Examining associations of social disconnectedness change patterns with subsequent healthcare utilization and costs among public healthcare users in Central Singapore

Abstract

Social disconnectedness contributes to adverse health outcomes, yet how patterns of change in social disconnectedness are associated with healthcare utilization and costs in Singapore remains underexplored. Data from a longitudinal population health survey were linked with an institutional administrative database. Four patterns of change in social disconnectedness were derived from social isolation (Lubben Social Network Scale-6) and loneliness (three-item UCLA Loneliness Scale) statuses measured at two annual waves. Healthcare utilization and cost outcomes were extracted for 1- and 3-year (cumulative) after wave 2. Participant characteristics at wave 1 formed the baseline and were compared using Chi-square or Kruskal–Wallis tests. The associations between change patterns and subsequent healthcare utilization and costs were respectively examined using two-step hurdle models, adjusting for covariates. Persistent social disconnectedness (16.9% out of 1182 participants) was associated with increased odds of hospitalizations and emergency department (ED) visits and costs, and polyclinic visits, at both follow-ups, compared to participants who remained socially connected (56.7%). Participants who became socially connected (15.2%) exhibited higher odds of hospitalization at 3-year follow-up and higher odds of ED visits and costs at both follow-ups. Among public healthcare users, promoting social connectedness may help reduce costly hospitalizations and ease resource strains on polyclinic services.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-56960-1.

Introduction

Definitions and global prevalence of social isolation and loneliness

Social isolation captures objective measures of social disconnectedness, including the number and frequency of contacts an individual has with family and friends, whilst loneliness reflects the subjective experience of social disconnectedness such as the emotional distress caused by perceived discrepancies between existing and desired social relationships¹. In 2025, the World Health Organization Commission on Social Connection issued a landmark report which recognized social disconnectedness as a pressing public health challenge and urged policymakers to prioritize addressing these issues². Social disconnectedness has become a growing concern, particularly in industrialized societies which have become increasingly fragmented due to a multitude of factors, including aging populations and shifts towards nuclear family structures^3–6. The scale of this challenge is evident in recent prevalence data showing that social isolation affects approximately 25% of older adults⁷, while loneliness is experienced by approximately one-third to over half of adults in developed countries including the United States, European Union nations, Korea, and Japan^8–11.

As social isolation and loneliness share some common risk factors (e.g., living alone, being widowed or divorced)¹², and evidence suggests that both should be addressed concurrently to achieve marked improvements in overall health and wellbeing^12,13, this study examined social isolation and loneliness together under the broader concept of social disconnectedness. This approach is also aligned with how population-level intervention strategies in Singapore, such as Age Well SG¹⁴, tend to promote social connectedness broadly instead of targeting social isolation or loneliness individually.

Social disconnectedness, health impacts and healthcare utilization

Social disconnectedness is a critical social determinant of health, with extensive research noting the associations of social disconnectedness with adverse physical and mental health outcomes, including elevated risk of premature mortality¹⁵, depression severity^16,17, and engagement in unhealthy lifestyle behaviors¹⁸.

Social disconnectedness may in turn influence healthcare-seeking behavior and utilization patterns. The stress-buffering hypothesis suggests that a lack of social connections contributes to physiological and psychological stress responses¹⁹, which could increase the risk of illness onset and progression and the subsequent need for acute care. On the other hand, social disconnectedness could lead to healthcare avoidance through factors such as lower health literacy²⁰ and financial constraints²¹. These multi-faceted and competing mechanisms underlying healthcare-seeking behaviors could help illuminate the mixed findings noted below.

Generally, prospective studies utilizing administrative data have consistently found positive associations between social disconnectedness and hospitalizations²², emergency department (ED) visits^22,23, and general practitioner (GP) visits²³. In contrast, studies relying on retrospective, self-reported healthcare utilization have produced more variable results, with some noting increased hospitalizations²⁴, ED visits^24,25, and GP visits^24,26,27, while others have found no significant associations with hospitalizations^27,28 , primary care utilization²⁹, or even reduced likelihood of utilizing primary care services³⁰. These inconsistencies may also be reflective of several context-specific factors such as population characteristics, healthcare system structures, and selected measures of social disconnectedness and healthcare utilization outcomes.

