## Introduction

The virus SARS-CoV-2 emerged in December 2019, in Wuhan, China. This unknown respiratory disease developed into the pandemic, termed COVID-19, as declared by the World Health Organization on March 2020 (Bojdani et al., 2020). One of the main approaches worldwide for combating the disease is social isolation and distancing, at least until a protective vaccine is available (Koo et al., 2020; Lewnard and Lo, 2020; Bavel et al., 2020). Social isolation may prevent the spread of the disease, but it may also lead to other concerns. One of the greatest concerns regarding the influence of social isolation is its psychological effect on humans. Extended social isolation may lead to a significant decrease in quality of life and well-being, and high levels of stress, in both the infected and non-infected populations (Xiao et al., 2020; Bavel et al., 2020). Social isolation is an additional stressor to an already highly stressful world environment and people’s extensive fear of the novel COVID-19 pandemic threat (Bavel et al., 2020; LeDoux, 2012; Mobbs et al., 2015). In addition, social distancing included full lockdowns in many countries, as well as in Israel, with dramatic economic effects (Anser et al., 2020; Sangar et al., 2019). Adverse local and global economic impacts, in addition to drastic personal income reduction, may be detrimental to people’s psychological health and general well-being (Xiao et al., 2020).

Interestingly, the mental health benefits of owning a companion animal, such as a dog or a cat, have been shown by several scientific studies (Serpell, 1991; Beetz et al., 2012; Powell et al., 2019). The majority of studies indicate that interactions with animals may help with depression, anxiety, and stress, in particular under stress-prone conditions (Beetz et al., 2012). On the one hand, companion animals provide companionship, improve mood, and may ease loneliness; human–animal interactions may even improve peer-to-peer social relationships, as well as enhance feelings of respect, trust, and empathy between people (Powell et al., 2018; Beetz et al., 2012; Powell et al., 2019). On the other hand, it has also been shown that stress and poor well-being of owners negatively affect the stress and well-being of their companion animals (Buttner et al., 2015; Sumegi et al., 2014; Ryan et al., 2019). For example, there has been some indication that the stress of the owner could influence their dog’s cognitive ability (Sumegi et al., 2014). Moreover, changes in the attention of owners to their dogs may affect the behavior of the dogs (Kaminski et al., 2009; Payne et al., 2016). Therefore, we hypothesized that the COVID-19 pandemic might lead to dramatic changes in human–dog bidirectional relationships. On the one hand, owning a dog may assist the owner in coping with the stressful world situation, and therefore, more people may decide to adopt a dog during this pandemic. On the other hand, behavioral problems in dogs were reported to be one of the main reasons for the abandonment of dogs to shelters (Patronek et al., 1996; Salman et al., 2000); if changes in the lives of owners occurred during the COVID-19 pandemic, and indeed, if behavioral problems in their dogs developed as was shown under other circumstances (Sumegi et al., 2014), then this might increase the risk of dog relinquishment.

Another potential risk factor for dog abandonment and relinquishment during the COVID-19 pandemic was their suspected epidemiological role in the spread of SARS-CoV-2. There was a worldwide growing concern that companion animals, specifically dogs and cats, could transmit the disease to humans (Goumenou et al., 2020; Parry, 2020; Leroy et al., 2020). Although the anecdotal reports were inconclusive, it could lead to an increase in the number of dogs relinquished by their owners. Thus, overall, the inconclusive reports of companion animals being potential carriers of the COVID-19 virus, the economic crisis, and the general stress and panic during this pandemic, could potentially cause a dramatic increase in dog abandonment numbers. Since such a scenario might incur high costs and present considerable risk to public health, it should be explored. Relinquishment and abandonment of companion animals is a global problem. It is estimated that millions of pets are abandoned each year (Fatjo et al., 2015), even without a pandemic in the background. It results in increasing numbers of free-roaming animals, overcrowded animal shelters, impaired animal welfare, and it carries high costs to tax payers (Fatjo et al., 2015). Moreover, it is a severe public health issue due to the potential transmission of zoonotic diseases (such as rabies) and attacks on people (Carter, 1990; Burgos-Caceres, 2011). All of these threats also carry remarkable economic consequences, which affect national and local governments, humane organizations, as well as individuals (Carter, 1990).

