A consistent pet ownership dataset is now available for this birth cohort. ALSPAC is a well-characterised resource that contains a number of physical and mental health outcomes that can be used by researchers concerned with HAI. The access to possible confounding variables and longitudinal nature of the data will allow future researchers to investigate the possible mechanisms underlying potential associations between pets and well-being.
3.3. Univariable Analysis
Ownership of each pet type reported by the mother across all time points up to child age 10 years, were tested for association with the variables of interest. Univariable results are not presented in full due to the later use of multivariable modelling which accounts better for confounding effects. However, we have described univariable findings in short here, because it is important to indicate factors which may appear to be associated with pet ownership, on first glance. Although some findings for ownership of ‘any pets’ are presented here, for interest, we feel that it is more appropriate to focus on factors associated with individual pet types, as these can differ.
3.3.1. Association between Pet Types
On univariable analysis, ownership of some pet types was often shown to be associated with an increased likelihood of ownership of other pet types (data not shown). This is to be expected, as there are likely to be some people who like pets and own a number of different types. For cats, a negative relationship with dog or bird ownership was seen at gestation (data not shown).
3.3.2. Gender of Child
For all pet types except dog and tortoise/turtle, a positive association was seen for ownership by the mother and the child being female compared to male (data not shown). The effect appeared as the children got older.
Pet ownership reported by the mother differed by ethnicity of the child at all time points: at 7 years, 72% of category ‘white’ owned pets, compared to 59% ‘mixed’, 33% ‘asian’, 15% ‘black’ and 38% ‘other’ (P < 0.001). When categories were combined into ‘white’ and ‘other’, children that were ‘white’ were more likely to own any pets (OR = 2.15, 95%CI 1.63–2.83, P < 0.001). When analysed by separate pet types, the same relationship was seen for most years, except for birds and tortoise/turtle which did not appear to be associated with ethnicity (data not shown).
3.3.4. Number of People in the Household
Larger households (5 or more people) were more likely to own all pet types except cats, and this relationship persisted throughout childhood (data not shown).
3.3.5. Presence of an Older Sibling
Presence of an older sibling was assessed at 18 months of age and assumed to remain constant. It appeared to be consistently positively associated with the mother reporting ownership of all pet types except cats, for which there was a negative association at gestation which then disappeared by 33 months (data not shown). When stratified by whether the child had older siblings or not, the pet ownership trends over time remained similar to that reported above (Figure 1
), but generally lower rates of pet ownership were observed for those without an older sibling. However, there was now no evidence for a linear trend (Chi-squared test for trend) for birds by those with an older sibling. For both those with, and without older siblings, there was now stronger evidence of a linear trend for dog ownership than there was before stratification (P < 0.001 compared to P = 0.05).
3.3.6. Maternal and Paternal Education
The general pattern seen over most years was for those who had achieved higher education levels to be less likely to own each pet type than the lowest education level ((CSE)Certificate of Secondary Education/None), however, there was a positive association between higher education level and cat ownership (data not shown).
3.3.7. Maternal and Paternal Social Class
In nearly all years and for most pet types, professional occupations were least likely to own pets and an increasing gradient across decreasing social class could be seen (data not shown). However, for maternal social class, skilled manual were the least likely to own cats.
3.3.8. Maternal Age at Delivery
The mean age at delivery of mothers of children in each pet owning history category (never, sometimes, always—up to 7 years) was compared using one-way analysis of variance. Children whose mothers ‘sometimes had a pet’ had the youngest mothers (P < 0.001). Individual differences were also seen between ownership of different pet types at different years and maternal age (t-tests, data not shown), although direction of the association varied. Maternal age was categorised into <21, 21–30, >30 yrs; mothers <21 yrs being the greatest proportion of ‘sometimes had a pet’ (P < 0.001). Older mothers had the highest proportion of ‘never had dogs’, however the youngest mothers also had the least tendency to have ‘always had dogs’ (P < 0.001).
