Abstract
Guided by the Convoy Model of Social Relations, this study examines the longitudinal association between homebound status and social isolation in the United States. Data were drawn from the National Health and Aging Trends Study (Rounds 1–9, N = 6,464). Participants were categorized as non-homebound, semi-homebound, or homebound. The level of social isolation was measured across four domains: marital status, religious participation, club participation, and family and friends. Mixed-effects regression models were used to examine the longitudinal associations between homebound status and social isolation over 9 years. In Round 1, both semi-homebound and homebound older adults reported higher levels of social isolation compared to non-homebound older adults. However, compared to non-homebound older adults, neither semi-homebound older adults nor homebound older adults differed significantly in their trajectories of social isolation. Findings are discussed through the lens of the Convoy Model of Social Relations.
Introduction
In 2022, there were approximately 55.8 million older adults in the United States. The number of older adults is expected to reach 78.3 million by 2040 and 88.8 million in 2060 (Administration for Community Living [ACL], 2024). In response to these unprecedented demographic changes, policies that support older adults remaining in their homes and communities have received increasing emphasis (Yarker et al., 2024). These policy priorities align with the preferences of most community-dwelling older adults, who prefer to live independently in the community. However, a substantial proportion of older adults is homebound or semi-homebound. Specifically, approximately 4.9% are homebound older adults, who report never or rarely having gone outside their home in a given month, and approximately 6.5% are semi-homebound older adults who can go outside but experience difficulties or need support to go outside (Choi et al., 2025a). Given the increasing number of older adults in the United States, the number of homebound older adults is expected to increase (Ornstein et al., 2020). About 2%–3% of community-dwelling older adults became homebound every year, and about 12.7% of non-homebound older adults became homebound over 7 years of follow-up (Ornstein et al., 2020).
Social isolation, defined as a lack of social contact due to limited interaction and restricted social relationships, is often associated with homebound status (Holt-Lunstad, 2021; Kawai et al., 2023). Even though social isolation is related to homebound status, the two do not necessarily coexist (Holt-Lunstad, 2021; Mitsutake et al., 2021; Sakurai et al., 2019). For example, Xiang et al. (2020b)explored trajectories of homebound status among older adults in the United States over 7 years, finding that about 8.3% of older adults were persistently homebound, and about 58.6% of individuals in this group experienced social isolation. In contrast, about 65% of older adults were never homebound, and only 17.8% of this group reported social isolation. Similarly, a study of older adults by Sakurai et al. (2019) in Tokyo examined the prevalence of social isolation and homebound status, finding that 18.4% of older adults were homebound but were not socially isolated. In the same study, 13.8% of older adults reported experiencing social isolation, but they were not homebound. According to this study, only 7.6% of older adults experienced both homebound status and social isolation, and the remaining 60.2% of older adults experienced neither condition.
Several studies using data from the National Health and Aging Trends Study (NHATS) reported associations between social isolation and being homebound (Ankuda et al., 2022; Cudjoe et al., 2022; Szanton et al., 2016; Xiang et al., 2020b). Social isolation may increase the likelihood of becoming homebound. For example, Cudjoe et al. (2022) found that over a 9-year period, socially isolated older adults were more likely to become homebound compared to socially connected older adults. On the other hand, homebound status may pose a risk of social isolation (Ankuda et al., 2022; Szanton et al., 2016). For example, a cross-sectional study by Szanton et al. (2016) found that both homebound and semi-homebound older adults participated less in various in-person social activities such as religious participation, club participation, and going out for joy compared to non-homebound older adults. Similarly, a cross-sectional study by Ankuda et al. (2022) documented that homebound older adults tended to have less contact with their family members and friends across various communication modalities—in-person, telephone, and video—compared to non-homebound older adults.