Social disconnectedness and healthcare planning in Singapore

Recent demographic trends in Singapore (e.g., population ageing, rising singlehood and solo living) are likely to have contributed to growing social disconnectedness³¹. For instance, age-related social network reduction and adverse life events such as unemployment or disability heighten vulnerability to social disconnection^32,33. Moreover, despite enhanced connectivity, digital transformations may paradoxically exacerbate social disconnectedness by reducing meaningful in-person interactions^34–36. While most existing research has examined social disconnectedness at a single time point, social disconnectedness is inherently dynamic, fluctuating over time due to life transitions, health changes, or environmental factors, and different patterns of change may have differential associations with healthcare utilization. Given Singapore’s imperative to meet rising healthcare demands while containing costs, understanding these change patterns and their associations with healthcare utilization and costs is essential for informing evidence-based policy decisions and resource allocation within initiatives like Healthier SG³⁷ that emphasize preventive care and population health management.

Study aims and objective

As Singapore-based studies examining associations of patterns of change in social disconnectedness with subsequent utilization and costs in public healthcare settings are scarce, this study aims to address two key evidence gaps. First, there is limited knowledge on healthcare utilization and cost patterns among socially disconnected individuals who became socially connected over time, which is essential for evaluating the potential benefits of social connectedness interventions. Second, few studies have examined associations between social disconnectedness and specialist outpatient clinic (SOC) visits and costs, despite SOCs being a significant component of Singapore’s public healthcare delivery.

The objective of this exploratory study is to investigate how four patterns of change in social disconnectedness (encompassing social isolation and loneliness) are associated with subsequent 1-year and cumulative 3-year healthcare utilization and gross cost (defined as the total cost before subsidies, including consultation fees, diagnostic tests, procedures, and medications dispensed) among healthcare users in Central Singapore at four public healthcare settings: in-patient wards, EDs, SOCs, and polyclinics. This study employed a prospective design to minimize possible reverse causality, whereby healthcare utilization might have contributed to changes in social disconnectedness.

Methods

Study design and participants

Data were derived from the Population Health Index (PHI) Study Phase 1, a longitudinal population health survey initiated in 2015 that collected data annually over three consecutive years from a random sample of Singapore citizens and permanent residents aged 21 years and above residing in the Central region. The Central region was strategically selected as the study area because the National Healthcare Group (NHG Health), one of Singapore’s three public healthcare clusters, operates its network of healthcare facilities, including hospitals, polyclinics, and specialist centers, in this region. This geographical alignment facilitated comprehensive linkage between survey on social disconnectedness and administrative records of healthcare utilization and costs within the same healthcare system. Detailed descriptions of the PHI survey instrument development, stratified sampling methodology, and data collection procedures have been detailed elsewhere^16,38.

The PHI study received ethical approval from the National Healthcare Group Domain Specific Review Board (Reference Number: 2015/00269). Written informed consent was obtained from all individuals for survey participation and for optional linkage of their responses with administrative database. Participants were informed of their right to withdraw from the study at any time. To ensure participant confidentiality, all data were de-identified prior to analysis and stored securely with access restricted to authorized research personnel. In accordance with institutional data retention policies, all study data will be retained for a minimum of 6 years following study completion.

A total of 1942 adults completed the first survey (response rate of 53.3%), of whom 1704 (87.7%) provided additional consent to link their survey responses with administrative database for research purposes. For individuals who consented, their PHI survey responses were linked with healthcare utilization data extracted from the institution’s administrative database. This study’s sample comprised participants who completed both waves 1 and 2 surveys and provided consent for data linkage. Participants were excluded if they: (1) had missing social isolation or loneliness scores in either wave 1 or wave 2 survey, or (2) could not be linked to the institutional administrative database. This resulted in a final sample of 1182 for analysis. The participant selection process is illustrated in Fig. 1.

