This observational cross-sectional study utilises data from the National Quality of Life and Health Survey (NQoLHS-2023/2024), conducted by Chile’s Ministry of Health and the Institute of Sociology at the Pontifical Catholic University of Chile. It targets non-institutionalised Chileans and foreign residents aged 15 and older living in private homes for at least six months. The study assesses perceptions of health-related quality of life across all regions of Chile using a stratified sampling method. Personal interviews were conducted between October 2023 and February 2024 with 16,590 participants, resulting in a sample of 6755 individuals aged 60 and over who responded independently ([43]).

Based on the theoretical framework proposed by [6], which suggests that social ties and relationships are crucial in determining health status ([6]) and addressing the limitations of current tools for measuring social isolation in community-dwelling populations, as noted by [32], the social isolation concept was operationalized using a 6-item measure adapted from the approach of [79], which evaluates domains from both external and internal perspectives of social isolation ([79]).

External indicators of social isolation include: 1) Current marital or civil status (1 = married, civil partner, or cohabitant; 0 = other statuses), 2) Social participation (1 = participates in one or more social organizations, including religious groups; 0 = does not participate), 3) Social network support (If you were in financial trouble, would you have relatives or friends you could count on for help? 1 = yes; 0 = no), 4) Emotional support (When you have problems, do you have anyone you trust to ask for help or advice? 1 = yes; 0 = no). Internal indicators of social isolation include: 1) Level of satisfaction with time shared with family (Are you satisfied with how you and your close family share time together? 1 = yes; 0 = no), and 2) Sense of belonging to the community (How often do you feel like you belong to your neighbourhood, a social group, or similar? 1 = always/often; 0 = never/rarely). Because the SI is a negative construct, the value of each indicator was reversed before counting the total score. The total score ranges from 0 to 6, and because the distribution of total scores was negatively skewed, the cut-off score was set at the 67th percentile. A score of 3 or more indicates social isolation. With these criteria, 2222 individuals were identified as socially isolated. SI was categorised as a binary variable (1 = socially isolated; 0 = not socially isolated).

Loneliness was measured using the Three-Item Lonely Scale (TILS) from [27], which is part of the Revised UCLA Loneliness Scale (R-UCLA; [53]) ([27]; [53]). It includes questions about lack of companionship, feeling left out, and feeling isolated. Responses are scored on a 1 - 3 scale, yielding total scores from 3 to 9. A cut-off score at the 75th percentile (5 or higher) was operationally defined to better capture loneliness. Only those who scored 5 points were classified as lonely if they answered “often” to the first question. This method identified 1785 individuals experiencing loneliness.

2.3.1. Socio-Demographic Factors

The models for social isolation and loneliness examined various factors, including age, gender, education, income, employment, retirement status, and household size. Participants were categorised by age into the following groups: 60 - 64 years (reference), 65 - 69 years, 70 - 74 years, and 75 years and older. In the social isolation model, gender was binary, with women as the reference group; in the loneliness model, men served as the reference group. Education level was categorised into four groups: illiterate/elementary, high school, technical, and graduate, with the latter used as the reference group. Household income was categorised into quintiles, with the fifth quintile as the reference. Employment status and retirement status were treated as binary in their respective models, with individuals who were neither employed nor retired as reference groups. Lastly, household size was categorised into four groups: 1 person, 2 people, 3 people, and 4 or more people, with the latter serving as the reference group.

2.3.2. Health and Lifestyle Factors

Both social isolation and loneliness models considered factors like self-rated health, sensory deficits, multimorbidity, disability level, depression/anxiety or other mental health issues, physical activity, and nutritional status. Physical mobility—the ability to move from one place to another—was included only in the social isolation model.

