Companion animal demography and population management in Pinhais, Brazil

Accepted Manuscript Title: Companion animal demography and population management in Pinhais, Brazil Authors: Oswaldo Santos Baquero, Solange Marconcin, Adriel Rocha, Rita de Cassia Maria Garcia PII: DOI: Reference:

S0167-5877(18)30077-1 https://doi.org/10.1016/j.prevetmed.2018.07.006 PREVET 4498

To appear in:

PREVET

Received date: Revised date: Accepted date:

29-1-2018 4-7-2018 4-7-2018

Please cite this article as: Baquero OS, Marconcin S, Rocha A, de Cassia Maria Garcia R, Companion animal demography and population management in Pinhais, Brazil, Preventive Veterinary Medicine (2018), https://doi.org/10.1016/j.prevetmed.2018.07.006 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Companion animal demography and population management in Pinhais, Brazil Oswaldo Santos Baqueroa*, Solange Marconcinb, Adriel Rochab, Rita de Cassia Maria Garciac a

Department of Preventive Veterinary Medicine and Animal Health, School of

IP T

Veterinary Medicine and Animal Science, University of São Paulo, Av. Prof. Orlando Marques de Paiva, 87, Cidade Universitária, São Paulo, SP, CEP: 05508-270, Brazil. B

SC R

Seção de Defesa e Proteção Animal, Secretaria de Meio Ambiente, Prefeitura

Municipal de Pinhais.

Department of Veterinary Medicine, Federal University of Paraná, Curitiba, Paraná,

U

c

M

A

N

80035-050, Brazil.

Abstract

ED

We used a two-stage cluster sampling design to estimate the population sizes of owned

PT

dogs and cats in Pinhais, Brazil. For dogs, we simulated the population dynamics using a compartmental model of coupled differential equations, incorporating uncertainties in

CC E

a global sensitivity analysis and identifying the most influential parameters through local sensitivity analysis. The calibration with the known human population improved precision in population size for dogs but not for cats. Population pyramids had a wide

A

base, and the apparent population turnover was lower than the net population gain. A minority of animal immigrants came from other states, and most came from the state capital. Projected dog and human growth rates between 2017 and 2027 were positive and similar, while the projected number of sterilized dogs decreased over the same

period. The main reason provided for not sterilizing animals was the cost of the procedure, even though there were free alternatives. The demographic characterization made in the present study will serve for future comparisons and as a reference in epidemiological contexts. The simulations indicated what to expect in specific scenarios

IP T

and stressed the need to increase current sterilization rates. Keywords: dog, cat, demography, population dynamics, population management,

SC R

mathematical modeling, abandonment, sterilization, adoption.

U

Introduction

N

Population size is the most basic demographic parameter used to monitor population

A

management programs and estimate measures of disease frequency. Unfortunately, dog

M

and cat population sizes are unknown in most cities. Some previous efforts to estimate population sizes of dogs and cats have validity problems and do not report the error of

ED

estimates (Downes et al., 2013).

PT

Despite their use in monitoring population management programs, population sizes are not sufficient for the study of population processes. For example, a population

CC E

management program can be effective in reducing the proportion of fertile animals without a concomitant reduction in the total population size (Baquero et al., 2016).

A

Furthermore, transmission rates of a given disease can be reduced if reduction of freeroaming animals diminish contact rates or if the main sources of immigration are identified for surveillance purposes. In any of these examples, population size must be used together with other demographic parameters (number of sterilized, free-roaming and immigrants, respectively) to calculate relevant quantities.

Survey designs are expensive and must be used to collect as much relevant information as possible, given resource constrains. Therefore, the objective of this study was to collect dog and cat demographic data using a validated sampling design, and to use these data to create population management indicators for future assessments, simulate population dynamics, and suggest which population management interventions should

SC R

IP T

be prioritized, based on their simulated influence.

