Companion animal bereavement: Psychometric properties of the Spanish version of the Pet Bereavement Questionnaire

Depósito de Investigación de la Universidad de Sevilla https://idus.us.es/ This is an Accepted Manuscript version of the following article, published by Taylor & Francis in Death Studies: López-Cepero, J., Ferrer, M., Mori, M., & Español, A. (2024). Companion animal bereavement: Psychometric properties of the Spanish version of the Pet Bereavement Questionnaire. Death Studies. https://doi.org/10.1080/07481187.2024.2363477. It is deposited under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits noncommercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way. 1 Information Classification: General Companion animal bereavement: Psychometric proper�es of the Spanish version of the Pet Bereavement Ques�onnaire Running head: Valida�on of Spanish PBQ Javier López-Cepero1, Marta Ferrer2, Marina Mori1, and Alicia Español3 1 Universidad de Sevilla. Departamento de Personalidad, Evaluación y Tratamiento Psicológicos; 2 Universidad de Jaén. Departamento de Psicología; 3 Universidad de Sevilla, Departamento de Psicología Experimental Contact informa�on: Dr. Javier López-Cepero Borrego. Facultad de Psicología, C/Camilo José Cela S/N, despacho S502. 41018-Sevilla (España-Spain). [email protected] Acknowledgements: The authors want to show their gra�tude to Dr. Melissa G. Hunt for allowing us to use the Pet Bereavement Ques�onnaire in this study; and to Dr. Ines Testoni and Dr. Lucia Ronconi for providing unpublished data to complete comparisons to Italian sample. Abstract This work provides psychometric data on the validity and reliability of the Spanish adapta�on of the Pet Bereavement Questionnaire (PBQ), as well as informa�on on the intensity of bereavement in the Spanish popula�on. The study evaluated 333 Spanish par�cipants of legal 2 Information Classification: General age (M=31.5; SD=11.6), mostly women (76.3%). Confirmatory factor analysis (CFA) tested the adequacy of three different structures present in literature, finding beter fit indexes for a model that kept the original three-factor structure (grief, guilt, and anger) but rearranged 2 of the 16 items. Around 70% of par�cipants reported signs of intense bereavement on the grief scale, with higher means among women. The results confirm adequate psychometric quali�es of the PBQ, offering healthcare professionals a tool to evaluate bereavement intensity a�er the loss of a companion animal in Spanish samples. Keywords: Bereavement, Grief, Valida�on, Companion animal, Pet 3 Information Classification: General Companion animal bereavement: Psychometric proper�es of the Spanish version of the Pet Bereavement Ques�onnaire The number of companion animals in Spain has steadily increased in the recent years. These animals live at home with no other purpose than to provide companionship to the human being with whom they live. The Na�onal Associa�on of Pet Food Manufacturers reports more than 29 million registered animals, with dogs being the preferred animal, followed by fish, birds, and cats (Veterindustria & ANFAAC, 2021). According to the same report, 40% of Spanish households have at least one registered companion animal. Considering certain animals to be companion animals or pets is not due to any of their behavioral or phylogene�c characteris�cs but is rather a status granted by the human beings with which they live. Thus, while many animals are considered to be “just an animal”, it is increasingly common to find non-humans enjoying the status of family member (Bouma et al., 2022) in the so called more-than-human households (Irvine & Cilia, 2017). In fact, various studies point out that a relevant por�on of adults consider their companion animals to be like a child (Volsche, 2018, 2021). This sociological phenomenon is not widespread to all households with animals but warrants aten�on due to its possible implica�ons for family well-being, given that most of the species considered companion animals have a shorter life expectancy than humans. Thereby, the loss of animals is a phenomenon that many people will face -perhaps, in many occasions - along their lives, making it advisable to learn more about the experience of grieving for companions in this popula�on. To date, various studies have analyzed the experience of bereavement a�er the loss of companion animals. A shared conclusion is that this bereavement can be similar (i.e. intensity of grief, coping strategies, etc.) to that experienced by the loss of human