Supplementary Table 2.

Summary of Studies Used to Address Question 2, Presented According to Strategy and Study Type (Observational or Intervention)

1. Manage Pain
Observational studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Study Design	Exposure	Social Outcome	Study Finding
Almenkerk, 2015	The Netherlands	Nursing home residents with chronic stroke (N=274)	None specified	Cross-sectional	Pain, using Resident Assessment Instrument- Minimum Data (RAI-MDS)	Social engagement, using RAI-MDS Revised Index for Social Engagement (RISE)	Substantial pain was associated with low social engagement (OR 4.25, 95% CI 1.72-10.53; P < .05), but only in residents with no/mild or severe cognitive impairment; this relation disappeared adjusted for Neuropsychiatric Inventory Questionnaire score (OR 1.95, 95% CI 0.71-5.39)
Klapwijk, 2016	The Netherlands	Nursing home residents with dementia (N=288)	Inclusion: moderate to very severe dementia, using the Reisberg Global Deterioration Scale (Reisberg GDS) 5-7	Cross-sectional	Pain, using the Pain Assessment Checklist for Seniors with Limited Ability to Communicate (PACSLAC-D)	Social relations, using the QUALIDEM Social isolation, using the QUALIDEM	In unadjusted analysis, pain was associated with social relations (OR 0.88, 95% CI 0.83-0.94; P < .01) and social isolation (OR 0.88, 95% CI 0.82-0.94; P < .01). Associations were no longer statistically significant in multivariable analysis.
Lai, 2015∗	Hong Kong	Nursing home residents (N=125)	None specified	Cross-sectional	Pain	Social relationships, using the WHOQOL-BREF	Pain associated with lower social relationships score (P < .001)
Lood, 2017	Sweden	Nursing home residents (N=4451)	None specified	Cross-sectional	Pain, using the Pain Assessment in Advanced Dementia Scale	Social engagement, using a list of study-specific items on participation (eg, going on an outing/excursion, having everyday conversations with staff not related to care)	Pain was correlated with less participation in social occupations (P < .01); however, it was no longer statistically significant in the adjusted model
Tse, 2013	Hong Kong	Nursing home residents (N=535)	Exclusion: mental disorder or cognitive impairment	Cross-sectional	Pain, using an 11-point numeric rating scale (NRS)	Loneliness, using the UCLA Loneliness Scale	In unadjusted analysis, pain was not associated with loneliness (P = .557).
Tse, 2012	Hong Kong	Nursing home residents (N=302)	None specified	Cross-sectional	Pain, using the Geriatric Pain Assessment	Loneliness, using the UCLA Loneliness Scale	In unadjusted analysis, pain associated with higher loneliness (P = .05).
Van Kooten, 2017	The Netherlands	Nursing home residents (N=199)	Inclusion: diagnosis of dementia Exclusion: Parkinson disease dementia, alcohol-related dementia, cognitive deficits due to psychiatric disorders	Cross-sectional	Pain, using the Mobilization Observation Behavior Intensity Dementia (MOBID-2) Pain Scale	Social relations, using the QUALIDEM	The association between pain and social relations was not statistically significant for mild (P = .25) or moderate-severe pain (P = .25)
Won, 2006	United States	Nursing home residents with persistent pain (N=10,372)	Exclusion: moderate to severe cognitive impairment based on a Cognitive Performance Scale (CPS) score of >2 (equivalent of <19 in MMSE)	Cohort	Analgesic use, standing long-acting opioids (vs standing-acting opioids; standing nonopioids; and no analgesics)	Social engagement, using RAI-MDS Index of Social Engagement	Standing long-acting opioids (vs standing nonopioids) were associated with improvements in social engagement (propensity adjusted rate ratio 1.60; 95% CI, 1.02-2.48)

