voith, V.L., Borchelt, P.L. (Eds.), Readings in Companion Animal
Behavior. Veterinary Learning Systems, Trenton, New Jersey, pp. 217–
229.
Chen, S.C., Tang, F.C., Lee, H.S., Lee, C.T., Yen, C.H., Lee, M.C., 2000.
An epidemiologic study of dog bites among postmen in central
Taiwan. Chang Gung Medical Journal 23, 277–283.
Chomel, B.B., Trotignon, J., 1992. Epidemiologic surveys of dog and cat
bites in the Lyon area, France. European Journal of Epidemiology 8,
619–624.
Cornwell, J.M., 1997. Dog bite prevention: responsible pet ownership and
animal safety. Journal of the American Veterinary Medical Association 210, 1147–1148.
Davis, B.W., Alie, K., Fielding, W.J., Morters, M., Galindo, F., 2007.
Preliminary observations on the characteristics of the owned dog
population in Roseau, Dominica. Journal of Applied Animal Welfare
Science 10, 141–151.
De Keuster, T., Lamoureux, J., Kahn, A., 2006. Epidemiology of dog
bites: a Belgian experience of canine behaviour and public health
concerns. The Veterinary Journal 172, 482–487.
Drobatz, K.J., Smith, G., 2003. Evaluation of risk factors for bite wounds
inflicted on caregivers by dogs and cats in a veterinary teaching
hospital. Journal of the American Veterinary Medical Association 223,
312–316.
Fielding, W.J., Mather, J., 2001. Dog ownership in the West Indies: a case
study from the Bahamas. Anthrozoos 14, 72–80.
Fielding, W.J., Plumridge, S.J., 2005. Characteristics of owned dogs on the
island of New Providence, The Bahamas. Journal of Applied Animal
Welfare Science 8, 245–260.
Fishbein, D.B., Robinson, L.E., 1993. Rabies. New England Journal of
Medicine 329, 1632–1638.
Frangakis, C.E., Petridou, E., 2003. Modelling risk factors for injuries
from dog bites in Greece: a case-only design and analysis. Accident
Analysis and Prevention 35, 435–438.
Gershman, K.A., Sacks, J.J., Wright, J.C., 1994. Which dogs bite? A casecontrol study of risk factors. Pediatrics 93, 913–917.
Gisle, L., Buziarsist, J., Van der Heyden, J., Demarest, S., Miermans, P.J.,
Sartor, F., Van Oyen, H., Tafforeau, J., 2002. Health Enquiry by
Interview, Belgium, 2001. Epidemiology Department. Institute of
Public Health. IPH/EPI Report No. 2002-22 (Written in Dutch).
Greenland, S., 1989. Modeling and variable selection in epidemiologic
analysis. American Journal of Public Health 79, 340–349.
Greenland, S., Pearl, J., Robins, J.M., 1999. Causal diagrams for
epidemiologic research. Epidemiology 10, 37–48.
Guy, N.C., Luescher, U.A., Dohoo, S.E., Spangler, E., Miller, J.B.,
Dohoo, I.R., Bate, L.A., 2001a. A case series of biting dogs:
characteristics of the dogs, their behaviour, and their victims. Applied
Animal Behaviour Science 74, 43–57.
Dohoo, I.R., Bate, L.A., 2001b. Demographic and aggressive characteristics of dogs in a general veterinary caseload. Applied Animal
Behaviour Science 74, 15–28.
Dohoo, I.R., Bate, L.A., 2001c. Risk factors for dog bites to owners in
a general veterinary caseload. Applied Animal Behaviour Science 74,
29–42.
Hart, B.L., Hart, L.A., 1988. The Perfect Puppy: How to Choose Your
Dog by its Behavior. W.H. Freeman, New York.
Horisberger, U., Stark, K.D.C., Rufenacht, J., Pillonel, C., Steiger, A.,
2004. The epidemiology of dog bite injuries in Switzerland –
Characteristics of victims, biting dogs and circumstances. Anthrozoos
17, 320–339.
Kahn, A., Bauche, P., Lamoureux, J., 2003. Child victims of dog bites
treated in emergency departments: a prospective survey. European
Journal of Pediatrics 162, 254–258.
Kobelt, A.J., Hemsworth, P.H., Barnett, J.L., Coleman, G.J., 2003. A
survey of dog ownership in suburban Australia – conditions and
behaviour problems. Applied Animal Behaviour Science 82, 137–
148.
Kumar, S., 1999. New Delhi – when the threat to public-health is in the
shape of a dog. Lancet 353, 2219.
Lockwood, R., 1995. The ethology and epidemiology of canine aggression. In: Serpell, J. (Ed.), The Domestic Dog: Its Evolution, Behaviour
and Interactions with People. Cambridge University Press, Cambridge,
pp. 131–138.
Mathews, J.R., Lattal, K.A., 1994. A behavioral analysis of dog bites
to children. Journal of Developmental and Behavioral Pediatrics 15,
44–52.
Mendez Gallart, R., Gomez Tellado, M., Somoza Argibay, I., Liras
Munoz, J., Pais Pineiro, E., Vela Nieto, D., 2002. [Dog bite-related
injuries treated in a pediatric surgery department: analysis of 654 cases
in 10 years]. Anales Espanoles de Pediatria 56, 425–429.
Morgan, M., Palmer, J., 2007. Dog bites. British Medical Journal 334,
413–417.
Moyer, K.E., 1968. Kinds of aggression and their physiological basis.
Communications in Behavioral Biology. Part A 2, 65–87.
Ortega-Pacheco, A., Rodriguez-Buenfil, J.D., Bolio-Gonzalez, M.E.,
Sauri-Arceo, C.H., Jimenez-Coello, M., Forsberg, C.L., 2007. A
survey of dog populations in urban and rural areas of Yucatan,
Mexico. Anthrozoos 20, 261–274.
Ozanne-Smith, J., Ashby, K., Stathakis, V.Z., 2001. Dog bite and injury
prevention–analysis, critical review, and research agenda. Injury
Prevention 7, 321–326.
Peters, V., Sottiaux, M., Appelboom, J., Kahn, A., 2004. Posttraumatic
stress disorder after dog bites in children. Journal of Pediatrics 144,
121–122.
Reisner, I.R., Houpt, K.A., Shofer, F.S., 2005. National survey of ownerdirected aggression in English Springer Spaniels. Journal of the
American Veterinary Medical Association 227, 1594–1603.
Royston, P., Ambler, G., Sauerbrei, W., 1999. The use of fractional
polynomials to model continuous risk variables in epidemiology.
International Journal of Epidemiology 28, 964–974.
Savino, F., Gallo, E., Serraino, P., Oggero, R., Silvestro, L., Mussa, G.C.,
2002. Dog bites in children less than fourteen years old in Turin.
Minerva Pediatrica 54, 237–242.
215
Serpell, J., Jagoe, J.A., 1995. Early experience and the development of
behaviour. In: Serpell, J. (Ed.), The Domestic Dog: Its Evolution,
Behaviour and Interactions with People. Cambridge University Press,
Cambridge, pp. 79–102.
Spiegelman, D., Hertzmark, E., 2005. Easy SAS calculations for risk or
prevalence ratios and differences. American Journal of Epidemiology
162, 199–200.
Szpakowski, N.M., Bonnett, B.N., Martin, S.W., 1989. An epidemiological investigation into the reported incidents of dog biting in the city of
Guelph. Canadian Veterinary Journal-Revue Vétérinaire Canadienne
30, 937–942.
The American Veterinary Medical Association, 2001. A community
approach to dog bite prevention. Journal of the American Veterinary
Medical Association 218, 1732–1749.
The American Veterinary Medical Association, 2002. US Pet Ownership
& Demographics Sourcebook. Membership & Field Services, American Veterinary Medical Association, Schaumburg, IL.
Thompson, P.G., 1997. The public health impact of dog attacks in a major
Australian city. Medical Journal of Australia 167, 129–132.
Van Eeckhout, G.P.A., Wylock, P., 2005. Dog bites: an overview.
European Journal of Plastic Surgery 28, 233–238.
Weiss, H.B., Friedman, D.I., Coben, J.H., 1998. Incidence of dog bite
injuries treated in emergency departments. Journal of the American
Medical Association 279, 51–53.
Zou, G., 2004. A modified Poisson regression approach to prospective
studies with binary data. American Journal of Epidemiology 159, 702–
706.
Preventive Veterinary Medicine 107 (2012) 110–120
Contents lists available at SciVerse ScienceDirect
Preventive Veterinary Medicine
journal homepage: www.elsevier.com/locate/prevetmed
Risk factors for dog bites occurring during and outside of play: Are
they different?
Locksley L. McV. Messam ∗ , Philip H. Kass, Bruno B. Chomel, Lynette A. Hart
School of Veterinary Medicine, Department of Population Health and Reproduction, 1114 Tupper Hall, University of California Davis, Davis, CA 95616, USA
a r t i c l e
i n f o
Article history:
Received 30 December 2011
Received in revised form 10 April 2012
Accepted 13 May 2012
Keywords:
Dog
Non-play bite
Bites during play
Human–canine contact
Risk factor
a b s t r a c t
The aim of this study was to determine whether the effects of selected human–canine
interaction/environmental factors on bites occurring when the victim was and was not
playing with the dog differed from each other. A veterinary clinic-based retrospective cohort
study was conducted in Kingston, Jamaica (709), and San Francisco, USA (513) to compare
the effects of selected exposures on non-play bites (161) relative to bites preceded by play
with the dog (110) as reported by veterinary clients. Additionally, 951 non-biting dogs
were used for a risk factor analysis of bites occurring during play. Using directed acyclic
graphs and the change-in-estimate procedure to select and adjust for confounders, modified
Poisson regression was used to estimate (a) the ratios of proportions of non-play bites out
of all bites comparing exposed to unexposed dogs (proportionate bite ratios) and (b) risk
ratios for bites occurring during play for each factor of interest.
Proportionate bite ratios ranged from 0.84 to 1.29, with most 95% confidence intervals
including one, thus implying a lack of specificity of effects of the examined factors on nonplay bites relative to bites occurring during play with the dog. Consistent with this lack
of specificity, risk ratios for bites occurring during play were similar in magnitude and
direction to risk ratios previously published for non-play bites using the same non-biting
dogs as a reference group. No country-specific differences in proportionate bite ratios were
detected.
Each human–canine environmental factor showed similar levels of association with
both types of bites. One possible explanation is that both types of bites have a common
causal pathway leading from each factor up to the point of human–canine contact. If the
human–canine contact then leads to either play or non-play interactions with dogs and
subsequently to both types of bites, the presence of such a common pathway would make
the factor non-specific to either type of bite. As some of the examined factors are associated with increased frequencies of both types of bites, this could explain high percentages
of bites occurring during play with the dog as reported in various case series of dog bites.
If so, dog bite prevention strategies targeting these factors will simultaneously reduce the
incidence of both types of bites.
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
∗ Corresponding author. Present address: Department of Public Health
Sciences, School of Medicine, University of California Davis, 1616 Da Vinci
Court, Davis, CA 95616, USA. Tel.: +1 530 754 8824; fax: +1 530 752 3118.
0167-5877/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.prevetmed.2012.05.007
Research on dog bites has revealed that they can
result in disfiguring and psychologically scarring injuries
(Mathews and Lattal, 1994; De Keuster et al., 2006;
Kesting et al., 2006; Morgan and Palmer, 2007) sometimes ending in death (Sacks et al., 2000). These
L.L.McV. Messam et al. / Preventive Veterinary Medicine 107 (2012) 110–120
injuries financially burden the public health system
(Weiss et al., 1998), often hurt owner–dog relationships (Reisner et al., 1994; Hart, 1995; Burt, 1997),
and have a negative impact on society’s view of dog
rearing (Serpell, 1995; Hunthausen, 1997).
Research focusing on the circumstances of dog bites
mentions that some bites occur during play (Parrish et al.,
1959; Thompson, 1997; Guy et al., 2001; Ozanne-Smith
et al., 2001; Horisberger et al., 2004), which is both a
type and a trigger of aggression (Beaver, 1983; Wright
and Nesselrote, 1987; Landsberg et al., 1997). Overall,
various studies report between 5 and 29% of dog bites
occurring either as a sequel to or during play with the
dog (Beck et al., 1975; Szpakowski et al., 1989; Shewell
and Nancarrow, 1991; Ashby, 1996; Guy et al., 2001;
Maragliano et al., 2007; O’Sullivan et al., 2008). One possible explanation for the high percentages of bites occurring
during play among reported dog bites is that certain
exposures might be risk factors both for bites occurring during play as well as for bites occurring outside
of the context of play. Nevertheless, systematic study
of bites occurring while the victim is playing with the
dog has largely been neglected. Distinguishing between
the two types of bites in studies will help to uncover
the specificity, or lack thereof, of underlying associations with potential risk factors, help to clarify aspects
of the causal web of dog bites and thus provide a basis
for intervention aimed at reducing total dog bite incidence. It should also help explain the high proportion of
bites occurring during play among the total numbers of
bites.
