Relocation Patterns and Predictors of Relocation and Mortality in Swedish Sheltered Housing and Aging in Place

Abstract A reported objective of Swedish sheltered housing is to postpone care needs and relocation. The aim of this study was to describe migration patterns and explore predictors of relocation to nursing homes and mortality, in a sample of residents in sheltered housing and aging in place. To explore longitudinal differences between groups, study data were combined with registry data. The results showed that a higher percentage of residents in sheltered housing had relocated to a nursing home and deceased over a three-year time period, compared to aging in place, implying further interventions may be required to promote health in sheltered housing.


Introduction
Sheltered housing was established in Sweden in 2008, as an accommodation option for older people who feel lonely, anxious, or unsafe aging in place (Ministry of Health and Social Affairs, 2008).Swedish sheltered housing provides residents an opportunity to either rent, or own, an apartment in a building with accessible spaces and close to services, and in which increased opportunities to social participation are provided.Each sheltered housing accommodation has a host/hostess that arranges social activities during certain hours of the week, as well as common facilities to encourage social interaction.Swedish sheltered housing was intended to be available to older people capable of independent living and health-care services are not provided as part of residency in sheltered housing; however, residents may be eligible to seek home care services.Increased social participation while residing in Swedish sheltered housing is anticipated to promote resident health and wellbeing, postponing the need for health-care services and institutionalization, and ultimately decreasing the expenses related to the care of older people.The expectation of increased health of older people living in sheltered housing is based on older studies, which have shown that social support has a positive effect on one's health, that social interaction provides a protection from different forms of stress, and that there is a correlation between social relationships and reduced mortality, sickness and psychological illness (Ministry of Health and Social Affairs, 2008).Both those residing in sheltered housing and those aging in place in Sweden live independently and have access to public health-care services; however, sheltered housing is specifically intended for older people, and those residing in sheltered housing have increased possibilities to attend social events and engage in social interactions, as well as live in apartments with accessible spaces.
To date, studies concerning Swedish sheltered housing are few and have been mainly qualitative, focusing on aspects related to perceptions of safety.These studies have shown that in general residents of Swedish sheltered housing feel that the housing model contributes to their sense of safety, although residents had safety concerns related to declining health, as well as unease due to the behavior of neighbors with cognitive impairments, not knowing who enters the building, and being potential victims of crime (Berglund-Snodgrass & Nord, 2019;Lindahl et al., 2018).Previous quantitative studies on Sweden sheltered housing have shown that residents in sheltered housing report higher depressive mood, and lower levels of health, health-related quality of life, and functional status, when compared to those aging in place, implying potential health-care needs existing or arising among residents in sheltered housing (Corneliusson et al., 2019b).Study on the same sample also revealed that while experiencing increasing level of depressive mood, and decreasing levels of health and functional status, residents in sheltered housing generally reported higher levels of thriving, when compared to those aging in place, implying that there are features within sheltered housing that support thriving (Corneliusson et al., 2019a).Thriving, which has recently been defined as a concept that denotes lived experiences of situated contentment, provides novel insight into the interplay of different features that may affect wellbeing in a specific context (Baxter et al., 2021).
Directly comparable forms of housing to Swedish sheltered housing within the international context are difficult to identify, largely due to differences in welfare systems and fiscal policies.Nonetheless, significant similarities can be found to sheltered housing in other parts of Europe, namely, in the UK and the Netherlands, such as the intention of independent living, an emphasis on social participation, providing opportunities to social participation, and accessible spaces; the amount of care services provided in international sheltered housing may, however, vary, with some sheltered housing accommodations offering access to 24/7 alert systems or near-by staff that can check in on residents (Cook et al., 2017;Egbu et al., 2011;Herbers & Meijering, 2015;Howe et al., 2013;Iecovich, 2016;Van Bilsen et al., 2008).A longitudinal study from the UK which explored the relationship between housing pathways and mortality, based on data from the British Household Panel Survey from 1993 to 2008 observed at three consecutive points, showed no statistically significant difference in mortality for those residing in sheltered housing compared to those who remained in private housing (Robards et al., 2014).A study based on the same data observed at two consecutive points, however, showed that contact with health-care services during the year before was an indicator of transition to sheltered housing (Vlachantoni et al., 2016).A cross-sectional study also from the UK with a sample of 978 residents living in sheltered housing showed 57% of residents had no hospital admissions during the past year, the mean length of elective hospital admissions was 1 day, the mean length of emergency hospital admissions 8.2 days, and 95% returned to their usual place of residence (Cook et al., 2017) implying that residents in general did not suffer from severe health issues.Studies from the UK, Ireland and the Netherlands, which were based on smaller samples, showed little difference in mobility and health between those living in sheltered housing and aging in place (Fox et al. 2017;Van Bilsen et al. 2008); however, one study found higher rates of activity limitation and depression among residents living in sheltered housing (Field et al., 2002).As previous studies on the same sample residing in Swedish sheltered housing has shown statistically significant differences between groups in relation to in functional status, depressive mood, thriving in relation to the fore mentioned, and living alone (Corneliusson et al., 2019a), it can be hypothesized that these may be predictors for relocation to nursing home.