Intervention studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Chibnall, 2005	United States	Nursing home residents with moderate-to-severe dementia (N=25)	Inclusion: moderate-to-severe dementia indicated by a stage 5 or 6 on the Functional Assessment Staging (FAST)	Yes	Randomized controlled trial, crossover	Analgesic medication, 4 weeks of acetaminophen (3000 mg/d) (vs placebo)	Social interaction (direct and passive social involvement), using Dementia Care Mapping (DCM) Social withdrawal, using DCM	Acetaminophen intervention group exhibited significant increases in direct social interaction (P = .05) and passive social involvement (P = .006)
Husebo, 2019	Norway	Nursing home residents (N=723)	None	Nursing homes randomized	Cluster-randomized controlled trial	Staff education and training on communication, systematic pain management, medication review, and activities (vs usual care)	Social relations, using the QUALIDEM Social isolation, using the QUALIDEM	During the follow-up (month 4-9), there was an intervention effect for social relations (P < .05)
Tse, 2012	China	Nursing home staff (N=147) and residents (N=535)	Exclusion: cognitive impairment and history of mental disorders	Nursing homes randomized	Cluster-randomized controlled trial	Integrated pain management program including a physical exercise program and multisensory stimulation art and craft therapy, 1 h/wk for 8 wk (vs usual care)	Loneliness, using the Chinese version of Revised UCLA Loneliness Scale	Intervention group showed significantly lower loneliness after the program (P < .001). There was no change in the control group.
Tse, 2013	China	Nursing home staff (n=60) and residents (n=90)	Inclusion: oriented to time and place	Nursing homes randomized	Pretest-posttest (2 groups)	Integrated pain management program that included garden therapy and physiotherapy exercise for the residents, 1 h/wk for 8 wk (vs usual care)	Loneliness, using the Chinese version of Revised UCLA Loneliness Scale	Intervention group showed significant improvement in loneliness after the program (P < .05) but not in the control group
Tse, 2016	China	Nursing home residents (N=50)	Inclusion: score ≥6 in the Abbreviated Mental Test. Exclusion: cognitive impairment or mental disorders	Nursing homes randomized	Pretest-posttest (2 groups)	Group-based pain management program that included physical exercise, interactive teaching and sharing of pain management education, 1 h twice per wk for 8 wk (vs usual care)	Loneliness, using the Chinese version of Loneliness Scale	Loneliness decreased in both intervention and control groups; no significant difference in loneliness between the 2 groups at baseline or week 12

2. Address Vision and Hearing Impairments
Observational studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Study Design	Exposure	Social Outcome	Study Finding
Achterberg, 2003	The Netherlands	Newly admitted nursing home residents (N=562)	None specified	Cross-sectional	Vision impairment, using the Resident Assessment Instrument–Minimum Data Set 2.0 (RAl-MDS) Hearing impairment, using RAI-MDS	Social engagement, using RAI-MDS Index of Social Engagement	Vision impairment associated with low social engagement (OR 1.7, 95% CI 1.1-2.5; P = .011) but not hearing impairment (OR 1.0, 95% CI 0.7-1.6; P = .85)
Bliss, 2017∗	United States	New nursing home residents followed to 1 y (N=15,927)	None specified	Cohort	Vision impairment, using RAI-MDS	Social engagement, using RAI-MDS Index of Social Engagement 1 y after admission	Vision impairment associated with lower social engagement at 1-y follow-up (P < .001)
Branco, 2007∗	United States	African American and white nursing home residents (N = 1667)	None specified	Cross-sectional	Vision impairment, using RAI-MDS Hearing impairment, using RAI-MDS	Social engagement, using RAI-MDS Index of Social Engagement	Stratified by racial/ethnic group: impaired vision was associated with lower social engagement among whites (P < .001) but not African Americans; the associations with hearing impairment were not statistically significant
Guthrie, 2018	Canada	Long-term care (LTC) residents (N= 110,578)	None specified	Cross-sectional	Vision impairment, using RAI-MDS Hearing impairment, using RAI-MDS Dual sensory impairment, using RAI-MDS Deafblind Severity Index (DbSI)	Social engagement, using RAI-MDS Index of Social Engagement	Residents with cognitive impairment and dual sensory impairment (DSI) experienced the lowest rates, based on the raw proportions, on 5 of 6 Index of Social Engagement items
Kang, 2012	United States	Nursing home residents with dementia (N=153)	Inclusion: diagnosis of Alzheimer's disease or other dementia, as recorded in their medical charts	Cross-sectional	Vision impairment, using RAI-MDS Hearing impairment, using RAI-MDS	Social engagement, using the MDS-NH section F1e Sense of Involvement/Initiative	Vision impairment inversely associated with social engagement (P = .039); the association with hearing impairment was not statistically significant.
Li, 2014∗	United States	Nursing home residents (N = 868,011)	None specified	Cross-sectional	Vision impairment, using RAI-MDS Hearing impairment, using RAI-MDS	Social engagement, using RAI-MDS Index of Social Engagement	Results suggest, when stratified by racial/ethnic group, highly or severely impaired vision and more than minimal difficulty hearing had lower social engagement for all groups
Owsley, 2007	United States	Nursing home residents with cataracts (N=45)	Exclusion: moderate or severe cognitive impairment (MMSE score < 14)	Cohort	Cataract surgery	Social interaction, using the Nursing Home Vision-Targeted Health-Related Quality of Life Questionnaire (NHVQoL)	Cataract surgery group exhibited significant score improvements in social interaction (P = .033)
Resnick, 1997∗	United States	Nursing home residents (N=18,873)	None specified	Cross-sectional	Vision impairment, using RAI-MDS Hearing impairment, using RAI-MDS	Social engagement, using RAI-MDS Index of Social Engagement	Vision: minimal (OR 1.19, 95% CI 1.10-1.29), moderate (OR 1.40, 95% CI 1.19-1.63), and severe vision impairment (OR 1.51, 95% CI 1.23-1.86) were all associated with low social engagement. Hearing: Only severe hearing impairment (OR 1.42, 95% CI 1.10-1.83) was associated with low social engagement.