To determine whether the effects of selected
human–dog environmental factors on the risk of nonplay bites differed from their effects on the risk of
bites occurring during play with the dog, a veterinary
clinic-based retrospective cohort study was conducted
in both San Francisco (SF), USA, and Kingston, Jamaica,
directly comparing both types of bites as reported
by clients. The premise was that if a particular factor
specifically causes (or prevents) non-play bites but not
bites occurring during play with the dog, then there
should be proportionately more (or fewer) non-play
biters among all biting dogs exposed to that factor
when compared to those unexposed to the factor. This
translates to the ratio of proportions of non-play biters
among exposed and unexposed biters being greater
(or less) than one. Correspondingly, the associated 95%
confidence intervals should predominantly include
values greater (or less) than one. Additionally, the
greater (lesser) the specificity of the effects, the further (closer) will be the 95% confidence limits from (to)
one.
Though little has been published on cross-cultural
differences in dog keeping worldwide (Wan et al.,
2009), previous research points to different cultural
attitudes to dogs between the United States and the
Caribbean (Fielding and Mather, 2001; Davis et al., 2007;
Fielding, 2008). The goal of the bi-national component of the study was to assess whether the effects of
the factors being studied differed in these two countries.
111
2. Materials and methods
2.1. Study protocol
The protocol for this study constitutes a part of a large
cohort study on human–animal interactions as risk factors for dog bites approved by the Human Subjects Review
Committee at the University of California, Davis, USA. The
methods used are identical to those previously described
in detail elsewhere (Messam et al., 2008) and thus only a
brief description is provided here.
2.2. Study participants and study sites
Study participants were clients in the waiting rooms
of eight veterinary clinics in Kingston, Jamaica, from May
30 to August 9, 2003, and from three veterinary clinics in
San Francisco (SF), USA, from October 20, 2003 to January
10, 2004. Persons were eligible to participate if they had
owned the dog for ≥24 h, were living 7 days a week in the
same home as the dog, had the dog at the clinic during
the interview, and were at least 18 years old. Data were
collected by interviewer-administered questionnaire and
canine information pertained to only the dog present at the
time of the interview. Identical data collection protocols
were observed in both countries.
2.3. Outcome determination
A dog bite was defined as sudden pressure from a dog’s
teeth to a part of a person’s body and/or clothing. Outcome
categories were determined based on responses to the following questions: (a) “During play, in the last 2 years, did
the dog ever hold onto or catch a part of any person’s body
with its teeth and cause a wound?”; (b) “Not during play, in
the last 2 years, did the dog ever hold onto or catch a part of
any person’s body with its teeth and cause a wound?”; and
(c) “Not during play, in the last 2 years, did the dog ever
hold onto or catch a part of any person’s body or clothes
with its teeth but not cause a wound?” The outcome was
a bite occurring during play with the dog if the respondent answered “yes” to (a) but “no” to both (b) and (c),
and a non-play bite if the respondent answered “yes” to (b)
and/or (c) but “no” to (a). The dog was considered a nonbiter if the respondent answered “no” to all three questions.
If the respondent answered in the affirmative to (a) and (b),
(a) and (c) or to all three, the event that occurred earliest
was chosen as the outcome. Using two questions in parallel
to determine outcome status for non-play bites increased
the sensitivity of detecting instances when a dog attacked
and made contact with a person with its teeth. In defining
bites occurring during play, outcomes included those circumstances in which the victim was playing with the dog
but were restricted to those resulting in wounds to exclude
cases of playful mouthing where a dog might grasp a person’s body without applying sudden pressure. The use of
the term “play bites” has been avoided in preference to
“bites occurring during play with the dog” as it was not
ascertained whether the dog was playing at the time of the
bite. The type of bite, where only the human is playing, is
hereafter referred to as a “play” bite to distinguish it from
112
C
Exposures acting both
before and after contact
Reason for acquisition
Means of acquisition
(dog’s origin)
Breed
Children (5-15 years)
present in home
Joint humancanine play
Play
bite
Bites during
play
Human play
Canine non-play
aggression
“Play”
bite
Canine non-play
aggression
Non-play
bite
Human-canine
contact
GENETIC
FACTORS
A
Exposures acting before contact
Neuter-gender status
Sight/Hearing problems
Housing (yard space)
Hours per day inside house
Hours per day chained
Hours per day locked up
Sleep in bedroom
Allowed around visitors/strangers
Point of action
B
Exposures acting after contact
Removed when growls
Removed when fearful
Causal pathway
Fig. 1. Diagram indicating hypothesized points of action of exposures along the causal chain of dog bites.
a situation in which the dog was also known to be playing
(Fig. 1).
or a verbal stimulus from the person that elicits a play bow.
(Play with the dog may also be a component of certain types
of contact.)
2.4. Exposure assessment
2.5. Data analysis
Exposures for this analysis were identical to those previously examined for their effect on the risk of non-play bites
(Messam et al., 2008) (Fig. 1). Based on their hypothesized
role in leading to dog bites, they were placed into three
categories. First, certain exposures were thought to exert
their effects only prior to the initiation of human–canine
contact (Fig. 1, box A) and thought to determine contact. Second, other exposures were thought to exert their
effects only subsequent to the initiation of human–canine
contact (Fig. 1, box B) and third, some exposures were
thought to exert their effects both prior and subsequent
to the initiation of human–canine contact (Fig. 1, box C).
Because the essential etiological distinction between both
types of bites hinges on whether or not the event occurred
while the victim was playing with the dog, it was assumed
that the causal pathway from an exposure acting prior to
human–canine contact would be the same for both types of
bites up to the point of the human–canine contact, and then
after contact, diverge before play with the dog occurred.
Human–canine contact is used in this context to mean any
habitual or intermittent human–canine interaction facilitated by occupation of the same space and time and need
not involve tactile contact. For example, a playful interaction may start by touching the dog, a play bow by the dog,
During data collection, 41 persons (11 in Kingston and
30 in San Francisco) elected not to participate. In total, 1235
(718 in Kingston, and 517 in SF) interviews were conducted.
During data entry, thirteen questionnaires were found to be
ineligible and thus data from 1222 participants (110 biters
during play, 161 non play biters and 951 non-biters) were
used for final data analysis.
Modified Poisson regression (Zou, 2004) in SAS version 8.2 was used to estimate proportionate bite ratios
for non-play bites and associated 95% CIs (Spiegelman
and Hertzmark, 2005). The proportionate bite ratio for
E−
non-play bites is given by: PBRNP = PBE+
NP /PBNP , where
E−
PBE+
and
PB
were
the
proportions
of
non-play
biters
NP
NP
among all biting dogs for exposed and unexposed categories, respectively, of the factor. Functional forms of
continuous variables were determined using fractional
polynomials (Royston and Altman, 1994) and directed
acyclic graphs (DAGs) were used to create a subset of potential confounders to control for each exposure of interest
(Greenland et al., 1999) (Table 1). The owner’s age, gender and method of response were included in each of these
subsets and for each exposure of interest, identical subsets
to those previously reported for analyses of non-play bites
113
Table 1
Variables included in each hypothesized sufficient set of confounders during the modified Poisson regression procedure analysing risk factors for non-play
bites relative to bites occurring during play.
Exposures
Code
Sufficient set of potential confoundersb
Dog’s combined sex and neuter status
Housing
Dog chained (h/day)
Sleep in family member’s bedrooma
Allowed in the presence of strangers
Can leave premises unaccompanieda
Reason for acquiring dog
Dog’s origin
Breed
Children (5–15 years) living at home
Dog removed/allowed to retreat when growlsa
G
J
O
P
Q
R
A, B, C, D, E, F, L, M, N
A, B, C, D, F
A, B, C, D, K
A, B, C, D, F, H, I, M, O
A, B, C, D, F, H, I, M, U
A, B, C, D, F, H, I, M, U
A, B, C, D, F, H, I, P, M, O
A, B, C, D, F, H, I, J, P, M, O
A, B, C, D, F, M, O, Q, R
A, B, C, D, K, O
A, B, C, D, O
A, B, C, D, K, M, O
A, B, C, D
A, B, C, D, F, H, I, K, M, O, U
A, B, C, D, F, H, I, K, M, O, U, V
U
M
N
H
K
V
A, country; B, respondent’s age; C, respondent’s gender; D, method of response; E, age at acquisition; F, current age; G, dog’s sex/neuter status; H, breed; I,
dog breed weight based on breed standard; L, length of ownership.
a
Not included in any sufficient set of potential confounders.
b
Initial set of potential confounders chosen for model selection.
were used (Messam et al., 2008). Initially, to detect differences in PBRNP attributable to city of residence, interaction
terms consisting of the factor of interest with country
were added to each model. These terms were retained if
the corresponding regression coefficients were statistically
significant (p < 0.05). Finally, a set of confounders for each
factor of interest was selected from its respective DAG via
forward selection using the change-in-estimate method
(Greenland, 1989) with a ≥10% change in the estimated
PBRNP required for retention in the model.
Finally, the above mentioned procedures were repeated
in order to estimate risk ratios (RR) and 95% CIs for bites
occurring during play with the dog and informally compare
them to results previously reported for non-play bites with
the same non-biting dogs as a reference group (Messam
et al., 2008).
a home with a child 5–15 years old. Of the victims for which
relevant information was collected, 70% (91% of those bitten during play vs. 57% of those bitten outside of play) were
family members and/or lived in the same home as the dog.
Overall, 42% of victims were family members and/or lived
with the dog, while 18% were not, and most persons in this
latter category were familiar with both dog and owner. The
relationships of the remaining 40% of victims with the dog
was not specified.
Twelve percent of biters in Kingston were born at their
current home compared to 1% in SF, and more biters in
Kingston (97%) had access to yard space than in SF (59%).
In SF, more biters were neutered (56% compared to 7% in
Kingston) and fewer (1%) were acquired for protection than
in Kingston (17%).
3.2. Proportionate bite ratios for non-play bites
3. Results
3.1. Study population characteristics
Distributions of the exposures of interest with respect
to biters and non-biters by country are presented in
Tables 2–4. Forty percent of all biters, bit while being played
with by the victim, with the proportion in SF (43%) slightly
greater than in Kingston (37%). Most respondents were
female (58%), especially among Kingstonians (62%) and
owners of non-play biters (63%). Eighty percent of owners of non-play biters and 72% of owners of dogs that bit
while being played with answered questions without assistance from a spouse, child or other accompanying person
and similar proportions of owners of dogs that bit during
(89%) and outside of play (83%) witnessed the respective
biting incidents. Dogs that bit during play with the victim, were younger than non-play biters with inter-quartile
ranges of ages of 10 weeks to 1 year and 11 months to
6.5 years, respectively. Fifty-eight percent of both types of
biters were acquired for companionship and not protection and approximately 25% of both types of biters lived in
No interaction with country of residence was found for
any of the factors examined, and thus PBRNP estimates
based on the pooled data from both countries are reported.
Estimates of PBRNP for non-play bites were close to one
(ranging from 0.84 to 1.29) and associated 95% CIs tended
to be narrow (Tables 5 and 6, Fig. 2(a)–(c)). Ninety-five
percent confidence intervals for effects of the exposures
thought to act (i) prior to contact and (ii) the exposures
thought to act both prior to and subsequent to contact
all included 1 (Tables 5 and 6, Fig. 2(a)–(c)). For example, having yard space was associated with a 3% increase
in the proportion of non-play bites (PBRNP = 1.03; 95% CI:
0.93–1.14), showing that the data is compatible with both
increases and decreases in the proportion of non-play bites
as well.
Proportionate bite ratios with 95% CIs excluding one
were found only for exposures thought to act exclusively
subsequent to contact. For instance, the proportion of
non-play bites among dogs that were always removed
after growling in the presence of visitors and strangers to
the home compared to the proportion of non-play biters
114
Fig. 2. Proportionate bite ratio estimates and associated 95% confidence intervals for the effects on non-play bites of human–canine interaction/environmental factors hypothesized to act (a) only prior to human–canine contact, (b) both prior and subsequent to human–canine contact and
(c) only subsequent to human–canine contact.
115
Table 2
Distribution of bites and non-bites by respondent’s characteristics and country of origin: Kingston (Kgn.), Jamaica and San Francisco (SF), USA (2003–2004).
Exposure
Exposure categories
Total
Non-play bitesa
Bites during play
Kgn.
n (%)b
SF
n (%)b
Kgn.
n (%)b
Non-bitesa
SF
n (%)b
Kgn.
n (%)b
SF
n (%)b
Respondent’s
age (years)
≤20
21–30
31–40
41–50
51–60
61–70
≥71
Total
45
2 (4)
254 14 (6)
338 14 (4)
244
8 (3)
175
9 (5)
108
2 (2)
49
0 (0)
c
1213
49
2 (4)
14 (6)
25 (7)
11 (5)
6 (3)
0 (0)
2 (4)
60
6 (13)
19 (7)
21 (6)
19 (8)
12 (7)
3 (3)
2 (4)
82
0 (0)
13 (5)
31 (9)
14 (6)
12 (7)
7 (6)
1 (2)
78
30 (67)
100 (39)
136 (40)
113 (46)
91 (52)
70 (65)
35 (71)
575
5 (11)
94 (37)
111 (33)
79 (32)
45 (26)
26 (4)
9 (18)
369
Respondent’s
gender
Male
Female
Total
533 18 (3)
689 31 (4)
1222c
49
35 (7)
26 (4)
61
32 (6)
50 (7)
82
28 (5)
51 (7)
79
270 (51)
308 (45)
578
150 (28)
223 (32)
373
Method of
response
962 33 (3)
Alone
5 (5)
Spouse/companion helped 105
99
8 (8)
Child helped
56
3 (5)
1222c
49
Total
46 (5)
7 (7)
5 (5)
3 (5)
61
63 (7)
4 (4)
12 (12)
3 (5)
82
67 (7)
5 (5)
4 (4)
3 (5)
79
444 (46)
45 (43)
59 (59)
30 (54)
578
309 (32)
39 (39)
11 (11)
14 (25)
373
a
b
c
Messam et al. (2008).