In order to control the pandemic, gradual social restrictions were initiated during March 2020 in Israel, while in April, a total lockdown was implemented for a full month by the Israeli government, as marked on the timeline in Fig. 1. During this period, walking the dog and veterinary care were exceptions for the lockdown restrictions, as well as dog adoptions from animal welfare organizations and municipal shelters. Therefore, whereas it was not allowed to be outside of a 100 m radius from your home, dog adoption and dog walking were permitted throughout these periods.

The objectives of this study were to investigate: (1) how the COVID-19 pandemic affected adoption and abandonment of dogs at shelters, and the public’s general interest in adopting a dog; (2) the association between the quality of life of owners and their dogs during the pandemic; as well as (3) the effect of the pandemic on the development of new behavioral problems and on the relinquishment rate of dogs by their owners.

## Results

This study focused on a new aspect of the COVID-19 pandemic by investigating the human–dog relationship during this crisis. Dog adoptions, abandonment, as well as the association between the well-being of the owners and their perceptions of the quality of life of their dogs, were examined. Overall, in contrast to some of the initial concerns, all dog adoption measures significantly improved as the social restrictions became stricter. Furthermore, there was a clear association between an individual’s quality of life and their perceptions of their dog’s quality of life and behavior, as well as the probability of their relinquishing their pet.

### Changes in dog adoption and abandonment

Another option available for the public on Yad4 website was to fill in a request to serve as a foster family, as an alternative to adoption. Usually, the demand for foster families among the organizations is very high, but the number of available foster families is low. Therefore, typically, there are no available foster families since the organizations use them all. During the pandemic period, the number of foster families was higher than the demand. Accordingly, from the reports about the outbreak in China until the end of the lockdown in Israel, as well as during the gradual opening, the number of available foster families increased significantly. For example, as described in Fig. 3a, b, by the end of April 2019 there were no available foster families on Yad4 website, since they were all occupied and used by the organizations; contrarily, at the time of the outbreak in China, 226 foster families were available but did not receive a dog to foster, and by the end of April 2020, there were 844 available foster families.

### Local and global online searches for adoptable dogs

The daily number of visitors on the Yad4 website since the first COVID-19 patient diagnosed in Israel until the end of the full lockdown was significantly higher, as compared to the whole period before the pandemic (Fig. 3). The effect of year and month were controlled in the models (Supplementary Table S1). The linear regression model revealed that there was a significant increase in daily visits online when the outbreak emerged in Israel during March by 657.9 ± 80.8 (coefficient ± SE) visits, and by 2311 ± 82.1 daily visitors online during the total lockdown period (Fig. 3a–c; P < 0.05). For example, the absolute number of visits online in April 2020 was 221,959 visits, as compared to 72,703 in April 2019, and 91,920 visits in October 2019, which is typically the busiest season of the website. Interestingly, according to global non-scientific media reports, the demand for adoptable dogs worldwide was also high in other countries. Pictures of empty cages from many countries were published, but until now, to the best of the authors’ knowledge, no scientific data has yet been published documenting this phenomenon. Thus, the global trend was investigated by analyzing Google Trends data for searches all around the world, as well as specifically in the USA. In order to do so, the timeline was divided to three periods: (1) before the outbreak in China; (2) from the first media reports about the outbreak in China on December 27th until March 13th—when the World Health Organization (WHO) announced Europe to be the epicenter of the pandemic; (3) the main lockdown worldwide—from the announcement of the WHO until the gradual opening on May; and (4) during May. The effect of year and month were controlled in the models (Supplementary Table S2). Interestingly, the world trends, according to the Google Trends data, were found to be similar to that we report herein for Israel (Fig. 3). The trends of worldwide searches online for “adopt a dog” were significantly higher during the periods of the outbreak in China, as well as during the period many countries declared lockdowns, as compared to the year of 2019 (Fig. 3d–f).