Aim and research questions
The aim of this study was to describe relocation patterns and explore predictors of relocation to a nursing home and mortality, in a sample of older adults living in Swedish sheltered housing and aging in place.The research questions were: 1. What are the patterns of relocation to a nursing home/mortality in a sample of older adults living in Swedish sheltered housing and aging in place?
2. How does the distribution of home care services differ in a sample of older adults living in Swedish sheltered housing and aging in place, and how does receiving home care services influence patterns of relocation/mortality? 3. What are the predictors of relocation to a nursing home/mortality in a sample of older adults living in Swedish sheltered housing and aging in place?

Sample
Residence in a sheltered housing accommodation was the inclusion criteria for this study.The data for this study consist of data from the U-age sheltered housing study, combined with registry data on social services resource use and mortality.For the U-age sheltered housing study, contact information to sheltered housing properties in Sweden was obtained by contacting the National Board of Housing, Building and Planning, whom hold registries on sheltered housing accommodations in Sweden which have received governmental financial assistance.Property owners of both public and private sheltered housing accommodations were thereby contacted by either mail/phone and requested to assist in obtaining the addresses of sheltered housing residents.A matching control person was sought out on a ratio of 1:1 for each resident living in sheltered housing, with the matching criteria being age, sex and municipality of residence.Statistics Sweden sent the survey to both residents living in Swedish sheltered housing, and the matched control group.

Data collection
The U-age sheltered housing self-report survey, an information letter and a postage-paid envelope was sent to residents living in sheltered housing and the control group aging in place in October 2016.In total two reminder letters were sent, one in mid-November 2016, and another during the first week of December 2016.The data collection was ended on the 15th of January 2017.In total 8,163 people were eligible for the survey; 4,087 surveys were posted to people living in sheltered housing, and 4,076 to people aging in place.The final sample consisted of 1,955 responses to the survey from residents in sheltered housing, and 1,850 aging in place, with the response rate being 47.8% and 45.5%, respectively.The total sample size was thereby 3,805 surveys, with an overall response rate of 46.6%.
To enable longitudinal analysis, registry data on social services resource use (including data on relocation to a nursing home) and mortality were sought and obtained from the Department of Health and Welfare in Sweden and Statistics Sweden.The registry data included the complete sample from the U-age sheltered housing study, consisting of data from the years 2017-2019.As the sample of this study consists of an aging demographic, a three-year time interval was selected in this study, as shifts in health and functional ability may occur expeditiously.

Study variables
The U-age sheltered housing study survey consisted of questions related to demographic data as well as established instruments.Scores to the instruments from the U-age sheltered housing study related to functional status, health, wellbeing, thriving, and depressive mood were included as independent variables to explore the influence these factors have on relocation to a nursing home/mortality.Demographic data such as age, sex, and living alone were also included as background variables.The independent variable "type of residence" was utilized to explore the significance of living in sheltered housing vs. aging in place.The registry data variables relocation to a nursing home and mortality were included as dependent variables.