Intervention studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Owsley, 2007	United States	Nursing home residents (>55 y old) with uncorrected refractive error (N=142)	Exclusion: moderate or severe cognitive impairment (MMSE score < 14)	Yes	Randomized controlled trial	Immediate refractive error correction (vs delayed correction)	Social interaction, using the Nursing Home Vision-Targeted Health-Related Quality of Life Questionnaire (NHVQoL)	At follow-up, refractive error group exhibited higher social interaction (P = .03)

3. Sleep at Night, Not During the Day
Observational studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Study Design	Exposure	Social Outcome	Study Finding
Garms-Homolovà, 2010	Germany	Nursing home residents (N=2577)	None specified	Cross-sectional	Sleep disturbances, “sleepless, has difficulty falling asleep or staying asleep” (insomnia) and “non-restful sleep/tired in the morning” (NRS), using RAI-MDS	Social engagement, using RAI-MDS Index of Social Engagement (ISE)	Compared to those with no sleep disturbances, those with pronounced sleep disturbances had lower social engagement (P < .01)
Lai, 2015∗	Hong Kong	Nursing home residents (N=125)	None specified	Cross-sectional	Sleep difficulty	Social relationships, using the WHOQOL-BREF	Sleep difficulty was not associated with social relationships score

Intervention Studies
First Author, year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Alessi, 2005	United States	Nursing home residents with daytime sleepiness and nighttime sleep disruptions (N=118)	None specified	Yes	Randomized controlled trial	Multiple nonpharmacologic efforts to improve sleep, ie, decreased daytime in-bed time, 30 min of outdoor sunlight exposure, increased physical activity, and structured bedtime routine (vs usual care)	Social engagement, using observations of participation in social activities and calculated as percentage of observations per day	Intervention group exhibited significant increases in social engagement (P < .001)

4. Find Opportunities for Creative Expression, Like Art, Music, and Storytelling
Intervention Studies
First Author, year	Country	Population (N=)	Inclusion/Exclusion Related To Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Boersma, 2018	Netherlands	Residents with dementia (n=141) and their professional caregivers (n=136)	Inclusion: cognitive problems due to dementia	No	Pretest-posttest (2 groups)	Veder contact method, a person-centered method using theatrical, poetic, and musical communication for application in 24-h care that encourages social interaction (vs usual care)	Social relations, using the QUALIDEM Social isolation, using the QUALIDEM	Implementation of VCM led to significant positive improvements in the residents' social relations (P = .002). The association with social isolation was not statistically significant
Fritsch, 2009	United States	Nursing home residents with dementia and staff (2088 ten-minute observation periods that were conducted in 20 nursing homes)	Inclusion: dementia	Nursing homes randomized	Post only (2 groups)	TimeSlips program, a group storytelling program that encourages creative expression among persons with dementia, 1 h/wk for 10 wk (vs usual care)	Social engagement, using 10-min, coded observations	There were higher levels of social engagement among residents in TimeSlips homes (P = .003)
Roswiyani, 2019∗	Indonesia	Older adult nursing home residents (N=267)	Exclusion: moderate or severe cognitive impairment (MMSE score < 18)	Yes	Randomized controlled trial (4 groups)	Art activities + qigong exercise (intervention integration); art activities only; qigong only, 90 min twice per week for 8 wk (vs control)	Social relationships, using the WHOQOL-BREF	There was an increase in social relationships in the intervention integration and art groups (but the increase was larger in the art group); the comparison of the art group to the control group was statistically significant (P = .019)
Van Dijk, 2012	Netherlands	Nursing home residents (N=169)	Inclusion: diagnosed dementia	No	Pretest-posttest (3 groups)	Veder Method for group theater living-room activities with trained professional caregivers (group 1) or professional actors (group 2) [vs regular reminiscence group activity (group 3)]	Social relations, using the QUALIDEM Social isolation, using the QUALIDEM	At post-test, group 2 showed less socially isolated behavior (P = .04); no difference was observed in social relations
Weiss, 1989	United States	Nursing home residents (N=49)	None specified	No	Pretest-posttest (2 groups)	Textile art classes, 1 h 3 times per week for 8 wk (vs control)	Social network, using number of other residents in the nursing home a subject reported knowing by name Social interaction	Quality of social interaction was significantly higher in posttest (P = .01) No difference was observed in social network (P = .14)

