Randomized controlled study on the effectiveness of animal‐assisted therapy on depression, anxiety, and illness perception in institutionalized elderly ORIGINAL ARTICLE Randomized controlled study on the effectiveness of animal-assisted therapy on depression, anxiety, and illness perception in institutionalized elderly Caterina AMBROSI ,1 Charles ZAIONTZ,1 Giuseppe PERAGINE,2 Simona SARCHI2 and Francesca BONA2 1Psychology Unit, IES Abroad c/o Università Catto- lica del Sacro Cuore, Milan and 2Fondazione Santa Chiara Long-Term Geriatric Care, Centro Multiser- vizi per la Popolazione Anziana, Lodi, Italy Correspondence: Dr Caterina Ambrosi PhD, Psychology Unit, IES Abroad c/o Università Cattolica del Sacro Cuore, Via Carducci 26, 20123 Milan, Italy. Email: caterinambrosi@gmail.com Disclosure: None declared. Received 9 January 2018; revision received 28 February 2018; accepted 29 June 2018. Key words: animal-assisted therapy, depression, dog, elderly, illness perception. Abstract Aim: The aim of this study was to verify dog-assisted therapy’s effective- ness on depression and anxiety in institutionalized elderly. Patients’ illness perception was examined to identify core beliefs regarding mood, personal control, and illness coherence because they can affect treatment compli- ance. Subjective perception of pain, social interaction, and setting-bound observable variables were also studied. Methods: This study involved a randomized sample of institutionalized patients 65 years of age and older; the treatment group had 17 subjects and the control group had 14 subjects. All patients were administered the Mini-Mental State Examination, 15-item Geriatric Depression Scale, Positive and Negative Affect Schedule, Generalized Anxiety Disorder 7, Illness Per- ception Questionnaire, and Numeric Pain Rating Scale. Intra- and inter- group data analysis was performed before and after treatment. Over the course of 10 weeks, patients participated in individual 30-min sessions. An observational methodology was developed to record verbal and non-verbal interactions between the elderly, the dog, and the dog handler. Results: A large effect size and a statistically significant decrease in 15-item Geriatric Depression Scale scores were identified in the treatment group. No significant differences were detected in the Generalized Anxiety Disorder 7, Positive and Negative Affect Schedule, and Numeric Pain Rating Scale. However, the Positive and Negative Affect Schedule and the Numeric Pain Rating Scale showed a moderate decrease. The Illness Perception Questionnaire’s timeline (acute/chronic) and treatment control subscales showed a clinically relevant, large effect size. Conclusions: Dog-assisted therapy has proved to be effective in reducing symptoms of depression in institutionalized elderly. The increase in verbal interactions with the handlers throughout the study suggests the dog acts as a facilitator of social interaction, eliciting positive emotional responses. Dog-assisted therapy shows promising results in the perception of illness timeline and treatment control, indicating potential enhancement of the sense of treatment-related empowerment. However, further study is required. INTRODUCTION Major depression and clinically significant depressive symptoms represent a consistently corroborated clin- ical reality among elderly in long-term care as observed by Thakur and Blazer in a systematic litera- ture review.1 Nonetheless, such clinical conditions tend to go undetected because of co-morbidities such as cognitive functional impairment; multiple © 2018 Japanese Psychogeriatric Society 55 doi:10.1111/psyg.12367 PSYCHOGERIATRICS 2019; 19: 55–64 https://orcid.org/0000-0002-1937-9781 http://crossmark.crossref.org/dialog/?doi=10.1111%2Fpsyg.12367&domain=pdf&date_stamp=2018-09-16 medical conditions, often coupled with physical pain, act as inter-related variables that increase functional impairment.2 Non-pharmacological interventions, such as music therapy, have proved to be effective complementary modalities in reducing depression and delaying the deterioration of cognitive functions.3 In contrast, despite receiving increasing attention in the last few years as reflected in the research and international lit- erature, dog-assisted therapy (DAT) for the elderly still shows mixed results with regard to be an effec- tive treatment of depression and anxiety. However, DAT has been shown to improve social interactions in the patient population according to limited studies of insufficient methodological quality.4,5 It should be noted that promising data on the cor- relation between animal-assisted therapy (AAT) and physical well-being of patients have been observed, particularly in relation to cardiovascular disorders.6 However, the literature has paid less attention to the study of DAT’s effects on mood disorders and illness perception in institutionalized elderly populations, despite the biopsychosocial implications