beings (ClementeRicolfe & González-Navarro, 2020; Lavorgna & Huton, 2019). Although not all companion animal losses result in intense or complicated bereavement, they can some�mes meet the 4 Information Classification: General criteria to diagnose prolonged grief disorder, according to the Diagnos�c and Sta�s�cal Manual of Mental Disorders (DSM-5-TR) and the Interna�onal Classifica�on of Diseases (ICD-11, 11th revision) (American Psychiatric Associa�on, 2022; Lee, 2020; World Health Organiza�on, 2022). Currently, the assessment of bereavement for companion animals is conducted using various tools. Some studies (Adrian & S�t, 2017; Yüksel et al., 2022) have chosen to adapt instruments intended to assess bereavement for human losses, such as the Grief and Meaning Reconstruction Inventory (Gillies et al., 2015), the Inventory of Complicated Grief (Prigerson et al., 1995), or the Traumatic Grief Inventory (Boelen & Smid, 2017), although the literature offers op�ons specifically developed for the death of companion animals, among which the Pet Bereavement Questionnaire (PBQ) stands out. The PBQ was originally validated in a US sample (Hunt & Padilla, 2006). These authors created an ini�al set of 20 items referring to bereavement behaviors that follow the loss of a companion animal, aimed to measure four factors: grief, anger, guilt, and trauma. Their exploratory factor analysis (EFA) recommended preserving 16 items, distributed in three factors (trauma was not retained). Since its appearance, the PBQ has been validated in different countries, in some of which the original structure has been corroborated by a Confirmatory Factor Analysis (CFA): Italy (Testoni et al., 2019) and Portugal (Silva et al., 2023). However, other valida�on studies recommended alterna�ve structures. In Turkey, the combina�on of EFA and CFA retained the 16 items present in the original valida�on, but two items of the anger scale loaded on the grief scale (Yüksel et al., 2022). These items referred to the experience of nightmares and haun�ng memories, reac�ons that fell under the trauma factor in the original set of 20 items. The fit for the model improved by including some covariates (errors for items 3 and 5, and items 1 and 13). In a recent valida�on in a Korean sample, the authors carried out an EFA with the original 20 item set, ge�ng a reduced solu�on of 12 items distributed in three factors: grief, anger, and trauma, which replaced guilt (Jeon et 5 Information Classification: General al., 2022). These differences in the structure of the PBQ may be related to cultural context, given that expressions of bereavement are socially constructed (Neimeyer et al., 2014). Also, gender of the respondent is one of the most relevant variables in this regard. A recent review (Cleary et al., 2022) corroborates that the studies dedicated to the death of companion animals have a clear gender bias, with 85% of par�cipants repor�ng being women. These results are in line with those noted in other reviews within the field of human-animal interac�on (Herzog, 2007), as most of the results available in the literature are more representa�ve of women than of men. In summary, the PBQ is the most commonly used measure of bereavement for the loss of companion animals. However, the original factorial solu�on has been both supported and challenged by later studies, especially around items referring to trauma�c experience. Given that the PBQ has not been validated in Spanish so far, the present study tested the adequacy of preexis�ng solu�ons in the Spanish popula�on, providing psychometric informa�on on its structure and reliability. Second, it presents findings on the intensity of bereavement in the Spanish popula�on, measuring the rela�ve effect of gender. Method Participants The present study included a total of 333 Spanish par�cipants of legal age (M = 31.5 years; SD = 11.6 years). The majority of par�cipants reported being women (76.3%; 1.8% iden�fied as non-binary). Regarding household composi�on, 8.4% lived with no other humans; 12.3% lived with friends or roommates; 38.1% lived with a partner; and, 46.2% lived with other family members such as parents and children (these last two condi�ons could overlap, so the percentages exceed 100%). All reported experience living with companion animals, with dogs and cats being the main companions (79.0% and 36.6%, respec�vely), followed by birds 6 Information Classification: General (18.9%), rodents (16.8%), rabbits (12.6%), rep�les (8.1%), and other op�ons below ten cases. All par�cipants