Self-rated health was categorised into three groups: very poor/poor, fair, and good/very good, with the latter group serving as the reference. Sensory deficiency, including blindness and/or deafness, was a binary variable, and individuals without sensory deficits served as the reference group. Multimorbidity was assessed by the number of self-reported physician diagnoses of 14 chronic diseases. These conditions include hypertension, diabetes mellitus, acute myocardial infarction, heart failure/arrhythmias, stroke, arthritis/osteoarthritis, depression, anxiety or other mental health disorders, chronic obstructive pulmonary disease (COPD), liver cirrhosis/chronic liver damage, cataracts/glaucoma, chronic pain (lasting more than 3 months), urinary incontinence, chronic kidney failure and other chronic diseases. Multimorbidity was defined as having at least two of the fourteen selected conditions mentioned above, and was categorised as a binary variable (1 = multimorbidity; 0 = none or one chronic disease).

The disability level was a binary variable, with those with moderate or severe disability as the group of interest, and those with mild or no disability as the reference group. Depression/anxiety or other mental health disorders were included as a binary variable (1 = yes; 0 = no). According to WHO guidelines, older adults should engage in 150 to 300 minutes of moderate-intensity or 75 to 150 minutes of vigorous-intensity aerobic activity, or an equivalent combination of both throughout the week ([66]). Physical activity was included as a binary variable: individuals who met the WHO recommendations were classified as physically active, whereas those with no activity or lower levels of activity were classified as the control group. Physical mobility was measured with the question, “How well do you get from one place to another?” Responses ranged from “very bad” to “very good”. Responses were grouped into three categories: “very bad/bad”, “neither good nor bad”, and “good/very good”, with the last category as the reference group.

The WHO defines obesity in adults as a BMI ≥ 30 kg/m²; however, given the issue of sarcopenic obesity, this study uses a cut-off of BMI ≥ 32 kg/m² for older adults, per Chile’s Ministry of Health guidelines ([42]). Due to the exploratory nature of the study, all four nutritional categories—underweight (BMI < 23.0), normal weight (BMI 23.1 - 27.9), overweight (BMI 28.0 - 31.9), and obesity (BMI ≥ 32)—were included in the loneliness model, with the normal weight group as the reference group.

2.3.3. Individual Trait Factors

The study examines the psychological and sociological factors that affect human interactions. Key variables included in both models were: discrimination, general life satisfaction, family life satisfaction, personal relationship satisfaction, and caregiver activity. The social isolation model incorporates perceptions of living in a safe neighbourhood, smartphone use, and feelings of loneliness, whereas the loneliness model includes satisfaction with oneself, a sense of belonging to the neighbourhood, acceptance of one’s physical appearance, pet ownership, and social isolation.

Discrimination was measured with the question, “Have you felt discriminated against in the last year?” Response options were yes or no. The no option was the reference group. General life satisfaction was measured with the question, “How do you feel about your life in general?” Responses ranged from 1 to 7, where 1 corresponded to “very bad” and 7 to “very good”. Responses were grouped into three categories: “very unsatisfied/unsatisfied” (scores 1 - 3), “neither satisfied nor unsatisfied” (score 4), and “satisfied/very satisfied” (scores 5 - 7), with the last category as the reference group. Family life satisfaction was measured with the question, “How do you feel about your family life?” Responses ranged in a scale from 1 to 7, where 1 corresponding to “very bad” and 7 to “very good”. Responses were grouped into three categories: “very unsatisfied/unsatisfied” (scores 1 - 3), “neither satisfied nor unsatisfied” (score 4), and “satisfied/very satisfied” (scores 5 - 7), with the last one category as the reference group.

Satisfaction with oneself was measured by the question: “In the last 2 weeks, how satisfied are you with yourself?” Responses ranged from 1 (“Very dissatisfied”) to 5 (“Very satisfied”) and were categorized as follows: “Very dissatisfied/Dissatisfied” (scores 1 - 2), “Neither satisfied nor dissatisfied” (score 3), and “Satisfied/Very satisfied” (scores 4 - 5), with the last group as the reference. Personal relationship satisfaction was measured with the question: “In the last 2 weeks, how satisfied are you with your personal relationships?” Responses ranged from 1 (“Very unsatisfied”) to 5 (“Very satisfied”). They were grouped into three categories: “Very unsatisfied/unsatisfied” (scores 1 - 2), “Neither satisfied nor unsatisfied” (score 3), and “Satisfied/very satisfied” (scores 4 - 5), with the last group as the reference.