for patients in long-term care facilities. Popularly referred to as ‘pet therapy’, this kind of clinical intervention has played an important role in strengthening the therapeutic alliance between patient and treater. It also activates cognitive circuits and communication channels, eliciting thoughts and memory patterns that tend to remain silent, unex- plored, and inhibited.7 As observed by Moretti et al.,8 the ‘frequent co- occurrence of cognitive and mood disorders, psy- chotic and anxiety symptoms make elderly persons especially suitable for treatment based on affective and emotional motivations and psychological stimulations’. A study by Stasi et al. regarding DAT showed that patients included in the treatment group had decreased depressive symptoms and blood pressure variables.9 In a report by Kawamura et al.,10 AAT cor- related with improvements in mental and emotional functions, steadily increasing patients’ emotional well-being. Conversely, Phelps et al. indicated that dog visits did not significantly improve depression scores, mood, or social interactions in their sample.11 Notably, in 2015, the Italian Ministry of Health determined that DAT is a clinical intervention directed at the ‘treatment of disorders of neuro-psychomotor cognitive, emotional and relational spheres’ and issued new norms and regulations for its implemen- tation. DAT requires a medical prescription and is entirely patient-centred. According to the current leg- islation, DAT must be carried out in a manner consis- tent with evidence-based criteria and requires a clinically stringent treatment plan, a multidisciplinary team, quantifiable objectives, and verification of achieved results.12 The dog acts as a catalyst, facilitator, and motiva- tor of social attention in the intervention. The inter- vention depends on the relationships forged among the patient, the clinician, the dog, and the dog han- dler. The latter has the fundamental task of identify- ing which dog is best suited to the patient based on morphological and behavioural characteristics. It is also the dog handler’s job to determine the most suitable methods for the patient to interact with the dog, ‘translating’ both in an ethological and anthro- pomorphic way what the dog wants to communicate to the patient and vice versa. In light of the mixed findings on DAT, this present study aimed to investigate its effects on mood, affect, and illness perception. The study was con- ducted in compliance with the substantial changes introduced by the Italian Ministry of Health on the implementation of AAT in clinical settings. Our main objectives were to identify whether the presence of the dog effectively elicited noteworthy changes in the psychopathological framework, to determine in which specific dimensions of depressive and anxiety disor- ders these changes occurred, and to assess patients’ sense of empowerment and control over their per- ceived sense of illness and their spontaneous verbal and non-verbal communication. METHODS Study population This study was conducted in a National Health Service-accredited long-term care facility for the elderly in northern Italy from March 2017 to September 2017. The study involved two fully ran- domized groups from this facility: a treatment group of 17 subjects and a control group of 14 subjects. Patients were selected from the general patient population subject to the following inclusion criteria: • Age: 65–90 years • Institutionalized for at least 2 months C. Ambrosi et al. © 2018 Japanese Psychogeriatric Society56 • A score of 5 or higher on the 15-item Geriatric Depression Scale (GDS-15)13 • A score of 19 or higher on the Mini-Mental State Examination14 • Willingness to interact with the dog and absence of animal allergies. The main exclusion criterion was being unable to interact with the clinical team because of multisen- sory impairment. Locomotor disability was not an exclusion factor: single or double amputees were not excluded. Existing records were examined by the facility’s clinical team to identify which patients met the inclu- sion criteria. Fifty-three patients met the Mini-Mental State Examination threshold, and of these, 36 met the GDS-15 threshold. Three were excluded because they refused to participate in any type of activity or leave the unit, and two more were excluded because their clinical condition had severely deteriorated. This left 31 subjects to participate in the study. Seventeen subjects were randomly assigned (by drawing their names written on pieces of paper from a bowl) to the treatment group, leaving 14 sub- jects in the control group. One member of the control group died during the study so that person’s assess- ments were excluded. There was no statistically significant difference in the ages of the participants in the treatment and con- trol groups (median: 85 vs 88 years), and there was no statistically significant difference between the two groups on any of the