Sample and procedure
The sample included all dog owners (n = 1113; 41.3 %) participating in the Pet Connections Study. Pet Connections is a cross-sectional study designed to examine the relationship between pet ownership and social capital, including between and within country (US vs. Australia) differences, and to identify pet-related factors that precipitate and maintain social connectedness in neighborhoods. The methodology is reported elsewhere [14, 20], and is briefly described here. To be eligible, participants were required to be aged ≥ 18 years and to have lived in their neighborhood for at least one year. Overall, 2692 adults participated in a cross-sectional telephone survey across four study sites (San Diego, US n = 690, Response Rate (RR) = 45.8 %; Nashville, US n = 664, RR = 44.1 %; Portland, US n = 634, RR = 42.1 %; Perth, Australia n = 704; RR = 60.2 %), yielding an overall response rate of 47.3 %. Study site samples were representative of the wider population in terms of sex, age group, neighborhood socio-economic status and pet ownership rates [20]. The survey was administered to coincide with the autumn and early winter months in the US (September to December 2012) and Australia (April to June 2012).
Socio-demographic measures
Socio-demographic variables included: age, gender, highest education level attained (secondary or less, vocational training, bachelor degree or higher or other), ethnicity (US only – White/Caucasian, Hispanic or Latino Descent, Black/African American, Asian and Other), country of birth (Australia only – Australia or Overseas), number of dependents <18 living at home, number of years lived in neighborhood and type of residence (house, duplex, townhouse or villa, apartment or flat or other). Participants were also asked whether they owned a dog [21].
Physical activity measures
Number of days per week participants reported ≥30 min of moderate-vigorous physical activity were recorded [22]. Participants reported the frequency and location (e.g., park, streets, beach/river) of where they walked or jogged in their neighbourhood in a usual week. These items were based upon the Neighborhood Physical Activity Questionnaire (NPAQ), which has acceptable reliability [23]. Items from the Dogs and Physical Activity (DAPA) tool were used to collect frequency and duration of dog walking per usual week [21]. These items have excellent test-rest reliability (frequency of dog walking/week intra-class correlation (ICC) = 0.98; duration of dog walking/week ICC = 0.94) [21]. Dog owners who reported that they walk or jog with their dog (s) were classified as ‘dog walkers’.
Perceptions of safety and the neighborhood
General neighbourhood perceptions of safety were based on existing scales. The ‘Feel Safe In Neighborhood’ scale was based on four items (i.e., Feel safe: walking alone in daytime in neighborhood; walking alone at night in neighborhood; using parks in neighborhood; and in own home) (Cronbach’s α = 0.78) [24, 25]. The ‘Neighborhood Natural Surveillance’ scale included three items (i.e., see people out walking and jogging; people in neighborhood feel it is a safe place to live; and few people walk down my street) (Cronbach’s α = 0.50) [14, 24]. Participants also reported whether their perception of neighborhood safety was enhanced by three items relating to natural surveillance (i.e., I feel safer when I see people walking; I feel safer when I see people out walking with their dog; and I feel less safe when I see deserted streets and parks) (Cronbach’s α = 0.60) [24]. All items in scales were measured using a 4-point Likert scale (1 = strongly disagree; 4 = strongly agree). Items in each scale were summed and then averaged.
The perceived potential ‘Neighborhood Problems’ scale measured graffiti and/or vandalism, crime, traffic, neighborhood maintenance (6 items: inadequate lighting at night, houses or yards not well looked after, poor upkeep of parks and public open space, trash or litter in public areas, vacant or run-down buildings, and poor street lighting) and social incivilities (3 items: noisy neighbours or loud parties, drug dealing or drug use, people not cleaning up after dogs) [24–28]. All items in this scale were measured using a 4-point Likert scale (1 = strongly disagree; 4 = strongly agree). Items were dichotomised (agree vs. disagree) and summed (range 0–12) (Cronbach’s α = 0.89). Details of this measure are reported elsewhere [29].
Two items measured neighborhood perceptions specifically in relation to dog walking. Owners reported whether being with a dog helped them to feel safer when out walking or jogging and whether walking or jogging with their dog enabled them to get to know their neighborhood (dichotomous: true/false) [14].
Statistical analyses
Analyses were restricted to dog owners only (N = 1113). Chi-square tests were used to examine socio-demographic differences between dog walkers and non-dog walkers for each study site and overall. Linear regression was used to examine the relationship between dog walking status (independent variable) and days/week of ≥30mins of moderate-vigorous physical activity and frequency of walking in the neighbourhood/week (dependent variables). Logistic regression was used to examine the relationship between dog walking status (independent variable) and if participants walked to their local park in a usual week (dependent variable). Two models were run for each physical activity-related dependent variable; unadjusted and then adjusted for age group, sex, highest education level, ethnicity (US), country of birth (Australia), number of children in household, housing type, and time lived in neighborhood. Descriptive analyses (mean and standard error) were conducted for frequency and duration of dog walking/week. All analyses were conducted separately for each study site with non-dog walkers as the reference group. Between city comparisons were performed for dog walkers only (reference site = Perth), using the Bonferroni correction procedure for hypothesis testing involving multiple comparisons.
Linear regression was used to examine the relationship between dog walking status (independent variable) and perceptions of safety dependent variables (Feel safe in neighborhood; Neighborhood problems; Neighborhood surveillance; Feel safe if have neighborhood surveillance). All models were adjusted for age group, sex, highest education level, ethnicity (US), country of birth (Australia), number of children in household, housing type, and time lived in neighborhood. All analyses were conducted separately for each study site with non-dog walkers as the reference group. Between city comparisons were examined for dog walkers only (reference site = Perth) using Bonferroni correction of p-values for multiple comparisons.
Finally, descriptive analyses were conducted by study site and overall to examine the percentage of dog walkers who reported that they ‘Feel safer walking with dog’ and ‘Got to know neighborhood through walking dog’. Non-dog walkers were excluded from analyses as our analyses were focused on dog walking only. Logistic regression was used to examine the relationship between gender (independent variable; reference group = males) and the proportion of participants reporting they ‘Feel safer walking with dog’ and ‘Got to know neighborhood through walking dog’ (dependent variables). All models adjusted for age group, highest education level, ethnicity (US), country of birth (Australia), number of children in household, housing type, and time lived in neighborhood. All analyses were conducted separately for dog walkers from study site and overall. Between city comparisons were performed for a) all dog walkers (reference site = Perth) and b) female dog walkers only (reference site = Perth) correcting our p-values for multiple comparisons using the Bonferroni procedure.