Age differences in associations of different indicators of socioeconomic status with social isolation: A cross-sectional study

Background Socioeconomic status is a crucial determinant of social isolation. However, little is known whether the associations between different indicators of socioeconomic and social isolation vary across age groups. This study examined the association of individual socioeconomic status indicators with social isolation in three age groups: young (21-44 years), middle-aged (45-64 years), and older adults (≥65 years). Methods Cross-sectional data for 1,930 representative community-dwelling adults aged 21 and above in the Central region of Singapore was used. The 6-item Lubben Social Network Scale was used to assess social isolation. Socioeconomic status was measured using education level, employment status, personal income, housing type and self-perceived money sufficiency). Separate logistic regression analyses were conducted to examine the association between each SES indicator and social isolation in each age group. Results Each socioeconomic indicator showed a clear gradient with social isolation and significant age disparities were found in their relationship. Socioeconomic status indicators significantly associated with social isolation were income (R2 change=2.5%) and self-perceived money insufficiency (R2 change=1.5%) in young adults, education (R2 change=0.5%), employment status (R2 change=1.3%), income (R2 change=0.8%), housing type (R2 change=1.9%) and self-perceived money insufficiency (R2 change=2.0%) in middle-aged adults, and housing type (R2 change=1.3%) and self-perceived money insufficiency (R2 change=3.7%) in older adults when adjusting for demographics and other indicators. Conclusions The influence of individual socioeconomic status indicators on social isolation varied across age groups. This study provides a rationale for the choice of socioeconomic status indicator and specific interventions need to target different socioeconomic status groups for different age groups.


Abstract
Background Socioeconomic status is a crucial determinant of social isolation. However, little is known whether the associations between different indicators of socioeconomic status and social isolation vary across age groups. This study examined the association of individual socioeconomic status indicators with social isolation in three age groups: young (21-44 years), middle-aged (45-64 years), and older adults (≥65 years).
Methods Cross-sectional data for 1,930 representative community-dwelling adults aged 21 and above in the Central region of Singapore was used. The 6-item Lubben Social Network Scale was used to assess social isolation. Socioeconomic status was measured using education level, employment status, personal income, housing type and self-perceived money sufficiency). Separate logistic regression analyses were conducted to examine the association between each SES indicator and social isolation in each age group.
Conclusions The influence of individual socioeconomic status indicators on social isolation varied across age groups. This study provides a rationale for the choice of socioeconomic status indicator and specific interventions need to target different socioeconomic status groups for different age groups.

Background
Social isolation reflects a lack of quantity and quality of social relationships that provide positive feedback, and are meaningful to the individual [1]. Evidence suggests that people around the world are more socially isolated now than ever before [2][3][4][5]. In Singapore, population ageing and shift towards nuclear families increases the likelihood of social isolation [6]. Social isolation has increasingly recognized as a global social issue that casts significant and growing influence on people's health across life course [7], including but not limit to physical and psychological health [8][9][10], morbidity [11], and mortality [12,13]. A meta-analytic review suggests that the influence of social isolation on mortality is comparable with well-established risk factors like cigarette smoking and alcohol consumption [14]. Being socially isolated also results in higher spending in health and social care services [15,16]. Social isolation is not only prevalent and significant for older adults, but it also affects younger individuals at different stages of life [17][18][19].
Socioeconomic status (SES) describes one's combined economic and social status, reflecting one's access to collectively desired resources including material goods, money, power, friendship networks, healthcare, leisure time, or educational opportunities [20]. Research has suggested that many risk factors of social isolation are unequally distributed in society and are more prevalent among economically or socially disadvantaged individuals [1,21]. This implies that SES plays a non-ignorable role on social isolation, whether it is direct or indirect. Education, employment status or occupational class, and income [22,23] are commonly used as proxy indicators of SES.
The highest level of education is usually attained and fixed in early adulthood. It is positively associated with a range of health outcomes across all ages. Prior research reported inconsistent findings in the association between the highest level of education and social isolation. While more educated individuals could have larger confidant networks than those less well educated [4,7], two other studies found that individuals with lower level of education had lower likelihood of being social isolated in mid and older life [24].
Unemployment is a stressful life event, which may affect various aspects of health [25,26], including social isolation. Unemployment at different stages of life would have different degree of impacts on people's social life, financial status, physical and psychological health [27]. One large European study found that unemployed young and middle-aged adults were more prone to be socially isolated than their employed counterparts [28] while another study demonstrated that the retired, unemployed, sick/disabled and homemakers had poorer social engagement compared to employed older adults [29].
Personal or household income indicates the availability of economic and material resources. A small body of literature has documented the relationship between low-income and social isolation. A mixed method study showed that low-income individuals were more likely than their respective counterparts to feel isolated [30]. Two large scale population studies also demonstrated that low income was independently associated with social isolation [7,24]. The relationship between income and social isolation among different age groups was rarely documented.
Other than the above-mentioned objective measures, indicators of subjective SES are also found to be strong predictors for several aspects of health [31,32]. However, the association of self-perceived SES and social isolation is scantly explored. A study on elderly residents in Japan showed that having a low self-perceived SES was strongly associated with isolation [21]. However, whether the subjective indicator of SES is associated with social isolation in younger adults or whether it has stronger association with social isolation compared to objective indicators is still unclear.
While the associations between different indicators of SES and social isolation are extensively explored, little is known about whether and how they are differently associated at different stages of life. There is increasing awareness of the instability of an individual's SES through the life course [33], and different socioeconomic factors at different stages of the life course could have varied influence on health [34,35]. This has aroused interest in exploring how different indicators of SES affect health at different life stages [36]. As people age, their social networks may change for a variety of reasons including change in living arrangement, migration of family members or friends, change in social roles, physical illnesses, decline in physical or cognitive abilities, and death of social network members [7]. It implies that the relationship between different indicators of SES and social isolation