Despite established relationships between homebound status and social isolation in cross-sectional studies, a longitudinal association has not been explored. While Mehrabi et al. (2024) documented that frail older adults, who often experience homebound status, were more likely to experience social isolation, the extent to which homebound status is associated with trajectories of social isolation remains unexplored. In addition, while objective social isolation is a multidimensional concept, previous studies have rarely examined social isolation holistically, focusing primarily on contact frequency or social engagement alone (Ankuda et al., 2022; Szanton et al., 2016). Building on previous studies (Ankuda et al., 2022; Szanton et al., 2016), the present study examines the longitudinal association between homebound status and multidimensional social isolation over time using data from the NHATS, a nationally representative panel survey.
Factors Associated With Social Isolation in Later Life
Various sociodemographic, health-related, housing, and neighborhood features are associated with social isolation in later life (Cudjoe et al., 2020). Regarding sociodemographic and health-related variables, a systematic review by Wen et al. (2023) identified being aged 80 or over, having less than a high school education, not having a spouse, having functional limitations (e.g., activities of daily living [ADLs]), having chronic diseases, and cognitive decline as risk factors of social isolation in later life. Usage of mobility-assistive devices was negatively associated with social engagement (Latham-Mintus et al., 2022). Regarding income, a low household income level was associated with social isolation both from family members and friends (Chatters et al., 2018). Another study found that low income was a risk factor for social isolation in terms of living arrangement, identification of social relations who can talk about important things, religious attendance, and other social engagement (Cudjoe et al., 2020). The link between race/ethnicity and social isolation was mixed. For example, compared to African Americans, non-Hispanic White older adults were more likely to live without a partner and had limited social interaction with their congregational members. However, African American and non-Hispanic White older adults were no different in other aspects of social contact, such as maintaining contact with family members, maintaining contact with friends, and attending group activities in a neighborhood (Taylor et al., 2019). The subjective value placed on social engagement, including visiting friends or family members, attending religious events, attending club events, and going out for enjoyment, was also positively associated with social interaction among older adults (Latham & Clarke, 2018).
Environmental factors encompass physical aspects of housing, residential types, and neighborhood environment that affect social isolation. Regarding physical aspects of housing, features such as stairs and lack of ramps were risk factors for social isolation in later life (Clarke, 2014; Sixsmith & Sixsmith, 2008). Concerning residential types, about 25% of community-dwelling older adults experience social isolation (National Academies of Sciences, Engineering, and Medicine, 2020), and residing in long-term care facilities has been identified as a risk factor for social isolation in later life (Boamah et al., 2021; Simard & Volicer, 2020). Lastly, previous studies have shown that both objective and subjective neighborhood characteristics are linked with the level of social connections in later life (Finlay et al., 2023; Latham & Clarke, 2018; Moore et al., 2011; Yang & Moorman, 2021). For example, Latham and Clarke (2018) found that objective neighborhood characteristics such as the presence of physical disorder and subjective neighborhood characteristics such as a low level of neighborhood cohesion were associated with fewer meetings with families or friends and less participation in clubs or organizations. However, only the presence of physical disorder was associated with less going out for enjoyment. Neither physical disorder nor low levels of neighborhood cohesion were associated with religious attendance (Latham & Clarke, 2018).
Conceptual Framework
Guided by the Social Convoy Model of Social Relations, this study examines the extent to which homebound status is associated with the level of social connection in later life (Antonucci, 1986; Antonucci et al., 2019; Kahn & Antonucci, 1980). According to the Social Convoy Model of Social Relations, people's social relations are composed of a very close inner circle, a fairly close middle circle, and a role-based outer circle. While inner circle social networks, such as immediate family members, tend to be maintained over time, the fairly close middle circle, such as extended relatives, and the role-based outer circle, such as friends in clubs or in religious organizations, can be relatively easily changed by situational characteristics or may change over time (Antonucci, 1986; Antonucci et al., 2019; Kahn & Antonucci, 1980). This theory also posits that social relations are a multifaceted concept that includes structural, support, and quality domains, and the characteristics of social relations can change over time due to situational characteristics. For example, the degree of confinement to one's housing, a situational context, may influence the nature of the structural dimension of social relations such as contact frequency with family and friends and participation in church and club activities, as well as the support they receive from their social relations, including family members and friends, such as sharing important matters (Antonucci et al., 2019).