Functional status
Functional status was measured using the combined Katz Index of Independence in Activities of Daily Living (Katz ADL) and Lawton Instrumental Activities of Daily Living (Lawton IADL) instruments.The Katz ADL consists of six categories related to activities of daily living (such as bathing, feeding, continence), with each response scored either 1 or 0, and has a total range from 0 (dependent) to 6 (independent) (Katz et al., 1963;Shelkey & Wallace, 2012).The Lawton ADL consists of 8 categories related to instrumental activities of daily living (such as using the phone, preparing food, doing laundry), and each response is scored either 0 or 1, with a total range of 0 (dependent) to 8 (independent) (Graf, 2013).As data on medication use were obtained via registry data, the question on medication use was omitted from the Lawton IADL, leading to a total possible range of 0-7 in this study.This study utilized screening questions for both the Katz ADL and Lawton ADL; the highest scores were imputed for those with missing items due to positive responses to the screening questions.The total range of the combined scales used to measure functional status in this study was 0 (dependent) to 13 (independent).Previous studies on the combined ADL and IADL have concluded that the combined scale may be used as a one-dimensional construct to measure functional disability among older people, with the advantages of providing a larger and more parsimonious scale (Spector & Fleishman, 1998).The combined ADL and IADL scale has shown good validity and reliability for predicting future unfavorable health outcomes in community-dwelling older people (Laan et al., 2014).

Thriving
The Thriving of Older People Assessment Scale, sheltered housing version (TOPAS-SH) was utilized to measure thriving.The original 32-item TOPAS, scored on a six-point scale, was originally developed within the nursing home context (Bergland et al., 2015).Thereby, for this study, a modified version of TOPAS, the TOPAS-SH, was utilized.The TOPAS-SH consisting of 20 items, is similarly to the 32-item version scored on a sixpoint scale, and has a total range of 20-120; higher values indicate higher thriving.Thriving was included as an independent variable, as previous studies have shown an interaction between type of residence, functional status, self-rated health and depressive mood in relation to thriving (Corneliusson et al., 2019a).

Depressive mood
The Geriatric Depression Scale 4-item version (GDS-4) was utilized to measure depressive mood.The GDS-4 consists of 4 yes/no questions related to mood and wellbeing, and each item may score either 0 or 1 points (van Marwijk et al., 1995).The total range of GDS-4 is 0-4; higher values indicate a higher possibility of depressive mood.

Ethical considerations
The Regional Ethical Review Board in Umeå, Sweden has granted ethical approval of this study (Dnr 2016/40-31).

Data analyses
Descriptive statistics, such as frequencies, means, and chi-square tests, were utilized to describe the demographic data and to analyze the differences between the groups (sheltered housing/aging in place) in respect to relocation to a nursing home and mortality.Major neurocognitive disorder was calculated based on the amount of individuals that had either received medication for, or a diagnosis of, a major neurocognitive disorder in 2016 or 2017.To explore factors that influence relocation to a nursing home/ mortality, logistic regression models with the dependent variable relocation to a nursing home/mortality and the independent variables sex, age, type of residence (sheltered housing/aging in place) living alone, functional status, thriving, depressive mood and received home care services (in 2017) were performed.Models with the interaction terms type of residence by all other explanatory variables were also performed to explore potential differences between groups (living in sheltered housing/aging in place).Statistical significance was defined at p .05.The data were analyzed using IBM SPSS version 25 (for Windows; IBM Corp., Armonk, NY, USA).

Validity and reliability
A correlation matrix was calculated for the continuous independent variables included in the logistic regression models.All variables showed statistically significant values, with the Pearson's r values varying between .159 to À.425, indicating low correlations.For the logistic regressions, linearity of the continuous variables with respect to the logit of the dependent variable was assessed using the Box-Tidwell procedure.Based on this assessment, all continuous independent variables (age, thriving, functional status, depressive mood) were found to be linearly related to the logit of the dependent variable.Goodness of fit of the logistic regression model was assessed using the Hosmer-Lemeshow test (p !.05),and the standardized residuals were examined to identify potentially influential outliers.The internal consistency reliability of the three utilized scales were explored using Cronbach's alpha.The Cronbach's alpha of the TOPAS-SH, the combined ADL-scale, and the GDS-4 were deemed satisfactory at a ¼ 0.927, a ¼ 0.855 and a ¼ 0.640, respectively.For the TOPAS-SH, values for up to two missing items were replaced with the mean values of the total score for the individual.