reported having suffered the loss of at least one companion animal. Instruments Sociodemographic data included respondent gender and age, the composi�on of the household) and the history of companion animals. The Pet Bereavement Questionnaire (Hunt & Padilla, 2006) is 16 items in three scales: grief (7 items; I have had nightmares about the death of my pet); anger (5 items; I feel angry with the vet for not being able to save him/her); and guilt (4 items; I feel bad for not being able to save him/her). Each item is answered from 1 = strongly disagree to 4 = strongly agree. In addition to the direct score obtained from the sum of the responses, the correction rule of Botchway et al., (2022) was applied, considering cases with averages higher than the midpoint (2.5 points per item) as positive. Procedure We obtained permission from the original study author (Dr. Hunt) to validate the PBQ in Spanish. We posted a message on social networks (Twiter and Facebook) giving the objec�ves of the study and a link with the instruc�ons, informed consent, and a batery of instruments hosted in a web applica�on. A total of 369 people agreed to par�cipate, although 36 par�cipants reported not having suffered any companion animal loss, so we eliminated them from the final database (9.8% of the total). In the responses to the PBQ, only two missing data were located (out of a total of 5,328 possible responses), which were replaced by the sta�s�cal mode. The study was approved by the Ethics Commitee of the University of Sevilla, Spain (ref. 1080-N-2023). 7 Information Classification: General The CFA was developed using the JAMOVI so�ware, version 2.3.28 (The JAMOVI project, 2020), which offers Kaiser’s Measure of Sampling Adequacy (MSA > 0.5), as well as fit indicators as Chi Square (*p < .05), Compara�ve Fit Index (CFI ≥ .95 considered good; ≥.90 adequate); Root Mean Square Error of Approxima�on (RMSEA < .05 considered good; .05 to .10 moderate; >.10 inadequate) and Standardized Residual Root Mean Square (SRMR < .09) (cut-off points taken from Hu & Bentler, 1999; Lorenzo-Seva & Ferrando, 2021). The reliability of the factors was es�mated using Bayes Expected -A-Posteriori Alpha (EAP-Alpha > .700), offered by FACTOR 10 (Ferrando & Lorenzo-Seva, 2016). For descrip�ve analyses we used SPSS so�ware, version 26 (IBM, 2022).The calcula�on of the effect size was conducted using Cohen's d sta�s�c (mean difference between standard devia�on, small effect for d > .20, moderate effect for d > .50, large effect for d > .80) (Cohen, 2013). Results Before CFA was carried out, MSA values were obtained. MSA ranged from .864 - .952, indica�ng adequate discriminant power of all the items (informa�on available in supplemental materials). Of the three models (original structure of three factors by Hunt & Padilla, 2006; and two alterna�ve structures proposed by Yüksel et al., 2022), indicators showed a greater parsimony of the 3-factor structure proposed by Yüksel et al. (2022), including rearrangement of two items (#4 and #14 moved from Anger to Grief factor) and covariates between of some measure errors. This model showed the best fit indicators, as well as the highest reliability indexes (Table 1). Factor loadings are presented in supplemental materials. On the grief scale, women scored significantly higher than men on both the original, t(325) = 4.522; p < .001; d = .58, and modified structures, t(325) = 4.882; p < .001; d = .63. On the anger scale, women scored significantly higher than men on the original scale, t(325) = 3.037; p = .003; d = .40, but on the modified scale women and men were not significantly different, t(325) = 1.460; p = .073; d = .19. On the guilt scale, women and men were not 8 Information Classification: General significantly different, t(325) = 1.048; p = .295; d = .14. Using a dichotomous correc�on, which considered posi�ve cases those that exceeded the midpoint of each PBQ scale (Botchway et al., 2022), we found that more than 64% of the par�cipants showed signs of intense bereavement on the grief scale; more than 35% did so on the anger scale; and 40% on the guilt scale (See Table 2). People who iden�fied themselves as nonbinary were not included in means contrasts analyzes due to their small numbers (N=6). However, all of these par�cipants exceeded the cutoff point to be categorized as experiencing high intensity bereavement on the grief scale, and four out of the six exceeded this cut-off point on the anger and guilt scales. Finally, we compared present means to those in other European