U

Methods

N

Data collection

A

The survey was conducted in Pinhais, Paraná State, Brazil. The estimated human

M

population of Pinhais in 2017 was 129,445, making it the 224th most populated city

ED

among a total of 5570 Brazilian cities (Instituto Brasileiro de Geografia e Estatística, 2017) The city ranks 370th in mean wage, 3382th in literacy rate, and 454th in per capita

PT

GDP (Instituto Brasileiro de Geografia e Estatística, 2017). A group of 25 interviewers was trained to administer a questionnaire with variables to identify visited households

CC E

and characterize owned dog and cat demography. Appendix 1 has the Portuguese and English-translated questionnaire with instructions on administration. In the present

A

study we only used the data from questionnaire sections 1 to 4. All interviews were completed between April 2016 and June 2017. Survey design

The study population was based on the urban households of Pinhais identified in the last census, from 2010, in which 36,149 households grouped in 136 census tracts and 117,008 persons were counted (Instituto Brasileiro de Geografia e Estatística, 2010). We used a validated two-stage cluster design (Baquero et al., 2018a) with census tracts as primary sampling units and households as secondary sampling units. The census tracts

IP T

were selected with probability proportional to size and with replacement; households

were selected through systematic random sampling. We selected a total of 41 census

SC R

tracts and 30 households per census tract to ensure a sample size greater than the

minimum sample size required to estimate the owned dog population size with a precision of 8% – 11% (Baquero et. al., 2018). We calibrated the estimates using a

U

linear regression calibration model (Lumley, 2010) with the estimated human

N

population total for 2017 as the auxiliary variable. This calibration adjusted the original

A

weights to make the Horvitz-Thompson estimator match the known human population

M

total of 2017. Each estimate was characterized by its point estimate, confidence interval,

ED

relative error and design effect. The last is the ratio of the variance of the estimate in the complex survey to the variance of the same estimate in a simple random sample (Kish,

PT

1965).

CC E

We calculated the Pearson correlation coefficient for the number of humans and dogs per household and for the number of humans and cats per household. To take into account the survey design, the Pearson correlation coefficient was the estimated

A

variance-covariance matrix, scaled to the correlation. For other demographic variables we calculated their frequencies and did not take into account the sampling design because they referred to animal characteristics, not household parameters. Mathematical modeling of population dynamics

We modeled the dog population dynamics utilizing the compartmental model and sensitivity analysis (global and local) described by (Baquero et al., 2016). However, we included the adoption rate as a fraction relative to the owned dog population size (instead of the unowned dog population size) to estimate the rate from data (instead of from expert opinions). This changed the system of coupled differential equations;

IP T

Appendix 2 shows the updated equations, highlighting the changes. Figure 1 is a graphical representation of model compartments and we refer readers to the manuscript

SC R

of the original model, which describes the model assumptions.

U

We solved the differential equations numerically using the Runge–Kutta fourth-order

N

method (Soetaert et al., 2012) for a period ranging from 2017 to 2027. We used data to

A

estimate all initial conditions and parameters for owned dogs, except the carrying

M

capacity and mean owned females per harem. We assumed the owned dog carrying

ED

capacity as five times the current population size and the mean owned females per harem as one. For unowned dogs we assumed that population size was equivalent to 5%

PT

the owned population size, 10% of females and 5% of males were sterilized, the mortality rate was 20%, the carrying capacity was two times the current population size

CC E

and the mean owned females per harem was 0.5. With the owned population as reference, we assumed that birth and death rates in the unowned population were 50%

A

and 20% higher, respectively, and sterilization rates were 70% lower. Current evidence suggests that unowned dogs are not sustained independently of direct human support and when present represent a small fraction of the owned population (M.K. Morters et al., 2014; Michelle K. Morters et al., 2014). Therefore, our modeled

unowned population was small relative to the owned population size. It was completely hypothetical and used only to allow for adoption and abandonment under the conceptual framework underlying the mathematical model (Baquero et al., 2016). We were not interested in inferences about the unowned population and the potential biases in its parameter values were of no concern because they did not have relevant influences on

IP T

the owned population due to the small unowned population size. We did not model the cat population because there were not enough observations to distribute among the

SC R

model compartments.

The initial conditions and parameters of the present study are described in Appendix 3. For global sensitivity analysis, we perturbed all parameters in a magnitude equivalent to

U

20% of the parameter nominal value (10% above and 10% below). We used the Latin

N

hypercube algorithm (Soetaert and Petzoldt, 2010) to sample the parameters of the

M

A

respective intervals in each one of the 100 population dynamics simulations. We compared the owned dog population growth rate during the period from 2017 to

ED

2027, with the human population growth rate for the same period. Since we were unable to find population projections for Pinhais in 2027, we first calculated the Pinhais/Paraná

PT

human population ratio in 2017 and then multiplied the ratio by the projected human

CC E

population of Paraná in 2027. The human populations of Pinhais and Paraná in 2017 were 129,445 and 11,320,892, respectively; the projected human population of Paraná in 2027 is 11,929,009 (Instituto Brasileiro de Geografia e Estatística, 2017).