Demographic data
The characteristics of the sample are presented in Table 1.The mean age was 83.09 in sheltered housing and 82.86 in the matched control group, and the majority of participants were females.Analyses of functional status, depressive mood and thriving showed that those living in sheltered housing reported lower functional status (10.89/11.47,p .001) and higher depressive mood (.80/.57, p .001) compared to those aging in place.Thriving showed no significant difference between groups.In 2017, 25.5% of residents living in sheltered housing and 15.4% of residents aging in place received home care services (p .001).In 2019, 26.3% of residents living in sheltered housing and 19.0% of residents aging in place received home care services (p .001).The difference between those living in sheltered housing and those aging in place whom did not receive home care services in 2017, but did in 2019, was not statistically significant (12.3% and 10.4%, respectively, p ¼ .065).The mean amount of annual home care hours between the years 2017-2019 showed a statistically significant difference between groups, with those living in sheltered housing receiving 360.85 h of home care services, and those aging in place 284.30h of home care services (p .001).

Relocation patterns
The results showed that there were statistically significant differences between the groups sheltered housing/aging in place in relation to risk of relocation and mortality.Between the years 2017-2019, 219 (11.2%) residents living in sheltered housing at baseline (end of 2016), and 151 (8.2%, p ¼ .002) of those aging in place at baseline, had relocated to a nursing home.Between the same time period, 422 (21.6%) residents of sheltered housing at baseline and 324 (17.5%, p ¼ .002) of those aging in place at baseline had deceased.In total, 1,314 (67.2%) residents remained in sheltered housing, while 1,375 (74.3%) remained aging in place (Figure 1).The results showed that those whom had received home care services (in 2017) had a higher rate of relocation to a nursing home/mortality in both groups (living in sheltered housing/aging in place), compared to those who did not receive home care services over a three-year time period.Of those living in sheltered housing that had received home care services in 2017, 9% relocated to a nursing home, while 3.8% of those living in sheltered housing without home care services in 2017 relocated to a nursing home between the years 2017-2019 (p .001).Mortality was also higher among those receiving home care services, as 37.6% of those living in sheltered housing receiving home care services in 2017 had deceased, while 16.1% of those not receiving home care services in 2017 had deceased, between the years 2017-2019 (p .001).Similar patterns concerning the influence of receiving home care services were found among those aging in place.Please see Figure 2 for more details.

Variables explaining relocation to a nursing home
A logistic regression model was performed to explore how sex, age, living alone, type of residence, functional status, thriving, depressive mood and receiving home care services affect risk of relocation to a nursing home.The interaction terms type of residence by all other explanatory variables were also explored, but were not statistically significant, and therefore not included in the final model.The full model containing eight independent variables was statistically significant, v 2 (8, N ¼ 3,805) ¼292.626,p .001), indicating that the model was able to distinguish between respondents who moved and those who did not move to a nursing home.As shown in Table 2, three variables made a statistically significant contribution to the model: age, functional status, and receiving home care services.The odds ratios indicate that increasing age, decreasing functional status and receiving home care services increase the odds of relocation to a nursing home in both groups (living in sheltered housing/aging in place).

Variables explaining mortality
A logistic regression model was performed to explore how sex, age, living alone, type of residence, functional status, thriving, depressive mood, and receiving home care services affect risk of mortality.The interaction terms type of residence by all other explanatory variables were also explored but were not statistically significant and therefore not included in the final  model.The full model containing eight independent variables was statistically significant, v 2 (8, N ¼ 3,805) ¼441.397,p .001), indicating that the model was able to distinguish between respondents who deceased and those who did not decease.As shown in Table 3, five variables made a statistically significant contribution to the model: sex, age, functional status, living alone and receiving home care services.The odds ratios indicate that being male, increasing age, decreasing functional status, living alone and receiving home care services increases the odds of death in both groups (living in sheltered housing/aging in place).