pretreatment assessments. Informed consent was obtained from the patients or their legal guardians according to Italian law. Patients’ right to privacy was preserved, and patient anonymity was guaranteed so that none of the partic- ipants would be identifiable to third parties. The study was approved by the local ethical committee, and it conforms to the Declaration of Helsinki as revised in 2013. The CONSORT 2010 guidelines and checklist were followed. The study started with a 4-hour training session on AAT for the clinical staff who participated in the research project; they were an integral part of the multidisciplinary team conducting the study. The staff were trained on the foundations of AAT according to the new Italian ministerial guidelines and on the experimental design of this study. The study involved two professional dog handlers from a fully registered non-profit pet therapy organization, a veterinary surgeon, a clinical psychol- ogist, and six dogs (five golden retrievers and one flat-coated retriever). The facility’s team comprised a geriatrician, a professional nurse, a professional reha- bilitation therapist, and a clinical supervisor. Dogs were professionally trained and had passed certification aptitude tests for therapy dogs. All cri- teria in the Italian Ministry of Health’s guidelines for animal-assisted interventions and animal welfare were satisfied. Assessments At baseline, assessment tests were administered to all the participants in the study before the DAT began. The same set of tests was administered to all the participants after they completed 10 weeks of treatments. Baseline results were compared with post-treatment results, and the results of the treat- ment group were compared with those of the control group. To assess baseline depression, mood, affect, and illness perception, patients completed the GDS-15, Generalized Anxiety Disorder 7 (GAD-7),15 Positive and Negative Affect Schedule (PANAS),16 and Illness Perception Questionnaire-Revised (IPQ-R).17 They also filled out a Satisfaction Questionnaire and Numeric Pain Rating Scale (NPRS).18 The NPRS asks respondents to rate their level of pain from 0 to 10, with 0 indicating no pain, 1–3 slight pain, 4–7 moderate pain, and 8–10 severe pain. The battery of tests included the IPQ-R,19 a quan- titative measure of five components of illness repre- sentation according to Leventhal’s self-regulatory model. The IPQ-R now also includes subscales assessing cyclical timeline perceptions, illness coher- ence, and emotional representations. While illness perceptions have previously been found to correlate with mood,20 functional adaptation,21 and compliance to treatment,2 the IPQ-R components show that ‘beliefs in treatment and personal control and a sense of illness coherence [are] inversely related to pes- simistic beliefs about the timeline and consequences of the illness as well as negative emotional representa- tions’.2 Pessimism may result from factors that interfere with therapy in the elderly patient population, leading to poor treatment adherence. Our study investigated how DAT could impact the IPQ-R components and whether it represented a via- ble and effective form of treatment for the patients. DAT for institutionalized elderly © 2018 Japanese Psychogeriatric Society 57 DAT DAT sessions took place once a week for 10 weeks. Each patient in the treatment group participated in one 30-min session each week. Each DAT session involved one patient, one dog, one dog handler, and one observer (trained health volunteers from the civil ser- vice). Patients in the control group did not participate in these sessions nor did they participate in any partic- ular substitute activity. Patients in both groups contin- ued with their usual care, including pharmacological treatments and voluntary participation in social activi- ties. The dog, dog handler, and observer rotated each week. The objective was to reduce the confounding effect of a specific dog, dog handler, or observer. Social interaction focused on verbal and non- verbal cues towards the dog or the dog handler dur- ing the DAT. Verbal interaction with the dog was defined as the patient speaking with the dog or emit- ting sounds or vocalizations towards the dog. Non- verbal interaction with the dog was defined as the patient petting or stroking the dog or giving or throw- ing the dog a bit of food or a toy (e.g. a ball, rope). Verbal interaction with the dog handler was defined as the patient making a comment or posing a ques- tion to the dog handler. Non-verbal interaction was defined as the patient tapping, touching, or carrying out an act directed at the dog handler. At two-minute intervals, the observer would record an interaction as being in one of these four categories of interaction if one or more interactions of that type occurred. Then the total number of interactions during each DAT ses- sion in each category was calculated. The resulting figure is the