A

Software

We used the R packages capm 0.12.8 (Baquero et al., 2018b), tidyverse 1.1.1 (Wickham, 2017), sf 0.5-5 (Pebesma, 2017), ggsn 0.4.6 (Baquero, 2017) and gridExtra 2.2.1 (Auguie, 2016). Appendix 4 has the data and code to reproduce the results.

IP T

Results Of the 1,230 visited households, 78.6% participated in the interview, 5.5% were not

SC R

home on two visits, and 15.9% refused to participate in the interview. Demography

U

The uncalibrated and calibrated estimates for the total number of individuals of each

N

species are presented in Table 1. The uncalibrated estimate of the total human

A

population, contained by the confidence interval in its error relative to the true value,

M

was 1.8%.

ED

The Pearson correlation coefficient between dogs and humans per household was 0.12 (CI (95%): 0.06 – 0.18, p < 0.01), whereas the correlation between cats and humans per

CC E

PT

household was 0.02 ( CI (95%): -0.04 – 0.08, p = 0.52).

The human/dog ratio was 2.57 whereas the human/cat ratio was 16.7. The mean number

A

of dogs, cats and humans per household was 1.3, 0.2, and 3.3, respectively. Most households did not have animals but there were some with up to 12 dogs and 7 cats (Figure 2). Table 2 shows the absolute and relative distributions of dogs and cats according to their sex, sterilization status and free-roaming status.

The mean age was 5.09 (median = 4, interquartile range = 1 – 8) years for dogs and 2.77 (median = 2, interquartile range = <1 – 4) years for cats, and most animals were less than one year (Figure 3).

IP T

Litter sizes ranged from 1 to 16 (median = 5) for the 6% of female dogs that gave birth during the year prior to the interview. For cats that gave birth during the previous year

SC R

(also 6.2%), litter sizes ranged from 1 to 5 (median = 2.5). The total births represented 13% and 8% of the dog and cat population sizes, respectively.

U

For both species, the most frequent type of acquisition was adoption (Table 3). Almost

N

all dogs (99%) were acquired in Paraná state, and those acquired in other states came

A

from Santa Catarina, São Paulo, Federal District, Mato Grosso or Rio Grande do Sul.

M

There was one dog acquired in Portugal. The main city of acquisition was Pinhais (81.7%), followed by Curitiba (9.3%) (Figure 4). Cats were also mainly acquired in

ED

Paraná Sate (99%), and those acquired in other states came either from Santa Catarina or São Paulo. The main city of acquisition for cats was Pinhais (85.7%), followed by

CC E

PT

Curitiba (8.9%) (Figure 4). A total of 36 cities contributed with immigrants (Figure 5).

Cities which contributed less than 10 animals were grouped in the category “Other x

A

cities”.

During the year before the interview, 4% of dogs and 6% of cats were acquired, and 14% dogs and 13% cats left the household, most commonly due to death (Table 4).

Taken together births and acquisitions accounted for 17% and 14% of dogs and cats respectively, that were acquired by households during the year prior to the interview. Therefore, there was a net gain of animals, and the apparent annual population turnover was equal to the percentage of animals that left the households. In households in which at least one dog had left the household during the year before the acquisition, 15% of

IP T

dogs were acquired. For cats, this percentage was 14%.