Present Study
Even though the number of homebound older adults is expected to increase, they remain an overlooked population (Ornstein et al., 2015; Qiu et al., 2010). Homebound older adults are particularly vulnerable to social isolation, a serious public health issue. Existing longitudinal studies have identified social isolation as a risk factor for being homebound (Cudjoe et al., 2022; Xiang et al. 2020b). However, while homebound status may pose a risk of social isolation, the longitudinal association between homebound status at baseline and social isolation has not been explored using a nationally representative sample of older adults. Furthermore, given their lower frequency of going out and greater functional limitations among semi-homebound older adults, though to a lesser degree than homebound older adults, they may be susceptible to social isolation similar to homebound older adults. However, to date, many previous studies have not differentiated between semi-homebound and homebound status (Cudjoe et al., 2022).
To fill a gap in the current literature, this study examines whether homebound status (non-homebound, semi-homebound, and homebound) is associated with changes in social isolation over 9 years using data from a nationally representative panel survey of older adults in the United States. Building on the Convoy Model of Social Relations and the previous studies on homebound status and social isolation (Cudjoe et al., 2022; Kahn & Antonucci, 1980; Szanton et al., 2016; Xiang et al. 2020b), this study proposes the following hypotheses.
Compared to non-homebound older adults, both semi-homebound and homebound older adults will report higher levels of social isolation at baseline.
Compared to non-homebound older adults, both semi-homebound and homebound older adults will report a steeper increase in social isolation over time as the two groups tend to experience poorer health and experience more functional limitations compared to non-homebound older adults.
Methods
Data and Sample
Data from the NHATS (Rounds 1–9, 2011–2019) were utilized to examine the association between homebound status and social isolation. The National Institute of Aging initiated the NHATS to investigate various issues among older adults in the United States (Freedman & Kasper, 2019). The NHATS started in 2011 and includes a nationally representative sample of Medicare beneficiaries aged 65 and older in the United States (Montaquila et al., 2012). The study participants were restricted to individuals who provided data on homebound status at Round 1 and data on social isolation for at least two rounds. Since questions related to social isolation were administered to both sample persons and proxy respondents, proxy respondents were included. A total of 7.46% of respondents were proxies at Round 1. The range of proxy respondents ranged from 12.58% to 14.88% between Rounds 2 and 9. The final analytic sample size was 6,464. On average, study participants completed six annual surveys, and there were 34,122 person-years of observation between 2011 and 2019.
Measures
Homebound Status
To determine participants’ homebound status, the following questions were utilized: “How often did you go out last month?,” “did anyone ever help you?,” “how often did you go outside by yourself?,” and “how much difficulty did you have when leaving your home by yourself?” Using responses to these questions, homebound status was categorized into three groups: homebound, semi-homebound, and non-homebound (Ornstein et al., 2015). Specifically, being homebound includes people who have never or rarely left their home in the last month. semi-homebound includes people who left their homes more than 2 times per week: (1) only with someone who helps them, or (2) have a helper but left home by themselves, or (3) without a helper but experience difficulty when going out. Lastly, non-homebound status includes people who left their home at least twice per week without any difficulty or a helper. Homebound status was defined based on participants’ Round 1 status (baseline).
Social Isolation
Based on Berkman and Syme’s Social Network Index (SNI), the level of social isolation is measured across four aspects: (1) marriage, (2) religious participation, (3) club participation, and (4) family and friends (Berkman & Syme, 1979; Pohl et al., 2017). Specifically, the following five questions were used to measure the level of social isolation across the four domains. (1) “Are you currently married, living with a partner, separated, widowed, or never married?” (marriage domain), (2) “In the past month, did you ever attend religious services?” (religious domain), (3) “In the past month, did you ever participate in clubs, classes, or other organized activities?” (club participation domain), (4) “In the last month, did you ever visit in-person with friends or family not living with you either at home or theirs?” (family and friend domain), and (5) “Looking back over the past year, who are the people you talked with most often about important things?” (family and friend domain). For the first four questions, if respondents’ response was “no,” they received one point for each question: no marriage/no partner (1 point), no religious participation (1 point), no club participation (1 point), and no in-person visit with friends or family in the past month (1 point). For the last question, if respondents’ response was “no,” they could receive up to two points. Specifically, they received one point if they could not identify any friends who can talk about important things in the past year, and an additional one point if they could not identify any family member who can talk about important things in the past year. The scores were added, and a higher score indicates a higher level of social isolation (range: 0–6). The psychometric properties of this measure were documented by Pohl et al. (2017), and the measure has since been used in subsequent studies on social isolation (Pohl et al., 2022).