percentage obtained by dividing the total number of interactions by the number of two-minute intervals in the session. After sessions 3, 6 and 9, participants were asked to complete a short questionnaire about their emo- tional state using a 5-point Likert scale and their desire to see the dog in the next session of DAT. Statistical analyses Statistical analysis was performed using Real Statis- tics release 5.5 (http://www.real-statistics.com). Our principal investigation was to see whether there was a significant difference (based on P-value) and/or important difference (based on effect size) between GDS-15, PANAS, GAD-7, IPQ-R, and NPRS scores before and after the 10 weeks of DAT. Because some of the respondents gave a range of responses on the NPRS, this was analyzed separately. Also, because of the large number of IPQ-R subscales (eight), these were analyzed separately. Hotelling’s paired T-square test was used to deter- mine whether there was a significant difference between the scores after DAT, and then multiple paired t-tests were performed using a Bonferroni cor- rection factor as a follow-up. A significance level of α = 0.05 was employed (subject to the Bonferroni experiment-wise error correction based on at most 5 + 1 + 8 = 14 tests). Cohen’s dav (based on a pooled variance equal to the average of the pre- and post-treatment variances) was used to measure effect size. Cohen’s dz (based on the standard devia- tion of the differences between the pre- and post- treatment scores) was also reported. Because of the limitations in previous DAT studies, we were not able to evaluate the effect size based on previous studies. As such, we used the usual rough guidelines that d = 0.20 represents a small effect, 0.50 a medium- sized effect, and 0.80 a large effect. Confidence intervals of the effect size were calculated by using a noncentral t distribution approach. Statistical power analysis was performed based on the paired t-test. Assessment scores between the treatment and control groups before DAT were compared to ensure that the separation of patients into the two groups was unbiased. No significant differences were found based on the independent sample versions of Hotell- ing’s T-square test and t-tests. The mean age of the treatment group participants was 82.6 years, and the median was 85 years. The mean age of the control group participants was 87.1 years, and the median was 88 years. This differ- ence was not statistically significant based on the Mann–Whitney test (because the data were not normally distributed): U = 70, P = 0.086. Except for two men in the treatment group, all the participants were women. The independent sample versions of Hotelling’s T- square test and t-tests were also employed to deter- mine whether there was a significant difference between the treatment and control groups regarding the change in the test scores over the treatment period. Cohen’s effect size d for independent samples as well as effect size confidence intervals (again based on a noncentral t distribution approach) were also obtained. Social interactions are represented in Figures 1-3. C. Ambrosi et al. © 2018 Japanese Psychogeriatric Society58 http://www.real-statistics.com RESULTS Effectiveness of DAT in the treatment group GDS-15, PANAS, and GAD-7 The paired sample Hotelling’s T-square test showed that there was a significant difference between the scores before and after DAT in the treatment group. The differences occurred across six scales—one GDS- 15 scale, one GAD-7 scale, and four PANAS scales (T2 = 95.55, F6,11 = 10.95, P = 0.0004). The follow-up analysis demonstrated that the significant difference was attributable to the GDS-15 test (Table 1). The scores on the GDS-15 test decreased by an average of 33.5% after DAT, indicating a significant decrease in depression (t16 = 6.52, P = .000007, dz = 1.58, dav = 2.05). Even with a Bonferroni correc- tion that brought alpha down to 0.05/14 = 0.00357, this result was still highly significant. The decrease in GDS-15 score was very large, with an effect size much bigger than 0.80. It should be noted that for a sample of size 17, a paired t-test can detect an effect size of 0.85 (or higher) with power of 90% for α = 0.05 (or an effect size of 1.2 with power of 90% for α = 0.00357). This is sufficient for the GDS-15 result. Of the 15 items on the GDS-15 scale, all but items 2, 5, and 9 improved in the treatment group. The per- centage change in scores for all 15 items is shown in Table 2. PANAS state negative affects showed a 21.3% improvement, a medium-sized effect (dz = 0.59, dav = 0.65), and P = 0.026, which, although smaller than α = 0.05, is not significant after a Bonferroni or similar correction is taken into account. NPRS All patients rated their pain on a scale of 0 to 10, with 10 representing the most severe pain. Three of the patients in the treatment group gave a range of pain values (5–8 or 0–5); the mean was used for these