Acceptance of physical appearance was measured by the question: “In the last two weeks, have you been able to accept your physical appearance?” Responses were scored from 1 to 5, with 1 indicating “Not at all” and 5 meaning “Totally.” The scores were categorised as “Not at all/little” (1 - 2), “Moderate” (3), and “Quite a bit/totally” (4 - 5), with the last category serving as the reference group.

Safety perception in the neighbourhood was assessed with the question: “How safe do you feel walking alone in your neighbourhood after dark?” The response options were: very unsafe, somewhat unsafe, moderately safe, and very safe. For analysis, these were grouped into a dichotomous variable: 0 for “very unsafe/somewhat unsafe”, and 1 for “moderately safe/very safe”. A sense of belonging to the neighbourhood was measured by the question: “How much do you agree with the statement that you feel like you belong in your neighbourhood?” Responses were scored on a 1 - 5 scale, where 1 = “Strongly agree” and 5 = “Strongly disagree”. The responses were categorised as follows: “Very agree/agree” (scores 1 - 2), “Neither agree nor disagree” (score 3), and “Disagree/strongly disagree” (scores 4 - 5). The first category serves as the reference group.

The study highlights that caregiving among older adults has notable psychological and social effects, particularly influencing mental health and overall well-being ([9]; [78]). Caregivers were identified with the question: “Apart from your usual activities, are you responsible for caring for children, elderly individuals with moderate to severe dependency, or chronically ill individuals?” Response options were: “Yes, as the sole caregiver”, “Yes, shared with another person”, and “No, I am not in charge of anyone”, with the last option as the reference category.

Research indicates that smartphone use can help older adults strengthen social connections and alleviate feelings of loneliness ([58]; [39]). To measure usage, participants were asked, “In general, how many hours a day do you typically use your smartphone?” Responses varied from 0 to 24 hours. Due to the skewed nature of the data, usage was categorised based on the 90th percentile cut-off, resulting in two higher usage categories: moderate use (1 - 4 hours per day) and excessive use (5 or more hours per day). A “no use” category served as the reference group. Similarly, research indicates that pet ownership, particularly of dogs, is associated with lower loneliness among older adults, especially among women living alone ([28]; [40]). In this study, pet ownership was measured with the question, “Do you have pets?” Response options were yes or no. The no option was the reference group.

Descriptive statistics were used to characterise the sample. The sample design was integrated into the analysis to ensure valid inferences about the population. Multicollinearity was assessed using a Spearman correlation matrix and variance inflation factors (VIFs). Multivariate logistic regression models were used to examine predictors of social isolation and loneliness. Additionally, separate models were developed for men and women to investigate gender differences. To improve understanding of the results, average marginal effects (AMEs) were calculated to estimate average adjusted probabilities (AAPs).

A univariate analysis was performed for each variable, focusing on those with p-values < 0.25 from the Wald test of logistic regression. A purposeful approach was used to refine covariate selection, including variables with significance levels of 0.1 or 0.15, and excluding non-significant covariates that were not considered confounders ([26]). Multivariate logistic regression models for social isolation and loneliness were developed from survey data, utilising svyset and svy commands to adjust for sampling design. Odds ratios, standard errors, and 95% confidence intervals were estimated. Model goodness-of-fit was assessed using the Hosmer-Lemeshow test, and predictive ability was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC). Average marginal effects (AMEs) were calculated using the dx/dy command and the delta method to standardise standard errors, showing changes in predicted probabilities as predictors moved from the reference category, while controlling for other variables ([72]). Statistical significance was assessed at the 10% level using Wald’s chi-square test, with analyses performed in Stata 14.0.