Row percentages. Not all percentages sum to 100 due to rounding error.
Differences in totals reflect differences in the number of responses to each question.
among dogs that had never growled in the presence of visitors and strangers was PBRNP = 1.29 (95% CI: 1.17–1.42).
Using the same comparison group, the PBRNP was 1.26 (95%
CI: 1.16–1.37) for dogs that were either never or only sometimes removed after growling in the presence of visitors
and strangers to the home.
3.3. Risk factors for bites occurring during play with the
dog
The magnitudes and directions of the RRs for bites
occurring during play were similar to the RRs previously reported for the same exposures on non-play bites
(Messam et al., 2008). There was also substantial overlap
of their respective 95% CIs with the major difference being
in the precision of the estimates (Tables 5 and 6).
4. Discussion
With few exceptions (Beck et al., 1975; Messam et al.,
2008), bites occurring during and outside of play are routinely pooled together for analysis. To our knowledge, this
is the first report of data collected with the express purpose
of distinguishing between both types of bites. Previously,
in case series conducted in urban areas, Ashby (1996)
(5%), Beck (1975) (9.6%), Shewell and Nancarrow (1991)
(12%), Szpakowski et al. (1989) (12.5%), Guy et al. (2001)
(28.6%) and Horisberger et al. (2004) (14%), reported lower
percentages of dog bites occurring (among total reported
bites) while the victim played with the dog than the 40%
reported in this study. This difference is difficult to explain
as published reports yield little information on exactly
how bites occurring during play with the dog are distinguished from bites occurring outside of the context of play.
Others, Georges and Adesiyun (2008) (32%) and Parrish
et al. (1959) (33%) reported estimates more similar to this
study’s, but combined petting and playing with the dog
into one exposure category. Thus, apart from differences in
study populations, differences in case definitions are likely
to explain these differences in estimates.
While owners should be capable of accurately identifying and reporting human play, it has been suggested
that they are unreliable in distinguishing between play
and non-play signals in the dog (Moss and Wright, 1987;
Mathews and Lattal, 1994; Maragliano et al., 2007; Reisner
and Shofer, 2008; Tami and Gallagher, 2009). This might
particularly be true if the dog’s behaviour changes during
the interaction. Thus, apart from indicating a situation in
which a true play bite occurred, i.e., while both dog and
victim were playing (Fig. 1), a dog bite “during play with
the dog” might have other interpretations: at the time of
the bite, unbeknownst to the respondent, the dog may not
have been playing. This may have been the case for the
duration of the entire person–dog interaction or just during its latter stages. Alternately, the respondent might have
known that the dog was not playing but may have misrepresented (intentionally or otherwise) the dog’s actions to
minimize the circumstances of the bite. This is consistent
with a tendency among some owners to “display a dog positivity bias” (Rajecki et al., 1998) and excuse perceived dog
misbehaviour (Sanders, 1990; Rajecki et al., 1999).
The low proportionate bite ratio estimates for non-play
bites, coupled with 95% confidence limits close to the null
(Tables 5 and 6, Fig. 2(a)–(c)), suggest that the selected
human–dog environmental exposures are associated with
very slight changes in the proportions of bites occurring
either during or outside of the context of play with the
dog. Consequently they are not specific to either of the two
types of bites. In addition, this lack of specificity of effects
is consistent in both countries, though for some factors,
low numbers of exposed dogs in either of the countries
might have contributed to a lack of power to detect interactions. Considering that some of the exposures examined
in this analysis are both positively associated with bites
116
Table 3
Distribution of bites and non-bites by exposures of interest: Kingston (Kgn.), Jamaica and San Francisco (SF) USA (2003–2004).
Exposure
Exposure categories Total
Kgn.
n (%)b
By exposures thought to act prior to contact
Dog’s sex and neuter Male intact
Male castrated
status
Female intact
Female spayed
Total
441
222
336
214
1213c
Non-play bitesa
Bites during play
Non-bitesa
SF
n (%)b
Kgn.
n (%)b
SF
n (%)b
Kgn.
n (%)b
SF
n (%)b
24 (5)
1 (<1)
23 (7)
0 (0)
48
18 (4)
15 (7)
18 (5)
9 (4)
60
40 (9)
4 (2)
34 (10)
4 (2)
82
14 (3)
33 (15)
11 (3)
20 (9)
78
298 (68)
19 (9)
221 (66)
36 (17)
574
47 (11)
150 (68)
29 (9)
145 (68)
371
Sight/hearing
problems
Yes
No
Total
57
1138
1195c
0 (0)
49 (4)
49
1 (2)
59 (5)
60
2 (4)
80 (7)
82
2 (4)
76 (7)
78
18 (32)
549 (48)
567
34 (60)
325 (29)
359
Housing
Yard space
No yard space
Total
1017
200
1217c
47 (5)
2 (1)
49
34 (3)
26 (13)
60
80 (8)
2 (1)
82
47 (5)
31 (16)
78
569 (56)
6 (1)
575
240 (24)
133 (67)
373
Dog in house (h/day) 19–24 h
13–18 h
7–12 h
1–6 h
0h
Total
537
128
80
162
312
1219c
20 (4)
4 (3)
5 (6)
9 (6)
11 (4)
49
42 (8)
13 (10)
5 (6)
0 (0)
0 (0)
60
34 (6)
4 (3)
6 (8)
11 (7)
27 (9)
82
56 (10)
13 (10)
9 (11)
0 (0)
1 (<1)
79
114 (21)
35 (27)
26 (33)
132 (81)
269 (86)
576
271 (50)
59 (46)
29 (36)
10 (6)
4 (1)
373
Dog chained/leashed 1–24 h
0h
(h/day)
Total
111
1108
1219c
17 (15)
32 (3)
49
2 (2)
57 (5)
59
13 (12)
69 (6)
82
0 (0)
79 (7)
79
69 (62)
508 (46)
577
10 (9)
363 (33)
373
2 (3)
4 (10)
13 (9)
10 (12)
30 (3)
59
1 (2)
3 (7)
9 (6)
5 (6)
63 (7)
81
1 (2)
2 (5)
9 (6)
7 (9)
59 (7)
78
49 (84)
27 (66)
64 (43)
27 (33)
410 (46)
577
3 (5)
5 (12)
47 (31)
19 (23)
299 (34)
373
Dog locked up
(h/day)
19–24 h
13–18 h
7–12 h
1–6 h
0h
Total
58
41
150
81
887
1217c
2 (3)
0 (0)
8 (5)
13 (16)
26 (3)
49
Sleep in family
member’s room
Yes
No
Total
507
709
1216c
12 (2)
37 (5)
49
43 (8)
16 (2)
59
34 (7)
47 (7)
81
61 (12)
17 (2)
78
84 (17)
492 (69)
576
273 (54)
100 (14)
373
876
251
1127c
30 (3)
12 (5)
42
56 (6)
3 (1)
59
61 (7)
15 (6)
76
74 (8)
3 (1)
77
295 (34)
206 (82)
501
360 (41)
12 (5)
372
223 36 (16)
1002
13 (1)
1215c
49
54 (24)
5 (<1)
59
16 (7)
65 (6)
81
8 (8)
70 (7)
78
95 (43)
480 (48)
575
4 (2)
369 (37)
373
Yes/sometimes
Allowed in the
presence of strangers No
Total
Can leave premises
unaccompanied
a
b
c
Yes
No
Total
occurring during play and previously have been identified
as positively associated with non-play bites (Tables 5 and 6)
(Messam et al., 2008), this lack of specificity of effects
would explain high proportions of bites during play in case
series of dog bites.
Growling in the presence of visitors/strangers to the
home showed some specificity of effects on non-play bites
compared to bites occurring during play. Because aggression in dogs is often preceded by growling (Messent, 1983;
Wright and Nesselrote, 1987; Bradshaw and Nott, 1995;
AVMA, 2001; Rooney et al., 2001), this finding is not surprising. However, the low specificity may have resulted
from some growls having occurred while the dog was playing as opposed to being signs of aggravation.
This study has primarily focussed on environmental factors which in exerting their effects prior to human–canine
contact (Fig. 1, Groups A and C) are not immediate causes
of dog bites. Nevertheless, they provide the environmental
context for different types of dog bites to occur by determining the type of human–canine contact which precedes
the bite. As has previously been noted (Westgarth et al.,
2008), it is expected that the precise nature (duration and
character) of each human–canine contact be exposure specific. For instance, the excitable prancing around of some
dogs, while waiting to be unleashed or let out of a kennel
and the fact that some dogs come running at the sound of
their owner’s car entering the driveway or the back door
opening, will all depend on the behaviour of the individual
dog and human as well as the type of household it lives in.
Thus, for example, human contact with a dog chained for a
portion of the day will likely be different from human contact with a dog that usually sleeps in a family member’s
bedroom. Possible sequential pathways for bites following human–canine contact are: (a) human–canine contact,
human-play with dog, joint human–canine play and then
a play-bite, (b) human–canine contact, human-play with
117
Table 4
Distribution of bites and non-bites by exposures of interest: Kingston (Kgn.), Jamaica and San Francisco (SF) USA (2003–2004).
Exposure
Exposure categories
Non-play bitesa
Total Bites during play
Kgn.
n (%)b
By exposures thought to act both prior and subsequent to contact
173
6 (3)
Protectionc
Reason for acquiring
623 31 (5)
Companionshipd
dog
425 12 (3)
Othere
49
Total
1221f
Non-bitesa
SF
n (%)b
Kgn.
n (%)b
SF
n (%)b
Kgn.
n (%)b
SF
n (%)b
1 (1)
44 (7)
16 (4)
61
16 (9)
38 (6)
28 (7)
82
1 (1)
56 (9)
22 (5)
79
144 (83)
194 (31)
239 (56)
577
5 (3)
260 (42)
108 (25)
373
Dog’s origin
Born at home
Acquired
Total
144
4 (3)
1077 45 (4)
1221f
49
0 (0)
61 (6)
61
12 (8)
70 (6)
82
1 (<1)
78 (7)
79
125 (87)
452 (42)
577
2 (1)
371 (34)
373
Breed
German shepherd
Rottweiler
Labrador
Shih Tzu
Other
Total
47
3 (6)
43
2 (5)
40
0 (0)
47
4 (9)
1045 40 (4)
1222f
49
2 (4)
0 (0)
6 (15)
1 (2)
52 (5)
61
9 (19)
4 (9)
1 (3)
9 (19)
59 (6)
82
2 (4)
0 (0)
2 (5)
2 (4)
73 (7)
79
29 (62)
33 (77)
4 (10)
24 (51)
488 (47)
578
2 (4)
4 (9)
27 (68)
7 (15)
333 (32)
373
Child (5–15 years)
living in home
Yes
No
Total
311 17 (5)
911 32 (4)
1222f
49
10 (3)
51 (6)
61
33 (11)
49 (5)
82
8 (3)
71 (8)
79
199 (64)
379 (42)
578
44 (14)
329 (36)
373
By exposures thought to act only subsequent to contact
Yes
141
5 (4)
Dog removed/allowed to
58
3 (5)
retreat when fearful in the No/sometimes
Situation never occurred 1005 40 (4)
presence of visitors or
1204f
48
Total
strangers to home
5 (4)
7 (12)
46 (5)
58
7 (5)
5 (9)
70 (7)
82
8 (6)
7 (12)
62 (6)
77
67 (48)
20 (34)
485 (48)
572
49 (35)
16 (28)
302 (30)
367
Yes
116
4 (2)
247 13 (5)
No/sometimes
Situation never occurred 790 32 (4)
1203f
49
Total
3 (2)
10 (4)
46 (6)
59
25 (15)
38 (15)
18 (2)
81
8 (5)
28 (11)
41 (5)
77
110 (66)
110 (45)
345 (44)
565
16 (10)
48 (19)
308 (39)
372
Dog removed/allowed to
retreat when growls in the
presence of visitors or
strangers to home
a
b
c
d
e
f
Data reproduced from Messam et al. (2008).