patients. Pain levels went down by 11.4%, which was small and not significant (P = 0.46, d = 0.18). Because the Table 1 Improvements in GDS-15, PANAS, and GAD-7 in the treatment group GDS-15 State PA State NA Trait PA Trait NA GAD-7 Sample size 17 17 17 17 17 17 Min increase −8 −11 −15 −14 −11 −9 Max increase 0 17 7 9 9 8 % Increase −33.5% 6.5% −21.3% −8.8% −8.2% −8.6% Mean increase −3.35 1.53 −3.76 −2.06 −2.12 −0.82 SD 2.12 6.62 6.35 5.67 6.11 4.81 P-value 0.00 0.36 0.03 0.15 0.17 0.49 Lower 95% CI −4.44 −1.88 −7.03 −4.98 −5.26 −3.30 Upper 95% CI −2.26 4.94 −0.50 0.86 1.03 1.65 Cohen’s dz −1.58 0.23 −0.59 −0.36 −0.35 −0.17 Hedges’ gz −1.51 0.22 −0.56 −0.35 −0.33 −0.16 Lower dz −2.29 −0.25 −1.10 −0.85 −0.83 −0.65 Upper dz −0.85 0.71 −0.07 0.13 0.15 0.31 Cohen’s dav −2.05 0.19 −0.65 −0.30 −0.34 −0.17 Hedges’ gav −1.95 0.18 −0.62 −0.28 −0.32 −0.16 Lower dav −2.97 −0.21 −1.20 −0.69 −0.81 −0.63 Upper dav −1.10 0.58 −0.07 0.11 0.15 0.30 All measurements reflect post-treatment scores minus pre-treatment scores. 95% CI is the 95% confidence interval of the mean increase. dz is Cohen’s effect size based on the standard deviation of score differences. dav is Cohen’s effect size based on the average of the standard deviations pre- and post-treatment. Hedges’ d is a more unbiased statistic corresponding to Cohen’s d. Tests are based on paired t-tests. GAD-7, Generalized Anxiety Disorder 7; GDS, 15-item Geriatric Depression Scale; PANAS, Positive and Negative Affect Schedule; NA, Negative Attributes; PA, Positive Attributes. Table 2 Changes by item on the 15-item Geriatric Depression Scale in the treatment group Pre-treatment Post-treatment % Change Q1 0.471 0.294 37.5% Q2 0.706 0.765 −8.3% Q3 0.882 0.412 53.3% Q4 0.882 0.412 53.3% Q5 0.765 0.824 −7.7% Q6 0.294 0.235 20.0% Q7 0.588 0.353 40.0% Q8 0.882 0.471 46.7% Q9 0.412 0.471 −14.3% Q10 0.294 0.176 40.0% Q11 0.706 0.647 8.3% Q12 0.882 0.529 40.0% Q13 0.765 0.412 46.2% Q14 0.765 0.353 53.8% Q15 0.706 0.294 58.3% Total 0.667 0.443 33.5% All scores are the mean. DAT for institutionalized elderly © 2018 Japanese Psychogeriatric Society 59 normality assumption was violated, the Wilcoxon signed-rank test was used instead of a paired t-test. IPQ-R The paired sample Hotelling’s T-square test showed there was no significant difference between the scores on the eight subscales before and after ther- apy in the treatment group (T2 = 17.65, F8,9 = 1.24, P = 0.37). Based on this result, no follow-up was necessary, but it was performed anyway to get fur- ther information (Table 3). The mean of subscale 4 decreased by 30.4%, a large effect size (dz = 0.62, dav = 0.88), but it was not significant when a Bonferroni correction was taken into account (P = 0.021). Satisfaction The patients in the treatment group were asked to rate their satisfaction with DAT on a scale of 1–5 after weeks 3, 6, and 9. The mean rating was 4.89, with all rating being 4 or 5. Comparison between the treatment and control groups Comparison before DAT There was no significant difference between the treatment and control groups on any of the tests. A series of two-sample t-tests for 13 scales were performed as well as a Mann–Whitney test for the NPRS (because the normality assumption was violated). None of the P-values for these tests was less than 0.05. In fact, none was lower than 0.27, except for IPQ-R scale 2 (P = 0.069). These results gave us more confidence that patients were indeed randomly assigned to the treatment and control groups. Comparisons of the changes in each scale The two-sample Hotelling’s T-square test showed there was no significant difference in the net scores between the treatment and control groups on the GDS-15, the GAD-7, and the four PANAS scales (T2 = 16.51, F6,23.4 = 2.22, P = 0.077). Based on this result, no follow-up was necessary, but it was per- formed anyway using multiple two-sample t-tests to get further information (Table 4). We also added NPRS to this table (using the Mann–Whitney test because the normality assumption was violated). The GDS-15 result was significant and showed a large effect size (mean � SD: −3.10 � 2.18, t22.8 = 3.73, P = 0.0011, d = 1.42, power = 95%). This demonstrated a high degree of confidence that the treatment group had a much greater improve- ment on the GDS-15 than the control group. Similarly, the two-sample Hotelling’s T-square test showed there was no significant difference in the net scores between the treatment and control groups on the eight scales of the IPQ-R test (T2 = 14.13, F6,25.1 = 1.29, P = 0.29). Table 3 Improvement in the Illness Perception Questionnaire-Revised in the treatment group Scale 1 Scale 2 Scale 3 Scale 4 Scale 5 Scale 6 Scale 7 Scale 8 Sample size 17 17 17 17 17 17 17 17 Min increase −7 −11 −7 −20 −11 −5 −14 −12 Max increase 7 8 7 9 11 8 9 4 % Increase 14.4% −4.3% −1.1% −30.4% 9.5% 8.0% −16.1% −8.1% Mean increase 0.76 −1.18 −0.24 −4.82 1.35 1.47 −1.76 −1.88 SD 3.63 3.91 4.01 7.75 6.04 4.19 5.23 4.40 P-value 0.40 0.23 0.81 0.02 0.37 0.17 0.18 0.10 Lower 95% CI −1.10 −3.19 −2.30 −8.81 −1.75 −0.68 −4.45 −4.14 Upper 95% CI 2.63 0.83 1.83 −0.84 4.46 3.62 0.92 0.38 Cohen’s dz 0.21 −0.30 −0.06 −0.62 0.22 0.35 −0.34 −0.43 Hedges’ gz 0.20 −0.29 −0.06 −0.59 0.21 0.33 −0.32 −0.41 Lower dz −0.27 −0.78 −0.53 −1.14 −0.26 −0.14 −0.82 −0.92 Upper dz 0.69 0.19 0.42 −0.09 0.70 0.84 0.16 0.08 Cohen’s dav 0.23 −0.29 −0.04 −0.88 0.34 0.31 −0.40 −0.31 Hedges’ gav 0.22 −0.27 −0.04 −0.84 0.33 0.29 −0.38 −0.29 Lower dav −0.30 −0.75 −0.35 −1.61 −0.40 −0.13 −0.97 −0.66 Upper dav 0.76 0.18 0.28 −0.13 1.08 0.74 0.18 0.05 All measurements reflect post-treatment score minus pre-treatment score. 