key: cord-1040741-yzfd3o14 authors: Bowen, Jonathan; García, Elena; Darder, Patricia; Argüelles, Juan; Fatjó, Jaume title: The effects of the Spanish COVID-19 lockdown on people, their pets and the human-animal bond date: 2020-06-13 journal: J Vet Behav DOI: 10.1016/j.jveb.2020.05.013 sha: e799858e4ea23ee147363c836a100920b6160b3f doc_id: 1040741 cord_uid: yzfd3o14 The purpose of this study was to investigate the effects of the Spanish confinement for the control of the COVID-19 pandemic on the behavior of pet cats and dogs, and the support that pets provided to their owners. We found that the quality of life owners was strongly influenced by the lifestyle and emotional effects of the confinement, and that pets provided them with substantial support to mitigate those effects. However, pets showed signs of behavioral change that were consistent with stress, with dogs that had pre-existing behavioral problems being the most affected. Jonathan Bowen 1,2 , Elena García 1 , Patricia Darder 1 , Juan Argüelles 4 , Jaume Fatjo 1,3 * * * * 4 1 Introduction Introduction Introduction Introduction 28 On January 30th 2020, the World Health Organization declared a Public Health Emergency 29 related to the outbreak of a new coronavirus (SARS-CoV-2). Twelve days later the name for this 30 new disease was announced (COVID-19) and just four weeks after that, the outbreak was 31 upgraded to a pandemic. In response, many countries implemented unprecedented restrictions 32 on the movement, work and leisure activities of their citizens, with the aim of reducing the 33 reproduction number of the virus (R0). This study looked at the effects of the initial confinement period on Spanish pet owners, their 36 pet cats and dogs, and on the relationship between them. We were interested in understanding 37 how the human-animal relationship might help pet owners to cope with the effects of the 38 confinement, given that approximately 24% of Spanish households have at least one dog and 39 11% have at least one cat (FEDIAF, 2018) . Previous studies have looked at the psychological consequences of different degrees of 42 quarantine and self-isolation related to infectious disease outbreaks, such as SARS, MERS and 43 Ebola. Separation from friends and relatives, the loss of freedom, fear of the disease and 44 boredom can all have negative effects on quality of life and health. Commonly reported effects 45 include stress, anxiety, low mood, depression, irritability, insomnia, and difficulties with the 46 resumption of normal life after the end of the period of confinement (Brooks et al, 2020; 47 Hawryluck, 2004) . Research on human-animal relationships suggests that companion animals can be a source of 50 social support for people and can help them to cope with difficult situations (McNicholas and 51 Collis, 2006; Wood et al, 2015) . Most studies have focused either on very specific scenarios (e.g. 52 people suffering from specific illnesses, going through a process of bereavement, or animal 53 assisted interventions), or on pet owners, using scales that measure overall attitudes about 54 human-animal relationships. The ongoing COVID-19 crisis offers an opportunity to explore the 55 role of companion animals as sources of social support at a time when the vast majority of the 56 population is being exposed to the same social and environmental challenges. 57 58 However, companion animals can also experience the negative consequences of a period of 59 home confinement; the quality of life of dogs and cats is highly influenced by the characteristics 60 of their physical and social environment, and the behavior and lifestyle of their owners, all of 61 which would be substantially changed during an official lockdown (Fatjó and Bowen, 2020) . Spain was chosen as the subject of this survey because it had a well-defined confinement policy 64 that was strictly enforced. On March 14, 2020, an official lockdown act entered into force in 65 Spain, which included the following measures; social distancing, the closure of schools and 66 universities, banning of mass gatherings and public events and the suspension of all non-67 essential economic activities (BOE-A-2020-3692). The lockdown act did allow dog owners to 68 walk their dogs, but only one person could walk the dog at a time, the animal had to be on a 69 leash at all times, and dogs were not allowed to interact with people or other dogs. The effects of confinement in Spain were therefore likely to be more consistent than in other 72 countries. For example, in the UK there was a period of advisory isolation and social distancing, 73 followed by a gradual shutdown as businesses chose to close and furlough their staff, and finally 74 an official lockdown when all but essential shops and businesses were closed. However, 75 although there was a very stringent lockdown in Spain, many people chose to self-isolate in 76 Spain before the official announcement. The study used an online survey to collect a convenience sample of respondents, which was 81 deemed the safest approach, given the movement restrictions and the risk of infection that 82 would result from a more traditional face-to-face public survey. A link to the online 83 questionnaire was circulated through social media and online forums for pet owners, veterinary 84 clinics, animal shelters and charities. 85 86 At the start of the survey respondents were asked to confirm that they were residents of Spain, 87 and that they agreed to the terms of the study. In addition, the survey tool was able to provide 88 information about the country of the respondent, and this information was used to exclude any 89 responses that were not from Spain. This was deemed to be important because the confinement 90 regulations differ markedly between countries. Respondents who were aged under 18 were also 91 excluded. Materials Materials Materials Materials 94 A questionnaire was developed by the authors to collect information from dog and cat owners 95 on the effect of confinement on both the family and the pet. The questionnaire included 96 multiple choice and Likert scale questions with options to enter additional text information for 97 some of the items. A full copy of the questionnaire can be found in the supplementary material 98 (Appendix A). The time required to complete the questionnaire was estimated to be 16 minutes. 