The projected human population in Pinhais in 2027 is 136,398, which corresponds to an

SC R

increase of 5%. According to the mathematical model, the total owned dog population

would increase by 4% (quantiles 5% – 95%: -2% – 10%) in 2027 (Figure 6), and the

U

most influential parameter during that period would be the birth rate. Other parameters

N

that had an influence of at least 5% of the number of births were: male death rate

A

(51.5%), female death rate (46%), abandonment rate (17.1%), adoption rate (13.6%),

M

carrying capacity (13.3%) and immigration rate (8.4%) (Figure 7). The sterilized owned dog population was predicted to decrease by 6% (quantiles 5% – 95%: -13% – 2%) in

ED

2027 (Figure 6), and the most influential parameter in this projection was the female death rate. Excepting the mean females per harem and the proportion of sterilized

PT

immigrants, all parameters for the owned population had an influence of at least 5% of

CC E

the influence of the female death rate: male death rate (86.2%), female sterilization rate (65.7%), male sterilization rate (65.7%), birth rate (45%), abandonment rate (29.2%), adoption rate (23.2%), carrying capacity (11.4%), and immigration rate (7.5%) (Figure

A

7). If female and male sterilization rates were increased by 30%, the sterilized owned dog population would increase by 8% (quantiles 5% – 95%: 2% – 16%) in 2027. sd: standard deviation; Qt: quantile.

L1: norm of the columns of the sensitivity matrix. q: immigrants; f: intact females; m: intact males; fs: sterilized females; ms: sterilized males; 1: owned; 2: unowned; z: proportion of sterilized immigrants; w: birth function; c: mortality function; a:

IP T

abandonment rate; α’: adoption rate; s: sterilization rate.

The most frequent reason given by participants for not sterilizing their owned animals

SC R

was the high cost of the procedure (Table 5). Of the total 36 given reasons, only 5 had a

U

percentage higher than 4%.

N

Discussion

A

We estimated the dog and cat population sizes, quantified and compared demographic

M

parameters, and simulated dog population dynamics.

ED

Baquero et al. (2018) found that 30 x 30 (census tracts x households), 50 x 20 and 65 x 15 are two-stage sample compositions that estimate the owned dog population size with

PT

an approximated error of 4 – 7% if the census tracts are homogeneous (low variance of

CC E

the number of dogs per census tract). When census tracts are highly heterogeneous, errors increase and the more precise sample compositions are those with a higher number of census tracts; 65 x 15 is the best with an approximated error of 11 – 13%.

A

Based on prior experiences, we assumed that the heterogeneity of the census tracts was low and that was a reason to use a sample composition similar to but larger than 30 x 30. Although we could have used the 65 x 15 to be more conservative, the 41 x 30 was more convenient; 30 households per census tract might support better inferences about

census tracts in future studies. Of course, 65 x 30 would have been the best, but time constraints impeded us to increasing the sample size even more. As shown by the uncalibrated estimate of the owned dog population size, our assumption for Pinhais turned out to be true.

IP T

The calibrated estimate decreased the error in population size for dogs due the correlation between dogs and humans per household. The uncalibrated estimate of cat population size had a considerable error that will only serve to detect strong relative

SC R

changes in population size; calibration did not bring improvements due to the lack of

correlation between cats and humans per household. The association between the

U

number of humans, dogs and cats per households is typically studied in the framework

N

of logistic regression, to identify factors associated with the presence of these animals;

A

it has been found that age and sex structure of human residents are factors associated

M

with the ownership of both species (Downes et al., 2009; Westgarth et al., 2007), and that household characteristics (location, type, social class) are associated with dog

ED

ownership (Downes et al., 2009). The identification of factors associated with dog and cat ownership was not an objective of the present study but we calculated correlations to

PT

show why calibration only improved the precision of the estimated dog population size.

CC E

We do not know of any biological reasons to expect a male/female ratio to be different from 1, and our results conform to this expectation. Nevertheless, studies from Mali and South Africa suggest that human preference for one sex might influence that ratio

A

(Mauti et al., 2017; Michelle K. Morters et al., 2014). Free-roaming was reportedly less common than in a previous Mexican questionnaire-based report (Kisiel et al., 2016) and it might have been influenced by other factors such as sterilization status, sex, age

among others. We are exploring the effect of these factors and the details will be addressed in a separate manuscript.

The population pyramids had years aggregated in pairs, after the first year. They were

IP T

aggregated to smooth irregularities we attribute to an insufficient sample size for the total number of year–sex categories. The wide base showed that reproductive control

SC R

must be a preferential target of population management; this was further supported by the local sensitivity of dog population size to the number of births.