95% CI is the 95% confidence interval of the mean increase. dz is Cohen’s effect size based on the standard deviation of score differences. dav is Cohen’s effect size based on the average of the standard deviations pre- and post-treatment. Tests are based on paired t-tests. C. Ambrosi et al. © 2018 Japanese Psychogeriatric Society60 Follow-up testing using multiple t-tests was per- formed to get further information (Table 5). Although there were no significant differences between the treatment and control groups (based on a Bonferroni correction), the results for subscales 2 (timeline: acute/chronic) and 5 (treatment control) are interesting and worth further investigating. The scores for subscale 2 in the treatment group decreased, whereas those for the control group increased, with a difference of 3.54, P = 0.05 and d = 0.84, denoting a large effect. The scores for subscale 5 in the treatment group increased, but those for the control group decreased, with a differ- ence of 4.17, P = 0.07 and d = 0.73, denoting a large effect. Analysis: social interaction Social interaction was measured by participants’ ver- bal and non-verbal interactions with the dog and the dog handler during a DAT session. The highest level of participation was verbal with the handler (mean: 79.6%), followed by 68.0% for non-verbal with the Table 4 Comparison of GDS-15, PANAS, GAD7, and NPRS between treatment and control groups GDS-15 State PA State NA Trait PA Trait NA GAD-7 NPRS % Increase treatment group −33.5% 6.5% −21.3% −8.8% −8.2% −8.6% −11.4% % Increase control group −3.2% 16.8% −23.9% 1.8% −14.9% −18.4% −5.0% Size of treatment group (n) 17 17 17 17 17 17 17 Size of control group (n) 12 11 11 11 11 12 12 Treatment mean −3.35 1.53 −3.76 −2.06 −2.12 −0.82 −0.26 Control mean −0.25 1.36 −4.91 −1.55 −4.64 −2.25 −0.04 Mean difference −3.10 0.17 1.14 −0.51 2.52 1.43 −0.22 Treatment SD 2.12 6.62 6.35 5.67 6.11 4.81 1.45 Control SD 2.26 6.47 9.65 6.36 7.58 4.20 1.74 Pooled SD 2.18 6.57 7.79 5.95 6.71 4.57 1.57 P-value 0.00 0.95 0.73 0.83 0.37 0.40 0.40 Lower mean difference −4.79 −5.06 −5.05 −5.24 −2.82 −2.11 — Upper mean difference −1.42 5.39 7.34 4.22 7.86 4.96 — Cohen’s d −1.42 0.03 0.15 −0.09 0.38 0.31 −0.14 Hedges’ g −1.38 0.02 0.14 −0.08 0.36 0.30 −0.14 Lower d −2.24 −0.73 −0.61 −0.84 −0.39 −0.43 Upper d −0.58 0.78 0.91 0.67 1.14 1.05 All measurements reflect change in treatment scores versus change in control scores. 95% confidence intervals for the mean difference and Cohen’s d (two inde- pendent samples). Hedges’ d is an unbiased statistic corresponding to Cohen’s d. All tests are based on two independent sample t-tests, except the NPRS, which used the Mann–Whitney test. The effect size for NPRS is r = −0.16 (based on the Mann–Whitney test). GAD-7, Generalized Anxiety Disorder 7; GDS, 15-item Geriatric Depression Scale; NPRS, Numeric Pain Rating Scale; PANAS, Positive and Negative Affect Schedule; NA, Negative Attributes; PA, Positive Attributes. Table 5 Comparison of the Illness Perception Questionnaire-Revised between the treatment and control groups Scale 1 Scale 2 Scale 3 Scale 4 Scale 5 Scale 6 Scale 7 Scale 8 % Increase treatment group 14.4% −4.3% −1.1% −30.4% 9.5% 8.0% −16.1% −8.1% % Increase control group 2.3% 6.7% −11.1% −33.9% −20.8% 9.3% −14.4% −14.9% Size of treatment group (n) 17 17 17 17 17 17 17 17 Size of control group (n) 11 11 11 11 11 11 11 11 Treatment mean 0.76 −1.18 −0.24 −4.82 1.35 1.47 −1.76 −1.88 Control mean 0.18 2.36 −1.91 −6.27 −2.82 2.18 −2.45 −2.82 Mean difference 0.58 −3.54 1.67 1.45 4.17 −0.71 0.69 0.94 Treatment SD 3.63 3.91 4.01 7.75 6.04 4.19 5.23 4.40 Control SD 3.03 4.70 4.35 6.68 5.25 3.68 6.38 6.35 Pooled SD 3.41 4.23 4.14 7.36 5.75 4.00 5.70 5.24 P-value 0.65 0.05 0.32 0.60 0.07 0.64 0.77 0.68 Lower mean difference CI −2.13 −6.90 −1.62 −4.40 −0.40 −3.89 −3.84 −3.23 Upper mean difference CI 3.30 −0.18 4.97 7.30 8.74 2.47 5.22 5.10 Cohen’s d 0.17 −0.84 0.40 0.20 0.73 −0.18 0.12 0.18 Hedges g 0.17 −0.81 0.39 0.19 0.70 −0.17 0.12 0.17 Lower d CI −0.59 −1.62 −0.37 −0.57 −0.06 −0.94 −0.64 −0.58 Upper d CI 0.93 −0.04 1.17 0.96 1.50 0.58 0.88 0.94 All measurements reflect change in treatment scores versus change in control scores. CI represents the 95% confidence intervals for the mean difference and