99 100 The survey consisted of several sections. The first section collected information about the 101 respondent and their household; the respondent's sex, age group, family role, the composition 102 of the household (numbers of 18-64 year-old adults, 65+ year-old adults, and different ages of 103 children), and the number of resident pet cats and dogs. Respondents were also asked about 104 their type of home (house or flat/apartment), size of home (small, medium or large), outdoor 105 space (garden, patio or balcony), whether the home was large enough for the residents to carry 106 out activities independently, and whether it provided sufficient light and fresh air. A subjective 107 rating of size of home was chosen instead of the size in square meters, because this was 108 considered to be an easier question to answer that was more reflective of the respondent's 109 perception of their environment. The second section asked about the effect of the confinement on the respondent and their 112 household; the number of weeks they had been confined, for how much longer they expected 113 the official confinement period to continue, and which people in the household were able to 114 work from home or had permission to go to work. It also asked about the negative financial, 115 emotional, health and lifestyle impact of the official confinement on the household, and the 116 effect the confinement had had on the respondent's overall quality of life. It would have been possible to calculate the duration of a respondent's official confinement 119 from the date of completion of the survey and the date of the introduction of mandatory 120 confinement (official lockdown). However, it was expected that some people would have 121 engaged in voluntary self-isolation at home, in accordance with unofficial advice that preceded 122 the official lockdown by several weeks. Other people who initially had permission to work 123 outside the home might have entered confinement later (for example, due to emergence of 124 disease signs). So, it was decided only to use the respondent's declared confinement period in 125 the study. 126 127 In the third section, the survey focused on one of the pets in the household, and the 128 respondent's relationship with it. The major part of this section was a modified version of the 129 Cat/Dog-Owner Relationship scale (C/DORS), developed by the authors for the measurement of 130 the human-animal bond between cats or dogs and their owners (Howell, et al. 2017 which is based on Richard Emerson's social exchange theory (Emerson, 1976) . This theory 133 proposes that the stability of a relationship is the product of the balance between its perceived 134 costs and benefits. 135 136 MDORS and C/DORS measure three independent dimensions of the owner-pet relationship; 137 interaction between owner and pet, perceived emotional closeness and perceived costs. The 138 interaction dimension describes the way a person shares time with a pet in terms of day-to-day 139 activities like play, grooming and social activities. The emotional closeness dimension describes 140 how dependent on the pet the person is, and how much emotional support the person derives 141 from the relationship. The perceived costs dimension evaluates the degree to which the pet 142 affects the owner's financial and time budgets, and overall lifestyle; for example how much it 143 costs to care for, and how it prevents the person from doing things they would otherwise want, 144 or need, to do. 145 146 Some items from C/DORS were removed because they related to activities that were not 147 permitted during the official lockdown, such as taking the pet to visit people. The item asking 148 about the respondent's emotional reaction to the death of their animal was also removed, as this 149 was considered to be potentially distressing for people to answer during the present crisis. Respondents were also asked directly about the effect the confinement had had on their pet's 152 quality of life, their relationship with their pet, and how much their pet had helped them 153 during the period of the confinement. They were also asked about whether they had been 154 angry with their pet more often recently, since the confinement. We wanted to get information 155 about tension between the owner and the pet, but we chose not to ask a direct question about 156 the use of punishment. This was because, in our experience, respondents are put off by such 157 questions and may not answer truthfully. So, we chose to ask the question "Recently, how often 158 have you been mad at your dog/cat", with a 7-point response from "much less than before the 159 confinement" to "much more than before the confinement". 160 161 It might be expected that the confinement period could lead to the development, or worsening, 162 of problem behavior in cats and dogs. Respondents were therefore presented with a short list of 163 behavior problems that were common to both cats and dogs, including family-directed 164 aggression, aggression toward resident conspecifics, destructiveness, house-soiling, and noise 165 fear. They were asked to indicate which problems had got better, stayed the same or got worse 166 during the confinement (with the option to indicate that the animal had never shown each 167 problem behavior). In addition, respondents were asked about species-specific behavior 168 problems; for example, cat owners were asked about urine marking, and dog owners about 169 aggression to other dogs during walks and problems with being left alone at home. Apart from these specific behavioral problems, respondents were also asked to indicate which 172 of a set of general changes in behavior their cat or dog had exhibited, including being more 173 nervous, more stressed, more relaxed, more excitable, more calm, more attention-seeking, more 174 demanding, more frustrated and more irritable since the confinement. These are subtle changes, 175 some of which would be expected to lead to conflict between the pet and owner, and which 176 could, over time, lead to the development of behavior problems. It was expected that these 177 general changes would be more likely to be affected by the confinement than the prevalence of 178 the specific problem behaviors. 179 Respondents were not given specific guidance on how to assess these general changes, as this 180 would have lengthened the survey and we intended only to use the negative changes to 181 calculate a composite score to represent the owner's overall impression. Respondents were presented with a set of likely confinement-related concerns and asked which 184 of these were of concern to them. These included eight concerns that were common to both cat 185 and dog owners; concerns about the pet's health, being able to obtain food for the pet, accessing 186 medicine and veterinary care, concerns about prohibitions on the pet going outside, as well as 187 concerns about weight gain, children not respecting the pet's space and need for rest, the effect 188 that the loss of routine might have on their pet and how the pet might cope with going back to 189 normal life after the confinement. Prior to the study, the authors had been asked by their 190 clinical clients about how difficult it might be for their pets to adapt to normal life after the 191 confinement was over, so this was presented as one of the concerns in this section. One point 192 was given for each of the eight common owner concerns that were common to both dogs and 193 cats, and then summed to produce a composite score "owner concerns score" for each 194 respondent. For the questions on problem behavior, general changes in behavior and concerns, respondents 197 were also given the opportunity to write any additional comments into a text box. Dog owners were asked how often they walked their dogs each day before the confinement, as 200 well as for an approximate total duration of daily walks. Cat owners were asked about their cats' 201 outdoor access before and during the confinement. Respondents with multi-pet households were asked to answer the pet-specific questions about 204 the cat or dog whose name was first in alphabetical order. This was done in order to randomize 205 the selection of pet, to avoid bias toward a particular pet that the person felt strongly about or 206 was concerned about. 