Dog-assisted interventions and outcomes for older adults in residential long-term care facilities: A systematic review and meta-analysis

Objective: To comprehensively review studies on dog-assisted interventions (DAIs) among older people in residential long-term care facilities (RLTCFs) and to provide an overview of their interventions, outcomes and methodological quality. Method: We searched 18 electronic databases to identify English articles (published January 2000–December 2019) reporting on well-defined DAIs targeting older adults (≥65 years) in RLTCF. Data were extracted by two independent reviewers. Descriptive statistics were produced for quantitative studies, with key themes iden tified among qualitative studies. Where possible, estimates were pooled from ran domised controlled trials using random effects meta-analyses. Results: Forty-three relevant studies (39 quantitative; 4 qualitative)


| INTRODUC TI ON
Older adults living in residential long-term care facilities (RLTCFs) (Siegel et al., 2019), including nursing homes (Sanford et al., 2015), represent one of the largest high dependency care populations worldwide (Chatterji, Byles, Cutler, Seeman, & Verdes, 2015). In England, approximately 400,000 adults aged 65 years and over currently live in care homes (Care Quality Commission, 2018).
The impact of long-term conditions, multimorbidities and reduced opportunities for social contact in a closed environment often creates a complex range of needs and requires a comprehensive holistic approach (World Health Organization, 2015). Meeting these care needs is the responsibility of a great number of professionals, including RLTC staff and a range of visiting healthcare professionals. It is also important to improve partnership working between RLTCF and health care at individual, organisational and system levels to improve outcomes that matter most to residents and their relatives. Some research has started to explore the pivotal role of creating a culture in RLTCF that genuinely thinks about different and diverse ways of maximising resources and finding ways to support care so that it meets the needs of the care community (Killett et al., 2013).
Animal-assisted interventions (AAIs), which can include both animal-assisted activities (AAAs) and animal assisted therapy (AAT) (Society for Companion Animal Studies, 2019), have been identified as one complementary method of support that offers purposeful engagement and easy implementation as part of existing treatment programmes. Studies have shown that AAIs have a wide range of benefits on well-being (Bernabei et al., 2013), specifically in improving psychosocial and physiological functioning (Allen, Blascovich, & Mendes, 2002). Some studies have also highlighted the benefit of AAI in reducing stress, depression and compassion fatigue commonly experienced by carers of people with dementia (Coleman, 2016b;Islam, Baker, Huxley, Russell, & Dennis, 2017;Zimmerman et al., 2005). The research findings also hint at how AAI may facilitate connectedness between residents, relatives and staff in a RLTCF as an important part of good practice in addition to clinical effectiveness. Research and practice in RLTCF, however, needs to also include ways of valuing and supporting people working together, through interaction and shared activity that helps to develop meaningful relationships between people both inside and outside of the care community (Killett et al., 2013). Family carers can feel overwhelmed when their loved one moves into a RLTCF, and the care providers need to find ways evaluation is required for future research and practice in providing holistic care for older adults.

K E Y W O R D S
long-term care, quality of life, therapeutic nursing, well-being What does this research add to existing knowledge in gerontology?
• There is a paucity of high-quality empirical research on dog-assisted interventions (DAIs) in residential longterm care facilities (RLTCF) internationally and a lack of qualitative research that includes the experiences of older people themselves.
• Almost half (n = 18, 46%) of the 39 quantitative studies did not find any significant changes over time, or differences between experimental and control groups, among residents exposed to DAI.
• Twenty-one quantitative studies (54%) produced statistically significant findings on a range of benefits of DAI for residents, including improved social functioning, reduced depression and reduced loneliness.

What are the implications of this new knowledge for nursing care with older people?
• There is potential for older adults to benefit from the provision of DAI in RLTCF, yet the full extent of such benefits remains to be determined.
• RLTC providers should explore all avenues for providing high-quality, evidence-based care that is able to enhance the quality of life of residents through enriching personal and interpersonal relationships in their everyday experiences.