Cohen’s d (two independent samples). Hedges’ d is an unbiased statistic corresponding to Cohen’s d. Tests are based on two independent sample t-tests. DAT for institutionalized elderly © 2018 Japanese Psychogeriatric Society 61 dog, 48.1% verbal with the dog, and 13.4% for non- verbal with the handler (Fig. 1). In general, participation levels were highest at the beginning of the session and, except for verbal inter- actions with the handler, rose again at the end of the session (Fig. 2). Figure 3 shows how the percentage of interactions varied by week. In general, these increased for verbal interactions with the handler and decreased for the other three categories. The frequency of interactions showed little variation with each dog or handler, although many of the participants did favour one dog over another. DISCUSSION The purpose of our study was to assess the effec- tiveness of DAT in reducing depression and anxiety Figure 1 Interactions at 2-min intervals. Hand, dog handler; non-verb, non- verbal interactions; verb, verbal interactions. Figure 2 Interactions at 2-min intervals by category. In each of the four categories, there is a reduction in interaction, often with a slight rise in the last 2 min. C. Ambrosi et al. © 2018 Japanese Psychogeriatric Society62 and to investigate the impact of such a therapeutic approach on affect and illness perception through a randomized controlled study. Our findings corrobo- rate the notion that animals help patients shift the focus off their symptoms of distress and, through interactions with the animal, elicit a sense of serenity and tranquillity, positively affecting their mood and thereby enhancing social interaction.22 The participants in this study had an increasing willingness to engage in DAT as well as positive emo- tional responses towards their encounters with the dogs, as overwhelmingly reported in the satisfaction questionnaires. These responses are indicative of the achievement of a major indirect objective: a consis- tent level of participation and engagement. DAT seemed to counteract signs of social withdrawal and the lack of motivation often observed in depressed institutionalized elderly.23 Based on a significant and very large decline in GDS-15 scores after treatment, our results demon- strate the effectiveness of DAT in reducing depres- sive symptoms. Although changes in anxiety and positive and negative affect were not significant, they should be investigated further to explore possible correlations between the elicitation and unearthing of emotional and memory patterns, which may remain dormant and unexplored,7 and the multidimensional compounded construct between alexithymia, anxiety, and ageing.24 The large effect size on subscales 2 and 5 of the IPQ-R suggests a considerable shift in the perception of treatment effectiveness. This shift may be general- izable with regard to further treatment adherence and a stronger sense of empowerment over the future of one’s disorder, which deserves further investigation. The reduction in the perception of physical pain, although not sizeable, also merits further study in light of the correlation between pain and depression in institutionalized elderly.2 There are some potential limitations to the general- izability of these findings. The samples were selected from a single long-term nursing care facility. How- ever, the facility is a National Health Service- accredited clinical institution; unlike in privately run facilities, its patients come from a wide sociodemo- graphic and clinical population, suggesting that the findings have broad relevance. Additionally, the Mini- Mental State Examination cut-off score was 19, which Figure 3 Interactions each week by category. Note that verbal communication with the handler trends up while communication with the dog trends down. DAT for institutionalized elderly © 2018 Japanese Psychogeriatric Society 63 means subjects with severe cognitive deterioration in comorbidity with depression were excluded. Further study is therefore recommended. The approach used in this study emphasized patients’ active participation in the therapeutic pro- cess. The synergetic relationship among the clinical team promoted cohesiveness and helped facilitate patients’ spontaneous behaviour that emerged during therapy. For example, one patient had a strong pas- sion for photography and was granted permission by the facility to take pictures of the dogs; this became a pivotal aspect of participants’ sense of belonging that occurred spontaneously over time and promoted interactions within the treatment group. Psychotherapeutic treatment for institutionalized elderly should aim to help patients gain a sense of time and space continuity, as well as identity coher- ence, after the ‘rupture’ from the individual’s ordinary reality. The institution can and should represent a haven where multiple clinical and relational aspects of a patient’s life are taken into account