All other combinations.
dog, joint human–canine play, canine non-play aggression
and then a “play” bite, (c) human–canine contact, humanplay with dog, canine non-play aggression and then a
“play” bite and (d) human–canine contact, canine non-play
aggression and then a non-play bite (Fig. 1). As noted for
other outcomes (Terry et al., 2000), if each type of bite can
result from the same type of intermediate human–canine
contact, the exposure leading to that type of human–canine
contact will have the same effect on each of the resulting
types of bite. Consequently, no pathway leading from the
human–canine contact would be expected to result in proportionately more (or fewer) non-play bites (or bites during
play) among exposed compared to unexposed dogs. Thus, a
high percentage of bites during play with the victim among
case series of dog bites may be a result of a given exposure
creating the environment in which both play and non-play
interactions with a dog could occur. This in turn leads to
both types of bites. Additionally, the effects of these exposures on bites occurring while both victim and dog were
playing (play bites) should not differ from their effects on
bites occurring when only the victim was playing (“play”
bites). This is because play by the victim serves as a common intermediate along the pathway for both these types
of bites (Fig. 1). In the presence of common intermediates,
neither prevarication nor misclassification of dog bites, by
respondents, is likely to be the fundamental explanation
for the results regarding the effects of exposures in Groups
A and C.
Exposures acting subsequent to human–canine contact
(Fig. 1, Group B) are more likely to be immediate determinants of dog bites and therefore have separate pathways
to bites occurring in and outside of the context of play.
Consequently, associations of these exposures with both
types of bites are more likely to be distinct and there will
be either proportionately more or fewer non-play bites
among exposed vs. unexposed dogs. A 95% CI for the proportionate bite ratio which includes only values greater
than one for dogs which growled in the presence of visitors and strangers to the home, is consistent with this
expectation.
The relative importance of the causal pathway from
each exposure to a dog bite depends on the exposure’s
prevalence. For example, low proportions of biters born at
home and acquired for protection in San Francisco, suggest
that causal pathways involving these exposures result in a
lower frequency of bites in San Francisco than in Kingston.
Similarly, low proportions of neutered biters and biters
without access to yard space in Kingston suggest that there,
118
Table 5
Proportionate bite ratios (PBRNP,S ), risk ratios (RRs), 95% confidence intervals (95% CIs) and confounders (CF) causing ≥10% change in PBRNP s and RRs for
associations of selected factors with bites occurring during play, non-play bites, and non-play bites relative to bites occurring during play, Kingston, Jamaica
and San Francisco (2003–2004).
Exposure
Exposure categories
By exposures thought to act prior to contact
Male (intact)
Dog’s sex and neuter
Male (castrated)
status
Female (intact)
Female (spayed)
Non-play bitesd
Bites during play
Non-play bites vs.
bites during play
RR
95% CI
CF
RR
1.09
1.00
1.08
1
n: 257a
0.99–1.20
0.90–1.10
0.98–1.20
F
4.15
1.87
4.76
1
n: 976b
95% CI
CF
RR
95% CI
CF
1.87–9.20
0.83–4.21
2.10–10.81
A, F
M, N
2.56
1.52
3.22
1
n: 1026c
1.51–4.34
0.94–2.46
1.86–5.59
A, F
Sight/hearing
problems
Yes
No
1.14
1
n: 260a
0.93–1.39
None
0.26
1
n: 979 b
0.04–1.85
F
0.41
1
n: 1025c
0.15–1.09
F
Housing
Yard Space
No yard space
1.03
1
n: 269a
0.93–1.14
A, K
0.77
1
n: 1050b
0.48–1.24
A, B
0.86
1
n: 1101c
0.57–1.30
A
19–24 h
13–18 h
7–12 h
1–6 h
0h
0.94
0.90
0.95
0.92
0.85–1.04
0.78–1.04
0.82–1.10
0.78–1.09
None
3.40
3.21
4.04
1.25
1
n: 998b
1.59–7.28
1.40–7.39
1.63–10.00
0.52–3.01
A, C
F, M
1.97
1.90
2.18
1.00
1
n: 1044c
1.17–3.32
0.99–3.62
1.18–4.02
0.51–1.96
A, F
H, I
0.96
1
n: 248a
0.84
0.97
0.99
0.90
1
n: 247a
0.85–1.08
F
A, F
F
F
1.15
1
n: 974c
0.44
0.93
1.15
1.71
1
n: 973c
0.66–1.99
0.62–1.13
0.85–1.10
0.90–1.09
0.80–1.02
2.34
1
n: 926b
1.25
1.34
1.27
3.31
1
n: 926b
0.07–2.76
0.35–2.46
0.72–1.83
1.02–2.86
F, U
1.06
–
1
n: 259a
0.99–1.15
None
n: 262a
Dog chained (h/day)
1–24 h
0h
Dog chained (h/day)
19–24 h
13–18 h
7–12 h
1–6 h
0h
Sleep in family
member’s bedroom
Yes
No
1.45–3.77
0.48–3.28
0.59–3.03
0.77–2.07
2.06–5.32
A, F
H, I
U
1.04
–
1
n: 997b
0.70–1.56
F, P
2.54 (JA)
1.11 (USA)
1
n: 1042c
1.43–4.54
0.67–1.85
Ae , F
H, P
Allowed in the
presence of strangers
Yes/Sometimes
No
1.08
1
n: 246a
0.95–1.22
A
1.37
1
n: 902b
0.79–2.40
F, I
M
1.77
1
n: 948c
1.03–3.04
F, H
I
Can leave premises
unaccompanied
Yes
1.03
0.92–1.14
None
1.56–4.24
A, F
M
Ae , F
1
n: 258a
1.04 (JA)
3.40 (USA)
1
n: 1042c
0.63–1.72
1.98–5.85
No
2.57
–
1
n: 998b
A, country; B, respondent’s age; C, respondent’s gender; F, dog’s current age; H, breed; I, dog breed weight based on breed standards; K, children (5–15
years) living at home; M, major reason for getting dog; N, dog’s origin; P = dog in house; U, allowed in the presence of strangers.
a
Number of observations in final model = number of participants (271) minus the number of participants with missing data for at least one variable in
the necessary set of confounders.
b
c
d
Taken from Messam et al. (2008).
e
Interaction of exposure of interest with country.
these exposures cause fewer dog bites than in San Francisco.
A limitation of this study is the use of a simple causal
model which does not specifically define human–canine
contact for each of the examined exposures. Additionally
the study focussed primarily on environmental exposures
felt to exert effects early in the causal path to a dog bite.
Recent analyses of cases series of dog bites suggest that
immediately preceding the bites, a high percentage of child
victims interacted, with the dog, in ways which might have
triggered the incidents (Kahn et al., 2003; Reisner et al.,
2007, 2011). Thus it is possible that an investigation of the
effects of more proximate exposures might reveal specificity in their effects on non-play bites relative to bites
occurring during play. This type of investigation is particularly important as it pertains to human–canine play
because of its importance to the human–canine relationship in some cultures (Messent, 1983; Rooney et al., 2001;
119
Table 6
Proportionate bite ratios (PBRNP ), risk ratios (RR), 95% confidence intervals (95% CIs) and confounders (CF) causing ≥10% change in PBRNP s and RRs for
associations of selected factors with bites occurring during play, non-play bites, and non-play bites relative to bites occurring during play, Kingston, Jamaica
and San Francisco (2003–2004).
Exposures
Exposure categories
PBRNP
95% CI
By exposures thought to act both prior and subsequent to contact
Protection
1.03
0.91–1.17
Reason for acquiring
Companionship
0.94
0.87–1.02
dog
1
Other
a
n: 267
Non-play bitesd
Bites during play
Non-play bites vs.
bites during play
CF
RR
95% CI
CF
RR
95% CI
CF
None
0.64
2.02
1
n: 1049b
0.28–1.44
1.32–3.09
None
0.82
1.36
1
n: 1100c
0.49–1.38
0.99–1.99
None
Dog’s origin
Born at home
Acquired
1.11
1
n: 267a
0.99–1.26
None
0.34
1
n: 1049b
0.13–0.93
A
0.71
1
n: 1100c
0.41–1.25
A
Breed
German shepherd
Rottweiler
Labrador
Shih Tzu
Other
1.06
1.05
0.84
1.06
1
n: 267a
0.92–1.22
0.83–1.32
0.66–1.66
0.92–1.22
None
1.75
0.80
1.19
1.46
1
n: 1049b
0.74–4.10
0.20–3.15
0.54–2.59
0.62–3.43
A, M
2.27
0.86
0.54
2.09
1
n: 1100c
1.33–3.88
0.33–2.22
0.18–1.64
1.22–3.59
A
Children (5–15
years)c living in
home
Yes
No
0.99
1
n: 267a
0.91–1.09
A
0.98
1
n: 1053b
0.62–1.53
A, D
1.13
1
n: 1104c
0.80–1.58
A
0.87–1.13
0.87–1.15
None
0.86
1.97
1
n: 961b
0.46–1.58
1.15–3.38
A, F
M
0.78
1.71
1
n: 961c
0.46–1.30
1.06–2.76
F
1.17–1.42
1.16–1.37
None
0.80
1.49
1
n: 892b
0.37–1.72
0.95–2.36
A, F
2.97
3.55
1
n: 892c
1.95–4.52
2.54–4.97
F, G
U
By exposures thought to act only subsequent to contact
0.99
Yes
Dog removed/allowed
1.00
No/Sometimes
to retreat when fearful
1
Never occurred
n: 245a
Dog removed/allowed
to retreat when growls
Yes
No/Sometimes
Never occurred
1.29
1.26
1
n: 236a
A, country; D, method of response; F, dog’s current age; G, dog’s sex/neuter status; M, major reason for getting dog; U, allowed in the presence of strangers.
a
b
c
d
Taken from Messam et al. (2008).
Westgarth et al., 2008). In those contexts, it would be
important for dog owners to know which types of playful human–canine interactions might lead first to canine
non-play aggression and consequently result in human
injury.
5. Conclusions
This study suggests that certain human–canine factors
are equally associated with dog bites occurring during
and outside of the context of play with the dog. To correctly interpret these results, it is necessary to consider
that, in all likelihood, only some bites occurring during
play occurred while the dog was actually playing. This is
likely to be the case with dog bites reported to health
authorities as well. Nevertheless, if a particular exposure creates a human–canine environment and consequent
human–canine contacts resulting in either play or non-play
interactions with a dog, it will be a common risk factor for
all types of dog bites resulting from these two interactions.
This lack of specificity of effects could explain high percentages of bites occurring while the victim played with the
dog and suggests that dog bite prevention strategies targeting some of these factors could simultaneously reduce
the incidence of both types of bites.
The study also suggests that an investigation of factors acting during or subsequent to human–canine contact
might reveal more specific exposure effects on non-play
bites relative to bites occurring in the context of play with
the dog. This will elucidate further aspects of the causal
web of dog bites including the role of human–canine play
in its etiology. Given the importance of human–canine play
to many human–canine relationships, it is important to
understand which types of human–canine play are likely
to increase the frequency of dog bites and which are not,
with the ultimate goal of reducing the prevalence of the
former in favour of the latter. Apart from obvious public
health benefits, this type of information will contribute
positively to human–canine relationships and ultimately
enhance canine welfare.
120
Acknowledgements
The authors expressly acknowledge the assistance of
staff of the following veterinary clinics in facilitating data
collection: All Creatures, All Pets, Animal Care, Chang’s Veterinary Clinic, JSPCA, Noah’s Ark, Phoenix Veterinary Clinic,
Veterinary Medical (Kingston, Jamaica), Pet’s Unlimited,
SFSPCA and the Avenues (SF, USA).
References
Ashby, K., 1996. Dog bites. Hazard 26, 7–13.
AVMA, 2001. A community approach to dog bite prevention. J. Am. Vet.
Med. Assoc. 218, 1732–1749.
Beaver, B.V., 1983. Clinical classification of canine aggression. Appl. Anim.
Ethol. 10, 35–43.
Beck, A.M., Loring, H., Lockwood, R., 1975. The ecology of dog bite injury
in St. Louis, Missouri. Public Health Rep. 90, 262–267.
Bradshaw, J.W.S., Nott, H.M.R., 1995. Social and communication behaviour
of companion dogs. In: Serpell, J. (Ed.), The Domestic Dog: Its Evolution, Behaviour and Interactions with People. Cambridge University
Press, Cambridge, pp. 115–130.
Burt, M.R., 1997. Canine legislation: Can dogs get a fair shake in court? J.
Am. Vet. Med. Assoc. 210, 1139–1142.
Davis, B.W., Alie, K., Fielding, W.J., Morters, M., Galindo, F., 2007. Preliminary observations on the characteristics of the owned dog population
in Roseau, Dominica. J. Appl. Anim. Welf. Sci. 10, 141–151.
De Keuster, T., Lamoureux, J., Kahn, A., 2006. Epidemiology of dog bites:
A Belgian experience of canine behaviour and public health concerns.
Vet. J. 172, 482–487.
Fielding, W.J., 2008. Attitudes and actions of pet caregivers in New Providence, The Bahamas, in the context of those of their American
counterparts. Anthrozoos 21, 351–361.
Fielding, W.J., Mather, J., 2001. Dog ownership in the West Indies: a case
study from the Bahamas. Anthrozoos 14, 72–80.
Georges, K., Adesiyun, A., 2008. An investigation into the prevalence of
dog bites to primary school children in Trinidad. BMC Public Health 8,
85.
Greenland, S., 1989. Modeling and variable selection in epidemiologic
analysis. Am. J. Public Health 79, 340–349.
Greenland, S., Pearl, J., Robins, J.M., 1999. Causal diagrams for epidemiologic research. Epidemiology 10, 37–48.