U

Regarding acquisition, we restricted the meaning of adoption to only the acquisition of

N

animals from a shelter, or directly from the streets (Appendix 2). The predominance of

A

adoption meant an inferior contribution by pet markets, a desired situation for

M

companion animal population management. This situation could counterbalance abandonment because most adopted animals were abandoned or would have been

ED

abandoned had they not been adopted. Furthermore, some animals acquired in the pet marked are fertile and the acquisition might be driven more by market stimulus than by

PT

informed decisions. Thus, fewer acquisitions from the pet market entail less fertile

CC E

animals diluting the effect of sterilization programs, and less animals at risk of abandonment due to false expectancies (Baquero et al., 2016). The predominance of adoption as a source of owned dogs was also found by Mauti et al. (2017).

A

Migration patterns of dogs and cats are poorly understood; to our knowledge the present study is the first to estimate immigration rates and identify the cities of emigration. Curitiba, the capital of the state, was the source of approximately 10% of the dog and cat population and it is possible that other cities receive similar frequencies of

immigrants from capitals. If this is the case and migration does not totally depend on disease-free status of infectious disease and demographic characteristics, diseases in capitals could be dispersed to other cities, adoption programs in capitals would be benefited, and sterilization programs in capitals would contribute to reproductive control in other cities. Although immigration from other states were less frequent,

IP T

Pinhais received animals from 36 different cities (including one from abroad), which

could introduce infectious diseases. We hope that future studies explore migration

SC R

patterns in other cities including capitals, and that the data concerning migration be used in surveillance systems.

U

The predominance of death as the fate for dogs was also observed in Mauti et al. (2017).

N

Adoption was the second most common fate but it must be noted that an animal adopted

A

out by its owner and not by a shelter would be counted as being given as a gift in the

M

classification of types of acquisition utilized in the present study. We made this distinction because it allowed us to separately estimate the adoptions that removed

ED

animals from streets and those that increased the turnover rates in shelters. Disappearance was the least common fate, and we did not disregard the possibility that

PT

some interviewees had abandoned their animals but instead reported that the animals

CC E

had disappeared. Mauti el al. (2017) also found that abandonment and disappearance were the least common fates.

We believe that most disappearances result from

irresponsibility and, under this assumption, the disappearance could be used as an

A

indicator of irresponsible ownership. Disappearance might be more frequent among free-roaming and unsterilized animals (characteristics of irresponsible ownership) if reproductive behaviors stimulate dispersal.

We considered annual population turnover as apparent because some animals that left the households disappeared or were adopted out. Since some of these animals might have been adopted during the same period, a fraction of the apparent turnover would have been within-population fluxes instead of a real population turnover (in such cases the apparent turnover is higher than the real). We also calculated the proportion of

IP T

acquisitions that replaced animals within households over the period of a year. This is a

type of household turnover that helps to understand the potential effect of compulsory

SC R

removal of animals for disease control purposes (replacement counterbalance removal).

We modified the mathematical model developed by Baquero et al. (2016) and designed

U

a questionnaire to estimate the owned dog parameters of that model. The only owned

N

dog parameters we could not estimate from data were the mean females per harem and

A

the carrying capacity. The absence of data to estimate the mean females per harem was

M

not a problem because it is not an influential parameter, as shown by Baquero et al. (2016) and the results of the present study. On the other hand, carrying capacity might

ED

be highly influential and changes in its magnitude my lead to different results. When compared to the estimates of Baquero et al. (2016), the assumed carrying capacity we

PT

used in the present study was relatively higher. This higher capacity reduced its

CC E

influence because the greater the distance between the population size and the carrying capacity, the lower the density-dependent effect. We assumed that carrying capacity is determined by human demand for dogs, and in the absence of data to calculate it, we

A

decided to use a higher estimate – when compared to Baquero et al. (2016) – to simulate a new scenario where demographic processes are less influenced by the human demand for dogs. Unfortunately, we had no time series data to test the model predictive

performance. However, dog and human population growth rates were similar and this is a plausible result for us. The lack of data available to test predictive performance led us to make considerable perturbations in all model parameters. This perturbation replaced unique deterministic

IP T

results with trends that reflected uncertainties about the real dynamics. The trend for population size showed that although a positive growth rate is the most probable result, equilibrium cannot be ruled out. The decreasing trend of sterilized animals indicated

SC R

that it is highly probable that the current sterilization rate is insufficient to increase or even maintain at equilibrium the proportion of sterilized dogs.