207 208 Statistical Statistical Statistical Statistical analysis analysis analysis analysis 209 Normality of distribution was tested using a D'Agostino-Pearson test. When comparisons were 210 made between groups, an appropriate test was chosen; Chi square for binary variables, and 211 either a t-test or the non-parametric equivalent for continuous or ordinal data, depending on 212 the distribution of the data. Correlations were tested using Pearson correlation or Spearman 213 rank correlation, again depending on the distribution of data. All cluster analysis was carried 214 out using a two-step clustering method with the distance measure being log-likelihood, and the 215 clustering criterion being Akaike's information criterion. Silhouette measure was used to assess 216 model quality. For binary logistic regression, the "enter" method was used. Omnibus measures 217 of model quality and overall percentage of correct classification were used to assess model 218 quality. The software packages used for the statistical analysis were SPSS version 25 (IBM) and 219 Prism version 8.4 (GraphPad). 220 Results Results 222 All responses with incomplete information were excluded from the initial population of 1329 223 responses from adults located in Spain, leaving a total of 1297 complete responses for the 224 analysis. Demographics Demographics Demographics Demographics 227 Of the 1297 surveys, 794 owners reported about a dog (61.2%) and 503 about a cat (38.8%). As 228 is common in online surveys, the majority of respondents were female (90.5%). Figure 1 shows 229 the percentage of people in each age group. The majority of households had one or two 18-64 230 year-old resident adults, and 36.1% had one or more children from one of the age groups (see 231 Seventy-four-point three percent of households had at least one dog, and 57% at least one cat, 248 with the majority of households having one cat or dog (see Table 2 ). Four hundred and four people reported having at least one dog and one cat (31.15%). Respondents were free to choose whether to answer about a dog or a cat. Although the official lockdown applied to all residents of Spain, essential workers were 273 permitted to continue to go to work. In the present study, in only 8.1% of households were all 274 members still permitted to go out to work. In the largest proportions of households all residents 275 were confined and none were able to work from home (23.8%), and in 21.3% of households all 276 residents were confined and some were able to work from home (see Table 4 ). We are all confined and none of us work from home 23.8% We are all confined and some of us work from home 21.3% We are all confined and we all work from home 15.7% Some of us are confined, and some of us have permission to go out to work 31.1% We all have permission to go out to work 8.1% 279 280 Table 4 . Level of confinement for members of the household. Some individuals would have self-isolated before the official lockdown, and some would have 292 experienced a change of working status during the lockdown; for example due to a change in 293 working role that would affect their right to go to work, or because of illness that would require 294 them to be quarantined. So, the duration of a respondent's confinement period would not align 295 perfectly with the official date of lockdown. The mean duration of confinement reported was 296 3.2 weeks (SD 1.19), and the mean expected further duration of the lockdown was 4.6 weeks 297 (SD 2,37). The confinement period would be expected to have economic, emotional, health and lifestyle 302 impact on households. In response to the question "what negative impact has the official 303 confinement had on your household, 49.2% of respondents indicated that there had been "a lot" 304 or "quite a lot" of negative lifestyle impact on their household. The levels of economic, 305 emotional and health impact were somewhat lower (see figure 4) The most common response was that the respondent's quality of life was slightly worse (44.6%), 319 with 26.8% indicating no change. Comparing those who reported any change in quality of life, 320 61.8% said that their quality of life had got worse, and 11.4% that it had got better. In order to understand the contribution these different impacts on the household might have 323 on overall individual quality of life, Spearman Rank correlation was carried out between these 324 variables (the data was not normally distributed). While all correlations were significant, the strongest correlation was between household 327 lifestyle impact and overall quality of life (r= -0.38), followed by emotional impact (r= -0.34), 328 see Table 5 . Although significant, the correlation between negative economic impact and 329 quality Each item of C/DORs was scored from -2 to +2 ("much less than before the confinement" to 340 "much more than before the confinement"), with zero being "no more or less than before the 341 confinement". A score for the three subscales of C/CORS (interaction, emotional closeness & 342 perceived costs) was calculated from the average of items for that subscale. Unlike in the 343 original MDORS and C/DORS scoring protocols, "Perceived costs" was scored so that high 344 scores indicated an increased negative effect on the owner. The mean scores for the subscales were; interaction 0. respectively). Figure 6 shows the comparison of C/DORS subscale scores between cats and dogs. 360 361 Figure 6 . Comparing C/DORS subscale scores between cats and dogs. Columns show the mean 362 and whiskers the standard deviation. We also asked respondents to answer the question "How much has your pet helped you during 365 the confinement compared with before?", with a 7-point response from "much less" to "much 366 more". Forty-seven percent of people indicated that their pet had helped them moderately more 367 or much more. For 25% of respondents there was no change, and for only 0.7% was there a 368 perceived reduction in support from the pet (see fig 7) . Figure 7 . Responses to the question "How much has your pet helped you during the 372 confinement compared with before?" 