Discussion
The aim of this study was to describe relocation patterns and explore predictors of relocation to a nursing home and mortality, in a sample of older adults living in Swedish sheltered housing and aging in place.Analysis of the demographic data showed a relatively high mean age (83.09) among those living in sheltered housing, with those living in sheltered housing reporting lower functional status and higher depressive mood compared to those aging in place.In 2017, 25.5% of residents living in sheltered housing and 15.4% of residents aging in place received home care services (p .001).In 2019, 26.3% of residents living in sheltered housing and 19.0% of residents aging in place received home care services (p .001).Those residing in sheltered housing also received on average more home care hours (360.85),than those aging in place (284.30h; mean value for the three-year period).The results of the analyses related to relocation patterns showed that a higher percentage (11.2%) of residents living in sheltered housing had relocated to a nursing home over a three-year time period, compared to those aging in place (8.2%).Furthermore, a higher percentage of residents in sheltered housing (21.6%) had deceased during the same three-year time period, compared to those aging in place (17.5%).Analyses related to the influence of receiving home care services showed that those whom had received home care services had a higher rate of relocation to a nursing home/mortality in both groups (living in sheltered housing/aging in place) over a three-year time period, compared to those whom did not receive home care services.The analyses related predictors of relocation to a nursing home showed that age, functional status and receiving home care services were statistically significant predictors of relocation.The analyses related to predictors of mortality showed that sex, age, functional status, living alone and receiving home care services were statistically significant predictors of mortality.
The results of this study showed a higher percentage of those living in Swedish sheltered housing being recipients of home care services, receiving on average more hours of home care compared to those aging in place.Furthermore, predictors of relocation to a nursing home were age, functional status and receiving home care services, indicating extensive healthcare needs existing and arising among those residing in sheltered housing, especially as access to home care services and nursing home residency is rigorously means-tested in Sweden (Brodin, 2005;Larsson & Szebehely, 2006).Higher rates of relocation to a nursing home are likely explained by either lower functional status or higher rates of major neurocognitive disorder among those residing in sheltered housing, as roughly half of all people living with a neurocognitive disorder in Sweden live in a nursing home (Lennartsson & Heimerson, 2012).As analyses related to the influence of receiving home care services showed higher rates of relocation to a nursing home and mortality among those residing in sheltered housing and receiving home care services, it is possible that residence in sheltered housing does not postpone relocation or reduce mortality specifically among residents with more extensive care needs.It is therefore possible that the health and wellbeing promoting intentions are achieved among those with fewer health-care needs, although further studies specifically exploring the differences between these two groups are needed to better understand the significance of the support provided in sheltered housing in relation to health-care needs.The amount of older people that remained aging in place and received home care services in 2017, but no longer received home care services in 2019, may be explained by these individuals either needing only short term-care, or by these individuals receiving home nursing care services, which are intended for those in need of more extensive nursing care (National Board of Health and Welfare, 2020).
To date, only one study on the longitudinal trajectories of residents living in sheltered housing has been published.This study from the UK showed no statistically significant difference in mortality for those residing in sheltered housing compared to those who remained in private housing (Robards et al., 2014), differing from the results of this study which found higher rates of mortality among those residing in Swedish sheltered housing compared to those aging in place.The results of this study showed that predictors of mortality included sex, age, functional status, living alone, and receiving home care services.Other significant variables explaining mortality may include, but not be limited to, illnesses such as cancer, heart conditions or diabetes.These data were, however, not available and therefore not included in these analyses.These results nonetheless imply that the social support provided in Swedish sheltered housing may not be sufficient enough to compensate for the adverse effects of living alone, as the interaction term type of residence by living alone showed no difference between groups.Previous studies have shown that loneliness increases the probability of various physical illnesses and cognitive decline, as well as predicts depressive symptoms (Masi et al., 2011).Therefore, further study into the significance of social support while residing in Swedish sheltered housing could assist in evaluating how the provided support is experienced, and in uncovering which further services are potentially needed to better alleviate feelings of loneliness while residing alone in sheltered housing.Furthermore, being male was a predictor of mortality; this result may, however, be explained by various biological and behavioral differences such as, but not limited to, genetic responses, hormonal factors, risk taking behavior and reluctance to seek medical treatment (Oksuzyan et al., 2008).
The results of this study seem to illuminate that despite the stated health-promoting intentions of this form of accommodation (Ministry of Health and Social Affairs, 2008), residence in Swedish sheltered housing does not seem to postpone or decrease relocation to a nursing home, nor decrease mortality, specifically among residents with existing health-care needs.As Swedish sheltered housing provides social support and better accessibility, yet is considered independent living and is generally not means-tested, it is possible and also likely that older people seek residence in Swedish sheltered housing as an easily accessible alternative when experiencing declining functional abilities with rising support and care needs.This would then result in a demographic of older people relocating to Swedish sheltered housing being more dependent than initially envisioned by the providers and authorities, potentially compromising the intention of compressed morbidity and increased wellbeing.However, a previous study on the same sample has shown that, despite no differences in self-reported wellbeing, those residing in sheltered housing generally reported higher levels of thriving compared to those aging in place, when experiencing decreasing levels of self-rated health and functional status and increasing level of depressive mood (Corneliusson et al., 2019a).Furthermore, a qualitative study on thriving in Swedish sheltered housing found four levels of facilitators and barriers to thriving in Swedish sheltered housing; individual factors, social context, environmental factors and organizational context (Corneliusson et al., 2021).These results of previous studies imply that there are features within sheltered housing that support thriving and are appreciated by residents: however, further studies are needed to determine if, how and to what extent the features that facilitate thriving can potentially support residents despite decreasing health or increased loneliness and/or vice versa; that perceived accessibility of support is a strong predictor of thriving.Specifically, studies focusing on the differences in perceived support needs while residing in sheltered housing between those with fewer and those with more extensive health-care needs would provide valuable information on how to best support health and wellbeing among residents with differing levels of health.Furthermore, continued longitudinal studies on the health status and service use of residents living in sheltered housing would be valuable to better understand the specific needs of this population in relation to the variation in health and social care offerings provided in these accommodations internationally.