5. Exercise
Observational Studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Study Design	Exposure	Social Outcome	Study Finding
Vitorino, 2012	Brazil	Long-stay care facility residents (N=77)	None specified	Cross-sectional	Physical activity, yes or no	Social relationships, using the WHOQOL-BREF Social participation, using the WHOQOL-OLD	Physical activity was not associated with social relationships (P = .561)
Wójcik, 2017	Poland	Nursing home residents (N=58)	None specified	Cross-sectional	Participation in rehabilitation and satisfaction with its progress, via questionnaire	Social relationships, using the WHOQOL-BREF	The association between participation in rehabilitation and quality of life was not statistically significant

Intervention Studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Barthalos, 2016	Hungary	Nursing home residents (N=45)	Exclusion: moderate or severe cognitive impairment (MMSE score < 15)	No	Pretest-posttest (3 groups)	Physical activity, resistance training 45 min twice per week vs physical + mental activity (weekly lectures and discussions on aging and quality of life) vs control (no physical or mental training)	Social participation, using the WHOQOL-OLD	Both physical activity (P = .004) and physical + mental activity (P = .004) groups improved in social participation
Castilho-Weinert, 2014	Brazil	Nursing home residents (N=43)	None specified	No	Pretest, postest (1 group)	Physical therapy program, recreational dynamic activities and psychomotor circuits, 30 min/wk for 16 wk	Social relationships, using the WHOQOL-BREF	There was no change in social relations (P = .384)
Hsu, 2016	Taiwan	Long-term care residents (N=60)	Exclusion: cognitive impairment (MMSE score < 25)	Yes	Randomized controlled trial	Seated tai chi exercise, 40 min 3 times per week for 26 wk (vs usual activity control group)	Social relationships, using the WHOQOL-BREF	Seated tai chi intervention improved social relations (P < .005)
Lee, 2010	China	Nursing home residents (N=139)	Inclusion: intact cognitive function (abbreviated mental test score >6)	No	Pretest-posttest (2 groups)	Tai chi exercise, 1 h 3 times per week for 26 wk (vs control group)	Social support (network and satisfaction), using the Chinese version of the Social Support Questionnaire–Short Form (SSQ6)	No significant changes were detected regarding the effect of the tai chi program on social support
Roswiyani, 2019∗	Indonesia	Nursing home residents (N=267)	Exclusion: moderate or severe cognitive impairment (MMSE score < 18)	Yes	Randomized controlled trial (4 groups)	Art activities + qigong exercise (intervention integration); art activities only; qigong only, 90 min twice per week for 8 wk (vs control)	Social relationships, using the WHOQOL-BREF	There was an increase in social relationships in the intervention integration and art groups (but the increase was larger in the art group); the comparison of the art group to the control group was statistically significant (P = .019)
Tse, 2014	China	Nursing home residents with chronic pain (N=396)	Exclusion: cognitive impairment	Nursing homes randomized	Pretest-posttest (2 groups)	Physical exercise program, consisting of muscle strengthening, stretching, and massages, 1-h/wk for 8 wk (vs no treatment control)	Loneliness, using the Chinese version of UCLA Loneliness Scale	The intervention group showed significant decrease in loneliness (P < .05) and the control group did not show any significant improvement

6. Maintain Religious Observations
Observational Studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Study Design	Exposure	Social Outcome	Study Finding
Bliss, 2017∗	United States	New nursing home residents followed to 1 y (N=15,927)	None specified	Cohort	Spirituality, using RAI-MDS	Social engagement, using RAI-MDS Index of Social Engagement 1 y after admission	Spirituality not associated with social engagement at 1-y follow-up (P = .06)
Branco, 2007∗	United States	African American (n = 172) and white (n = 1595) nursing home residents	None specified	Cross-sectional	Religious activities, using RAI-MDS Strength from faith, using RAI-MDS	Social engagement, using RAI-MDS Index of Social Engagement (ISE)	Among both African American and white residents, strength from faith (P < .01) and religious activity preference (P < .001) were positively associated with social engagement
Koenig, 1997	United States	Nursing home residents (N=115)	None specified	Cross-sectional	Religious coping, using the Religious Coping Index (RCI)	Social support, using frequency of visitors, frequency of other contacts, intimacy with staff, and intimacy with visitors	Religious coping was positively associated with social support (P = .01)