Given the high demand for dogs to adopt during the pandemic, the second part of our study included questionnaires targeting people who had recently adopted a dog, as well as current general dog owners, to explore the motivation behind this increase in demand for adoptable dogs.

### The association between impaired quality of life of owners to their perception of the quality of life of their dogs

In order to study the association between the quality of life of owners and their companion dogs under the COVID-19 pandemic situation, a digital questionnaire for dog owners was active during the full lockdown and social isolation (April). Participants replied to questions regarding their own well-being, as well as the well-being of their companion dog. Questions such as the effect of the pandemic on their stress level and personal finances, their concern about their own health, and their perceptions regarding their dog’s well-being and behavior under the COVID-19 related lockdown, were included. The questionnaire also included questions regarding the characteristics of the owners and their dogs, as well as the care they provided to their dog during the pandemic. These variables were controlled in the statistical models (details in Supplementary Table S3). The outcome variables were set on a scale of 1–5 (for example, 1-low stress; 5-extremely stressed). Scores 4 and 5 were relabeled as “severe stress” for the analyses, and they were compared to scores 1–3: “none to moderate”. The questionnaire was answered by n = 3138 individuals. Overall, 25% of the participants were very concerned about their health (Fig. 5a), 25.6% stated they were extremely stressed (Fig. 5b), and 22.9% reported that their personal finances were severely affected (Fig. 5c). For further analysis, an impaired quality of life index was calculated as the mean of these scores (general stress; concern for their own health; and the damage to their personal financial situation; Fig. 5d). In addition, in the questionnaire, owners were asked to rank on a scale of 1–5 their assessment of the quality of life of their dogs during the COVID-19 lockdown, as well as their recognition of new behavioral problems, and whether they have considered relinquishing their dog.

As hypothesized, multivariate logistic regressions revealed that an increase in the impaired quality of life index of the owner was associated with lower quality of life of the dog, as was assessed by the owner (odds ratio: 0.887 times lower for every one-unit increase in owner’s quality of life index; Fig. 5e; P < 0.05). In addition, for a one-unit increase in impaired quality of life index of the owner, the odds ratio for recognition of new behavioral problems of the dog (as defined and recognized by the owner) was 1.397 times higher (Fig. 5e; P < 0.05). Moreover, for a one-unit increase in the impaired quality of life index of the owner, the odds ratio for relinquishment was 1.762 times higher (Fig. 5d; P < 0.05). Overall, the number of people that recognized behavioral problems in their dogs was low (11.6% of the dog owners), as well as the number of people that considered relinquishing their dog (1%). Still, according to these data, severely impaired quality of life of owners under the COVID-19 pandemic and lockdown was a significant risk factor associated with the quality of life of the dog, as well as for the recognition of the dog’s behavioral problems, and for dog relinquishment, as reported by dog owners. Characteristics of the dogs and owners, as well as ownership habits, were controlled in the statistical models, as fully detailed in the “Methods” section and Supplementary Table S3. Further questions that were included in the model regarding the type of behavioral changes in the dogs are presented in Supplementary Table S4.

## Methods

### Online digital questionnaire for dog owners during the COVID-19 related lockdown in Israel

An online digital questionnaire targeting dog owners in Israel was active during the COVID-19 related full lockdown in Israel. The questionnaire was designed by the researchers using Google Forms, and was distributed by a company who specializes in this purpose (Lead Marketing Ltd.; https://leadmarketingltd.com), in order to specifically and effectively reach dog owners in Israel. The online distribution of the questionnaire was based on targeting a pre-defined group of respondents, with a high level of accuracy, characterized by their interests and online behavior (e.g., users that shop online for dog food or who perform searches for information on dog care). Various digital platforms were used (i.e., Google Display Network, Facebook, Instagram, and others), and banner ads led users to the questionnaire, asking them to voluntarily and anonymously participate in the survey with their consent to be part of this research study.