The NQoLHS-2023/2024 survey interviewed 6755 older adults, with an average age of 70.6 years (SD: 7.65). Among them, 30.4% reported social isolation (95% CI: 11.07 - 14.08), and 23.4% reported loneliness (95% CI: 21.55 - 25.26). These figures were slightly lower than the 32.9% and 26.5% reported in the raw data, respectively. Women accounted for 56.0% of those socially isolated, with an average age of 71.3 years (SD: 7.91) and a prevalence of 30.0% (95% CI: 27.69 - 32.34). In contrast, men with an average age of 71.5 years (SD: 7.51) reported 31.0% (95% CI: 27.90 - 34.37) of social isolation. The gender difference was not statistically significant. Prevalence varied by age: 29.1% in individuals aged 60 - 64, 30.5% in those aged 65 - 69, 27.1% in those aged 70 - 74, and 33.8% in those aged 75 and older.

Regarding loneliness, 60.4% of the participants were women, with an average age of 70.6 years (SD: 7.63) and a prevalence of 25.5% (95% CI: 23.3 - 27.8). In men, with an average age of 70.9 years (SD: 7.49), the prevalence was 20.7% (95% CI: 18.5 - 23.1). The gender difference was statistically significant. Prevalence also varied by age: 21.7% among individuals aged 60 - 64, 23.8% among those aged 65 - 69, 23.6% among those aged 70 - 74, and 24.3% among those aged 75 and older.

Additionally, the results indicate that 53.5% of women do not experience social isolation and loneliness, and 13.2% experience both. Among men, 54.4% did not experience social isolation and loneliness, and 13.3% experienced both. Table 1 presents survey data on the sample characteristics related to social isolation and loneliness among older adults in Chile.

Table 1. Sample characteristics according to social isolation and loneliness in Chilean older adults. NQoLHS-2023/2024.

ref. = reference group; SI: Social Isolation; L: Loneliness. *: p-value < 0.10; **: p-value < 0.05; ***: p-value < 0.001; ^ns: no significant.

The multicollinearity diagnostic showed that most variables had low correlations (rho < 0.30), except for a few pairs with moderate correlations, including self-rated health and physical mobility, retirement status and age groups, retirement status and working status, personal relationship satisfaction and satisfaction with oneself, and general life satisfaction and family life satisfaction. The mean variance inflation factor (VIF) was low at 1.25 and 1.26 for the social isolation and loneliness models, respectively, indicating no multicollinearity. The goodness-of-fit tests confirmed the models’ significance, and the ROC curve area demonstrated predictive powers of 72% and 77%. Alcohol consumption and smoking were examined but removed from the models. Alcohol consumption was excluded due to 50% missing data, and smoking was not significantly associated in either univariate or multivariate models, which affected the models’ robustness and simplicity.

The analysis of the overall survey data model identified several key factors contributing to social isolation among older adults in Chile. These include being men, lower educational levels, living alone, work status, being underweight, experiencing loneliness, having sole caregiving activity, dissatisfaction with overall life, family life and personal relationships, and moderate use of smartphone. Conversely, excessive smartphone use and shared caregiving activities, were protective factors significantly associated with social isolation. It must be noted that age groups, household incomes, self-rated status, sensory deficiencies, depression, disability, multimorbidity, physical inactivity, physical mobility, discrimination and live in an unsafe neighbourhood did not show a significant association with social isolation in the survey model. Details on logistic regression and average marginal effects (AMEs) from survey data on overall social isolation among older adults in Chile are provided in the supplementary materials (Table S1).

In the following tables, only the significant factors are presented. Table 2 presents the odds ratios (ORs) and average marginal effects (AMEs) with 95% confidence intervals from the logistic regression model using survey data to examine social isolation in the overall sample of older women in Chile.