U

The insufficient sterilization rates, outstanding influence of birth rates and reasons given

N

for not sterilizing animals should be considered in future public and private sterilization

A

campaigns. The most frequent reason provided for not sterilizing owned animals was

M

the cost of the procedure, even though there are campaigns do not charge for the

ED

service. The public campaign begun in 2008 and is open to all Pinhais’ residents, provided they register their animals at the Municipal Secretary of the Environment of

PT

Pinhais and participate in responsible ownership talks before the procedure. Thus, the human population could be better informed about the benefits of sterilization and the

CC E

existence of free of charge campaigns to increase sterilization rates. We did not make a comprehensive simulation study to find the most efficient combination of female and male sterilization rates to reverse the declining trend in the number of sterilized dogs but

A

we found that an increase of at least 30% would be necessary to cause that reversal, if we take uncertainties into account.

The influence of sterilization did not vary with sex. It should be remembered that influences were given by local sensitivity analysis, so they referred to the influence of parameters at their point estimates, which in the case of sterilization were low and implied a large proportion of fertile females and males that could continue reproducing themselves. Thus, sterilization campaigns in Pinhais could pursue an increase of its

IP T

coverage without worrying about the sex of the sterilized animals and once the coverage increases, reassess the influence of sterilization to see if a more selective campaign in

SC R

terms of the targeted sex is warranted. However, note that this does not mean that both sexes must be sterilized at equal rate; this does not rule out the possibility of increasing

U

sterilization coverage with a campaign biased to females (or males) at initial stages.

N

Death rates were highly influential parameters and might have contributed to the

A

formation of a wide-base population pyramid. The immigration rate had modest

M

influences as did the pet market, since we considered the animals acquired in other

ED

cities or purchased in Pinhais as immigrants. Classifying the purchased animals as immigrants implied that they came from a separate population (probably with different

PT

vital rates) used for economic purposes. None of the visited households had sold any animals during the last year, so we ruled out the inclusion of animals from the “market

CC E

population” in the study population.

A

Abandonment and disappearance are different in terms of intention but both result in the flux of animals from the owned to the unowned population. In the mathematical model this flux was named abandonment but in the absence of reliable abandonment data we

used the number of disappeared dogs during the year prior to the interview as an estimate of the abandonment rate for the model.

We would have liked a lower nonresponse rate. Nonetheless, it should be noted that

IP T

nonresponse rate is not necessarily related to nonresponse bias (Groves, 2006), which occurs only when the causes of participation are associated with survey variables (for

SC R

example, if households which remain without people most of the day or which refuse interviews more frequently, systematically have less animals). This is an issue that can

U

be explored in future studies.

N

The comparison between the known value of the calibrator (known human population)

A

and its uncalibrated estimate gives an idea of the quality of the estimates. The

M

demographic characterization will serve for future comparisons and as a reference population in epidemiological contexts. The simulations in the present study

ED

demonstrated what to expect in specific scenarios and stressed the need to increase

Funding

PT

current sterilization rates.

CC E

The current study was funded by the World Animal Protection, which had no role in study design, data collection and analysis, decision to publish, or preparation of the

A

manuscript.

Acknowledgments We want to give special thanks to Rosangela Ribeiro Gebara of the World Animal Protection, employees of the Municipal Secretary of the Environment of Pinhais, and

professors, residents and students of Federal University of Paraná, Pontifical Catholic

A

CC E

PT

ED

M

A

N

U

SC R

IP T

University of Paraná and Tuitui University of Paraná, who helped us in the field work.

References Auguie, B., 2016. gridExtra: Miscellaneous Functions for “Grid” Graphics. Baquero, O.S., 2017. ggsn: North Symbols and Scale Bars for Maps Created with ggplot2 or ggmap.

IP T

Baquero, O.S., Akamine, L.L.A., Amaku, M., Ferreira, F., 2016. Defining priorities for dog population management through mathematical modeling. Prev. Vet. Med. 123, 121–127. https://doi.org/10.1016/j.prevetmed.2015.11.009

SC R

Baquero, O.S., Amaku, M., Dias, R.A.R.A., Grisi Filho, J.H.H.J.H.H., Ferreira Neto, J.S., Ferreira, F., 2018a. Validity of a two-stage cluster sampling design to estimate the total number of owned dogs. Prev. Vet. Med. 151, 40–45. Baquero, O.S., Amaku, M., Ferreira, F., 2018b. capm: an R package for Companion Animal Population Management.