7. Garden, Either Indoors or Outside
Intervention studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Brown, 2004	United States	Nursing home residents (N=66)	Inclusion: could cognitively comprehend and answer questions	No	Pretest-posttest (2 groups)	Indoor gardening program once per week for 5 weeks (vs twice a week for 2 wk)	Loneliness, using the UCLA Loneliness Scale Social support, using the revised Social Provisions Scale	There were no significant differences in social support or loneliness between participant groups
Chen, 2015	Taiwan	Nursing home residents (N=10)	Exclusion: diagnosed cognitive impairment	No	Pretest-posttest (1 group)	Indoor horticultural program once per week for 10 wk	Loneliness, using the UCLA Loneliness Scale, Version 3	Loneliness decreased from baseline to follow-up at weeks 5 and 10 (P < .001)
Chu, 2019	Taiwan	Nursing home residents (N=150)	Exclusion: cognitive impairment (MMSE score < 25)	Yes	Randomized controlled trial	Horticultural program for 8 wk (vs usual care)	Loneliness, using the 20-item UCLA Loneliness Scale, Version 3	Loneliness decreased over time in the experimental group (P < .001), but increased in the control group (P < .001)
Lai, 2018	Hong Kong	Frail and prefrail nursing home residents (N=111)	Inclusion: normal cognition (Chinese Abbreviated Mental Test score >5) or mild cognitive impairment (questionable or mild dementia according to the Clinical Dementia Rating scale)	Yes	Randomized controlled trial	Horticulture program for 1 h/wk for 8 wk (vs social activities)	Social engagement, using the Index of Social Engagement Social network, using the Lubben Social Network Scale	There was no statistically significant difference over time, in social outcomes, between the groups
Tse, 2010	Hong Kong	Nursing home residents (N=53)	Inclusion: cognitively intact	Nursing homes randomized	Pretest-posttest (2 groups)	Indoor gardening program for 8 wk (vs usual care)	Loneliness, using the Revised UCLA Loneliness Scale Social network, using the Lubben Social Network Scale	There were significant increases in social networks (P < .01) and reductions in loneliness (P < .01) for the experimental groups but not the control groups (P > .05)

8. Visit With Pets
Observational Studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Study Design	Exposure	Social Outcome	Study Finding
Calvert, 1989	United States	Nursing home residents (N=65)	Inclusion: pass mental screening (correctly answering 3 items from the Pfieffer's Short Portable Mental Status Questionnaire)	Cross-sectional	Pet interaction (in pet programs), categorized into high vs low pet interaction groups	Loneliness, using the UCLA Loneliness Scale	Those in the high pet interaction group were statistically and significantly less lonely than those in the low pet interaction group (P = .03)