Table 2. Predictors and marginal effects of social isolation from the older women’s survey data.

dy/dx: the marginal change of the outcome (dy) concerning the change from the ref. group of the factors (dx). *: p < 0.1; **: p < 0.05; ***: p < 0.0001; ^ns: no significant.

Key predictors of social isolation in the women’s survey data include lower educational attainment, membership in the first household income quintile, living alone, dissatisfaction with general life and personal relationships, neutral satisfaction with family life, neutral physical mobility, and feeling lonely. Conversely, protective factors against social isolation in older women included sharing caregiving, being overweight and moderate smartphone use.

Analysis of AMEs shows that women with technical school education are 16.9% more likely to experience social isolation than university graduates. Individuals who are illiterate or have only completed elementary school are 10.4% more likely, while attendees of high school face an 10.3% higher risk than university graduates. Belonging to the first household income quintile increases the likelihood of social isolation by 12.2% compared to those in the fifth quintile. Living alone increases the likelihood of social isolation by 8.5% compared to those living with four or more people. Women who report being highly dissatisfied and neutral with their overall life have a 15.5% and 4.4% higher risk of social isolation, respectively, compared with those who are satisfied. Furthermore, those who report being highly dissatisfied or neutral with their personal relationships have a 15.2% and 8.2% higher risk of social isolation, respectively, compared with those who are satisfied. Women who experience feelings of loneliness are 9.5% more likely to be isolated. Additionally, women with neutral physical mobility have a 6.0% higher risk of social isolation than the reference group. Conversely, women with moderate smartphone use, sharing caregiving and being overweight, are 9.0%, 6.5% and 5.8% less likely to feel socially isolated compared to their reference groups, respectively.

Table 3 presents the odds ratios (ORs) and average marginal effects (AMEs) with 95% confidence intervals from a logistic regression examining social isolation among older Chilean men.

Table 3. Predictors and marginal effects of social isolation from the older men’s survey data.

dy/dx: the marginal change of the outcome (dy) concerning the change from the ref. group of the factors (dx). *: p < 0.1; **: p < 0.05; ***: p < 0.0001; ^ns: no significant.

The analysis of men’s survey data identified several factors associated with social isolation. Key predictors include the lowest educational attainment, working status, living alone, sole caregiving, feeling loneliness, being underweight, moderate/severe disability and dissatisfaction with family life. No significant protective factors against social isolation were observed in older men.

Older men who are illiterate or have only completed elementary school are 18.6% more likely to be socially isolated than university graduates. Being underweight increases the likelihood of social isolation by 15.4% compared to those in normal weight. Living alone raises the likelihood of isolation by 14.4% compared to living with four or more people. Sole caregiver men are 13.2% more likely to feel socially isolated compared to those without caregiving activities. Men who experience feelings of loneliness are 11.7% more likely to be isolated. Furthermore, men who report being neutral regarding family life satisfaction have a 8.8% higher risk of social isolation, compared with those who are satisfied. Additionally, men with moderate/severe disability and working status have about a 6.0% higher risk of isolation compared to their reference group peers.

The analysis of the overall survey data model identified several key factors contributing to loneliness among older adults in Chile. These predictors include younger age groups, lower educational levels, retiree status, the number of people living in the household, moderate to severe disabilities, depression or other mental health issues, experiences of discrimination, social isolation, caregiving responsibilities, dissatisfaction with oneself, lack of acceptance of one’s physical appearance, dissatisfaction with family life, feelings of not belonging to the neighborhood, and pet ownership. Details on logistic regression and average marginal effects (AMEs) from survey data on loneliness among older adults in Chile are provided in the supplementary materials (Table S2). Table 4 presents the odds ratios (ORs) and average marginal effects (AMEs) with 95% confidence intervals from a logistic regression model using survey data to examine loneliness in a sample of older Chilean women.