Limitations
As this is the first follow-up study of the population living in Swedish sheltered housing, the relatively short follow-up time and limited amount of variables may have affected the results; there are likely more factors affecting relocation and mortality not captured within this study.As data on possible relocation to regular/sheltered housing were not available, it is possible that those that are listed as remaining in sheltered housing/aging in place in this study have relocated to other independent housing facilities.Although this study included all sheltered housing accommodations listed in the National Board of Housing, Building and Planning's registries, the results of this study may not be representative of privately owned and funded sheltered housing accommodations.

Conclusions
The results of this study provides essential knowledge on the relocation patterns and predictors of relocation to a nursing home and mortality for those living in Swedish sheltered housing, which may be used for further research into this population and residential phenomenon, as well as for examination and development of the services provided within sheltered housing.In light of the results of this study, it may be that further interventions beyond social support and increased physical accessibility are required to further promote population health in Swedish sheltered housing, especially among those with existing home care needs.As residents in Swedish sheltered housing report lower health, functional status and higher depressive mood, as well as utilize home care services to a higher extent compared to a matched group aging in place, implementation of healthcare-and nursing interventions for older people residing in Swedish sheltered housing may further assist in maintaining functional ability, and potentially assist in attaining the goal of compressed morbidity and postponed relocation to a nursing home.Further studies are, however, required to establish which specific services and interventions are needed to better support resident health and wellbeing.

Figure 2 .
Figure 2. The influence of receiving home care services to relocation patterns.

Table 1 .
Characteristics of the sample and mean values of study variables (2017).

Table 2 .
Results of the logistic regression model-variables explaining relocation to a nursing home.

Table 3 .
Results of the logistic regression model-variables explaining mortality.