Social isolation
The 6-item Lubben Social Network Scale (LSNS-6), a standardized measure of social isolation, was used to screen for the presence and extent of social isolation [37]. The LSNS-6 measures the size, closeness and frequency of kinship and non-kinship contacts within an individual's social network.
Three questions were asked "How many relatives do you see or hear from at least once a month?", "How many relatives do you feel close to such that you could call on them for help?" and "How many relatives do you feel at ease with that you can talk about private matters?". The word "relatives" was replaced with "friends" in these three questions to ascertain non-kinship ties. The total score of the 6item scale ranges from 0 to 30, with lower score indicating a greater extent of social isolation. The present study demonstrated good internal consistency of the LSNS-6 with Cronbach's alpha = 0.82. As suggested by Lubben [37], a score of 12 or less on the LSNS-6 indicates social isolation whereas those who scored 13 and above were considered to be not socially isolated.
Socioeconomic status 6 Objective measures of SES The objective SES of the participants was measured using four indicators, including education level, employment status, personal income, and housing type.
The level of education was measured based on the highest educational attainment using the nine categories described in the Singapore Standard Educational Classification (SSEC) 2015 [38]. For this analysis, we recategorized the variable into three groups: 1: Low (primary or lower), 2: Middle (lower secondary, secondary, post-secondary), 3: High (polytechnic diploma, professional qualification, bachelor or higher).
The type of housing in Singapore is positively correlated with household income [10,36], and is often used as a measure of SES as it broadly reflects the social positioning in the Singaporean society. More than 80 percent of Singaporean households reside in housing built by the government-owned Housing Development Board (HDB) with households in the lowest income percentiles residing in 1-and 2-room flats. More affluent households typically reside in privately built residential dwellings. In this study, we have defined three categories of housing: 1: HDB 1-and 2-room flats, 2: HDB 3-and 4-room flats, 3: HDB 5-room flats and above (including private properties).
Subjective measure of SES We measured one's self-perceived money insufficiency (0:No, 1:Yes) based on the question "Do you often run out of money, even with proper spending plan, to buy essential items or pay bills to maintain basic living needs (i.e. accommodation, food, transportation and healthcare)?" as a subjective measure of SES.

Covariates
We also adjusted for the following covariates in the subsequent analyses: age, sex (male or female), ethnicity (Chinese, Malay, Indian, Others), marital status (single, married, divorced/widowed) and living arrangement (living alone / with unrelated individuals, living with spouse with/without child(ren), living with child(ren) but no spouse, living with parent/friend/other relative).