Intervention Studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Banks, 2002	United States	Long-term care residents (N=45)	Exclusion: cognitive impairment, diagnosed or MMSE score < 24	Yes	Randomized controlled trial	Animal assisted (AAT) once per week vs AAT 3 times per week vs no AAT	Loneliness, using the UCLA Loneliness Scale	Both AAT intervention groups showed lower loneliness than control group (P < .05) both the 2 AAT groups did not differ from each other
Banks, 2005	United States	Long-term care residents (N=37)	Exclusion: cognitive impairment (MMSE score < 24)	Yes	Randomized controlled trial	Animal-assisted therapy (AAT) group vs AAT individual, all 30-min sessions once per week	Loneliness, using the UCLA Loneliness Scale	Loneliness decreased for AAT individual (P < .05) but the difference was not statistically significant for AAT group. Posttest scores did not differ between groups
Bernstein, 2000	United States	Long-term care residents (N=33)	None specified	No		Animal-assisted therapy (AAT) vs arts and crafts and AAT vs snack bingo	Social interaction, using observation (brief conversation, long conversation, touch)	Cognitively alert patients in AAT groups showed more brief conversation (P < .01) and long conversation (P < .01) but less touch. They also initiated brief conversation more frequently (P = .009) Semialert/nonalert patients in AAT showed less brief conversation, but more long conversation
Martindale, 2008	United States	Nursing home residents (N=20)	Included.	No	Pretest-posttest (2 groups)	Animal-assisted therapy, five 1-h sessions over 6 wk (vs traditional recreation therapy activities)	Social interaction, using observation of interacting with people and the Passivity in Dementia Scale	Interacting with people was significantly greater for the AAT group (P = .032)
Phelps, 2008	United States	Nursing home resident (N=5)	Exclusion: diagnosed dementia; cognitive impairment (MMSE score < 24)	No	Multiple baseline design	Dog visits, 5-10 min once per week for 6 wk	Social interaction, using observational frequency of verbal and nonverbal interaction with other residents and with the dog	Dog visits had no significant effect on social interaction
Richeson, 2003	United States	Nursing home residents with dementia (N=15)	Inclusion: diagnosed dementia; moderate or severe cognitive impairment (MMSE score < 16)	No	Pretest-posttest (1 group)	Animal-assisted therapy, 1 h 5 d per week for 3 wk	Social interaction, using a data collection tool to determine if social interactions increase after interactions with therapy dogs and their handlers	Social interaction increased pretest to posttest (P < .05)
Sollami, 2017	Italy	Nursing home residents (N=28)	Inclusion: mild or absent cognitive impairment (as assessed by MMSE)	Yes	Pretest-posttest (2 groups)	Animal-assisted intervention, 1 h 2 times per week for 16 sessions (vs control, usual care)	Loneliness, using the UCLA Loneliness Scale Social interaction, using the Quality of Life Scale in Late-Stage Dementia (QUALID)	Intervention group showed significantly decreased loneliness (P = .001) and improved positive social interactions (P = .001)
Vrbanac, 2013	Croatia	Nursing home residents (N=21)	None specified	No	Pretest-posttest (1 group)	Animal-assisted therapy, 90 min 3 times per week for 6 mo	Loneliness, using the UCLA Loneliness Scale	Loneliness decreased after animal-assisted therapy (P = .003)
Wallace, 1987	United States	Nursing home residents (N=8)	None specified	No		Pet visitation program, 15 min, 3 times per week for 8 wk (vs visits without dogs)	Social interaction, using a behavioral activity questionnaire	Only a significant effect of visitations (P < .01), indicating that visits, either with or without pets, increased social interaction
Wesenberg, 2019	Germany	Nursing home residents with mild to moderate dementia (N=19)	Inclusion: diagnosed Alzheimer's disease or vascular dementia	No	Pretest-posttest (2 groups)	Animal-assisted intervention with a dog, once per week for 6 mo (vs control intervention without dogs)	Social interaction, using observational frequency; divided into verbal interaction, touch, nonverbal interaction and body posture	During the animal-assisted intervention, significantly longer and more frequent periods of social interaction were observed than during the control intervention
Winkler, 1989	Australia	Nursing home residents (N=21)	None specified	No	Pretest-posttest (1 group)	Resident dog	Social interaction, using observation Sanson-Fisher behavioral observation instrument, including group behaviors	Six weeks after the dog's arrival, a significant increase in frequency of interactive behaviors was seen; by 22 wk, behaviors had reverted to baseline levels

Robotic Animals
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Banks, 2008	United States	Nursing home residents (N=38)	Exclusion: cognitive impairment (MMSE score < 24) or Alzheimer's disease	Yes	Randomized controlled trial (3 groups)	Weekly animal-assisted therapy (AAT) with a living dog or AAT with a robotic dog, 30 min/wk for 8 wk (vs control)	Loneliness, using the UCLA Loneliness Scale	AAT with either AIBO or a living dog resulted in similar improvements in loneliness when compared with control group (P < .05)
Robinson, 2013	New Zealand	Retirement home, hospital and rest home residents (N=40)	None specified	Yes	Randomized controlled trial (2 groups)	Activity sessions with robotic seal (PAIRO), 1 h, twice per week for 12 wk (vs control)	Loneliness, using the UCLA Loneliness Scale	Those in the intervention group decreased in loneliness over time, whereas those in the control group increased in loneliness; there was a significant difference between groups in loneliness change over time (P = .033)