3.3.9. House Type
There were differences in mother reported pet ownership by house type for ownership of most pet types (data not shown). The common trend was that children living in terraced houses, flats or rooms were less likely to have mothers that own pets than those living in detached houses.
3.3.10. Mother Had Pets as a Child
When the child was 33 months old (nearly 3 years) the carer (usually the mother) was asked, ‘during childhood (up to the age of 16years), did you have any household pets?’ Many (46%) responded that “yes they always had pets”, 44% responded “yes, part of the time”, and only 10% responded “no, not at all”. This was compared with pet ownership reported by the mothers of the children of ALSPAC at different ages (data not shown). Mothers who sometimes owned a pet as a child were more likely to now report owning them than those who never had pets, and the likelihood was even higher for those who had always had pets as a child. This was seen for every year data point and was unaffected by adjustment for maternal education and social class (data not shown). Using the groupings produced by cluster analysis for pet ownership up to 7 years (never, sometimes, always), there was a pattern for those who had pets as a child (sometimes, always) increasingly also having pets when they had children (P < 0.001).
3.3.11. Previous Pet Ownership by Child
On univariable analysis, pet ownership at nearly 4 years (47 months) was a strong predictor of pet ownership at 7 years (85 months) (full data not shown; for cat and dog the likelihoods were increased by 42 times and 49 times respectively). This is partly explained by the fact that dogs and cats are likely to still be alive 3 years later.
The ALSPAC sample is of mothers recruited in pregnancy and follows their children from 0–10 years only. It is a sample of one population of children in the UK, and thus may not be generalisable to other areas. Participants are predominately white, but at baseline the ALSPAC cohort was broadly representative of the UK population [47
]. Interpretation of findings must note the difference between this population and other UK pet ownership studies which mainly report findings from community or general population-based samples of adults [29
]. The dataset has some advantages over previous ones because of its large sample size, longitudinal design, and use of multivariable analysis including a variety of different factors. The resulting data are powerful, and in part responsible for the discovery of new pet ownership risk factors and interaction terms. It also means that variables with relatively small effect sizes have been identified, which may lead to questioning whether such small effects are substantially useful in predicting pet ownership. Some models were simpler to build and some had better model fit than others. Interaction terms suggested here require cautious interpretation and there is a need to investigate whether these models, and their interaction terms, replicate in other datasets. Interpretation of our data leads to the conclusion that factors contributing to ownership of different pet types are complex and despite our inclusion of factors hypothesised or previously indicated to be associated with pet ownership, there are probably other important factors that were unmeasured and unaccounted for in our analysis..
Models were built without the use of a variable to indicate ownership of the same pet type at the earlier time point, 47 months. Final models were tested with the addition of this variable (data not shown). As hypothesised it was a strong predictor but had the effect of removing some other important variables from the model; mostly factors such as ethnicity, maternal age at delivery, and socioeconomic indicators (education or social class). Although the decision was made to exclude this variable from the model for the reasons above, it may have resulted in over-adjustment due to its place on the causal pathway.
Family pet ownership was seen to increase during childhood; this may be in part due to the prompting effect of asking about more types of pets from 21 months onwards, however this seems unlikely due to the increases mainly being in rodents and rabbits, which were asked in all years. Dog ownership did not follow a linear trend and appeared to dip in the first few years of childhood, which concurs with findings of other studies in the UK and Ireland, namely dog ownership in general is associated with having children in the house, but this is due to the effect of older children (school age), whereas those with young children are less likely to own a dog [29
]. Stratification by presence of older siblings suggested that the association between dog ownership and age became stronger (and now linear), within those with, or without, siblings. This supports the suggestion that dog ownership increases with age of the child.