Covariates
Individual-level sociodemographic variables included a six-level categorical age (65–69, 70–74, 75–79, 80–84, 85–89, and 90 or older), a binary gender (male and female), a four-category race (White/non-Hispanic, Black/non-Hispanic, Hispanic, and Other), a four-level categorical educational attainment (less than high school, high school, some college but no degree, and college graduate), household income quartiles (1st = 0 to 14,329, 2nd = 14,337–26,981, 3rd = 27,000–50,940, and 4th = 51,000–5,000,000), and a binary proxy respondent status (yes and no) variables. Following Latham & Clarke (2018), the importance of four social engagement activities: (1) visiting friends or family members, (2) attending religious events, (3) attending club events, and (4) going out for enjoyment was treated as continuous variables (range: 1 = not so important, 2 = somewhat important, and 3 = very important).
Second, individual-level health and functioning-related variables included a binary measure of hospitalization in the last 12 months (yes or no). Mobility-assistive device measure included canes, walkers, wheelchairs, or scooters (yes/no). Using this information, a variable measuring usage of any mobility-assistive device was created (0 = no use, 1 = use). ADLs included eating, bathing, toileting, and dressing. Respondents were asked whether they had difficulty performing any of the four activities (eating, bathing, toileting, and dressing) by themselves or had assistance in carrying them out (0 = no difficulty, 1 = experienced difficulty). Using this information, the number of difficulties in ADLs was created and treated as a continuous variable (range: 0–4). Similarly, a measure of Instrumental Activities of Daily Living (IADLs) included laundry, grocery shopping, meal preparation, banking, and taking medication. Respondents were asked whether they had difficulty performing these five activities (include laundry, grocery shopping, meal preparation, banking, and taking medication) alone or required assistance from others (0 = assistance is not required, 1 = assistance is required). Using this information, the number of limitations in IADLs was created and treated as a continuous variable (range: 0–5). The number of chronic diseases was coded as a continuous variable based on information on hypertension, heart disease, arthritis, osteoporosis, diabetes, lung disease, stroke, and cancer (range: 0–8). Following NHATS's dementia classification (Kasper et al., 2013), dementia status was coded as three categorical variables (no dementia, possible dementia, and probable dementia). Third, home environment level variables included the presence of stairs (yes/no), the presence of a ramp (yes/no), and three-category residential status (1 = community, 2 = non-nursing residential facility, and 3 = nursing homes) variables.
Concerning neighborhood-level variables, first, objective neighborhood disorder was measured by the NHATS interviewers using the following questions: (1) “how much litter, broken glass, or trash on sidewalks and streets?” (2) “how much graffiti on buildings and walls?” and (3) “how much vacant or deserted houses or storefronts?” (1 = none, 2 = a little, 3 = some, and 4 = a lot). A three-item objective neighborhood context scale was created using the mean scores of the three questions (Cronbach's α = .724). Given the skewed distribution of this variable (Latham & Clarke, 2018), a dichotomous neighborhood disorder variable was created using the lowest 10% as a cutoff point (score = 1.33 out of 3), where 0 represented no physical neighborhood disorder and 1 represented the presence of any physical neighborhood disorder. Unlike neighborhood disorder, subjective neighborhood social cohesion was measured based on survey participants’ response to the following items: (1) “people in this community know each other very well,” (2) “people in this community are willing to help each other,” and (3) “people in this community can be trusted.” These items were measured on a three-point Likert scale (1 = do not agree, 2 = agree a little, and 3 = agree; Latham-Mintus et al., 2022). A three-item subjective neighborhood context scale was created using the mean scores of the three questions (Cronbach's α = .725). Following Latham & Clark (2018), a dichotomous neighborhood social cohesion variable was created, given the highly skewed distribution of the social cohesion variable using the lowest 10% as a cutoff point (score = 1.66 out of 3), where 0 represented high level of neighborhood social cohesion and 1 represented low to moderate level of neighborhood social cohesion.