373 374 The The The The concerns concerns concerns concerns of of of of owners owners owners owners 375 Owners were presented with a range of concerns about the effect of the confinement on pet 376 ownership. The commonest concern for dog owners was prohibition of going on walks (61.7), 377 and the commonest concern for cat owners was access to veterinary care and medication 378 (39.6%). See Table 6 and had a correct classification rate of 72.9% (see Table 7 ). In this case the full The commonest behavior problems that were getting worse were annoying or excessive 473 vocalization (24.7%), and fear of loud or unexpected noises (16.9%), see Table 9 for a summary 474 of behavior problems in the population and how they changed during the confinement. These 475 are presented graphically in figure 8 . Figure 8 . Illustration of the problematic behaviors of dogs and how they changed during the 488 confinement. "Never present" indicates cases for which the behavior was not observed in the 489 animal either before or during the confinement. 490 491 492 We also asked owners to provide information about general changes in behavior that were not 493 specific behavioral diagnoses. With respect to these broader changes, 29.5% respondents 494 reported no significant changes in the dog's behavior during confinement. The most common 495 general aspect of behavior reported to be higher during confinement was attention-seeking 496 (41.6%), followed by being more nervous ( Figure 9 . General behavioral changes in dogs during the confinement, showing the percentages 516 of owners who reported an increase in each behavior. 517 518 Prior to the confinement, dogs went on an average of 3 walks per day (SD=1.14) compared with 519 2.5 walks per day during the confinement (SD=1.19). This difference was, however, not 520 significant. There was a clear reduction in the duration of walks during the confinement 521 (Wilcoxon matched-pairs signed rank test, two-tailed p<0.0001, W=-75495). Figure 10 shows 522 the amount of time dogs spent outside before and during the confinement period. 523 524 525 Figure 10 . Amount of time each day that dogs spent outside before and during the confinement. 526 527 Factors Factors Factors Factors associated associated associated associated with with with with h h h how ow ow ow dogs dogs dogs dogs were were were were coping coping coping coping with with with with the the the the confinement confinement confinement confinement 528 A score for general negative changes in behavior was calculated, with one point awarded for 529 each of "more nervous", "more stressed", "more excitable", "more attention-seeking", "more 530 demanding", "more frustrated" and "more irritable". This created a score from 0 to 7 for each 531 pet. This score is reflective of underlying changes in behavior that could be indicative of the 532 pet's ability to cope. A two-step clustering procedure was performed using this general change score as the variable. 535 The model was forced to generate two clusters. Silhouette value was 0.7. Sixty-two-point eight 536 percent of dogs showed at least one general change in behavior. Those animals with high scores 537 were considered to be coping less well with the confinement. Table 542 10 Odds ratios indicate the likelihood of the dog being in the group that was coping less well with 548 the confinement (high score for general behavior change). For every one-point in health impact on the owner, from 0 to 5, the dog was 1.3 times more 551 likely to be in the coping-poorly group (higher score for general changes). For every one-point 552 increase in the pet's quality of life, as evaluated by the owner, the dog was 74% more likely to 553 be in the coping-better group. For every one-point increase in score for how often the 554 respondent was getting mad with their dog, the dog was 1.8 times more likely to be in the 555 coping-poorly group. For every additional behavior problem that was getting worse, dogs were 556 1.9 times more likely to be in the coping-poorly group, and for every one-point increase in 557 owner concerns, from 0 to 8, the dog was 1.5 times more likely to be in the coping-poorly 558 group. Factors Factors Factors Factors associated associated associated associated with with with with w w w worsening orsening orsening orsening behavior behavior behavior behavior in in in in dogs dogs dogs dogs 561 Dogs were given a score of 1 for every behavior problem that was worsening, giving a score 562 from 0 to 10. The list of behaviors included were aggression toward family members, aggression 563 to non-resident people, aggression to resident conspecifics, aggression to other species in the 564 home, aggression to other dogs on walks, destructiveness, elimination problems, problematic 565 vocalization, fear of loud of unexpected noises, and problems being left alone at home. Only 566 one dog obtained the maximum score of 10. Two-step clustering was used to split the 567 population, with two clusters emerging naturally (the model was not forced). The silhouette 568 value was 0.8, which was very good. The two clusters were dissimilar in size, with one 569 including 75.7% of dogs, and the other 24.3%. In the larger cluster the mean score for the 570 number of worsening behaviors was 0.28. For the smaller group the mean score was 3.04. A binary logistic regression model was created to compare these two groups. The model passed 573 an omnibus test of model coefficients (Chi square=312.77, df=43, p<0.0001), and had a correct 574 classification rate of 84.0%. The variables in this model were the same as in the previous model 575 of how dogs were coping, but behavior problems scores were excluded as they related to the 576 outcome variable. Only those variables which were significant are presented in Table 11 580 Table 11 . Summary of binary logistic regression results for worsening problems in dogs during 581 the confinement. (only significant associations are reported). Odds ratios indicate the likelihood 582 of the dog being in the group whose behavior problems were worsening more during the 583 confinement (high score for number of worsening behaviors). 584 585 In this model confinement status was influential. Dogs in homes with all family members at 586 home, either none working from home, some working from home or all working from home, 587 were 4.9, 4.8 and 4.1 times more likely to be in the group with more behavior problems that 588 were getting worse, respectively. 589 590 For every one-point increase in score for emotional closeness, from -2 to +2, a dog was 5.1 times 591 more likely to be in the group with more behavior problems that were getting worse. Likewise, 592 for every one-point increase in how often the respondent was getting mad with their dog, the 593 dog was 1.5 times more likely to be in the group with more behavior problems that were 594 getting worse. For every one-point increase in general changes score the dog was 2.2 times 595 more likely to be in the group with more behavior problems that were getting worse. 