Statistical analyses
Descriptive analyses were first conducted for each age group with weighted mean and standard deviation (SD) reported for continuous variables, and unweighted frequency and weighted percentage reported for categorical variables. Chi-squared or Fisher's exact tests were used to determine between-group differences for categorical variables.
To examine the relationship between each SES indicator and social isolation stratified by the three age groups, we ran logistic regression analyses using each SES indicator as the independent variable and social isolation (0 = not isolated and 1 = isolated) as the dependent variable. To isolate the influence of each SES indicator, we have also adjusted for covariates including age, gender, ethnicity, marital status and living arrangement (Model 1). The results were presented in terms of Average Marginal Effect (AME) multiplied by 100 (AME %). We computed the AMEs to ascertain the average change in the probability of being socially isolated as SES increases from the reference group while holding other covariates at their observed values [39]. Odds ratios (ORs) and p-values were also presented.
To study the independent association of the five SES indicators with social isolation stratified by the three age groups, we included all the five SES indicators simultaneously in the full model, adjusting for age, gender, ethnicity, marital status and living arrangement (Model 2). The AMEs were estimated for individual SES indicators.
We used McKelvey & Zavoina's pseudo-R 2 to compare estimates of explained variance from different models using the same dataset [40]. To study how each SES indicator contributed to model fit, we calculated the change in pseudo-R 2 by subtracting the pseudo-R 2 value of the basic model (including only social isolation as the dependent variable and the variables for adjustment) from the pseudo-R 2 value obtained by adding each SES indicator separately to the basic model (Model 1). In addition, for each SES indicator, we also calculated the absolute change in pseudo-R 2 associated to the exclusion of that indicator from the full model (Model 2). All analyses were performed using Stata/SE 16.0. P < 0.05 was set as the level of significance.

Association between SES indicators and social isolation in individual age groups
We observed significant differences in the proportion of isolated individuals across the three age groups where 13.8%, 27.5% and 45.4% among those aged 21-44 years old, 45-64 years old and 65 years and above were socially isolated.

Education level
Chi-squared test showed that education level had significant association with social isolation in young adults (p<0.001) ( Table 2). Compared to those with high level of education, people with low or middle level of education had higher probability of reporting isolation given the demographic variables were held constant. Including education level in the basic model increased the model fit by 6.1%, which was slightly lower compared to personal income (

Employment status
Employment status was not associated with social isolation in young adults (

Personal income
Personal income was associated with social isolation in young adults with or without adjustment for demographics or other SES indicators. Young adults with personal income of SGD5,000 or lower was associated with elevated probability of social isolation compared to those with income of SGD5,000 and above. It contributed to the model fit with higher magnitude than other SES indicators ( Table 3,

Model 1). With inclusion of other SES indicators in the model, personal income still contributed most
to the explained variance in Model 2, and young adults with income of SGD1,500 or lower had an increased probability of isolation than those with income of SGD5,000 or above (AME%=24.78) ( Table   3, Model 2).

Housing type
Including housing type in the basic model showed its significant association with social isolation (increased the model fit by 1.9%) in young adults (Table 3, Model 1). However, after adjusted for other SES indicators, no statistical difference across three housing types was observed (Table 3, Model 2).

Self-reported money insufficiency
Young adults who reported money insufficiency for basic living needs were more likely to be isolated than those did not, even after adjusted for other SES indicators (AME%=21.67 and 11.35, respectively) (

Association between SES indicators and social isolation in middle-aged adults Education level
Education level had significant association with social isolation in middle-aged adults (p<0.001) (  The probability of being isolated in unemployed middle-aged adults was significantly higher compared to their employed counterparts, even after adjusted for other SES indicators. Employment status contributed to the model fit by 2.5% in Model 1 and 1.3% to Model 2, respectively (Table 3).

Personal income
The relationship between personal income and social isolation was significant in middle-aged adults, contributing to the basic model fit by 6.0%. The association remained significant after adjusted for other SES indicators (R 2 change=0.8%). Middle-aged adults with income of SGD1,500 or lower had higher probability of being isolated than those with income of SGD5,000 or above (AME%=16.29) ( compared to those residing in HDB 5-room flats or private properties.

Self-perceived money insufficiency
Middle-aged adults who perceived money insufficiency were more likely to be socially isolated compared to those who perceived money sufficiency for basic living needs, even after adjusted for other SES indicators (AME%=32.70 and 18.65, respectively) (

Association between SES indicators and social isolation in older adults
Education level Older adults with low level of education had higher probability of being isolated (53.7%) compared to those with middle or high level of education (31.6% and 37.6%, respectively) ( Table 2). However, the inclusion of the variable did not improve the model fit significantly. In the fully adjusted model, there was also no significant association between education level and social isolation in older adults (Table   3, Model 2).

Employment status
Employment status and social isolation was not significantly associated in older adults (Table 2). It only contributed to the model fit by 0.3% in Model 1 and 0.1% to Model 2 (Table 3).

Personal income
Like employment status, the relationship between personal income and social isolation was not significant in older adults. Including it in the basic model contributed to the model fit by 0.2%, lower than other SES indicators ( Table 3, Model 1).