Analysis Plan
First, descriptive analyses were conducted to examine differences in sociodemographic, health, and neighborhood features by homebound status (non-homebound status, semi-homebound status, and homebound status). Second, mixed-effects regression models were used to examine the association between homebound status and social isolation over 9 years. Specifically, Model 1 (unconditional growth model) estimated the association between time (year) and social isolation over 9 years. Intraclass correlation coefficients were calculated to examine the level of variance in social isolation that can be explained by interindividual differences. Model 2 examined (1) the time trajectory (year) of social isolation and (2) the effect of homebound status on the intercept of the time trajectory without covariates. In Model 3, all time-invariant covariates (baseline age, gender, race, and education) and time-variant covariates (marital status, dementia, hospital stay, income, ADL count, IADL count, chronic diseases, usage of mobility assistance device, neighborhood disorder, social cohesion, residence type, subjective importance of social engagement, and proxy status) were added. In Model 4, the interaction between homebound status and time was added. All mixed-effects models incorporate a random intercept, a random slope for time, and an unstructured variance–covariance structure to account for random effects.
Results
Sample Characteristics by Homebound Status
Characteristics of the total sample and study participants across the three homebound status groups are presented in Table 1. About 74.2% of older adults were categorized as non-homebound. Compared to the other groups, non-homebound older adults tended to be younger, non-Hispanic White, more educated, have higher incomes, and report better health. They reported placing greater importance on social engagement activities. They tended to live in neighborhoods without physical disorder. They were also more likely to live in neighborhoods with high social cohesion. Second, about 7.6% of the study sample was categorized as homebound. Older adults in the homebound group tended to show the opposite characteristics of those of non-homebound older adults. Lastly, about 18.1% of the study sample was categorized as semi-homebound. The individual and neighborhood characteristics of the semi-homebound group generally fell between those of the non-homebound and the homebound groups (see Table 1 for more detailed information).
Unweighted Sociodemographic, Health, and Environmental Characteristics of the Sample by Homebound Status at Round 1.
Note. ANOVA = analysis of variance; ADL = activities of daily living; IADL = instrumental activities of daily living; any device = a cane, walker, wheelchair, or scooter.
***P < .001; **P < .01; *P < .05.
Multilevel Models for the Association Between Homebound Status and Social Isolation
Table 2 shows the results of the association between homebound status and social isolation using mixed-effects models. Model 1 showed that the average social isolation level was 2.50, and it increased by 0.01 every year (β = 0.01, P = .016). In Model 2, time (year) was associated with an increase in social isolation (β = 0.01, P < .001). Compared to non-homebound older adults, semi-homebound and homebound older adults reported higher levels of social isolation, respectively (β = 0.45, P < .001; β = 0.88, P < .001). In Model 3, compared to non-homebound older adults, semi-homebound and homebound older adults continued to report higher levels of social isolation, respectively (β = 0.06, P < .05; β = 0.27, P < .001). Lastly, Model 4 showed that, compared to non-homebound older adults, neither semi-homebound nor homebound older adults differed in the trajectories of social isolation over time (β = −0.00, P = .905; β = −0.00, P = .920; see Table 2 for more detailed information).
Unweighted Mixed-Effects Models of Homebound Status and Social Isolation.
Note. ADL = activities of daily living; IADL = instrumental activities of daily living; any device = a cane, walker, wheelchair, or scooter; Model 1 = basic model (random intercept and random slope model only); Model 2 = unadjusted model; Model 3 = adjusted model; Model 4 = homebound status and time interaction model. The intraclass correlation (ICC) was 65%, indicating that 65% of the variance in social isolation was explained by between-person variance. The ICC is not shown in this table.