596 597 Factors Factors Factors Factors associated associated associated associated with with with with w w w worsening orsening orsening orsening specific specific specific specific behaviors behaviors behaviors behaviors in in in in dogs dogs dogs dogs 598 Most behavior problems in dogs and cats were not common, and even fewer got worse during 599 the confinement. So, it was not possible to create meaningful models to assess associations with 600 worsening problems. However, 196 out of the 794 dogs in the study (24.7%) showed an increase 601 in problematic vocalization. This was a large enough group to merit further analysis. A binary logistic regression model was created to compare dogs with an increase in problematic 604 vocalization with those that showed no change. The model passed an omnibus test of model 605 coefficients (Chi square=266.45, df=43, p<0.0001), and had a correct classification rate of 82.4%. 606 The variables in this model were the same as in the previous model of how dogs were coping, 607 but with behavior problems scores excluded as they related to the outcome variable. Only those 608 variables which were significant are presented in For every additional person aged 18-64 yoa in the household, a dog was 1.4 times more likely to 618 be in the worsening problematic vocalization group. For every one-point increase in increased 619 emotional closeness, from -2 to +2, a dog was 2.3 times more likely to be in the worsening 620 vocalization group. For every one-point increase in score for how often the respondent was 621 getting mad with their dog, the dog was 1.4 times more likely to be in the worsening 622 vocalization group. A dog was 32% less likely to be in the worsening vocalization group for 623 every additional walk they went on each day (from 0 to "9 or more" walks each day). For every 624 one-point increase in general changes in behavior score, from 0 to 7, a dog was 2.1 times more 625 likely to be in the worsening vocalization group. The For 52.1% of participants the relationship with their cat did not change, for 46.3% it improved 632 and for 1.6% it became worse during the confinement. 633 634 Table 13 summarizes the behavior problems in the cat population and how they changed during 635 the confinement. Figure 11 presents Figure 11 . Illustration of the problematic behaviors of cats and how they changed during the 646 confinement. "Never present" indicates cases for which the behavior was not observed in the 647 animal either before or during the confinement. 648 649 650 651 Respondents were asked to report on the same general behavior changes for cats and dogs. In 652 cats, 46.3% respondents reported no general changes in the cat's behavior during confinement. 653 The most common general aspect of behavior that was reported to be higher during 654 confinement was attention-seeking (36.4%), followed by being more relaxed (21.7%), being 655 calmer (9.7%), being more demanding (7.4%) and being more nervous (7%). See figure 12 . Outdoor access for cats did not differ between before and during the confinement period. 664 Figure 13 shows the percentages of cats with not outdoor access, limited and free outdoor 665 access A two-step clustering procedure was performed using this general change score as the variable. 673 The model was forced to generate two clusters. Silhouette value for the model was 0.8, which is 674 very good. 675 676 Forty-three-point three percent of cats showed at least one general change in behavior. As with 677 dogs, those animals with high scores were considered to be coping less well with the 678 confinement. A binary logistic regression model was created with membership of the low or high general 681 changes group as the outcome variable. The model passed an omnibus test of model coefficients 682 (Chi square=150.6, df=42, p<0.0001), and had a correct classification rate of 71.0%. The same 683 variables were included in this model as the one for dogs, except that the variables relating to 684 dog walks were replaced with the equivalent variables for outdoor access during the 685 confinement and change in outdoor access (from prior to the confinement). is evidence that women are more willing to participate in online surveys than men (Smith, 711 2008), and that they may use social media differently (Joinson, 2008; Duggan & Brenner, 2012) . 712 In addition, in a study of communication between veterinarians and clients in companion 713 animal practice, a similar female bias was found in respondents, with the implication that 714 women were more engaged with issues related to the pet (Shaw, 2012 When the data was collected for this study, the average time of confinement was 3.2 weeks, 732 which may be regarded as quite short. However, previous studies indicate that periods of 733 quarantine and home confinement as short as 10 days have been associated with negative 734 psychological consequences (Hawryluck et al., 2004) . Respondents in this study also reported 735 having been confined for periods that did not match with the official lockdown. This indicates 736 that many people chose to self-impose restrictions on the amount of contact they had with 737 other people, which could be related to the feelings of anxiety and uncertainty surrounding the 738 disease. In addition, three weeks is sufficient time for people to get a sense of the effect of the 739 confinement on them, but without any certainty about when the confinement might end; on 740 average, people expected to be confined for a further month and a half, but more than ten 741 percent of people indicated that they expected to be confined for a further 8 or more weeks. So, 742 we would propose that even though the duration of confinement was quite short, it is highly 743 likely that people would already have been experiencing considerable stress. After only a few weeks, we might not expect a dramatic change in the behavior of pets, such as 746 the development of new behavior problems like owner-directed aggression, but we might 747 expect changes in existing problem behaviors and this is what we found. For example, owners 748 reported that 24.7% of dogs that already had a problem of excessive or annoying vocalization, 749 became worse, and this could be due to a number of reasons from territoriality to stress and 750 frustration. However, it is possible that the main reason for the perceived increase in 751 problematic vocalization was that people were at home to hear it. 752 753 However, we must accept that the lockdown will not suddenly end, with people going back to 754 their previous lifestyles. It will go on, in some form, for months. For example, according to 755 current estimates from the Spanish Government, children would not be able to return to school 756 out until September 2020, and it may be even longer before dogs are able to