Guy, N.C., Luescher, U.A., Dohoo, S.E., Spangler, E., Miller, J.B., Dohoo, I.R.,
Bate, L.A., 2001. A case series of biting dogs: characteristics of the
dogs, their behaviour, and their victims. Appl. Anim. Behav. Sci. 74,
43–57.
Hart, L.A., 1995. Dogs as human companions: a review of the relationship.
In: Serpell, J. (Ed.), The Domestic Dog: Its Evolution, Behaviour and
Interactions with People. Cambridge University Press, Cambridge, pp.
161–178.
Horisberger, U., Stark, K.D.C., Rufenacht, J., Pillonel, C., Steiger, A., 2004.
The epidemiology of dog bite injuries in Switzerland—characteristics
of victims, biting dogs and circumstances. Anthrozoos 17,
320–339.
Hunthausen, W., 1997. Effects of aggressive behavior on canine welfare. J.
Am. Vet. Med. Assoc. 210, 1134–1136.
Kahn, A., Bauche, P., Lamoureux, J., 2003. Child victims of dog bites treated
in emergency departments: a prospective survey. Eur. J. Pediatr. 162,
254–258.
Kesting, M.R., Holzle, F., Pox, C., Thurmuller, P., Wolff, K.D., 2006. Animal
bite injuries to the head: 132 cases. Br. J. Oral Maxillofac. Surg. 44,
235–239.
Landsberg, G.M., Hunthausen, W.L., Ackerman, L.J., 1997. Handbook of
Behaviour Problems of the Dog and Cat. Butterworth Heinenann
Oxford, Boston.
Maragliano, L., Ciccone, G., Fantini, C., Petrangeli, C., Saporito, G., Di Traglia,
M., Natoli, E., 2007. Biting dogs in Rome (Italy). Int. J. Pest Manag. 53,
329–334.
Mathews, J.R., Lattal, K.A., 1994. A behavioral analysis of dog bites to children. J. Dev. Behav. Pediatr. 15, 44–52.
Messam, L.L.M., Kass, P.H., Chomel, B.B., Hart, L.A., 2008. The
human–canine environment: a risk factor for non-play bites? Vet. J.
177, 205–215.
Messent, P.R., 1983. Social facilitation of contact with other people by
pet dogs. In: Katcher, A.H., Beck, A.M. (Eds.), New Perspectives on
Our Lives with Companion Animals. University of Pennsylvania Press,
Philadelphia, pp. 37–46.
Morgan, M., Palmer, J., 2007. Dog bites. Br. Med. J. 334, 413–417.
Moss, S.P., Wright, J.C., 1987. The effects of dog ownership on judgments
of dog bite likelihood. Anthrozoos 1, 95–99.
O’Sullivan, E.N., Jones, B.R., O’Sullivan, K., Hanlon, A.J., 2008. Characteristics of 234 dog bite incidents in Ireland during 2004 and 2005. Vet.
Rec. 163, 37–42.
Ozanne-Smith, J., Ashby, K., Stathakis, V.Z., 2001. Dog bite and injury
prevention—analysis, critical review, and research agenda. Inj. Prev.
7, 321–326.
Parrish, H.M., Clack, F.B., Brobst, D., Mock, J.F., 1959. Epidemiology of dog
bites. Public Health Rep. 74, 891–903.
Rajecki, D.W., Rasmussen, J.L., Modlin, S.J., Holder, A.M., 1998. Dog bites
boy: judgments of blame and shame. Anthrozoos 11, 66–73.
Rajecki, D.W., Rasmussen, J.L., Sanders, C.R., Modlin, S.J., Holder, A.M.,
1999. Good dog: aspects of humans’ causal attributions for a companion animal’s social behavior. Soc. Anim. 7, 17–34.
Reisner, I.R., Erb, H.N., Houpt, K.A., 1994. Risk-factors for behaviorrelated euthanasia among dominant-aggressive dogs—110 cases
(1989–1992). J. Am. Vet. Med. Assoc. 205, 855–863.
Reisner, I.R., Nance, M.L., Zeller, J.S., Houseknecht, E.M., Kassam-Adams,
N., Wiebe, D.J., 2011. Behavioural characteristics associated with dog
bites to children presenting to an urban trauma centre. Inj. Prev. 17,
348–353.
Reisner, I.R., Shofer, F.S., 2008. Effects of gender and parental
status on knowledge and attitudes of dog owners regarding
dog aggression toward children. J. Am. Vet. Med. Assoc. 233,
1412–1419.
Reisner, I.R., Shofer, F.S., Nance, M.L., 2007. Behavioral assessment of childdirected canine aggression. Inj. Prev. 13, 348–351.
Rooney, N.J., Bradshaw, J.W.S., Robinson, I.H., 2001. Do dogs respond to
play signals given by humans? Anim. Behav. 61, 715–722.
Royston, P., Altman, D.G., 1994. Regression using fractional polynomials
of continuous covariates—parsimonious parametric modeling. Appl.
Stat. J. Roy. St. C. 43, 429–467.
Sacks, J.J., Sinclair, L., Gilchrist, J., Golab, G.C., Lockwood, R., 2000. Breeds
of dogs involved in fatal human attacks in the United States between
1979 and 1998. J. Am. Vet. Med. Assoc. 217, 836–840.
Sanders, C.R., 1990. Excusing tactics: social responses to the public misbehavior of companion animals. Anthrozoos 4, 82–90.
Serpell, J., 1995. From paragon to pariah: some reflections on human attitudes to dogs. In: Serpell, J. (Ed.), The Domestic Dog: Its Evolution,
Behaviour and Interactions with People. Cambridge University Press,
Cambridge, pp. 245–256.
Shewell, P.C., Nancarrow, J.D., 1991. Dogs that bite. BMJ 303,
1512–1513.
Spiegelman, D., Hertzmark, E., 2005. Easy SAS calculations for risk or
prevalence ratios and differences. Am. J. Epidemiol. 162, 199–200.
Szpakowski, N.M., Bonnett, B.N., Martin, S.W., 1989. An epidemiological
investigation into the reported incidents of dog biting in the city of
Guelph. Can. Vet. J. 30, 937–942.
Tami, G., Gallagher, A., 2009. Description of the behaviour of domestic
dog (Canis familiaris) by experienced and inexperienced people. Appl.
Anim. Behav. Sci. 120, 159–169.
Terry, M.B., Neugut, A.I., Schwartz, S., Susser, E., 2000. Risk factors for a
causal intermediate and an endpoint: reconciling differences. Am. J.
Epidemiol. 151, 339–345.
Thompson, P.G., 1997. The public health impact of dog attacks in a major
Australian city. Med. J. Aust. 167, 129–132.
Wan, M., Kubinyi, E., Miklósi, Á., Champagne, F., 2009. A cross-cultural
comparison of reports by German shepherd owners in Hungary and
the United States of America. Appl. Anim. Behav. Sci. 121, 206–213.
Weiss, H.B., Friedman, D.I., Coben, J.H., 1998. Incidence of dog bite injuries
treated in emergency departments. JAMA 279, 51–53.
Westgarth, C., Pinchbeck, G.L., Bradshaw, J.W., Dawson, S., Gaskell, R.M.,
Christley, R.M., 2008. Dog–human and dog–dog interactions of 260
dog-owning households in a community in Cheshire. Vet. Rec. 162,
436–442.
Wright, J.C., Nesselrote, M.S., 1987. Classification of behavior problems in
dogs: distributions of age, breed, sex and reproductive status. Appl.
Anim. Behav. Sci. 19, 169–178.
Zou, G., 2004. A modified Poisson regression approach to prospective studies with binary data. Am. J. Epidemiol. 159, 702–706.
The Veterinary Journal 197 (2013) 378–387
Contents lists available at SciVerse ScienceDirect
The Veterinary Journal
journal homepage: www.elsevier.com/locate/tvjl
Age-related changes in the propensity of dogs to bite
L.L.McV. Messam ⇑, P.H. Kass, B.B. Chomel, L.A. Hart
School of Veterinary Medicine, Department of Population Health and Reproduction, 1114 Tupper Hall, University of California Davis, Davis, CA 95616, USA
a r t i c l e
i n f o
Article history:
Accepted 15 January 2013
Keywords:
Age
Bite
Dog
Non-play
Play
a b s t r a c t
This retrospective cohort study was aimed at describing the effects of age at acquisition, age, and duration
of ownership of dogs on the risk of (1) bites during play and (2) non-play bites to humans. Data were collected on 110 dogs that had bitten during play with a person, 161 dogs that had bitten outside of play and
951 non-biting dogs from veterinary clients in Kingston (KGN), Jamaica and San Francisco (SF), USA. Modified Poisson regression was employed to model the relationships of both types of bites to each variable
separately.
Effects of the variables on dog bite risk (1) during and (2) outside of play with the dog, differed from
each other and by type of bite. Effects varied with the dog’s age and age-related associations were strongest in dogs younger than 1 year old. Ages at acquisition of dogs at highest risk for bites during play were
substantially lower than those at risk for non-play bites. Ages and durations of ownership of dogs at highest risk for bites during play were also lower than those of dogs at highest risk for non-play bites. The
propensity of a dog to bite changes as it ages and relationships between dog bites occurring during
and outside of play and the dog’s age at acquisition, current age, and duration of ownership, differ from
each other.
Ó 2013 Elsevier Ltd. All rights reserved.
Introduction
Dog bites are frequent sequelae to human–canine interactions
(Overall and Love, 2001; The American Veterinary Medical Association, 2001). This has led to much interest in identifying human
and canine risk factors for both bites and aggression to humans
in many parts of the world (Cornelissen and Hopster, 2010;
Feddersen-Petersen, 1994; Georges and Adesiyun, 2008; Gershman
et al., 1994; Guy et al., 2001; Maragliano et al., 2007; Messam et al.,
2008; O’Sullivan et al., 2008; Rosado et al., 2009; Wake et al.,
2009). While age is accepted as a risk factor for canine aggression
(Borchelt and Voith, 1996a,b; Lockwood, 1995; Overall and Love,
2001), little is known about the age or ages at which dogs are most
likely to bite (Overall and Love, 2001). Similarly, while a few studies have examined the association between the age of dogs at their
acquisition and subsequent aggression (Appleby et al., 2002; Hsu
and Sun, 2010; Petersen and Deddens, 2006), there is still need
for an understanding of how age at acquisition is related to dog
bites.
Knowledge of how a dog’s age at acquisition and current age are
related to its aggressive behavior will help veterinarians to contextualize properly for dog–owners both human-directed aggression
in newly acquired dogs as well as aggression-related behavior
changes in dogs as they age.
⇑ Corresponding author. Tel.: +1 530 754 9516.
1090-0233/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.tvjl.2013.01.024
To investigate the relationships of dog age-related factors to the
risk of dog bites, a retrospective cohort study was conducted in
Kingston (KGN), Jamaica and San Francisco (SF), USA. The premise
of the investigation was that if the effects of age-related factors on
the risk of a dog biting were not constant over a dog’s lifetime, then
age–time periods corresponding to higher or lower dog bite risks
should be identifiable using analytic methods which permit data
to define the shape of the age–time–dog bite relationship.
The goals of the study were: (1) to describe the relationships of
age at acquisition, dog age, and duration of ownership to the risks
of bites occurring during and outside of play; (2) to identify the
ranges of these variables corresponding to the highest risks of
dog bites; (3) to identify the ranges of these variables during which
the change in dog bite risk is greatest, and (4) for each variable, to
compare its relationship to the risk of bites occurring during play
to its relationship to the risk of non-play bites. Age, age at acquisition and duration of ownership were used as surrogate measures
for (1) the cumulative effect of time-related social and biological
changes occurring in the dog since its birth; (2) the effect of the
timing of the most recent change in the dog’s ownership and living
environment occurring during this process of change, and (3) the
cumulative effect of these changes in the dog since the most recent
change in its ownership and living environment, respectively.
The bi-national component in this study provided an opportunity to investigate if the effects of dog age-related factors on dog
bite-risk differed between the two countries. Previous research
points to differences in cultural attitudes to dog rearing between
379
the United States and the Caribbean (Davis et al., 2007; Deddens
and Petersen, 2008; Fielding and Mather, 2001).
Materials and methods
curred 1 year prior to the date of the interview. For dogs owned for less than
2 years, it was assumed that the bite preceded the day of the interview by a time
period equal to half the duration of ownership.
Exposures of interest
Study protocol
This study constituted a part of a cohort study on dog bites approved by the
University of California Davis Institutional Review Board. Most aspects of the materials and methods are identical to those previously described in detail (Messam
et al., 2008, 2012) and so only a brief description is provided here.
Study participants
Study participants were clients interviewed in the waiting rooms of eight veterinary clinics in KGN and three veterinary clinics in SF from May 2003 to January
2004. Clients were eligible to participate if they were at least 18 years old. Additionally, the dog in question had to be present at the time of the interview, owned for at
least 24 h and living 7 days/week in the same home as the client. Whenever more
than one dog was present, their names were ranked alphabetically and the first
ranked chosen for participation.