***P < .001; **P < .01; *P < .05; †P < .10.
Discussion
While the current study suggests that homebound status may be a risk factor for experiencing social isolation in later life, most previous studies considered the association in the opposite direction: social isolation as a risk factor for becoming homebound. To the best of our knowledge, this study is among the first to examine the longitudinal association between homebound status and social connectedness using the Social Convoy Model of Social Relations.
First, compared to non-homebound older adults, semi-homebound older adults reported higher baseline social isolation. The findings of this study are consistent with the cross-sectional study by Szanton et al. (2016), which found different participation rates between non-homebound and semi-homebound older adults across different social activities. Specifically, Szanton et al. (2016) found that a proportion of semi-homebound older adults did not have in-person meetings with family members and friends, attend religious events, participate in club activities, or go out for enjoyment, at rates of 15.8%, 36.7%, 54.0%, and 19.3%, respectively. In contrast, the percentages reported by non-homebound older adults were lower, 6.5%, 20.0%, 30.9%, and 8.4%, respectively. Furthermore, this study found that homebound older adults tended to report a higher level of social isolation at baseline compared to non-homebound older adults. Given the unique situational context of homebound older adults, that of being confined to their housing, this population is particularly susceptible to social isolation. This finding aligns with previous studies (Cheng et al., 2022; Marti & Choi, 2022; Szanton et al., 2016). For instance, according to the study by Marti and Choi (2022), only 20% of homebound older adults go out to meet their families, only 11.2% meet their friends, and only 11.2% attend religious events. Similarly, a qualitative study by Cheng et al. (2022) revealed that the majority of homebound older adults report difficulties in maintaining interaction with their family members and friends. It is important to note that previous studies focused solely on social interaction. Building on previous studies, this study examines the level of social isolation by utilizing a holistic social isolation measurement, which, in addition to social engagement, includes marital/partner status and identification of family members and friends who can discuss important things from the past year (Berkman & Syme, 1979; Pohl et al., 2017).
Regarding the trajectories of social isolation over time, neither semi-homebound nor homebound older adults significantly differed from non-homebound older adults. There are several plausible explanations for this finding. First, in this study, homebound status was determined by the frequency of going outside. Yet, homebound older adults may not be a homogeneous group in terms of resources, physical health, mental health status, and living arrangements (Mather et al., 2023). Given their heterogeneity, certain types of homebound older adults may be more susceptible to social isolation. Differentiating homebound older adults based on these conditions may identify more nuanced trajectories of social isolation.
Furthermore, previous studies have indicated that the importance of activities acts as a protective factor for social engagement (Latham & Clarke, 2018). Consistent with previous studies, the importance of social engagement activities may play a protective role in reducing social isolation. Older adults tend to prioritize emotionally meaningful relationships in later life, and they may prioritize discussing important matters within emotionally fulfilling relationships, instead of expanding their relationships. Because homebound and semi-homebound older adults have greater difficulty in going outside and maintaining in-person interactions, they might place a significant value on maintaining their current social relations with individuals who can talk about important things, as it is not necessary to leave home to talk to a close family member or friend. While the NHATS provides information on the perceived importance of social engagement activities, which is positively associated with actual participation rates (Latham & Clarke, 2018), it does not provide information on the value placed on sharing important things with social relations.
The Convoy Model of Social Relations posits that while some social relations, such as nonkin relations, are easily changed by situational characteristics, some social relations, such as kin ties, are stable (Antonucci, 1986; Kahn & Antonucci, 1980). Utilizing this framework, Nilsen et al. (2018) examined social relations of home and community-based service users who are mostly confined to their dwellings. Nilsen et al. (2018) found that while kin networks (family and relatives) accounted for 80% of the inner circle, the most essential networks, nonkin networks, accounted for only 20% of the inner circle. Similar to the study by Nilsen et al. (2018), our findings suggest that the lower baseline level of social connections among homebound older adults may be partly attributable to the absence of nonkin social relations at clubs or churches, which may be harder to maintain because of their confinement. Nonetheless, similar trajectories of social isolation over time imply that some essential social relations (e.g., family and close friends who can share important things) are maintained despite their confinement, a finding that aligns with the postulate of the Convoy Model that older adults gravitate toward those people in their inner circle.