interact with each 757 other, or with people, on walks. We should be aware that more general changes in the animals' 758 disposition could, over time, lead to more serious problems. So, we included a panel of questions 759 about changes in the pet's general behavior. As a result, we observed an increase in underlying, general dimensions of behavior such as 762 being more excitable, nervous, irritable, demanding or attention-seeking that could easily lead 763 to other problems if the lockdown continued or these changes were mishandled by owners. Animals that were showing more of these signs could be considered to be coping worse, and at 766 greater risk of worsening behavior or the development of new problems, so we compared 767 groups with more versus less general behavioral changes, and more versus less problem 768 behaviors using binary logistic regression. In the BLR model of general behavioral changes in dogs, there was a positive association 771 between the dog not coping well and negative health impact of the confinement on the owner, 772 frequency of the owner getting mad with the pet, the number of behavior problems getting 773 worse and the owner concerns score. There was a negative association with the effect of the 774 confinement on the pet's quality of life, as evaluated by the owner (i.e., pets with improved 775 quality of life were less likely to be in the group that was coping less well). It is interesting that negative health impact on the household was a factor in general changes, as 778 health would be a substantial source of stress for households. Not only are people concerned 779 about the risk of infection and any potential signs of infection they might observe in household 780 members, but also about how to deal with existing health problems and new, non-COVID, 781 related health problems. The implication is that such household stresses are having an effect on 782 pets. The owner concerns score relates to a range of potential problems that could arise from the 785 confinement, such as difficulty accessing veterinary care, obtaining food for the animal, the 786 animal gaining weight, and having difficulty adapting to normal life after the confinement. 787 Apart from pointing to specific problems that pet owners might face, it could be considered that 788 these are also an indication of an underlying state of worry or anxiety; people who have a 789 greater number of concerns, and therefore have a higher score for this variable in the analysis, 790 could be suffering from increased anxiety. The current COVID-19 outbreak has created levels 791 of uncertainty unparalleled in our recent history, being a reflection of a combination of fear of 792 the disease and anxiety about its short and long-term potential consequences. Fear of the 793 unknown has been described as one of the basic elements of anxiety and a fundamental 794 component of anxiety-related disorders (Carleton, 2016) . Our results point to a pattern of increased general behavioral changes, that probably result from 797 household stress and a reduced quality of life, which could lead to greater conflict with the 798 owner, a potential increase in anger and punishment from the owner, and therefore to an 799 increased likelihood of worsening behavior over time. The owner's psychological status and use 800 of punishment have already been found to be associated with problem behavior (Dodman et al., 801 2018). Overall, cats seem to be coping much better than dogs with the situation of confinement. One 804 reason may be that most cats in our sample were already indoor cats; the confinement had little 805 or no effect on their physical environment. Dogs, on the other hand, have experienced a 806 significant reduction in the duration of walks and, due to the confinement act, cannot interact 807 with people and dogs when they are outside. However, both cats and dogs are now sharing 808 their homes with people for a much greater amount of time, and the range of people they 809 interact with is much reduced. An alternative explanation is related to the salience of effects: 810 behavior changes in cats, particularly those related to stress, are often expressed as a reduction 811 in the frequency and/or intensity of certain behaviors, which may be less obvious to owners 812 (van der Leij, 2019). In the BLR model for cats, there was a positive association between the cat not coping well and 815 an increase in emotional closeness (C/DORS), the total number of problem behaviors getting 816 worse, and owner concerns score. There was a negative association with the number of resident 817 cats, meaning cats were more likely to be doing well if there were other cats in the household. Evidence from a study by Ramos et al (2013) suggests that many cats find certain forms of 820 human contact stressful, which would support the finding that increased emotional closeness 821 was associated with cats coping less well in the present study. The emotional closeness subscale 822 includes items like "How often do you kiss your pet", "I would like to have my pet near me all 823 the time", and "My pet is there whenever I need to be comforted". The majority of cats in the 824 present study live entirely indoors, making them unable to avoid this increased human contact. The association between cats doing better and the number of resident cats is puzzling but in the 827 same study by Ramos, the authors found no difference in glucocorticoid metabolites between 828 cats living in single, double or group housing. There is also the possibility that some of the signs 829 of not coping that we included in the composite measure are behaviors that are inhibited in 830 stressful situations. Taken as a whole, our findings suggest that general changes in behavior such as excitability and 833 being more attention seeking or demanding, could be seen as risk factors or even early 834 indicators of more serious future behavior problems. 835 836 Our study did not attempt to look at behavioral changes in detail, only as part of an overall 837 impression of the situation; a detailed study on the behavioral effects of the confinement is 838 definitely needed. However, in the BLR model there was a positive association between worsening behavior and 841 confinement level of the household, increase in emotional closeness (C/DORS), increase in 842 frequency of getting mad with the pet and score for general behavioral change. Again, there were associations with the frequency of the owner getting mad with the dog and 845 general behavioral changes. However, in this model we also see a link with level of 846 confinement; dogs were more likely to show worsening behavior if everyone in the household 847 was confined at home, with the effect being strongest in households in which nobody was 848 working from home. This may reflect the level of household tension due to all household 849 members being confined with little to do, but it may simply be the result of increased 850 opportunities for interaction, and therefore misbehavior. 