Pet ownership by the mothers during gestation was fairly high already considering that data were collected during 1990/1991; for comparison, pet ownership in general in the UK has increased over the years and by 2004/2005 it was estimated that approximately half of all UK households owned a pet and a quarter owned a dog [29
]. Cat ownership was the most commonly reported pet owned throughout, which again was unusual as dogs are often thought to be the most common pet type owned by households in the UK [29
Previous studies in the UK and Ireland suggested associations between cat ownership and dog ownership, although their results are conflicting [30
]; we found no association with dog ownership in our final model of cat ownership, and vice versa. An association between dog ownership and fish is interesting as it has been found in a previous study where it disappeared on multivariable analysis [29
]. Westgarth [29
] also reported an association between dog ownership and horse ownership; horses, if considered a ‘pet’ by respondents, are likely to have been included in our ‘other pets’ category. Birds were found to be associated with multiple compared to single dog ownership in Westgarth et al.
’s 2007 study [29
], but not dog ownership in general, so we suggest that this finding may be related to the multiple dog households in our study.
The relationship between cat or dog ownership and education level of the owners might be more complex than previously reported. Murray [31
] reported that cat owners had higher education levels than those without cats, but Eller [34
] reported the opposite association; however neither considered whether the respondents had owned pets as a child, with which we found an interaction term. Both Murray [31
] and Eller [34
] reported that dog ownership decreased as education level increased, however, again they did not report the inclusion of interaction terms. We found an interaction between the mother owning pets as a child and paternal education; where the mother had never owned pets, as paternal education increased the likelihood of dog ownership decreased (similar to that reported in [31
]), but we found that this relationship was not seen when the mother had owned pets as a child. Associations between dog ownership with social class were reported in Downes et al.
’s 2009 study [30
]. No previous studies have modelled both education and social class (occupation) as predictors of pet ownership, they usually choose one or the other, but our model of dog ownership suggests that they might have independent effects.
After multivariable modelling, maternal age at delivery was only independently associated with dog or rabbit ownership, with likelihood decreasing as maternal age increased. It is unlikely to be due to socioeconomic differences between mothers who give birth when they are older or younger, as these were also included in the model. When analysed using the categories never, sometimes, or always owned a pet (up to age 7 years) it appeared that the youngest mothers were more likely to ‘sometimes’ own a pet, however this detail could not be included in the final models due to the ordinal outcome used.
Previous research suggests that children with younger siblings have fewer pets than those with no younger siblings or singletons [52
]. It has also been suggested that the youngest sibling plays more with a pet [53
]. In our study, presence of an older sibling appeared to be an independent predictor of dog, rodent, bird and fish ownership by the mother. We defined pet ownership as any pet in the household reported by the mother, whereas previous findings were concerned with pets ‘owned’ by individual children, of which we did not have information. Our results are likely due to the effect of presence of older siblings having an additive effect on the likelihood of pets being in the house, as pet ownership increases as children get older. It may also be explained if having an older sibling increases your likelihood of being a youngest child (with no younger siblings) or, conversely, having no older siblings makes it more likely (over time) that you will acquire at least one younger sibling. In contrast to our findings, a meta-analytical study of birth cohorts on asthma and allergy reported that elder siblings reduced the odds to own cats, but not dogs [34
Although gender appeared to be associated with a number of pet types on univariable analysis, it did not always remain in the model as it did for cat, rabbit and rodent ownership only (females more likely). Girls may influence their parents to own certain types of pets. The finding that females were more likely to own cats than males concurs with other studies in the UK and Ireland [30
], however this relationship has never previously been demonstrated to also apply to children.