Outcome definition
Dog bite categories were determined using the following questions:
(a) During play, in the last 2 years, did the dog ever hold onto or catch a part of
any person’s body with its teeth and cause a wound?
(b) Not during play, in the last 2 years, did the dog ever hold onto or catch a
part of any person’s body with its teeth and cause a wound?
(c) Not during play, in the last 2 years, did the dog ever hold onto or catch a
part of any person’s body or clothes with its teeth but not cause a wound?
The outcome was considered a bite during play if the respondent answered ‘yes’
to (a) but ‘no’ to both (b) and (c); a non-play bite if the respondent answered ‘yes’ to
(b) and/or (c) but ‘no’ to (a), and a non-bite if the respondent answered ‘no’ to all
three questions. Bites occurring during play were restricted to those resulting in
wounds to exclude cases of playful mouthing where a dog might grasp a person’s
body without applying sudden pressure (Messam et al., 2012). ‘Bite during play,’ instead of ‘play bite’ was used whenever the victim was playing with the dog at the
time of the bite, as no distinction was made between when the dog was and was not
playing. For dogs owned for 2 years or more, it was assumed that the dog bites oc-
The exposures of interest were the dog’s age at acquisition, the dog’s current
age, and the duration of ownership (Table 1), with each recorded both as categorical
and continuous variables. In the absence of exact dates of birth and acquisition, the
following decision rules were used: when an exact age or time period was given,
that number was used; when a range was provided, the midpoint of the range
was used, and when fractions of weeks, months and years were given, the value
was rounded to the nearest week, month or year, respectively. If a respondent could
not provide one of the age or time periods, the value was estimated using the values
of the other two variables of interest if possible. When no age or time period was
obtained from the respondent, the value was omitted. Twenty-eight per cent of
the ages at acquisition and 18% of dog ages were estimated, respectively, for the
continuous variable analysis. No estimation of age–time variables was performed
when these exposures were recorded as categorical variables.
Statistical analysis
For analyses, modified Poisson regression (Zou, 2004) in SAS version 8.2 was
used. Initially, each exposure of interest was used as a continuous variable to model
play and non-play bites with functional forms (of the exposures of interest) separately, determined using fractional polynomials (Royston et al., 1999). This was necessary to allow the data, in addition to the statistical model, to define the shape of
each age (–time) variable–dog bite relationship. Directed acyclic graphs (DAGs;
Greenland et al., 1999) were used to choose a set of potential confounders of the
relationships of age at acquisition to bites occurring during and outside of play. This
initial set included city of residence, presence of yard space, source of the dog and
reason for the dog’s acquisition (Table 2).
A priori, no canine characteristics were believed to be confounders of the relationships of current age or duration of ownership to either type of bites, as both
these variables represent slightly different surrogates for aging in the dog. Since
aging is an inherent characteristic of the animal, its effect was not believed to be
confounded by other individual-level characteristics or variables. For model selection, the change-in-estimate criterion (Greenland, 1989) was employed to select
confounders from the DAG-based subset with a P10% change in the estimated
RR required for a potential confounder to be retained in the model. To detect differences in RRs attributable to city of residence, an interaction term consisting of the
exposure of interest and city was added to each model and retained if the corresponding regression coefficient was statistically significant (P < 0.05). If no statisti-
Table 1
Distribution of biting and non-biting dogs by selected exposures and city of origin: Kingston (KGN), Jamaica and San Francisco (SF), USA.
Exposure
Age at acquisition
Exposure categories
Birth
62 months
>2 months to 66 months
>6 months to 61 year
>1 year to 62 years
>2 years to 65 years
>5 years
Total
Current age
62 months
>1 year to 62 years
>5 years
Total
Duration of ownership
62 months
>1 year to 62 years
>5 years
Total
a
b
Total
149
481
317
84
48
43
34
1156b
123
326
145
153
184
233
1164b
425
183
139
95
154
193
1189b
Bites during play
Non-play bites
Non-bites
KGN
n (%)a
SF
n (%)a
KGN
n (%)a
SF
n (%)a
KGN
n (%)a
SF
n (%)a
5 (3)
35 (7)
8 (3)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
28 (6)
21 (11)
4 (5)
2 (4)
3 (7)
3 (9)
12 (8)
38 (8)
19 (6)
4 (5)
1 (2)
0 (0)
1 (3)
1 (<1)
22 (5)
27 (9)
9 (11)
8 (17)
5 (12)
6 (18)
129 (87)
242 (50)
111 (35)
24 (29)
10 (21)
7 (16)
3 (9)
2 (1)
116 (24)
131 (41)
43 (51)
27 (56)
28 (65)
21 (62)
48
61
75
78
526
368
7 (6)
28 (9)
9 (6)
3 (2)
1 (<1)
0 (0)
2 (2)
30 (9)
7 (5)
8 (5)
6 (3)
8 (3)
2 (3)
9 (3)
15 (10)
15 (10)
23 (12)
13 (6)
1 (<1)
4 (1)
9 (6)
16 (10)
22 (12)
26 (11)
106 (86)
200 (61)
65 (45)
58 (40)
53 (29)
52 (22)
5 (4)
55 (17)
40 (28)
53 (35)
79 (43)
134 (58)
48
61
77
78
534
366
27 (6)
13 (7)
4 (3)
3 (3)
1 (<1)
0 (0)
29 (47)
12 (20)
8 (13)
2 (3)
4 (7)
6 (10)
10 (2)
8 (4)
16 (12)
7 (7)
22 (14)
13 (7)
8 (2)
8 (4)
18 (13)
7 (7)
18 (12)
20 (10)
273 (64)
97 (53)
48 (35)
42 (44)
47 (31)
48 (25)
78 (18)
45 (25)
45 (32)
34 (36)
62 (40)
106 (55)
48
61
76
79
555
370
Differences in totals reflect differences in number of responses to each question.
380
Table 2
Distribution of biting and non-biting dogs by selected exposures and city of origin: Kingston (KGN), Jamaica and San Francisco (SF), USA.
Exposure
Respondent’s age (years)
Exposure categories
45
254
338
244
175
108
49
620
21–30
31–40
41–50
51–60
61–70
P71
1213c
Total
Respondent’s gender
Male
Female
533
689
1222c
Total
Method of response
Alone
Spouse/companion helped
Child helped
Male (intact)
Male (castrated)
Female (intact)
Female (spayed)
Housing
Yard space
No yard space
441
222
336
214
1213c
Total
1017
200
1217c
Total
d
Reason for acquisition
Included protection
Included companionshipe
Included protection and companionshipf
Love dogs
Take care of dog
Otherg
Dog’s origin
Born at home
Obtained from friend/acquaintance
SPCA or shelter
Purchased
Stray/found
Total
Total
b
c
d
e
f
g
962
105
99
56
1222c
Total
a
Total
173
623
75
208
49
93
1221c
144
423
158
455
39
1219c
Bites during play
Non-play bitesa
Non-bitesa
KGN
n (%)b
SF
n (%)b
KGN
n (%)b
SF
n (%)b
KGN
n (%)b
SF
n (%)b
2 (4)
14 (6)
14 (4)
8 (3)
9 (5)
2 (2)
0 (0)
2 (4)
14 (6)
25 (7)
11 (5)
6 (3)
0 (0)
2 (4)
6 (13)
19 (7)
21 (6)
19 (8)
12 (7)
3 (3)
2 (4)
0 (0)
13 (5)
31 (9)
14 (6)
12 (7)
7 (6)
1 (2)
30 (67)
100 (39)
136 (40)
113 (46)
91 (52)
70 (65)
35 (71)
5 (11)
94 (37)
111 (33)
79 (32)
45 (26)
26 (4)
9 (18)
49
60
82
78
575
369
18 (3)
31 (4)
35 (7)
26 (4)
32 (6)
50 (7)
28 (5)
51 (7)
270 (51)
308 (45)
150 (28)
223 (32)
49
61
82
79
578
373
33 (3)
5 (5)
8 (8)
3 (5)
46 (5)
7 (7)
5 (5)
3 (5)
63 (7)
4 (4)
12 (12)
3 (5)
67 (7)
5 (5)
4 (4)
3 (5)
444 (46)
45 (43)
59 (59)
30 (54)
309 (32)
39 (39)
11 (11)
14 (25)
49
61
82
79
24 (5)
1(<1)
23 (7)
0 (0)
18 (4)
15 (75)
18 (5)
9 (4)
40 (9)
4 (2)
34 (10)
4 (2)
14
33
11
20
578
373
(3)
(15)
(3)
(9)
298 (68)
19 (9)
221 (66)
36 (17)
47 (11)
150 (68)
29 (9)
145 (68)
48
60
82
78
574
371
47 (5)
2 (1)
34 (3)
26 (13)
80 (8)
2 (1)
47 (5)
31 (16)
569 (56)
6 (3)
240 (24)
133 (67)
49
60
82
78
575
373
6 (3)
31 (5)
5 (7)
2 (1)
1 (2)
4 (4)
1 (1)
44 (7)
1 (1)
4 (2)
5 (10)
6 (16)
16 (9)
38 (6)
7 (9)
15 (7)
0 (0)
6 (6)
1 (1)
56 (9)
1 (1)
10 (5)
5 (10)
6 (6)
144 (83)
194 (31)
56 (75)
126 (61)
8 (16)
49 (53)
5 (3)
260 (42)
5 (7)
51 (25)
30 (61)
22 (24)
49
61
82
79
577
373
4 (3)
19 (4)
1 (1)
22 (5)
3 (8)
0 (0)
14 (3)
12 (8)
33 (7)
2 (5)
12 (8)
38 (9)
0 (0)
31 (7)
1 (3)
1 (1)
22 (5)
25 (16)
28 (6)
3 (8)
125 (87)
259 (61)
11 (7)
166 (36)
15 (38)
2 (1)
71 (17)
109 (69)
175 (38)
15 (38)
49
61
82
79
576
372
Acquired for both protection and companionship.
Acquired for reasons which did not include those listed above.
cally significant interaction was detected, pooled RRs were calculated and city of
residence retained in the final model if it caused a P10% change in the estimated
RR. Overall, from 110 biters during play, 161 non-play biters and 951 non-biters,
data for 1061 and 1112 dogs were used for bite during play and non-play bite analyses, respectively. Thus the same group of non-biting dogs was used for both
analyses.
Analyses, using data for the same dogs, were then repeated using age at acquisition and current age as categorical variables to compare results with the continuous variable analysis. Additionally, Spearman’s rank correlation (r) between age
of dog and duration of ownership for both play and non-play bites was estimated.
Finally, a sensitivity analysis was performed in order to determine the robustness of the choices of functional forms of the age–time variables (Table 3) by omitting the observations with estimated exposure values and repeating the fractional
polynomial procedure.
From final models, RRs and 95% CIs for comparisons of interest (Tables 4 and 5)
were estimated using model-based variances and covariances (Table 3). From each
of the six final models, a range of dog bite risks corresponding to the range of its
respective exposure of interest was generated. In this way, age–time values corresponding to the 95th percentile of dog bite risks were obtained for each exposure of
interest.
Results
Approximately 50% of KGN respondents were 40 years or younger, compared to 60% of SF respondents, with most respondents in
both countries being female (Table 2). Compared to dogs in SF,
dogs in KGN were acquired at a younger age (92% vs. 77%
<6 months of age), were younger (53% vs. 19% <6 months old)
and owned for less time (46% vs. 23% owned for <2 months; Table 1). The relationships between the age–time variables and dog
bites were non-linear, with the exception of the relationship of
duration of ownership to bites during play (Figs. 1 and 3).
Age at acquisition
Dogs that bit while being played with were acquired at a younger age than non-play biters. Medians (M) and inter-quartile
381
Table 3
Final modified Poisson regression equations modeling the natural log relative risk, (ln(RR)), of (1) bites occurring during play and (2) non-play bites, as separate functions of age at
acquisition (X1), current age (X2) and duration of ownership (X3).
Exposure
Outcome
Age at acquisition
Bites during
playb,c
Non-play bites
Current age
Bites during
playc,d
Non-play
bitese
Duration of
ownership
Bites during
playc
Non-play
bitesc,f,g
Regression equationsa
lnðRRÞ ¼ 2:86 þ 0:52ð0:14Þ
1
X 1 þ1
10
0:05ð0:01Þ
2
X 1 þ1
10
0:55ð0:18ÞðCÞ
1
lnðRRÞ ¼ 0:99 0:07ð0:03Þ X 110þ1
lnðRRÞ ¼ 3:18 þ 1:40ð0:25Þ
1
X2
10
lnðRRÞ ¼ 1:11 þ 0:57ð0:11Þ ln
lnðRRÞ ¼ 1:47 0:22ð0:06Þ
X2
10
X3
10
þ 0:61ð0:13Þ
1
X2
10
ln
X2
10
0:96ð0:19ÞðCÞ
2
0:22ð0:06Þ ln X102
0:79ð0:18ÞC
lnðRRÞ ¼ 0:76 0:17ð0:20ÞC þ 0:07ð0:07Þ ln
X3
10
2
2
0:10ð0:04Þ ln X103
þ 0:30ð0:10Þ ln X103 ðCÞ 0:02ð0:06Þ ln X103
ðCÞ
a
Equations are of the form ln(RR) = b1 + b2A2+ +bn, An, where bn = nth regression coefficient in the equation (from left to right), An = variable in the equation, standard
errors of regression coefficients are in parentheses and cov(bi, bj) = covariance of the ith and jth parameters in the regression equation.
b
cov(b2, b3) = 0.002.
c
C = City (Kingston = 1, San Francisco = 0).
d
cov(b2, b3) = 0.032.
e
cov(b2, b3) = 0.006.
f
cov(b3, b4) = 0.0006, cov(b3, b5) = 0.005, cov(b3, b6) = 0.0006, cov(b4, b5) = 0.0006, cov(b4, b6) = 0.002, cov(b5, b6) = 0.001.
g
Interaction with city (C). P = 0.0047 and P = 0.76 for first and second interaction terms, respectively.
ranges (IQR) were M=2 months (IQR 6 weeks to 3 months) and
M=2.25 months (IQR = 6 weeks to 5.75 months), respectively.