How could the findings be used to influence policy or practice or research or education?
• The findings support development of a standardised format for designing, implementing and evaluating DAI in RLTCF, which would promote stakeholder inclusion and consistent methodology to determine its benefits.
• This research contributes to the evidence base for developing further guidance on how to expand and improve the quality of RLTCF services more generally.
• The research could lead to further exploration of potential partnerships between providers of animal-assisted interventions, RLTC and local community pet owners working together to improve the lives of local residents.
to make them feel welcome and involved (Nolan, 2001), and involving animals is one way to bridge these relationships essential to maintaining people's identities.
This review expands on these previous AAI reviews by comprehensively reviewing studies reporting on DAI in RLTC populations with a range of care needs. We are also interested in documenting how DAI programmes have been evaluated and what are considered to be appropriate methods and measures to inform future research.
Hence, this systematic review aims to (a) describe the methods and outcome measures that have been used to measure the impact of DAI among older people in RLTCF; (b) synthesise the reported benefits of DAI among older people in RLTCF; and (c) assess the quality of existing empirical evidence on DAI for older people in RLTCF.

| ME THODS
This is a mixed-design systematic review which integrates quantitative and qualitative studies to examine intervention outcomes in conjunction with 'real life' experiences (Mays & Pope, 2000). The inclusion of qualitative studies offers a more holistic approach to our attempt to understand how, why and what effect DAI can have given some of the challenges in 'measuring' it. This also has the potential for exploring aspects of care outside of causality and positivistic factors. The protocol was registered in the PROSPERO registry prior to full commencement (CRD 42,018,098,799), and the review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Moher, Liberati, Tetzlaff, & Altman, 2009) and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines (Stroup et al., 2000).

| Study selection
Using Covidence's online review software (Covidence, 2017), article titles, abstracts and full-texts were independently reviewed by at least two researchers (S.S, S.O'F and T.H-L). Any disagreements over study inclusions were reviewed by a third reviewer (B.J) and resolved through discussion with the team. Applying the PICO format, we included any quantitative or qualitative study reporting on older adults (aged 65 years and over), including those identified as having high dependency needs (e.g. physical and cognitive impairment) (population); participating in any well-defined dog-assisted intervention delivered individually or by group through any means, regardless of duration and number of treatment sessions (intervention); while residing in any residential setting where older aged individuals have access to on-site care services or personal care (context); and measuring any change in psychosocial well-being related to the dog-assisted therapy or any psychosocial health outcome measure or instrument used to measure benefits or effectiveness of dogassisted therapies (outcomes). Case studies, study protocols, studies with non-living dogs as the primary intervention (e.g. robotic or synthetic) and other animal-assisted interventions were excluded.
We did not exclude specific clinical populations or studies based on sample size, given the limited study inclusions of previous reviews (Bernabei et al., 2013), and due to our inclusion of qualitative studies.

| Data extraction and quality assessment
Using a piloted and standardised form, two reviewers (S.S and B.J.) extracted study information in duplicate, including study design, country, sample size, setting (as described by authors), main type of disorder or health problems among population, study aims, primary outcome instrument/measure, recruitment strategy, response rate/attrition (%), demographics (mean age, gender, ethnicity), intervention descriptions, effect estimates and followup intervals. For qualitative studies, we followed the guidelines laid out by Thomas and Harden (2008) and extracted all themes and result sections relevant to each study. Missing data were requested from four authors and excluded if not received within one month (100% response rate).
In the light of the broad inclusion criteria, the overall quality of studies was evaluated using the mixed methods appraisal tool (MMAT) (Pluye et al., 2011). Qualitative and quantitative studies were assessed on four key areas: appropriateness of data collection (e.g. sample representativeness and sampling strategy), appropriateness of analysis and ascertainment method, appropriateness of study interpretations with clear relevance to practice, and adequate methodological reflexivity or response rate. Mixed design studies were assessed in three additional domains: the appropriateness of the overall design, adequate integration of qualitative and quantitative findings and appropriate consideration given to limitations. Each study received a maximum score of 4, with mixed design studies being assigned the lowest overall quality score of its study components (Pluye et al., 2011). Aligning with similar reviews, studies were classified into low-quality (<3 points) or moderate-/high-quality (3 ≥ points) studies.