and addressed. This can help to promote a sense of psy- chological well-being, as can interventions that stim- ulate patient interactions and bonds, which may be fostered by the quiet, sensitive, and thought- provoking presence of a dog. ACKNOWLEDGMENTS We are thankful to all the staff of the Fondazione Santa Chiara Onlus, Centro Multiservizi per la Popo- lazione Anziana (Lodi, Italy), for their support and kindness and to Dr Chiara Maiorani, PhD for her valu- able clinical contribution. We are also grateful to Dia- mante Blu Kennel, operated by Dolores Genco, for supplying highly trained and sensitive therapy dogs that made this study possible. REFERENCES 1 Thakur M, Blazer DG. Depression in long-term care. J Am Med Dir Assoc 2008; 9: 82–87. 2 Mossey JM, Gallagher RM, Tirumalasetti F. The effects of pain and depression on physical functioning in elderly residents of a continuing care retirement community. Pain Med 2008; 1: 340–350. 3 Chu H, Yang CY, Lin Y et al. The impact of group music ther- apy on depression and cognition in elderly persons with dementia: a randomized controlled study. Biol Res Nurs 2014; 16: 209–217. 4 Banks M, Banks W. The effects of animal-assisted therapy on loneliness in an elderly population in long-term care facilities. J Gerontol 2002; 57: 428–432. 5 Cherniack E, Cherniack A. The benefits of pets and animal- assisted therapy to the health of older individuals. Curr Geron- tol Geriatr Res 2014: 623203. 6 Cole KM, Gawlinski A, Steers N, Koylerman J. Animal-assisted therapy in patients hospitalized with heart failure. Am J Crit Care 2007; 16: 575–585. 7 Maiorani C, Cesa-Bianchi G, Crostini C. Un’esperienza di “Dog assisted therapy” in RSA. J Gerontol Geriatr 2012; 60: 51–55. 8 Moretti F, De Ronchi D, Bernabei V et al. Pet therapy in elderly patients with mental illness. Psychogeriatrics 2011; 11: 125–129. 9 Stasi M, Amati D, Costa C et al. Pet therapy: a trial for institu- tionalized frail elderly patients. Arch Gerontol Geriatr 2004; 38: 407–412. 10 Kawamura N, Niiyama M, Niiyama H. Long term evaluation of animal-assisted therapy for institutionalized elderly people: a preliminary result. Psychogeriatrics 2007; 7: 8–13. 11 Phelps K, Miltenberger G et al. An investigation of the effects of dog visits on depression, mood, and social interaction in elderly individuals living in a nursing home. Behav Interventions 2008; 23: 181–200. 12 Regine Lombardia. Bollettino Ufficiale Serie Ordinaria n.16, 2016, Interventi Assistiti con gli Animali (I.A.A.). Linee Guida Nazionali. D.g.r. 2016-nX/5059: 154. 13 Sheikh JA, Yesavage JA. Geriatric Depression Scale (GDS): Recent evidence and development of a shorter version. In: Brink TL, ed. Clinical Gerontology: A Guide to Assessment and Intervention. New York: Haworth Press, 1986. 14 Folstein MF, Folstein SE, McHugh PR. ‘Mini Mental State’. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975; 12: 189–198. 15 Spitzer RL, Kroenke K, Williams JB et al. A brief measure for assessing generalized anxiety disorder: the GAD-7. Arch Intern Med 2006; 166: 1092–1097. 16 Watson D, Clark LA, Tellegen A. Development and validation of brief measures of positive and negative affect: the PANAS scales. J Pers Soc Psychol 1988; 54: 1063–1070. 17 Moss-Morris R, Weinman J, Petrie K, Horne R, Cameron DL, Buick D. The Revised Illness Perception Questionnaire (IPQ-R). Psychol Health 2002; 17: 1–16. 18 Williamson A, Hoggart B. A review of three commonly used pain rating scales. J Clin Nurs 2005; 14: 798–804. 19 Fortune DG, Richards HL, Main CJ, Griffiths CE. Pathological worrying, illness perception and disease severity in patients with psoriasis. Br J Health Psychol 2000; 5: 71–82. 20 Scharloo M, Kaptein AA, Weinman JA, Willems LN, Rooijmans HG. Physical and psychological correlates of func- tioning in patients with chronic obstructive pulmonary disease. J Asthma 2000; 37: 17–29. 21 Weinman J, Petrie KJ, Sharpe N, Walker S. Causal attribution in patients and spouses following first time myocardial infarc- tion and subsequent lifestyle changes. Br J Health Psychol 2000; 5: 263–273. 22 Creagan E. Pet prescription: not just puppy love. J Active Aging 2003; 2: 46. 23 Cornwell EY, Waite LJ. Social disconnectedness, perceived isolation and health among older adults. J Health Soc Behav 2009; 50: 31–48. 24 Onor ML, Trevisiol M, Spano M, Aguglia E, Paradiso S. Alexthy- mia and ageing. J Nerv Ment Dis 2010; 198: 891–895. C. Ambrosi et al. © 2018 Japanese Psychogeriatric Society64 Randomized controlled study on the effectiveness of animal-assisted therapy on depression, anxiety, and illness perception... INTRODUCTION METHODS Study population Assessments DAT Statistical analyses RESULTS Effectiveness of DAT in the treatment group GDS-15, PANAS, and GAD-7 NPRS IPQ-R Satisfaction Comparison between the treatment and control groups Comparison before DAT Comparisons of the changes in each scale Analysis: social interaction DISCUSSION ACKNOWLEDGMENTS REFERENCES