851 852 Excessive or annoying vocalization was the only specific problem behavior that worsened in a 853 sufficiently large number of dogs that there was a large enough group to analyze statistically. In 854 the BLR model, there was a positive association between worsening problems of vocalization 855 and number of 18-64 year-old adults at home, increase in emotional closeness (C/DORS), 856 frequency of getting mad with the pet and general behavioral changes score. There was a 857 negative association with number of walks per day during the confinement; dogs that were 858 walked more often during the confinement were less likely to have worsening problems of 859 vocalization. It appears that a lack of frequency of exercise was a significant factor, indicating 860 that taking dogs for more walks, even if they are shorter than prior to the confinement, could 861 be a useful preventative intervention for excessive vocalization. In both the models for increased score for number of worsening behavior problems in dogs and 864 worsening vocalization in dogs, but not the model of poor coping, emotional closeness was a 865 factor. This suggests that an intensification of this aspect of the human-animal bond may place 866 additional stress on dogs that already have behavior problems. An additional factor that could 867 be important in this context is that behavior problems may be secondary to, or influenced by, 868 disease or suboptimal health (Fatjó and Bowen, 2020), particularly given that pets may be 869 lacking medication or veterinary care. This brings us to the issue of the effect of the confinement on the quality of life of owners and 872 how they use their pets are a source of support. As a crude measure of support, we asked respondents to answer the direct question "How much 875 has your pet helped you during the confinement compared with before?", on a 7-point Likert 876 scale from much less than before to much more than before. Seventy-four-point three percent 877 of respondents indicated that they had some level of increased support from their pet. With respect to the human-animal bond, there were significant increases in C/DORS subscales 880 scores for emotional closeness and interaction with the pet, but a decrease in perceived costs. "Perceived costs" measures the extent to which the presence of the pet interferes with the 882 owner's freedom of choice to perform other activities. During the confinement the person's 883 freedom of choice was already restricted, and we would expect the pet to have less effect. These 884 changes in C/DORS offer an insight into the dynamic nature of the human-animal bond, and 885 how it can be affected by changes of circumstance. 886 activates different coping strategies, including increased seeking of emotional support The psychological impact of quarantine and how to reduce it: rapid review of the 1040 evidence Into the unknown: A review and synthesis of contemporary models 1043 involving uncertainty The Italian perception of the ideal 1046 companion dog Associations between owner personality and 1049 psychological status and the prevalence of canine behavior problems The demographics of social media users-2012 Development of the Monash Dog Owner 1057 Relationship Scale (MDORS) Conflict and affiliative behavior 1060 frequency between cats in multi-cat households: a survey-based study Social Exchange Theory Making the Case for Multi-Axis Assessment of Behavioral Problems European Pet Food Industry Federation) Facts and Figures Social Relationships and Mortality Risk: A 1076 Meta-analytic Review Development of 1080 the cat-owner relationship scale (CORS) Looking at', 'looking up' or 'keeping up with' people? Motives and uses of 1083 Proceedings-Online Social Networks High income improves evaluation of life but not emotional 1086 well-being Animals as Social Supports: Insights for Understanding Animal-1089 Assisted Therapy Perceived social support helps, but does not buffer the 1093 negative impact of anxiety disorders on quality of life and perceived stress C-tactile afferent stimulating 1097 touch carries a positive affective value Are cats (Felis catus) from multi-cat households more stressed? Evidence from 1102 assessment of fecal glucocorticoid metabolite analysis Gender differences in veterinarian-1106 client-patient communication in companion animal practice Does gender influence online survey participation?: A record-linkage analysis 1110 of university faculty online survey response behavior The effect of a 1114 hiding box on stress levels and body The pet factor -Companion animals as a conduit for getting to know people, 1119 friendship formation and social support For this study, we needed to have a measure of change of QoL due to the confinement. 888 Although there are single question measures of QoL, such as Cantril's Self-Anchoring Scale, 889these measure the individual's current situation rather than QoL relative to a previous period. 890Also, measures like Cantril's Scale have been found to be more influenced by a person's income 891 rather than their emotional wellbeing (Kahneman and Deaton, 2010) . Had we used such a 892 measure in the current situation, this could have given a very limited perspective of the effect 893 of the confinement. However, people's perception of quality of life is likely to have been quite 894 distorted by the confinement situation, particularly after a period of several weeks. Even 895 measures like Cantril's scale may not represent the same thing now as they did before the 896 confinement. Given these problems, and the broad range of effects of the confinement, we 897 decided to use our own single question about quality of life change for the individual 898 respondent, supported by four additional measures of the negative effect of the confinement on 899 the household (economic, emotional, health and lifestyle impact). We believe that this provides 900 a good indication of the meaning of quality of life for people within the context of such a major 901 change of circumstances. When we looked at the correlation between the main question on QoL, and the four additional 904questions, the strongest associations were with the negative impact on lifestyle and emotional 905 impact, followed by negative health impact, and finally negative economic impact. It is perhaps 906surprising that economic impact was so poorly correlated with self-reported QoL, but this 907 supports our decision to try to characterize QoL; within this study, and at this moment in time, 908QoL is largely a measure of the effect the confinement has had on an individual's lifestyle and 909 emotional wellbeing. It is possible that as the confinement continues, the economic impact will 910increase and the perceived character of QoL will change. This requires further study. Most respondents (61.8%) indicated