Dogs acquired at 6 weeks of age were at higher risk for biting
during play than those born into their current owner’s home. For
dogs acquired between 6 weeks and approximately 1 year of age,
the risk of biting while being played with decreased slightly with
increasing age at acquisition, but for dogs acquired older than
1 year of age, risks were essentially the same (Fig. 1a). Correspondingly, while a dog acquired at 6 weeks was 3.4 (95% CI: 1.3–8.9)
times as likely to bite while being played with, than one born into
its owner’s home, dogs acquired at 3 and 6 months were 2.6 (95%
CI: 1.0–8.7) and 1.8 (95% CI: 0.7–4.9) times, respectively, as likely
to bite during play than those born at their current home. Dogs
acquired between 1 and 1.5 months of age had estimated risks of
Table 4
Adjusted relative risks (RRs) and 95% confidence intervals (CIs) for the associations between 6 month increases in age at acquisition, current age, and duration of ownership and
(a) non-play bites and (b) bites occurring during play with the dog when the exposures are modeled as continuous variables, Kingston (KGN), Jamaica and San Francisco (SF), USA.
Exposure
b
c
d
e
f
g
h
Non-play bites
Bites during play
RR
95% CI
RR
95% CI
6 vs. 0
8 vs. 2
12 vs. 6
18 vs. 12
24 vs. 18
1.8
1.2
1.0
1.0
1.0
1.1–3.0
1.0–1.3
1.0–1.1
1.0–1.03
1.0–1.02
n = 1033h
1.8
0.5
0.8
0.9
0.9
0.7–4.9
0.4–0.8
0.7–0.9
0.8–0.9
0.9–1.0
n = 989h
Current agec,d
8 vs. 2
12 vs. 6
18 vs. 12
24 vs. 18
30 vs. 24
3.8
1.6
1.2
1.1
1.0
2.1–6.9
1.3–1.9
1.1–1.2
1.0–1.1
1.0–1.1
n = 1029h
0.6
0.6
0.7
0.8
0.9
0.4–1.0
0.5–0.7
0.7–0.8
0.8–0.9
0.86–0.93
n = 986h
Duration of ownership (SF)e
8 vs. 2
12 vs. 6
18 vs. 12
24 vs. 18
1.4
1.1
1.0
1.0
1.2–1.7
1.0–1.2
0.9–1.1
0.9–1.0
0.9f,g
0.9f,g
0.9f,g
0.9f,g
0.8–0.9
0.8–0.9
0.8–0.9
0.8–0.9
(KGN)e
8 vs. 2
12 vs. 6
18 vs. 12
24 vs. 18
2.3
1.3
1.1
1.0
1.2–4.2
1.1–1.7
1.0–1.3
0.9–1.2
n = 1029h
Age at acquisition
a
Months
a,b
Unadjusted age at acquisition–non-play bite association. No variable caused P10% change in RRs.
Age at acquisition–bites during play association adjusted for city.
Unadjusted current age–non-play bite association. No variable caused P10% change in RRs.
Current age–bites during play association adjusted for city.
Duration of ownership–non-play bite association: interaction with city.
Duration of ownership–bites during play association (linear model): RR. for each additional 6 months of ownership.
Duration of ownership–bites during play association adjusted for city.
Differences in totals reflect missing data for each exposure of interest.
n = 986h
382
Table 5
Adjusted relative risks (RRs) and 95% confidence intervals (CIs) for the associations
between age at acquisition and non-play bites, age at acquisition and bites occurring
during play, current age and non-play bites and current age and bites occurring
during play, when the exposures are modeled as categorical variables: Kingston,
Jamaica and San Francisco, USA.
Exposure
Non-play bites
Bites during play
RR
RR
95% CI
95% CI
a,b
Age at acquisition
Birth
>2 months to <6 months
P6 months to <2 years
P2 years to <5 years
>5 years
1
1.4
1.5
1.5
1.4
1.8
Current agec,d
62 months
>2 months to <6 months
6 months to <2 years
P2 years to <5 years
>5 years
1
1.0
3.6
5.3
4.6
0.8–2.6
0.8–2.9
0.8–3.0
0.6–3.3
0.8–4.0
n = 1032e
0.2–4.4
0.9–14.6
1.3–20.9
1.2–18.3
n = 1026e
1
3.6
2.5
1.0
1.2
2.1
1
1.9
1.3
0.5
0.3
1.4–9.4
1.0–6.8
0.3–3.4
0.3–5.2
0.6–7.2
n = 989e
0.7–4.9
0.5–3.5
0.2–1.6
0.1–0.9
n = 984e
a
Age at acquisition-non-play bite association adjusted for city.
Age at acquisition–bites during play association adjusted for city and reason for
acquisition.
c
Unadjusted current age-non-play bite association. No variable caused P10%
change in RR.
d
Current age–bites during play association adjusted for city.
e
Differences in totals reflect missing data for each exposure of interest.
b
biting during play in the 95th percentile of the range of risks for
bites during play.
The risk of non-play bites increased sharply with increasing age
at acquisition for dogs acquired younger than 6 months old and
then was constant (Fig. 1a). Thus, for dogs acquired older than
6 months of age, later ages at acquisition did not appreciably
change the RR of a non-play bite (Tables 4 and 5; Fig. 2b) when
compared to dogs acquired at 6 months. Compared to dogs born
at the respondent’s home, dogs acquired at 2 months, 6 months
and 1 year old were 1.6 (95% CI: 1.1–2.4), 1.8 (95% CI: 1.1–3.0)
and 1.9 (95% CI: 1.1–3.3) times as likely to bite outside of play,
respectively. Dogs aged 1.5 years or older when acquired had estimated risks of non-play bites in the 95th percentile of the range of
non-play bite risks.
Current age
Dogs that bit during play were younger than non-play biters
with M = 4 months (IQR = 10 weeks to 1 year) and M = 2.5 years
(IQR = 11 months to 6.5 years), respectively. The risk of bites during play increased sharply until approximately 3 months of age
and declined thereafter with increasing age (Fig. 1b). Thus, dogs
that were 6, 12 and 24 months old were 0.7 (95% CI: 0.6–0.8), 0.4
(95% CI: 0.3–0.5) and 0.3 (95% CI: 0.2–0.4) times as likely to bite
during play as a 3 month old puppy, respectively. Two to 4 month
old dogs had estimated risks of biting during play in the 95th percentile of the range of risks for bites during play.
Regarding non-play bites, the risk increased with age but at a
diminishing rate from 2 months to approximately 3 years, after
which it declined gradually (Fig. 1b). Correspondingly, while an
8 month old dog was 3.8 (95% CI: 2.1–6.9) times as likely to bite
as a 2 month old dog, a 1 year old dog was 1.6 (95% CI: 1.3–1.9)
times as likely to bite as a 6 month old dog. The most rapid increases in risk occurred in the age range 2–12 months (Figs. 1b
and 2d), and 1–1.5 year old dogs had essentially the same risks
of biting. Thus an 18 month old dog was just 1.2 (95% CI: 1.1–
1.3) times as likely to bite as a 12 month old dog (Table 4). Dogs
that were 21–65 months old had estimated risks of non-play bites
in the 95th percentile of the range of non-play bite risks.
Age at acquisition vs. current age
For dogs acquired before 4–6 months of age, the effect of increases in age at acquisition on the magnitude of the risk of nonplay bites was greater than the effect due to an increase in the
dog’s age (Fig. 3). For dogs acquired after 6 months of age, this tendency was reversed (Fig. 3).
Duration of ownership
Dogs that bit during play were owned for a shorter period before the bite took place than non-play biters, with M=2 months
(IQR = 3 weeks to 7 months) and M = 21.5 months (IQR = 8 months
to 4.5 years), respectively. Dogs owned for 3 months or less had
estimated risks of biting during play in the 95th percentile of the
range of risks for bites during play.
The risk of non-play bites, as a function of duration of ownership, showed a similar pattern to the risk of non-play bites as a
function of current age. It was highest at 24–36 months and decreased gradually thereafter (Fig. 1b and c). There was evidence
of differences in the association of duration of ownership on
non-play bites between the two cities (Table 4 and Fig. 1c). After
being owned for 6 months in SF, a further 6 months of ownership
did not change the risk of biting. In KGN, this was the case after
being owned for 1 year. In SF and KGN, dogs owned for 6–33 and
20–97 months, respectively, had estimated risks of non-play bites
in the 95th percentile of the range of non-play bite risk.
Correlation between current age and duration of ownership
There were high correlations between the current age and duration of ownership for dogs that bit during play (r = 0.88; 95% CI
0.81–0.95) and for dogs that were non-play biters (r = 0.89; 95%
CI 0.82–0.95).
Continuous vs. categorized exposures
The sensitivity analysis confirmed the choices of functional
forms of the exposures of interest used for final models in the continuous variable analysis (Table 3). Results using the exposures of
interest as categorical variables (Table 5) were similar to the continuous variable analysis. This was confirmed by the overlap in 95%
CIs when the RR estimates for the continuous variable analyses calculated at the midpoints of each category were used for comparison with the categorical variable analyses (Fig. 4).
Discussion
In this study, age at acquisition, current age and duration of
ownership have been used as surrogates for unspecified socio-biologic factors believed to be associated with dog bites. Thus, for instance, while canine age (which is simply the amount of time that
has transpired since the birth of a dog), cannot in itself be a causative or protective factor with respect to dog bites, it is likely to be
correlated with canine socio-biological changes which might be
causative or protective.
Most biters during play were acquired younger than 6 months
old and bites occurring during play with the dog occurred relatively soon after acquisition (75% within 6 months of ownership).
If bites during play are likely to occur soon after acquisition, the increase in bite risk observed for dogs acquired at 1.5–2 months of
383
Fig. 1. Comparisons of the risk of non-play bites to bites during play plotted as separate functions of a dog’s (a) age at acquisition, (b) current age and (c), duration of
ownership.
age compared to those born at home might be attributable to more
physical interaction between the owner and a newly acquired puppy than between the owner and a puppy that he/she has seen develop from birth. Additionally, increased responsiveness by a
2 month old puppy, the eruption of its teeth, its increased strength
and tendency to playfully mouth are possible reasons for increasing risks of biting while being played with during the first 1.5–
2 months after birth. Progressive decreases in the risk of bites
occurring during play for dogs acquired older than 1.5–2 months
of age and for dogs older than 3–4 months might be a consequence
of a decreasing tendency of older dogs to play, or for their owners
to play with them, or both. This is consistent with dog age being
inversely associated with the frequency of owner–dog play (Rooney et al., 2000), as well as with a reported decline in social play
in dogs after 6–7 months (Feddersen-Petersen, 1991).
Dogs that never changed homes being at the lowest risk for
non-play biting is consistent with previous observations that dogs
bred at home (Serpell and Jagoe, 1995), or which remained longer
(adopted at 60 days vs. 30–40 days) with litter mates (Petersen and
Deddens, 2006), were under-represented among dogs with behavior problems. It is also consistent with a previous report which
found that while there was an overall positive association between
being born outside in a kennel, garage or barn (as opposed to in the
residential part of the home) and stranger-directed aggression,
there was no association observed among that subset of the same
dogs acquired before 8 weeks of age (Appleby et al., 2002). Recently, somewhat contradictory findings have been reported: dogs
acquired as puppies (vs. as adults) were at higher odds of showing
stranger-directed aggression (Hsu and Sun, 2010). However, the
authors explain that people might not adopt aggressive adult dogs
and also that they might be unable to recognize signs of future
aggressive tendencies in puppies.
This study suggests that the association between age at acquisition and the risk of dog bites (both during and outside of play with
the dog) primarily occurs over a limited time window, i.e. during
the first 6–12 months of a dog’s life. This lends support to the view
384
Fig. 2. Plots of the relative risk (RR) of bites for 6 month increases in ages at acquisition for (a) age at acquisition–bites during play; (b) age at acquisition–non-play bites, and
for 6 month increases in current ages for (c) current age–bites during play, and (d) current age–non-play bites associations assuming both linear and non-linear (polynomial)
relationships to dog bites. For example, in (a) the risk of biting while being played with for a dog acquired at age 12 months would be 0.8 times its risk of biting had it been
acquired at age 6 months and in (d) a 12 month old dog’s risk of non-play biting is 1.4 times that of a 6 month old dog. Regardless of which age categories are compared, RRs
are constant when linearity is assumed (dashed lines).
that the timing of events in a dog’s life is an important determinant
of dog bites (Lockwood, 1995; Stein et al., 1994; Wright, 1996) and
that early experiences are more important determinants of adult
dog behavior than later ones (Serpell and Jagoe, 1995). For instance, it is possible that the trauma of changing both home and
owner can have negative consequences on canine development
and behavior, manifesting itself in an increased risk of biting outside of play. It is logical that this could still contribute to aggression
in non-play biters, even if human-directed aggression caused previous relinquishment.