Table 4. Predictors and marginal effects of loneliness from older women’s survey data.

dy/dx: the marginal change of the outcome variable (dy) concerning the change from the reference group of the factors (dx). ^ns: no significant; *: p < 0.1; **: p < 0.05; ***: p < 0.0001.

Key factors linked to loneliness among women include dissatisfaction with family life, difficulty accepting their physical appearance, dissatisfaction with themselves, experiences of discrimination, mental health issues like depression, living arrangement, malnutrition status, low educational attainment, age groups, ambivalence about belonging to their neighborhood, sole caregiving responsibilities, and pet ownership. In contrast, being a retiree and rating one’s health as fair are protective factors against loneliness.

The AMEs indicate that older women who are very dissatisfied or neutral about their family life are 22.2% and 15.9% more likely to feel lonely, respectively, compared to those who are satisfied. Women who have difficulty with accepting their physical appearance, whether they feel not at all or a little satisfied or moderately satisfied, are 10.7% and 7.8% more likely to experience loneliness. Additionally, older women who are dissatisfied with themselves are more likely to feel lonely, with rates of 6.1% among dissatisfied women and 8.2% among neutral women, compared to those who are satisfied. Women facing discrimination are 16.8% more likely to feel lonely, while those with depression or social isolation are 15.7% and 7.5% more likely compared to their reference groups, respectively.

Individuals who are illiterate or have only an elementary school education, and those with a high school education, are 11.0% and 11.1% more likely to feel lonely compared to university graduates. Among age groups, those aged 65 - 69, 70 - 74, and 75 or older are 4.4%, 7.8%, and 6.5% more likely to feel lonely than those aged 60 - 64. Women living alone are 6.4% more likely to feel lonely, while those living with two or three others are 3.3% and 4.7% more likely, respectively, than those living with four or more people. Underweight and overweight individuals are 7.2% and 5.6% more likely to feel lonely, respectively. Women who share caregiving responsibilities, have moderate to severe disabilities, or are pet owners are 7.3%, 5.5%, and 6.4% more likely to feel lonely compared to their peers. Conversely, retiree women and those rating their health as fair are 4.1% and 3.8% less likely to feel lonely, respectively, than their respective reference groups. Table 5 presents the odds ratios (ORs) and average marginal effects (AMEs) with 95% confidence intervals from the logistic regression model using survey data to examine loneliness in a sample of Chilean older men.

Table 5. Predictors and marginal effects of loneliness from older men’s survey data.

dy/dx: the marginal change of the outcome variable (dy) concerning the change from the reference group of the factors (dx). ^ns: no significant; *: p < 0.1; **: p < 0.05; ***: p < 0.0001.

Key factors associated with loneliness among older men, according to survey data, include experiences of discrimination, living arrangements, lower educational attainment, feelings of not belonging to their neighbourhood, caregiving responsibilities, social isolation, depression or other mental health issues, and dissatisfaction with themselves or their family life. Interestingly, being underweight or overweight serves as a protective factor against loneliness in older men.

The AMEs show that older men facing discrimination are 19.5% more likely to feel lonely. Those who feel neutral or lack a sense of belonging in their neighbourhood are 15.8% and 11.9% more likely to experience loneliness, respectively. Education also impacts loneliness; individuals with illiterate/elementary education and those with high school or technical school attainments are 12.1%, 7.4%, and 7.5% more likely to feel lonely compared to university graduates. Men who are dissatisfied with themselves or their family life are 10.4% and 6.9% more likely to feel lonely, respectively. Men with sole caregiving responsibilities are 10.6% more likely to feel lonely compared to those who are not involved in caregiving. Additionally, socially isolated men are 8.3% more likely to feel lonely, and those dealing with depression or mental health issues are 6.5% more likely compared to their reference groups. Conversely, both underweight and overweight older men are 5.8% less likely to feel lonely compared to those of normal weight.