Downes (2009) reported differences in cat ownership by house type, but this did not remain in our model. House type appeared to affect only dog and rabbit ownership in our study. This could be explained by reasoning that dogs and rabbits are perceived to require more outdoor, and possibly indoor, space than other pet types. Number of people in the household was also reported to be associated with dog ownership in the UK in [29
] and [31
], but we did not see a similar association for cat ownership, again concurring with previous findings [31
]. Ethnicity other than ‘white’ appears to be associated with decreased cat, or rodent, ownership, as it remained in the models, contrary to our models for some other pet types. This may be due to the power to detect differences due to lower numbers of some other pet types, or real differences in pet keeping by different cultures. It must be noted that in ALSPAC, prevalence of ethnic minorities is relatively low [47
A general theme was that pet ownership by the mother as a child was a predictor in all but one model (bird); usually owning pets as a child increased the likelihood of pet ownership later on. Although past pet ownership is known to be linked to future pet ownership [54
], this variable has not always included in studies of factors associated with pet ownership. A limitation of our data set was that we do not know the individual pet types that were owned previously. Pet ownership in one generation appears, in part, to influence ownership in the next generation; that is, people who grow up with pets allow their children to grow up with pets. One explanation is that the characteristics that make some people like pets may be heritable. Alternatively, the will or desire to own a certain pet may overcome what might be perceived as barriers to ownership for other people. For example, if you really like and want a dog, you own one, regardless of lifestyle factors. A significant contributor to this may be because you had dogs as a child.
When describing pet ownership in a sample, a case definition of what constitutes ‘owning’ a ‘pet’ is also important. This interpretation is likely to vary between geographic locations and cultures. Some studies are not as clear as others as to which criteria were used to define ownership of a pet (or even describe how the question was worded in their questionnaire). A random telephone survey in Ireland defined a pet dog as a dog that was being fed by the household and considered a pet by the eligible participant, and a pet cat as a cat that was both fed by the household and allowed into the house [30
]. Some differences across study findings may be due to differences in exclusion and inclusion criteria to define a pet, which may or may not include stray, free-roaming, or part-owned animals. In our study, interpretation of what defined whether an animal was a ‘pet’ lay with the respondent (the mother), but they were prompted by a list of common animals considered as pets. The word ‘own’ was not used in the question, which could affect the interpretation. Somebody may live with a pet that is considered ‘owned’ by a different member of the household, so use of this word could confuse interpretation by some respondents. We have also not considered more complex family arrangements where a child shares time between two homes. Although we had to make the assumption that children in this study live with the pets that the mother reported, it would be interesting to investigate whether a child feels that they also own each pet reported by the mother. Due to our assumptions we may have over-estimated or under-estimated the effects of some variables as it is possible that some children, although likely a minority, do not actually live in the same household as the mother, thus do not really live with these pets.
Many people grow up with pets, thus for even a small effect size, the population attributable risk of any health benefit may be of significant value [28
]. Three possible explanations are proposed to explain the apparent positive effects of pets on human health [3
]: (a) there is no real association and it is due to co-factors that are linked to both pet ownership and health measures; (b) effects are indirect due to enhancement of social contact with people due to proximity to the animal; (c) effects are due to the nature of the relationship with the animal and the provision of emotional support. It is probable that any real effects are mediated through a combination of both (b) and (c), however, due to the quality of study design and data analysis used in some papers championing the positive health benefits of pets, it is difficult to argue that (a) is not the case.
This study demonstrates clearly that there are a number of socioeconomic and demographic factors that are associated with pet ownership, thus the relevant ones to specific pet types must also be accounted for in data analysis of any possible improved health outcomes from owning pets. Socioeconomic factors are known to be associated with the health of individuals, for example, in children of the ALSPAC sample, both fat mass [56
] and number of doctor’s consultations [57
] differed by socioeconomic indicators. In our analyses, education and social class often had independent effects on pet ownership, and should therefore both be included if possible, rather than one as a proxy for both.
It is clear that the factors contributing to the ownership of different pet types can be very different, and so it is very important to consider these separately rather than lumping as ‘pet ownership’. This is particularly relevant when investigating the possible ‘effects’ of pet ownership as a whole, which may be misleading. It has been common practice in HAI research to report the benefits of ‘pet ownership’, when only dogs were studied, or the type of pet studied was not even reported [7
]. Considering our findings, this is of concern.