that the confinement had negatively affected their QoL. 913However, 11.4% of people reported an improvement in their QoL, which is quite surprising in 914 the current situation. We did not explore the specific reasons why some people might have an 915improved QoL, but our measure of QoL is largely influenced by lifestyle and emotional factors; 916 so, perhaps these people lived in locations where the risk of disease was low, or they had fewer 917 family members and dependents to be concerned about, or perhaps they were able to do more 918 of the things they usually enjoyed because they had more time available to them. In the BLR model that compared factors between people who reported a negative change in 921QoL and those reporting no change or an improvement in QoL, there was a positive association 922 between owner QoL group and the amount of outdoor space available at home, and with the 923 pet's QoL. Negative lifestyle, emotional and health impacts on the household were all 924 negatively associated with owner quality of life. Of these, the strongest association was with 925 impact on lifestyle. However, as mentioned previously, this may reflect perception of what 926 makes up quality of life in the current circumstances. The degree to which the person gained 927 support from the pet during the confinement was negatively associated with quality of life, 928which we interpret as meaning that the more severely affected the person's quality of life, the 929 more they gained support from the pet. In studies of the effect of social support on the negative 930 effect of anxiety disorders on quality of life and perceived stress, a similar inverse association 931 was found between support and wellbeing. The implication was that, as in our study, distress 932 (Panayiotou and Karekla, 2013). Dog owners were 53.2% more likely to be in the same/better QoL change group. This could be 936 explained by the fact that in Spain, one of the only permitted reasons for someone to leave the 937 home was to walk a dog. Anecdotally, this has led to cases of people borrowing dogs from 938 neighbors and family members, so that they had an excuse to go outside. Dog ownership, as 939 opposed to cat ownership, would seem to have a specific advantage in the type of confinement 940implemented in Spain, that could explain the association between species of pet and owner 941QoL. Many specific variables that might be expected to be associated with the owner's quality of life 944were not; these included age group, sex, the numbers of different ages of people at home, the 945 level of confinement, and the duration of confinement. In the BLR model that examined factors relating to the support the person got from the pet, 948there was a positive association between getting more support from the pet and the respondent 949being female, increased emotional closeness (C/DORS) and interaction (C/DORS) with the pet, 950improvement in the pet's quality of life and improvement in the relationship with the pet. 951There was a negative association between getting more support from the pet and improved 952 owner quality of life and increased perceived costs. Of these, there was a very strong association 953 with increased emotional closeness (C/DORS). This subscale of C/DORS includes items such as 954 "My pet gives me a reason to get up in the morning", "My pet helps me get through tough 955 times", "My pet is there whenever I need to be comforted", "How often do you tell your pet 956 things you do not tell anyone else?", and "How often do you kiss your pet". The interaction 957subscale includes "how often do you talk to your pet", "how often do you cuddle/hug your pet", 958and "How often do you pet your pet". Taken together, the emotional closeness and interaction 959subscales include many aspects of contact that form part of social support. Social support is a It is likely that many people experience loneliness during the confinement. Loneliness can be 968 divided into two main dimensions: social and emotional. Social loneliness is related to an 969 impoverished or negligible social network, whereas emotional loneliness is linked to the 970 absence of access to close relationships. Both dimensions are important, but it is the latter that is 971 more strongly correlated with adverse health and QoL outcomes, and which may be relevant to 972 the situation of confinement when social and physical contact is limited. Loneliness is not a 973 trivial matter; the quantity and quality of social relationships can be considered, by itself, to be 974 a risk factor for mortality (Holt-Lunstad et al., 2010). Our results suggest that during the current 975 outbreak, the relationship people have with their dogs and cats is helping to compensate for the 976 dramatic reduction in their social and physical interactions with people. This is supported by a 977 study on social support (Sarason et al., 1983) , which found that pets fulfill many social support 978functions. Conclusions Conclusions Conclusions Conclusions 981Our study found that pet owners gained substantial support from their pets, and that support 982 was increased when the owner's quality of life was more impaired. Support was also associated 983with increased emotional closeness and interaction with the pet. We also found interesting 984associations between behavioral problems, general behavioral changes and aspects of the 985 confinement, but some indication that the increased emotional needs of owners could 986 negatively affect pets that had existing behavioral problems. The findings of the study point to 987ways in which we may be able to minimize the effects of the confinement period. 988 989However, this study represents a general snapshot of the effects of a particular kind of official 990 confinement in one country. Survey participants were fully informed about the purpose and background of the study. Although the survey was anonymous, informed consent was required to participate, and participants were able to abandon the survey at any point. Participants were also provided with a link to official information about issues relating to the official confinement, which they could consult should any questions or concerns arise out of the study.• During COVID-19 confinement in Spain 61.8% of respondents said that their quality of life had deteriorated. • Poorer self-reported personal quality of life was associated with the lifestyle and emotional impacts of the confinement on the household. • People whose quality of life had been more severely affected reported that their pet provided proportionately more help for them during confinement. • The emotional bond between people and their pets strengthened during the confinement. • Thirty-seven percent of dogs showed signs of difficulty coping with the confinement, but cats were less affected. • Dogs with pre-existing behavior problems were more likely to have difficulty coping with the confinement.