### Statistical analyses

Statistical analyses were performed using commercial statistical software (IBM SPSS Statistics, version 24.0; STATA, version 15.0). Linear regression analysis was utilized to evaluate the effects of the spread of COVID-19 and lockdown stages on adoption and abandonment outcomes for pet dogs, using data from the Yad4 adoption website. The general structure of the estimated regressions was as detailed in the Eq. (1).

$$\begin{array}{*{20}{l}} Y_t =\beta _0 + \delta _{{\mathrm{ChinaOutbreak}}} + \delta _{{\mathrm{LocalOutbreak}}} + \delta _{{\mathrm{LocalLockown}}} + \delta _{{\mathrm{OpenUp}}} + \gamma _t\cr \\ + \,\beta _1 \cdot {\mathrm{trend}}_t + \beta _2 \cdot {\mathrm{regulation}}_t + \varepsilon _t \end{array}$$
(1)

where Yt was the outcome of interest in month t, the δ’s were dummy variable effects of the stages of outbreak and lockdown, all compared to the baseline period, before the outbreak of COVID-19 in China. $$\delta _{{\mathrm{ChinaOutbreak}}}$$ is a dummy variable for the months between the outbreak of COVID-19 in China and the first confirmed case in Israel; $$\delta _{{\mathrm{LocalOutbreak}}}$$ is a dummy variable for the period between the first local case and the start of the full lockdown; $$\delta _{{\mathrm{LocalLockdown}}}$$—between the start of the lockdown and the start of gradual opening in Israel. γt are the calendar month fixed effects controlling for seasonality in adoption activities, and $${\mathrm{trend}}_t$$ controls for a linear annual time trend. $${\mathrm{regulation}}_t$$ controls for a change in governmental initiatives with regard to encouragement of responsible ownership and adoptions between 2018 and 2019, in order to make sure this does not drive our results. εt is a standard error term. Several outcomes were considered variables: the number of adoption requests received through the website and the number of dogs marked as adopted, as measures of the level of interest in conducting an adoption process and the final outcome of successful adoptions; the number of dogs uploaded to the website, as a measure of recent abandonment cases; and the number of users on the website, as a measure of general interest in adoption.

The major part of the analysis uses Israeli data, as detailed above. However, two outcome variables were added from Google Trends to compare the results to worldwide trends. Google Trends were used to construct two additional outcome variables: the number of web-searches, in the US, and worldwide, of the phrase “adopt a dog” during the same period. The regression analysis for these outcomes was similar to the main model, but the stages of shutdown were defined as: the outbreak in China, the declaration of World Health Organization on Europe as the epicenter of the pandemic, and the gradual opening in May 2020. Because the lockdown policies are not centralized in the US and worldwide, we did not include a specific separate shutdown time-period.

The digital questionnaire of the dog owners was analyzed using logistic regression. The binary outcome variables that were considered were: the quality of life of the dog, as assessed by the owner; the development of new behavioral problems if recognized and defined by the owners; and whether the owner was considering abandoning the dog. The general model for estimation was as detailed in Eq. 2.

$$\begin{array}{*{20}{l}} {\mathrm{Prob}}\left( {Y_i = 1} \right) = \beta _0 + \beta _1 \cdot {\mathrm{LifeChange}}_i + \beta _2 \cdot {\mathrm{CrisisIndex}}_i + \gamma \cdot Z_i \cr \\ + \,\delta \cdot W_i + \alpha \cdot D_i + \nu _i + \varepsilon _i \end{array}$$
(2)