9. Use Technology to Communicate
Intervention Studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Neves, 2018	Canada	Long-term care home residents (N=5)	Exclusion: dementia	No	Pretest-posttest (1 group), feasibility study	Accessible communication app	Social support, using the Abbreviated Duke Social Support Index Social interaction, using the Abbreviated Duke Social Support Index Loneliness, using the short revised UCLA Loneliness Scale	Increases in social support (P = .105) and social interaction (P = .097) were not statistically significant The association with loneliness was not statistically significant.
Siniscarco, 2017	United States	Long-term care facility residents (N=8)	Exclusion: cognitive impairment (MMSE score < 24)	No	Pretest-posttest (1 group)	Videoconferencing (1 or more times per week for 2 mo)	Loneliness (emotional), using DeJong Gierveld Loneliness Scale Social isolation, using PROMIS instruments Social support (emotional and informational), using PROMIS instruments	Emotional loneliness and social isolation decreased slightly, but not significantly Emotional support and informational support increased slightly, but not significantly.
Tsai, 2010	Taiwan	Nursing home residents (N=57)	Exclusion: moderate or severe cognitive impairment (MMSE score < 16 for those with no formal education or MMSE score < 20 for those with at least a primary school education)	Nursing homes randomized	Pretest-posttest (2 groups)	Videoconferencing (at least 5 min/wk for 3 mo) vs regular care only	Social support, using the Social Supportive Behavior Scale Loneliness, using the UCLA Loneliness Scale	Subjects in the experimental group had significantly higher mean emotional and appraisal social support scores at 1 wk and 3 mo after baseline (compared to control group). Subjects in the experimental group also had lower mean loneliness scores at 1 wk and 3 mo after baseline.
Tsai, 2011	Taiwan	Nursing home residents (N=90)	Exclusion: moderate or severe cognitive impairment (MMSE score < 16 for those with no formal education or MMSE score < 20 for those with at least a primary school education)	Nursing homes randomized	Pretest-posttest (2 groups)	Videoconferencing (at least 5 min/wk for 3 mo) vs regular care only	Social support, using the Social Supportive Behavior Scale and including emotional, informational, instrumental, and appraisal support Loneliness, using the UCLA Loneliness Scale	Videoconference program had a long-term effect in alleviating loneliness and improved long-term emotional social support and short-term appraisal support, but decreased residents' instrumental social support. There was no effect on informational social support.

10. Laugh Together
Intervention Studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Kuru-Alici, 2018	Turkey	Nursing home residents (N=50)	Exclusion: Alzheimer's disease or other dementia	No	Pretest-posttest (2 group)	Laughter therapy, 35-40 min twice per week for 5 wk (vs control, no intervention)	Loneliness (emotional and social), using the De Jong Gierveld Loneliness Scale	Intervention associated with decreased emotional and social loneliness with statistically significant difference from control group (P < .001)
Low, 2013	Australia	Nursing home residents (N=398)	None specified	Yes	Cluster randomized controlled trial	Humor therapy from professional performers (ElderClowns + Laughterbosses), for 2 h once per week for 9-12 wk (vs usual care)	Social disengagement, using the Multidimensional Observation Scale for Elderly Subjects (MOSES)	Groups did not differ significantly over time on social disengagement (P > .05)
Tse, 2010	China	Nursing home residents with chronic pain (N=70)	Inclusion: cognitively intact (indicated by a score ≤8 on the abbreviated mental test)	No	Pretest-posttest (2 group)	Humor therapy program, 1 h/wk for 8 wk (vs control)	Loneliness, using the revised UCLA Loneliness Scale	Intervention group showed significant decreases in loneliness (P < .001) but not for the control group; however, difference between groups was not statistically significant