| Data synthesis
For RCTs, we pooled estimates using fixed-effects (I 2 < 40%) or randomeffects meta-analyses (I 2 > 40%) when at least four studies were available with similar control groups and outcomes. Estimates were pooled using the inverse variance method, applying the DerSimonian-Laird estimator for the random-effects models (Borenstein, Hedges, Higgins, & Rothstein, 2011). As several studies used different outcome scales, we converted estimates into the Hedges' g, allowing for a common effect size (standardised mean difference; SMD, 0.2 to <0.5 = small, 0.5 to <0.8 = moderate, ≥0.8 = large effect) (Borenstein et al., 2011;Higgins et al., 2019). Estimates were selected based on the end point of each study's primary outcome. When studies were based on the same sample, we included only the most comprehensive study (e.g. higher quality and larger sample size) to minimise pooling of non-independent samples.
The between-study heterogeneity was evaluated using χ 2 test, the I 2 statistic (heterogeneity: 0%-40%=small, 30%-60%=moderate, >75%=considerable) (Borenstein et al., 2011) andprediction intervals (IntHout, Ioannidis, Rovers, &Goeman, 2016). We also investigated the impact of individual studies on the between-study heterogeneity by serially excluding each study from the overall estimate. Studies with significantly large effects in either direction were examined further and excluded in the overall analysis if deemed inappropriate for pooling. Subgroup analyses and examination of publication bias were not statistically possible due to the limited studies (Sterne, Gavaghan, & Egger, 2000). All analyses were conducted in R (R Core Team, 2013) (version 3.6.1) using the 'meta' package (Schwarzer, 2007).
For qualitative studies, we employed framework analysis (Gale, Heath, Cameron, Rashid, & Redwood, 2013) and followed the recommendations by Thomas and Harden (2008) to conduct thematic synthesis. This involved coding of text to develop 'descriptive themes', close to the original study and then the generation of 'analytical themes' where reviewers' interpretation attempts to generate new interpretive constructs, explanations or hypotheses.
Accordingly, two reviewers (T.H-L and S. O'F) coded line by line of the extracted qualitative data to form initial categories of each study's content. Codes were compared and refined in conjunction with a third reviewer (B.J) to achieve triangulation and then grouped into higher-order themes with relevance to our research question.
Any disagreements over study themes were resolved through discussion with the team.
The mean age of participants was reported in 30 studies, ranging from 55 to 88 years, and a pooled mean age of 83 years. Based on studies that reported gender frequencies, the majority of participants were female (71%). Quality assessment was conducted on the 39 quantitative studies using the MMAT assessment criteria (Table 2), and it was found that the vast majority of studies were considered low quality (n=26, 67%).

| Interventions
Dog-assisted interventions shared some common elements across studies. For example, the most common breed of dog were Retrievers (n = 14) and Labradors (n = 6), with the remaining being of multiple different breeds, ranging from small-to medium-sized dogs. Dog visits most frequently occurred one day per week (n = 17), for an intervention period of between 1 and 52 weeks (pooled mean of 13.8 weeks).
The duration of each visit ranged from 3 min (for individual one-on-one sessions) to 3 hr (for whole institution visits) across studies, with the most frequently used durations ranging from 30 to 90 min (n = 24). In most studies, the intervention group was compared to a control group (n = 28). The most frequently used control group was treatment as usual (n = 16), psychosocial group or social visits (n = 7) and interventions using robotic or plush toys (n = 4).

| Outcome measures and methods
The most common primary outcomes focused on reducing depression or low mood (n = 15), improving social functioning (n = 13) and improving overall cognitive functioning (n = 6). Some studies were more general and looked instead on resident's overall quality of life or the intervention's general effect on a range of health and social outcomes (n = 9). The most common ascertainment method was clinician or researcher interview (n = 18) using standardised tools such as the Mini-Mental State Exam (MMSE) (Folstein, Folstein, & McHugh, 1975), Geriatric Depression Scale (GDS) (Yesavage et al., 1982), Cohen Mansfield Agitation Inventory (CMAI) (Cohen-Mansfield & Billig, 1986) and the UCLA Loneliness Scale (Russell, 1996) (Table 1).
The results can be split into three groups based on study design:    and Olsen et al. (2016), respectively.
c Method of outcome ascertainment (e.g. self-report, observer ratings, clinical interview). Similarly, in the second group of studies that compared pretest to post-test results for the same group of older adults who had received DAI, 10 studies also found no significant changes. Among studies that did detect significant changes from pretest to post-test, DAI was found to reduce loneliness (Banks, 2005;Vrbanac et al., 2013); reduce agitation (Richeson, 2003;Sellers, 2006); improve social functioning (Sellers, 2006); and improve quality of life (Karefjard & Nordgren, 2018;   Overall, almost half of the quantitative studies (n=18, 46%) found no statistically significant changes over time or differences between groups across outcomes. Of those that did detect significant results, the main impacts for older adults as a result of the DAI included improved social functioning (n = 10), reduced depression (n = 6), and reduced loneliness (n = 5).