Previous studies assuming a constant effect of age on non-play
bite risk have reported odds ratios of 0.96 (95% CI: 0.89–1.03) (Guy
et al., 2001) and 1.1 (95% CI: 1.0–1.2) (Drobatz and Smith, 2003) for
1 year increases in age. When constant age effects were assumed,
in this study, a similar result (RR = 1.1; 95% CI: 1.0–1.1) was obtained. This result suggests that for every 1 year increase in age,
there is a 1.1-fold increase in the risk of biting, thus implying that
the risk of dog bites increases by a constant multiple throughout
the lifetime of the dog. These results differ from, and are less plausible, than the results obtained using fractional polynomials
385
Fig. 3. Plot comparing the effects of age at acquisition and current age on the risks of non-play bites.
Fig. 4. Plots of estimated relative risks (RR) and 95% confidence intervals from categorical- and continuous variable analyses for (a) age at acquisition–bites during play; (b)
age at acquisition–non-play bites; (c) current age–bites during play; and (d) current age–non-play bites associations. RR estimates for the continuous variable analysis are
calculated at the midpoints of categories used for the categorical variable analyses. Straight lines used to connect point estimates from continuous variable analyses are used
for comparison purposes only. In reality lines connecting these point estimates are not straight.
386
(Tables 4 and 5, Fig. 2b), which suggest that the relationship between age and risk of dog bites varies with the dog’s developmental stage.
The rapid increase in non-play bite risk observed in the first
year of the dog’s life corresponds to the period of most rapid sensory, motor and social development (Estep, 1996). Further increases in non-play bite risk up to approximately 3 years of age,
with little change for 2–6 year old dogs (Fig. 1c), are consistent
with the appearance, within the first 3 years of age, of various
types of canine aggression towards humans, as noted by others
(Borchelt and Voith, 1996b; Luescher and Reisner, 2008).
Comparisons between the effects of age at acquisition and age
at the time of biting suggest that effects of re-homing are more
important than correlates of age in determining non-play bites
for younger dogs. However, as the dog matures, correlates of age
become more determinant in whether a dog reacts by biting in a
given circumstance (Fig. 3).
The high positive correlation between duration of ownership
and age for both play and non-play bites explains the similarity
in their relationships to dog bite risk and supports a belief that
both are proxies for similar developmental processes. If so, both
duration of ownership and current age should be associated with
each type of bite through similar mechanisms, even if these mechanisms differ between non-play bites and bites occurring during
play. These results also suggest that changes in non-play bite risk
with increasing duration of ownership are greatest during the first
year (SF) to 1.5 years (KGN) of ownership. As no city-related differences in RRs were detected for current age, the observed city-related differences for duration of ownership might point to
underlying qualitative differences in norms for human–canine
interactions between the two countries.
It is possible that there was some misclassification of the age–
time exposures recorded. As most dogs were unregistered, documented dates of birth and acquisition were not available and owner recall remained the only practical source of age–time
information. Thus, estimated values of age–time variables based
on information provided by owners are not likely to be exact.
While this misclassification of the exposures of interest could
cause inaccurate RR estimates, consistency between the results
from the categorical and continuous variable data analyses
(Fig. 4a–d) inspire some confidence that the results obtained in this
study are not artefacts of the estimated values of the exposures of
interest. Nevertheless, greater importance should be attached to
the overall relationships that the results describe, as opposed to
the precise numeric values of RR estimates. Additionally, the low
prevalence of dogs born in their current home in SF (<1%) suggests
that comparisons involving dogs born at home were heavily influenced by KGN data (28%) and that the conclusions apply primarily
to dogs from KGN. Nevertheless, these results might still be relevant to other US localities, as one study based on US national estimates reported that 26.5% of newly acquired dogs were born in the
respondent’s home (New et al., 2004). Finally, breed-related differences are also likely to exist between groups of dogs, but this was
not investigated as it would require much larger breed-specific
sample sizes.
Conclusions
This study suggests that the associations of dog age at acquisition, current age and duration of ownership with the risk of bites
occurring during and outside of play differ from each other; that
these associations vary during the lifetime of the dog in an agedependent manner; that the association between these age–time
variables and dog bites is strongest in the first year of the dog’s life,
and that the dogs most likely to bite while being played with are
younger than those most likely to bite outside of play. Using fractional polynomials to model these age-time characteristics as continuous variables has been a valuable step in providing an insight
into how their relationships with dog bites change over the lifetime of a dog. Pending confirmation of these findings, it is to be
hoped that veterinarians can use this information to help owners
develop realistic expectations regarding changes in their dogs’
behavior over time. This is important, as incongruencies between
dog–owner expectations and canine aggressive behavior sometimes culminate in relinquishment and/or euthanasia.
Conflict of interest statement
None of the authors of this paper has a financial or personal
relationship with other people or organizations that could inappropriately influence or bias the content of the paper.
Acknowledgements
The authors thank the staff of the following clinics for assistance in data collection – All Creatures, All Pets, Animal Care,
Chang’s Veterinary Clinic, JSPCA, Noah’s Ark, Phoenix Veterinary
Clinic, Veterinary Medical (KGN), Pet’s Unlimited, SFSPCA, The Avenues (SF).
References
Appleby, D., Bradshaw, J., Casey, R., 2002. Relationship between aggressive and
avoidance behaviour by dogs and their experience in the first six months of life.
Veterinary Record 150, 434–438.
Borchelt, P.L., Voith, V.L., 1996a. Aggressive behavior in dogs and cats. In: Voith, V.L.,
Borchelt, P.L. (Eds.), Readings in Companion Animal Behavior. Veterinary
Learning Systems, Trenton, New Jersey, pp. 217–229.
Borchelt, P.L., Voith, V.L., 1996b. Dominance aggression in dogs. In: Voith, V.L.,
Borchelt, P.L. (Eds.), Readings in Companion Animal Behavior. Veterinary
Learning Systems, Trenton, New Jersey, pp. 230–239.
Cornelissen, J.M., Hopster, H., 2010. Dog bites in The Netherlands: A study of
victims, injuries, circumstances and aggressors to support evaluation of breed
specific legislation. The Veterinary Journal 186, 292–298.
Davis, B.W., Alie, K., Fielding, W.J., Morters, M., Galindo, F., 2007. Preliminary
observations on the characteristics of the owned dog population in Roseau,
Dominica. Journal of Applied Animal Welfare Science 10, 141–151.
Deddens, J.A., Petersen, M.R., 2008. Approaches for estimating prevalence ratios.
Occupational and Environmental Medicine 65, 486–501.
Drobatz, K.J., Smith, G., 2003. Evaluation of risk factors for bite wounds inflicted on
caregivers by dogs and cats in a veterinary teaching hospital. Journal of the
American Veterinary Medical Association 223, 312–316.
Estep, D.Q., 1996. The ontogeny of behavior. In: Voith, V.L., Borchelt, P.L. (Eds.),
Readings in Companion Animal Behavior. Veterinary Learning Systems, Trenton,
New Jersey, pp. 19–31.
Feddersen-Petersen, D., 1991. The ontogeny of social play and agonistic behaviour
in selected canid species. Bonner Zoologische Beiträge 42, 97–114.
Feddersen-Petersen, D., 1994. Social behavior of wolves and dogs. Veterinary
Quarterly 16, 51–52.
Fielding, W.J., Mather, J., 2001. Dog ownership in the West Indies: A case study from
the Bahamas. Anthrozoos 14, 72–80.
Georges, K., Adesiyun, A., 2008. An investigation into the prevalence of dog bites to
primary school children in Trinidad. BMC Public Health 8, 85.
Gershman, K.A., Sacks, J.J., Wright, J.C., 1994. Which dogs bite? A case-control study
of risk factors. Pediatrics 93, 913–917.
Greenland, S., 1989. Modeling and variable selection in epidemiologic analysis.
American Journal of Public Health 79, 340–349.
Greenland, S., Pearl, J., Robins, J.M., 1999. Causal diagrams for epidemiologic
research. Epidemiology 10, 37–48.
Guy, N.C., Luescher, U.A., Dohoo, S.E., Spangler, E., Miller, J.B., Dohoo, I.R., Bate, L.A.,
2001. Risk factors for dog bites to owners in a general veterinary caseload.
Applied Animal Behaviour Science 74, 29–42.
Hsu, Y.Y., Sun, L.C., 2010. Factors associated with aggressive responses in pet dogs.
Applied Animal Behaviour Science 123, 108–123.
Lockwood, R., 1995. The ethology and epidemiology of canine aggression. In:
Serpell, J. (Ed.), The Domestic Dog: Its Evolution, Behaviour and Interactions
with People. Cambridge University Press, Cambridge, pp. 131–138.
Luescher, A.U., Reisner, I.R., 2008. Canine aggression toward familiar people: A new
look at an old problem. Veterinary Clinics of North America: Small Animal
Practice 38, 1107–1130.
Maragliano, L., Ciccone, G., Fantini, C., Petrangeli, C., Saporito, G., Di Traglia, M.,
Natoli, E., 2007. Biting dogs in Rome (Italy). International Journal of Pest
Management 53, 329–334.
Messam, L.L., Kass, P.H., Chomel, B.B., Hart, L.A., 2012. Risk factors for dog bites
occurring during and outside of play: Are they different? Preventive Veterinary
Medicine 107, 110–120.
Messam, L.L.M., Kass, P.H., Chomel, B.B., Hart, L.A., 2008. The human–canine
environment: A risk factor for non-play bites? The Veterinary Journal 177, 205–
215.
New, J.C., Jr., Kelch, W.J., Hutchison, J.M., Salman, M.D., King, M., Scarlett, J.M., Kass,
P.H., 2004. Birth and death rate estimates of cats and dogs in U.S. households
and related factors. J Appl Anim Welf Sci 7, 229–241.
O’Sullivan, E.N., Jones, B.R., O’Sullivan, K., Hanlon, A.J., 2008. Characteristics of 234
dog bite incidents in Ireland during 2004 and 2005. Veterinary Record 163, 37–
42.
Overall, K.L., Love, M., 2001. Dog bites to humans – Demography, epidemiology,
injury, and risk. Journal of the American Veterinary Medical Association 218,
1923–1934.
Petersen, M.R., Deddens, J.A., 2006. Re: ‘Easy SAS calculations for risk or prevalence
ratios and differences’. American Journal of Epidemiology 163, 1158–1159;
author reply 1159–1161.
Rooney, N.J., Bradshaw, J.W.S., Robinson, I.H., 2000. A comparison of dog–dog
and dog–human play behaviour. Applied Animal Behaviour Science 66, 235–
248.
387
Rosado, B., Garcia-Belenguer, S., Leon, M., Palacio, J., 2009. A comprehensive study of
dog bites in Spain, 1995–2004. The Veterinary Journal 179, 383–391.
Royston, P., Ambler, G., Sauerbrei, W., 1999. The use of fractional polynomials to
model continuous risk variables in epidemiology. International Journal of
Epidemiology 28, 964–974.
Serpell, J., Jagoe, J.A., 1995. Early experience and the development of behaviour. In:
Serpell, J. (Ed.), The Domestic Dog: Its Evolution, Behaviour and Interactions
with People. Cambridge University Press, Cambridge, pp. 79–102.
Stein, D.J., Dodman, N.H., Borchelt, P., Hollander, E., 1994. Behavioral-disorders in
veterinary practice – Relevance to psychiatry. Comprehensive Psychiatry 35,
275–285.
The American Veterinary Medical Association, 2001. A community approach to dog
bite prevention. Journal of the American Veterinary Medical Association 218,
1732–1749.
Wake, A.A.F., Minot, E.O., Stafford, K.J., Perry, P.E., 2009. A survey of adult victims of
dog bites in New Zealand. New Zealand Veterinary Journal 57, 364–369.
Wright, J.C., 1996. Canine aggression: Dog bites to people. In: Voith, V.L., Borchelt,
P.L. (Eds.), Readings in Companion Animal Behavior. Veterinary Learning
Systems, Trenton, New Jersey, pp. 240–246.
Zou, G., 2004. A modified Poisson regression approach to prospective studies with
binary data. American Journal of Epidemiology 159, 702–706.

RecentResearch-Risk_Factors_for_Dog_bites

Transcription

Similar documents

Winter 2014 Newsletter

Children`s TV 1

Shopping List - The Wayward Hound

Did You Hear the One About? Woofstock Festivals: Days of

denio`s 2015 - Denios Market

Puttln`On Th - Greenwich Public Schools

x vendor parkin g

Lesson Plan: Dogs

pdf of project descriptive panel

AEC Dog Show flyer06.indd