N

U

Downes, M., Canty, M.J., More, S.J., 2009. Demography of the pet dog and cat population on the island of Ireland and human factors influencing pet ownership. Prev. Vet. Med. 92, 140–9. https://doi.org/10.1016/j.prevetmed.2009.07.005

M

A

Downes, M.J., Dean, R.S., Stavisky, J.H., Adams, V.J., Grindlay, D.J.C., Brennan, M.L., 2013. Methods used to estimate the size of the owned cat and dog population: a systematic review. BMC Vet. Res. 9, 121. https://doi.org/10.1186/1746-6148-9-121

ED

Groves, R.M., 2006. Nonresponse Rates and Nonresponse Bias in Household Surveys. Public Opin. Q. 70, 646–675. https://doi.org/10.1093/poq/nfl033 Instituto Brasileiro de Geografia e Estatística, 2017. Instituto Brasileiro de Geografia e Estatística [WWW Document]. URL www.ibge.com.br

PT

Instituto Brasileiro de Geografia e Estatística, 2010. Censo Demográfico 2010 Agregado por setores censitário do universo.

CC E

Kish, L., 1965. Survey sampling. John Wiley and Sons, Inc.

A

Kisiel, L.M., Jones-Bitton, A., Sargeant, J.M., Coe, J.B., Flockhart, D.T.T., Reynoso Palomar, A., Canales Vargas, E.J., Greer, A.L., 2016. Owned dog ecology and demography in Villa de Tezontepec, Hidalgo, Mexico. Prev. Vet. Med. 135, 37–46. https://doi.org/10.1016/J.PREVETMED.2016.10.021 Lumley, T., 2010. Survey sampling, a guide to analysis using R, First edit. ed. John Wiley and Sons, Inc, New Yersey. Mauti, S., Traoré, A., Sery, A., Bryssinckx, W., Hattendorf, J., Zinsstag, J., 2017. First study on domestic dog ecology, demographic structure and dynamics in Bamako,

Mali. Prev. Vet. Med. https://doi.org/10.1016/J.PREVETMED.2017.07.009

146,

44–51.

Morters, M.K., Bharadwaj, S., Whay, H.R., Cleaveland, S., Damriyasa, I.M., Wood, J.L.N., 2014. Participatory methods for the assessment of the ownership status of free-roaming dogs in Bali, Indonesia, for disease control and animal welfare. Prev. Vet. Med. 116, 203–208. https://doi.org/10.1016/J.PREVETMED.2014.04.012

IP T

Morters, M.K., McKinley, T.J., Restif, O., Conlan, A.J.K., Cleaveland, S., Hampson, K., Whay, H.R., Damriyasa, I.M., Wood, J.L.N., 2014. The demography of freeroaming dog populations and applications to disease and population control. J. Appl. Ecol. 51, 1096–1106. https://doi.org/10.1111/1365-2664.12279

SC R

Pebesma, E., 2017. sf: Simple Features for R.

Soetaert, K., Cash, J., Mazzia, F., 2012. Solving differential equations in R. Springer. Soetaert, K., Petzoldt, T., 2010. Inverse modelling, sensitivity and monte carlo analysis in R using package FME. J. Stat. Softw. 33, 1–28.

A

N

U

Westgarth, C., Pinchbeck, G.L., Bradshaw, J.W.S., Dawson, S., Gaskell, R.M., Christley, R.M., 2007. Factors associated with dog ownership and contact with dogs in a UK community. BMC Vet. Res. 3, 5. https://doi.org/10.1186/1746-61483-5

A

CC E

PT

ED

M

Wickham, H., 2017. Tidyverse: Easily Install and Load “Tidyverse” Packages.

Figure 1. Compartmental model of dog population dynamics. Figure 2. Sample distribution of humans, dogs and cats per household. Figure 3. Dog (left) and cat (right) population pyramids. Pinhais, Brazil, 2017.

Paraná cities, and location of Paraná state (bottom). Brazil, 2017.

IP T

Figure 4. Proportion of immigrant dogs (top left) and cats (top right) contributed by

least with one dog (top) or cat (bottom), to Pinhais. 2017.