Housing type
Older adults residing in HDB 1-and 2-room flats were more likely to be isolated (68.6%) compared to those residing in HDB 5-room flats or private properties (

Discussion
This study examined the association of four objective SES indicators (including education level, employment status, personal income, and housing type) and one subjective SES indicator (selfperceived money insufficiency) with social isolation stratified by three age groups and the results indicated that different SES indicators were differently associated with social isolation in different age groups. We found that personal income, education level, self-perceived money insufficiency housing type, and employment status were individually associated with social isolation in young adults; each of the five SES indicators had separate association with social isolation in middle-aged adults; self-perceived money insufficiency and housing type was individually associated with social isolation in older adults. The magnitude of association between each SES indicator and social isolation varied in each age group, consistent to what reported by Geyer and colleagues [41]. The associations remained significant and varied in magnitude in each age group after fully adjusted for other SES indicators except employment status and housing type in young adults.
Educational level and personal income were associated with social isolation in young and middle-aged adults and employment status was associated with social isolation in middle-aged adults but not in adults aged 65 years and above. The results indicate that these individual-level objective SES indicators may be more suitable for young and middle-aged adults than older adults when study social isolation. Education level is related to non-material resources such as knowledge and skills, and is a strong determinant of employment and occupation as well as income [42,43], especially during early adulthood. Unemployment presented consistent association with increased odds of social isolation in middle-aged adults where being productively employed is the norm among this life stage.
The loss of daily contact with colleagues could be one mechanism. It also can be explained by selfwithdrawal from families and friends due to feelings of shame and embarrassment and/or the need to cut down on expenses of socializing [44]. Low income, which financially or emotionally prevents people from participating in various social activities [45] or restrict one's capability to obtain social support, results in social isolation [46]. Unemployment, however, was not associated with isolation in young adults after adjusted for other SES indicators. This probably can be explained by the reason for unemployment as the impact of unemployment depends on the reason for unemployment [47,48].
For young adults, unemployment was mainly voluntary or temporary, which might have little impacts on social isolation.
Housing is an important social determinant of health [49]. Housing type, a proxy of household SES in Singapore, was found to be associated with social isolation in middle-aged and older adults in this study with those residing in HDB 1-and 2-room flats having higher probability of social isolation, even after adjusted for other SES indicators. This is consistent with a Singapore study which reported that older adults residing in HDB 3-room flat or smaller had significantly higher risk of loneliness [6]. This suggests that housing type is a sensitive SES indicator that can be used to estimate the SES effect on social isolation in middle-aged and older adults. Older adults residing in small-sized housing tend to have higher prevalence of isolation and loneliness [50]. It implies that specific interventions (i.e. a wide range of organized group activities) to be provided at small-sized housing estates should have considerable potential to tackle the isolation issue in the residents.
Similar to a recent study conducted in Japan [51], this study also showed that individuals who felt money insufficiency for daily living needs were consistently found to be strongly associated with the likelihood of being socially isolated regardless of age groups. In general, individuals' perception of their money insufficiency explained the variance in social isolation better than any of the other indicators, which indicated that the subjective SES indicator had more of an effect on social isolation than did individual objective SES indicators. This probably can be partially explained by the phenomenon that individuals who felt money insufficiency may have inferiority complex which affects their personal relationships or social interaction negatively [52].
This study contributes to the literature by investigating the most commonly used indicators of SES as well as a subjective indicator of SES in relation to probability of social isolation in three adult age groups. There are also a few limitations. First, the study used self-reported data obtained from a representative population health survey. Although the surveyors were well-trained for administering the survey questionnaire, we cannot exclude the likelihood of reporting errors. Second, we used housing type as a proxy of household SES, however, the ownership of the property or wealth was not captured. This may affect our findings as a small portion of participants were tenants or lived in HDB rental flats and staying in a rental flat was found to be associated with loneliness [6]. Third, the authors tried to disentangle each SES indicator's independent impact on social isolation by adjusting for other indicators. However, we acknowledge that the interrelationship between each SES indicator would potentially build complex pathway to social isolation and simply adjusting for other indicators might not identify one indicator's independent impact on social isolation.

Conclusions
The relationship of SES with social isolation is complex with different SES indicators having varied association with social isolation in different age groups.

Availability of data and materials
According to the Data Protection Act Commission Singapore-Advisory Guidelines for the Healthcare Sector, all the individual data collected for the Population Health Index study are protected under the Act. As such, the datasets analysed during the current study are not publicly available. However, minimal dataset underlying the findings in the manuscript is available from the corresponding author on reasonable request. of the manuscript; BHH: conception and design, acquisition of financial support, and approval of the manuscript. All authors read and approved the final manuscript.