samples (Silva et al. 2023; Testoni et al., 2019). Present Spanish par�cipants reported more intense grief, anger, and guilt a�er bereavement reported than the Portuguese par�cipants, and reported less grief, but more anger and guilt, than Italian par�cipants. This informa�on is included in Table 3. Discussion and Conclusions Social changes require measuring instruments adapted to the needs of the popula�on. The increasing number of households in Spain that have at least one companion animal, along with the increasing recogni�on of companion animals as family members, highlights the need to create validated tools to measure the various aspects of human and animal co-existence, including bereavement for the loss of an animal. As far as we know, it is the first study dedicated to validate a measure to assess bereavement for companion animals in Spanish, thus represen�ng a valuable resource for both prac��oners and researchers. The first objec�ve of the present study was to test the validity and reliability of the different PBQ structures in the Spanish popula�on. The obtained results challenge the validity of the structure proposed by Hunt and Padilla (2006). Although two fit indexes (Chi2 and SRMR) showed acceptable values, the other two (CFI and RMSEA) were poor, thus recommending 15 Information Classification: General World Health Organiza�on. (2022). International classification of diseases: Eleventh Revision (ICD-11) (11th ed.). World Health Organiza�on. Yüksel, Ö., Apak, B., & Demirci, Ö. (2022). Turkish version of the Pet Bereavement Ques�onnaire: Validity, reliability and psychometric proper�es. Death Studies, 0(0), 1–9. htps://doi.org/10.1080/07481187.2022.2034073 16 Information Classification: General Table 1. CFA adjustment indicators and reliability for the three alterna�ve models Fit index Original (16 items, 3 factors) Modified by Yüksel ( moved #4 and #14) Modified by Yüksel ( + error covariances) Chi 2 <.001 <.001 <.001 gl 101 101 99 CFI .863 .891 .903 RMSEA .112 .100 .095 CI RMSEA 90% .103-.122 .091-.110 .086-.105 SRMR .079 .064 .063 Reliability Cronbach’s alpha Grief .886 .891 .891 Anger .790 .836 .836 Guilt .862 .862 .862 EAP alpha Grief .934 .937 .937 Anger .892 .921 .921 Guilt .945 .945 .945 EAP = Expected-a-posteriori Bayes alpha. Grief scale has seven items in the original model and nine in the modified solu�on; anger scale includes five items in the original model and three in the modified solu�on. 17 Information Classification: General Table 2. Descrip�ve and classifica�on analysis for the total sample, women, men, and non-binary par�cipants. Total (N=333) Women (N=254) Men (N=73) Non-binary (N=6) Means (SD) % Posi�ve Means (SD) % Posi�ve Means (SD) % Posi�ve Means (SD) % Posi�ve Grief (original) 20.16 (5.18) 70.0% 20.79 (5.05) 74.0% 17.75 (5.05) 53.4% 22.67 (3.14) 100% Grief (modified) 24.71 (6.50) 64.3% 25.54 (6.32) 69.7% 21.45 (6.21) 42.5% 29.17 (4.17) 100% Anger (original) 10.89 (4.04) 35.1% 11.19 (4.07) 39.0% 9.58 (3.70) 19.2% 14.33 (2.58) 66.7% Anger (modified) 6.34 (2.88) 36.0% 6.44 (2.97) 38.2% 5.88 (2.58) 17.4% 7.83 (1.47) 50% Guilt 9.53 (3.68) 40.2% 9.59 (3.73) 41.7% 9.08 (3.48) 32.9% 12.33 (2.73) 66.7% 18 Information Classification: General PBQ Total 40.58 (11.31) 48.3% 41.57 (11.28) 53.1% 36.41 (10.55) 28.8% 49.33 (7.47) 83.3% Original = version by Hunt & Padilla (2006); modified = version by Yüksel et al. (2022). Cut-off points to consider the case posi�ve: Grief (original) > 17.5; Grief (modified) > 22.5: Anger (original) > 12.5; Anger (modified) > 7.5; Guilt > 10.0; PBQ Total > 40 (Botchway et al., 2022). 19 Information Classification: General Table 3. Contrast of means (t test for a sample) for studies with comparable methodology (df = 332; ***p < .001 in all contrasts) M SD Means differ. t Weighted SD d Portugal Grief (original) 17.64 3.99 2.52 8.87 4.78 .53++ Grief (modified) 21.03 7.19 3.68 10.31 6.73 .55++ Anger (original) 7.85 2.90 3.04 13.71 3.66 .83+++ Anger (modified) 4.46 1.10 1.88 11.91 2.28 .82+++ Guilt 7.92 2.03 1.61 7.99 3.12 .52++ Italy Grief (original) 21.66 4.93 -1.50 5.30 5.06 .29+ Grief (modified) 25.93 6.23 -1.22 3.44 6.37 .19 Anger (original) 8.52 3.10 2.37 10.69 3.57 .66++ Anger (modified) 4.25 1.96 2.09 13.23 2.42 .86+++ Guilt 8.68 3.63 0.85 4.22 3.66 .23++ The means and SD have been recalculated for Portuguese (N=169) and Italian (N=327) samples to allow comparisons with the present study. Nega�ve means differences indicate that Italian records surpassed Spanish means. Size effect + small; ++ medium; +++ large.