| Findings from qualitative synthesis
The four qualitative studies (

| Theme 1: Animals as effective transitional objects
Animals were often described as effective transitional objects to supplement missing interaction by either 'filling a void' or supplementing other human interactions and bonds (p.154) (Coleman, 2016a).
There were references to dogs providing a stimulus for conversation bled an opportunity for others to be able to reach the person on a cognitive level rather than simply responding in a reactionary manner to physical care needs (Swall et al., 2015). The potential offered through these developments to enhance and focus communication with the older people and towards a more person-centred approach to their care were posited as significant benefits and a means of addressing or rebalancing well-documented unequal power relations in institutions. Reminiscence was frequently reported as an outcome of DAI (Coleman, 2016a;Gundersen & Johannessen, 2018;McCullough, 2014;Swall et al., 2015), and the presence of a dog was reported to act as memory triggers and evoked feelings from 'time and places retold'. These memories could also be triggers for other memories that open up and are reflected upon in a coherent way (Swall et al., 2015). Swall et al.'s study of people with Alzheimers suggested that the dog provoked feelings of confidence and strength through its presence, and a means of promoting self-esteem where they acted to protect, care and take responsibility for the dog (Swall et al., 2015). These recollections and feelings could be negative as well as positive and occurred in the moment through senses and memories which served to enhance 'one's past and present existence' through an emotionally connected experience of living (p.19) (Swall et al., 2015).

| Theme 2: The value of pets as therapy and the nature of that therapeutic value
While many of the impacts of DAI described constituted 'naive' descriptions (p.21) (Swall et al., 2015), impacts such as reducing stress, spiritual connection, being in the moment and 'create a good moment' for those with affected cognitive function (Gundersen & Johannessen, 2018) were all cited as observed or perceived benefits for those who were not able to sustain other relationships. Sensatory comfort was described as significant in the absence of carers being able to meet some of the older person's unspoken needs such as personal loss and the need for physical comforting, particularly at the end of life (Coleman, 2016a). Existential perspectives on life and living (Swall et al., 2015) were attributed with the use of such words as 'love' (p.78) (McCullough, 2014) and 'communion' (p.1) (Swall et al., 2015) and 'harmony' (p.13) (Swall et al., 2015). These sensations provided a sense of release and tears (McCullough, 2014) and were observed to provoke heightened sensitivity in the dogs themselves. These were expressed through changed voice and body language as well as through facial expressions (Swall et al., 2015).
By focusing on the physical interaction (cuddling, touching, stroking) and the responsiveness of the dog and recipients of therapy, there were constant references which anthropomorphised the love that the dogs show and how they act it out and the independence of dogs in deciding who they approach. In addition to physical effects, several handlers described the emotional release that affection with their therapy dog can generate. This positioned the dog as a co-therapist. Participants' comments described the enduring connection of the human-animal bond, as well as the non-judgemental relationship that seemed to exist between therapy animals and the people they encounter. Participants also spoke about the seeming inherent ability for the animals to identify those individuals who need their attention the most (Coleman, 2016a). Two of the studies specifically addressed caregiver insights into the value of DAI (Coleman, 2016a;Gundersen & Johannessen, 2018;Swall et al., 2015) and the impact on their own roles and well-being. The need for psychosocial stimulation was specifically noted as a challenge to focus on in their otherwise busy day and the rewards where some were able to observe benefits such as calmer moods and behaviour (Coleman, 2016a;Gundersen & Johannessen, 2018;Swall et al., 2015). Two studies focused on the role of the dog handler (Gundersen & Johannessen, 2018;McCullough, 2014).
The studies reported some theorising about why and how DAI was of value-and reflected on how the methodologies enabled demonstration of this. Two domains of supportive behaviour were found to be commonly exhibited by the dogs during their visits: 'interest' and 'affection'. Swall et al. (2015) used a lifeworld approach and reflected on the use of phenomenological hermeneutics in which the researcher 'enters the hermeneutical circle with an ongoing movement between the parts and the whole in the text' (p.22) (Swall et al., 2015). They discussed how their structural analysis validated the naïve readings, and with the aim of the study in mind, the analysis moved back and forth to get a deeper understanding of the phenomenon to interpret the lived experience of the person with dementia in their encounters with a dog.

| Theme 3: The significance of the care environment and its stakeholders in facilitating dogassisted interventions
Purposive induction to the reasons, procedure and desired out- This could be passive such as nodding, smiling and giving encouragement as well as giving more active or directive encouragement by giving compliments or praise to the older person and the dog as interaction takes place, or elaborating on how the person could enhance the interaction and physically helping them to do this and also by contributing their own personal comments such as conveying good wishes for the person's health (McCullough, 2014).