It is important to note that the reciprocal relationship between homebound status and social isolation was not the focus of the current study. However, previous research suggests a stronger directional relationship from homebound status to social isolation. For example, Mehrabi et al. (2024) examined the bidirectional relationship between frailty and social isolation. They found that while frailty, which often accompanies homebound status, is a risk factor for social isolation, social isolation has no lagged effect on frailty. Similarly, Xiang et al. (2020a) explored the bidirectional relationship between homebound status and depression and found a stronger lagged effect from homebound status to depressive symptoms compared to the lagged effect from depressive symptoms to homebound status. Taken together, these findings suggest that homebound status may be a risk factor for social isolation in later life.
Limitations and Future Directions
The present study has several limitations. First, the way information about homebound status is collected and defined needs to be considered. Specifically, in the NHATS, participants’ self-report questions were used to obtain information about homebound status. While the self-report approach is often used in large surveys, this approach is vulnerable to recall bias (Qiu et al., 2010).
Second, while homebound status is defined in a given month in the NHATS, different studies have used different time periods (e.g., 1 year) to define homebound status (Ko & Noh, 2021). Future studies could examine whether the association between homebound status and social isolation varies depending on how homebound status is defined.
Third, this study uses baseline homebound status to examine the association between homebound status and changes in social isolation over 9 years. However, homebound status is dynamic (Ankuda et al., 2021; Xiang et al. 2020b). When homebound status was compared in Rounds 1 and 9, 69.37% of people were categorized into the same homebound status, whereas the remaining 30.63% of the study sample were categorized into a different homebound status. Future studies could examine how changes in homebound status are associated with changes in the levels of social isolation.
Fourth, while this study utilized data that were collected before COVID-19, COVID-19 may influence the association between homebound status and social isolation. For example, the study by Ankuda et al. (2022) found a disproportionate impact of COVID-19 on contact frequency with friends and family among homebound older adults compared to non-homebound older adults. Future studies could examine whether COVID-19 has a differential impact on other interpersonal interactions, such as club participation and religious attendance, based on the homebound status.
Fifth, in this study, social isolation was measured based on Berkman and Syme's SNI. SNI was developed before the advances in various communication technologies, so it primarily focuses on in-person social contact. Thus, the way social isolation is measured in this study does not capture social interaction through communication technologies (Nicholson et al., 2020). Future studies should incorporate social interaction via communication technology, as older adults are the fastest-growing population concerning technology adoption (Faverio, 2021).
Conclusion
Aging in place refers to people continuing to live in their current dwelling. Given the lack of social connections, even though homebound and semi-homebound older adults are able to stay in their current home, they cannot be considered to be aging in the right place. This study shows that homebound status is a risk factor for social isolation. However, homebound status is a modifiable factor. For instance, improving the quality of their neighborhood environment or providing reliable transportation for social connections may reduce the burden of going outside among homebound and semi-homebound older adults (Choi et al., 2025b; Choi et al., 2025c). Homebound older adults may also be unaware of resources that could increase their social interaction, as they are mostly confined to their home (Cheng et al., 2022). Social workers could refer homebound older adults to relevant local resources to mitigate social isolation. For example, the Village model is a community-based aging-in-place intervention that helps older adults live independently in their communities (Cho et al., 2023; Hou & Cao, 2021). The Village model could help homebound older adults to have more social interaction, as this model provides various services, such as friendly visits and transportation services (Szanton et al., 2016). Social workers could also refer homebound older adults to the Meals on Wheels program, as the meal delivery service provides an opportunity for homebound older adults to have a short interaction with volunteers (Timonen & O’Dwyer, 2010).
Footnotes
Acknowledgment
I would like to thank Dr Ruth E. Dunkle for her valuable feedback during the development of this paper.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.