11. Reminisce About Events, People, and Places
Intervention Studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Randomized (Yes/No)	Study Design	Intervention	Social Outcome	Study Finding
Chiang, 2010	Taiwan	Institutionalized residents (N=92)	Exclusion: moderate or severe cognitive impairment (MMSE score < 19)	Yes	Randomized controlled trial	Reminiscence therapy 90 min/wk for 8 wk (vs wait list control)	Loneliness, using the Revised UCLA Loneliness Scale	Intervention group showed a decrease in loneliness, with significant difference between groups (P < .001)
Lai, 2004	China	Nursing home residents with dementia (N=101)	Inclusion: dementia diagnosis	Yes	Randomized controlled trial	Individual life story book to encourage reminiscence once per week for 6 wk vs comparison (social contacts) vs control (no program)	Social engagement, using the Social Engagement Scale	There were no statistically significant differences in social engagement between the groups
Schafer, 1985	United States	Nursing home residents (N=185)	None specified	No	Pretest-posttest (4 groups)	Three intervention groups (1 h/wk for 12 wk): (1) audio tapes + structured group intervention; (2) structured group intervention, and (3) audio tapes + individual activity vs control (no treatment)	Social network, using the ratio of the number of other participants a subject knew divided by the total number of people Social engagement, using the spontaneous initiation of activities with other residents Social support, using whether resident reports there was someone in whom they could confide	There were statistically significant differences between the groups for social network (P = .02) and social engagement (P = .02) Group 2 associated with increased social network and the highest social engagement. The association with social support was not statistically significant.
Serrani-Azcurra, 2012	Argentina	Nursing home residents with dementia (N=135)	Inclusion: diagnosed with Alzheimer's disease and Folstein Mini Mental Exam Score above 10	Yes	Randomized controlled trial	Life-approach reminiscence therapy, 1 h biweekly for 12 wk vs active control (counseling and informal social contacts) vs passive control	Social engagement, using the Social Engagement Scale	Social engagement increased in the intervention group, with significant difference between groups (P < .01)
Siverova, 2014	Czech Republic	Hospitalized long-term care elderly patients (N=41)	Inclusion: mild or moderate cognitive impairment (9 < MMSE score < 24)	No	Pretest-posttest (1 group)	Narrative group reminiscence therapy, 40-60 min once per week for 6-8 wk	Social relationships, using the WHOQOL-BREF Social participation, using the WHOQOL-OLD	Intervention was not associated with change in social relationships (P = .63) but there was an increase in social participation (P = .002)
Siverova, 2018	Czech Republic	Older adults in institutional care (N=116)	Inclusion: mild or moderate cognitive impairment (10 < MMSE score < 24)	No	Pretest-posttest (2 groups)	Group narrative reminiscence therapy, 40-60 min/wk for 8 wk (vs standard care)	Social relationships, using the WHOQOL-BREF Social participation, using the WHOQOL-OLD	There were no statistically significant differences in social relationships. The intervention group showed an increase in social participation, with a significant difference with control group (P = .041).
Tabourne, 1995	United States	Nursing home residents (N=40)	Inclusion: diagnosis of Alzheimer's disease or other cognitive disorder	No	Pretest-posttest (2 groups)	Life review program, 2 sessions per week for 12 wk (vs control)	Social interaction, using observer ratings	There was a significant increase in social interaction for the experimental group (P < .001) but not for the control group; the pre-posttest differences between groups was statistically significant (P < .001)

12. Address Communication Impairments and Communicate Nonverbally
Observational Studies
First Author, Year	Country	Population (N=)	Inclusion/Exclusion Related to Cognition	Study Design	Exposure	Social Outcome	Study Finding
Ballard, 2001	England	Care facility (residential and nursing homes) residents (N=112)	Inclusion: dementia, using AGECAT (“organic disorder”) and the Clinical Dementia Rating Scale (CDR) category of 0.5 or greater	Cross-sectional	Language function, using Sheffield Screening Test for Acquired Language Disorders	Social withdrawal, using Dementia Care Mapping (DCM)	Greater impairment of receptive language was associated with increased social withdrawal (P = .03).
Bliss, 2017∗	United States	New nursing home residents followed to 1 y (N=15,927)	None specified	Cohort	Communication difficulty, using RAI-MDS	Social engagement, using RAI-MDS Index of Social Engagement 1 year after admission	Communication difficulty associated with low social engagement at 1-y follow-up (P < .001)
Li, 2014∗	United States	Nursing home residents (N= 868,011)	None specified	Cross-sectional	Communication difficulty, using RAI-MDS (assessing whether primary mode of expression was speech and the resident was able to make themselves understood by others)	Social engagement, using individual items from RAI-MDS Index of Social Engagement (ISE)	Communication difficulty associated with lower social engagement
Potkins, 2003	England	Nursing home and social care facility residents (N=315)	Inclusion: dementia, using AGECAT (“organic disorder”) and the Clinical Dementia Rating Scale (CDR) category of 0.5 or greater	Cross-sectional	Expressive and receptive language function, using the Sheffield Screening Test for Acquired Language Disorders	Social withdrawal, using Dementia Care Mapping (DCM) Social engagement, using participation in social activities and Dementia Care Mapping (DCM)	Both expressive (P = .04) and receptive aspects of language (P < .01) were correlated with decreased participation in social activities. Social withdrawal was only correlated with receptive language difficulties (P = .01).
Resnick, 1997∗	United States	Nursing home residents (N=18,873)	None specified	Cross-sectional	Communication difficulty, using RAI-MDS (assessing whether resident's primary mode of communication is defined and resident is able to be understood by others)	Social engagement, using individual items from RAI-MDS Index of Social Engagement (ISE)	Communication difficulty associated with low social engagement (OR 1.72, 95% CI 1.51-1.95)

^∗Study listed under more then one strategy.