F I G U R E 3
Effect of dog-assisted intervention (DAI) versus treatment as usual (TAU) on activities of daily living or physical functioning among older adults in residential long-term care facilities

Animals as effective transitional objects
The value of pets as therapy, and the nature of that therapeutic value Environmental conditions also included the provision of training and certification programmes from the DAI provider organisation and collaborative support from other dog handlers (Coleman, 2016a;Gundersen & Johannessen, 2018). Volunteers talked about the need for induction for volunteers on conditions such as dementia so that they could respond better to older residents behaviour (Gundersen & Johannessen, 2018). They also faced barriers if they had insufficient information about residents. All of the studies referred to controlling potential risk factors such as hygiene and allergens which were attended to, and none of these were seen to present any challenges.
Ethical issues were reflected upon such as the importance of reciprocity and mutually beneficial interaction for the dogs and need for ethical standards for the use of dogs in the field to ensure the dog's well-being during DAI and rest and recuperation after visits (McCullough, 2014). Also, in the research process itself, proxy consent was used when some participants were not able to consent either to the visit or to the observations. All of the studies had been given ethical approval.

| Summary of findings
This systematic review identified 43 peer-reviewed research articles examining the impact of DAI on older people living in RLTCF published between 2000 and 2018. The majority (~70%) of these were classified as low-quality studies according to the MMAT criteria.
The paucity of high-quality empirical research is surprising given the popularised use of DAI in RLTCF in many countries. Anecdotally, and perhaps quite obviously, a dog visiting an older person in a RLTCF is generally considered to be a good thing that makes people happy, and as such is often used as the go-to 'good news story' for local media (Oksman, 2015). Communities are now beginning to accept and formalise these assertions by developing protocols to support organisations considering working with dogs in care settings and allied health environments (Royal College of Nursing, 2018). A next logical step would be to identify how such protocols assist in the evaluation of DAIs as the challenge faced thus far, and has been illustrated by this review, is the variation in how DAI programmes are designed and delivered. Better quality evaluation may be possible when the practice becomes more formalised.
Almost half of the quantitative studies evaluating the impact of DAI for older people in RLTCF found no significant changes over time, or differences between experimental and control groups, in the outcomes measured. Most likely, this finding reflects the high proportion of low-quality studies in the field, coupled with the challenges associated with designing and conducting research in RLTCF (S. Hall, Longhurst, & Higginson, 2009;Lam et al., 2018). The remaining 21 quantitative studies did identify improved social functioning; reduced depression; and reduced loneliness as significant benefits of DAI for residents. In particular, the strongest impact of DAI seemed to be conferred through improved social functioning, observed both during the intervention and postintervention. It is also likely that  (Bowers, Esmond, & Jacobson, 2000).
In addition to identifying the benefits of DAI, this research was also interested in the design of DAI which typically involved a 30to 90-min visit from a small-to medium-sized dog accompanied by a handler once a week for a period of 13 weeks. Despite these commonalities, there was significant variation in how DAI is administered in RLTCF and how it has been evaluated. This leads to an unclear and somewhat patchy picture of how social care practice contributes to positive outcomes and how best practice models can be developed. We have not discussed the psychometric properties of any of the measurement tools used in the quantitative studies identified here. For example, the GDS was originally developed as a screening tool but has also been used as an outcome measure and the review has not been able to take account of these potential differences in relation to assessing the outcome of the interventions evaluated including the relevance and challenges in using these tools with peo-  & Lavender, 2015). Some of the themes from the review captured the unexplored potential of partnerships with DAI provider agencies and local community members coming in with their pets and that there may be wider benefits for those involved. These may reflect a more local approach dependent on community relationships and the move towards developing initiatives that facilitate person-centred care. However, as illustrated in this review, there is insufficient evidence to support a policy and commissioning response around AAT. While the evidence from the qualitative synthesis has demonstrated some positive impacts, much more work needs to be done to research and understand these impacts in such a way that any evidence can inform a more structured approach to commissioning AAI within RLTCF and to explore any specific therapeutic effects. Based on initial studies (Dayson & Bennett, 2016;Kimberlee, Jones, & Powell, 2013), there appear to be some synergies between older adults in RLTCF and patients who benefited through improvements in their quality of life and emotional well-being, mental and general well-being and levels of depression and anxiety (The Kings Fund, 2017).

| Strengths and limitations
Supporting community partnerships with RLTCF at individual, organisational and system levels may be key to achieving the outcomes that matter most to residents and their relatives and capitalise on the pivotal role of the RLTCF manager in creating a culture in homes that enables engagement and change (National Institute for Health Research, 2017).

CO N FLI C T O F I NTE R E S T
The authors confirm they have no conflict of interests to declare.

AUTH O R CO NTR I B UTI O N S
BJ was responsible for study conception and design, data collection and analysis, drafting and final approval of the manuscript and team coordination. SS was hired as an external contractor to assist with the design and conduct of the review, including data collection and analysis, initial drafting of the methods and results, critical revisions and final approval of the manuscript. THL was responsible for study conception and design, data collection and analysis, initial drafts of the introduction and discussion, critical revisions and final approval of the manuscript. SOFP was responsible for study conception, elements of data collection, critical revisions and final approval of the manuscript.