where Yt is the binary outcome of interest for respondent i. $${\mathrm{LifeChange}}_i$$ is a dummy variable depicting whether the respondent declared that his life changed following the COVID-19 outbreak. $${\mathrm{CrisisIndex}}_i$$ is an average of three responses addressing three aspects of negative effects of the outbreak: economy, health concerns, and stress (as reported by the responders). Zi are owner characteristics: gender, age and whether there are young children in the household. Wi are dog characteristics: age, whether the dog was adopted from a shelter, number of years with the owner. Di are characteristics of the care given to the dog: number of walks a day, the average duration of the walks, and a general measure of attention to the dog. vi are geographical area fixed-effects and εi is a standard error term. The logistic regressions, as they are based on the responses of the owners, which can be biased on their own, should be interpreted as descriptive analyses rather than being given causal interpretation. Descriptive statistics are given as mean ± SE, 95% confidence interval, or as frequency (n) with percentage (%), as applicable. A P < 0.05 was considered statistically significant. All reported P-values were based on a two-tailed hypothesis.

## Discussion

Humans and dogs are both social animals, and their bond can be traced back at least 15,000 years to the Bonn-Oberkassel dog that was found buried with two humans (Janssens et al., 2016). According to the 2019–2020 National Pet Owners Survey conducted by the American Pet Products Association (APPA), approximately 63.4 million households in the USA owned at least one dog, making them the most widely owned type of companion animal across the USA at this time. The advantages of raising a dog have been widely investigated. The human–dog bond has potential physical, psychological and mental benefits, and can improve the general well-being and happiness of owners (Lass-Hennemann et al., 2020; Tzivian et al., 2015; Barker and Barker, 1988; Wells, 2007). Despite all the known advantages, and the evidence that separation between a dog and its owner negatively impacts not only the dog but also the wellness of the owner (Lowe et al., 2015), millions of companion dogs are abandoned every year (Marder and Duxbury, 2008). Dog abandonment carries high costs and a significant risk for public health (Fatjo et al., 2015; Kumar, 2002; Carter, 1990). Prior to this study, it was unknown whether the COVID-19 pandemic was a risk factor for dog abandonment, as well as a risk for impaired well-being of the dogs as a reflection of the potentially impaired well-being of the owners. Therefore, the motivation to conduct this study was to explore the human–dog relationship during this pandemic, to benefit the welfare and well-being of both humans and animals, in accordance with the One Welfare approach. The One Welfare approach extents the One Health theme, suggesting that there is a strong connection between welfare and health of human and animals, including both physical and mental health, and that improving animal welfare often improves human welfare (and vice versa) (Pinillos et al., 2016; Mor et al., 2018; Panning et al., 2016; Lem, 2019; Jordan and Lem, 2014; Card et al., 2018). According to this approach, veterinarians, animal’s owners, animal welfare organizations, human psychiatrists, environmental scientists, and others, should collaborate and share expertize in order to care for the welfare of both animals and their owners. Accordingly, the rationale behind this study was the hypotheses that human perceptions and acts regarding dog ownership and adoption might be influenced by the COVID-19 pandemic and the related social isolation, as well as the stress and well-being of both species.

Our data indicate that not only is the concern of increased dog abandonment not justified, at least so far, the opposite has occurred. As social restrictions increased during the COVID-19 pandemic, the rates of dog adoptions improved significantly (Fig. 2); the demand for adoptable dogs and the requests to serve as foster families increased significantly, and accordingly, the length of stay of dogs at the shelter was significantly shorter. Previous reports following disasters, such as earthquakes or other situations that require immediate evacuation, were associated with a massive unintentional dog abandonment (Nagasawa et al., 2012). However, people may refuse to separate from their pet when needed due to disasters or extreme situations, as pet owners may find their pets closer or at the very least, as close as family (Chadwin, 2017; Barker and Barker, 1988). This may be the reason why, so far, the vast majority of people were reluctant to relinquish their dog during the COVID-19 pandemic. Still, further investigation is required, as the potential risk for dog relinquishment in the coming months cannot be completely excluded, due to the various social and economic impacts that this pandemic may yet bring. Furthermore, as our climate continues to change, more disasters, including additional pandemics, will likely occur, highlighting the need for more research into crisis-driven human behavior changes, including changes in the human–animal relationship.