SC R

Figure 5. Chord diagram of migration flows from Brazilian cities which contributed at

Figure 6. Compartmental model of dog population dynamics in Pinhais, Brazil. Global

N

U

sensitivity analysis of the total (left) and sterilized (right) owned population.

A

Figure 7. Compartmental model of dog population dynamics in Pinhais, Brazil. Local

A

CC E

PT

ED

M

sensitivity analysis of the total (left) and sterilized (right) owned population.

A ED

PT

CC E

IP T

SC R

U

N

A

M

A ED

PT

CC E

IP T

SC R

U

N

A

M

A ED

PT

CC E

IP T

SC R

U

N

A

M

A ED

PT

CC E

IP T

SC R

U

N

A

M

A ED

PT

CC E

IP T

SC R

U

N

A

M

A ED

PT

CC E

IP T

SC R

U

N

A

M

A ED

PT

CC E

IP T

SC R

U

N

A

M

Table 1. Calibrated and uncalibrated (inside parentheses) estimates of total number of dogs and cats, and of percentage of households (PHH) with dogs and cats. Pinhais, Brazil, 2017. Estimate

CI 95%

Deff

Error (%)

Dogs

50,444

46,232 – 54,656

1.5

8.4

(46,874)

(41,921 – 51,827)

(2.3)

(10.6)

7,722

5,746 – 9,697

1.7

25.6

(7,192)

(5,302 – 9,081)

(1.8)

(26.3)

63.1 – 70.2

1.4

3.5

(66.5)

(62.9 – 70.1)

(1.5)

(3.6)

12.7

10 – 15.3

1.6

2.6

(12.6)

(10 – 15.2)

129,445

129,445 – 129,445

(119,121)

(109,976 – 128,266)

SC R

1.6

(2.6)

–

0

7.3

(7.7)

M

Humans

U

Cats (PHH)

N

Dogs (PHH) 66.7

A

Cats

IP T

Species

ED

Deff: design effect; Error for totals: coefficient of variation * 1.96 * 100; Error for percentages: difference between the upper limit (lower limit is equivalent up to

PT

rounding) and the point estimate, expressed as percent points. In calibrated estimates,

A

CC E

weights were adjusted to estimate without error the known human population size.

Table 2. Absolute and relative frequencies of dogs and cats according to their sex, and if they were sterilized and free-roaming. Dogs (%)

Cats (%) Females

Males

Females

Total

644 (51)

606 (49)

110 (56)

85 (44)

Sterilized

120 (19)

191 (32)

38(36)

31 (37)

Free-roaming

114 (21)

73 (14)

42 (58)

31 (32)

U

SC R

IP T

Males

N

The mean age was 5.09 (median = 4, interquartile range = 1 – 8) years for dogs and 2.77

A

(median = 2, interquartile range = <1 – 4) years for cats, and most animals were less

A

CC E

PT

ED

M

than one year (Figure 3).

Table 3. Sample frequencies of types of acquisition of dogs and cat in Pinhais, Brazil, 2017.

Type

Dogs

Cats %

Total

%

Adoption

536

51

119

72

Received as gift

401

38

42

26

Pet market

317

11

4

02

SC R

U N A M ED PT CC E A

IP T

Total

Table 4. Sample frequencies of fates of dogs and cats that left the household during the year prior to the interview.

Fate

Dogs

Cats %

Total

%

Death

124

70

15

58

Adopted out

34

20

6

23

Disappeared

18

10

5

19

SC R

U N A M ED PT CC E A

IP T

Total

Table 5. Calibrated estimates for reasons not to sterilize their own animals.

Deff

Error*

Surgery cost

45

39 – 51

1.2

6

Lack of time

16

12 – 20

1.2

4

Wants offspring

10

6 – 14

1.4

4

Do not know

8

5 – 12

1.3

It is dangerous for the animal 5

2–8

1.4

Others

10 – 20

15

IP T

Estimate % CI 95%

4 3

SC R

Reason

1.3

5

U

Deff: design effect; Error: difference between the upper limit (lower limit is equivalent

N

up to rounding) and the point estimate, expressed as percent points. In calibrated

A

estimates, weights were adjusted to estimate without error the known human

A

CC E

PT

ED

M

population size.