Fig. 1.

Selection process of study participants.

Measures

Social isolation

Social isolation was assessed using the Lubben Social Network Scale-6 (LSNS-6), a validated instrument that measures social disconnectedness³⁹. The scale comprises two subscales (family and friend), with three questions each, scored on a 6-point Likert scale ranging from 0 (none) to 5 (nine or more): 0 = 0, 1 = 1, 2 = 2, 3 = 3–4, 4 = 5–8, 5 =  ≥ 9. The six questions were: (1) “How many [relatives/friends] do you see/hear from at least once a month?”, (2) “How many [relatives/friends] do you feel at ease with whom you can talk about private matters?”, and (3) “How many [relatives/friends] do you feel close to such that you could call on them for help?”. The scale demonstrated good internal consistency in this study (Cronbach’s alpha = 0.80 for the overall scale; 0.80 and 0.81 for the family and friend subscales respectively). The total score (range: 0–36) was calculated by summing up all six item scores. Following established thresholds, participants were categorized as socially isolated if the total score ranged between 0 and 12³⁹.

Loneliness

Loneliness was assessed using the University of California Los Angeles (UCLA) Loneliness Scale consisting of three questions: 1) “How often do you feel that you lack companionship?”, 2) “How often do you feel left out?”, and 3) “How often do you feel isolated from others?”⁴⁰.

Each question has three frequency-related response options: 1 = Hardly ever, 2 = Some of the time, 3 = Often. The scale demonstrated good overall internal reliability in this study (Cronbach’s alpha = 0.87). The total score (range: 3–9) was calculated by summing up all three item scores. Based on the summed scores, participants were categorized as either lonely (total score: 6–9) or not lonely (total score: 3–5)⁴¹.

Change patterns in social disconnectedness

First, a binary variable for social disconnectedness was created to classify participants as either socially connected (not socially isolated and not lonely, coded as 0) or socially disconnected (socially isolated and/or lonely, coded as 1 for each time point (Supplementary File, Table S1). Thereafter, this study’s independent variable, a four-level categorical variable, was constructed to capture four patterns of change in social disconnectedness between waves 1 and 2 surveys: remained socially connected (socially connected at both waves, coded as 0), became socially connected (socially disconnected at wave 1 but socially connected at wave 2, coded as 1), remained socially disconnected (socially disconnected at both waves, coded as 2), and became socially disconnected (socially connected at wave 1 but socially disconnected at wave 2, coded as 3).

Healthcare utilization and costs

Healthcare utilization was defined as the number of visits at each care setting. The number of visits and gross cost incurred by each participant at four care settings (in-patient wards, EDs, SOCs, and polyclinics) were extracted for three time points: 1 year prior to the wave 1 survey (establishing the baseline healthcare utilization and cost), 1 year after the wave 2 survey, and 3 years after the wave 2 survey (cumulative).

Covariates

Models adjusted for baseline healthcare utilization or costs (1 year prior to wave 1 survey) and the following baseline socio-demographics from wave 1 survey: age group (younger adults aged 21–59 years, older adults aged ≥ 60 years), sex (male, female), ethnicity (Chinese, non-Chinese), education (post-secondary school & above, secondary school & below), employment status (employed, not employed), marital status (married, single/divorced/widowed/separated), living alone (yes, no), and perceived money insufficiency for basic living needs (yes, no); health-related data included the number of chronic conditions (0–1, ≥ 2) out of a list of 17 conditions⁴² and depressive symptoms measured using the Patient Health Questionnaire-9 (PHQ-9)⁴³.

Multicollinearity among covariates was assessed using Variance Inflation Factors (VIF). With all adjusted generalized VIF values below 2, there are no significant multicollinearity concerns^44,45.

Sample size calculation

To determine the minimum required sample size, the following formula for regression analyses was utilized, with an assumption of a medium effect size (significance level of 5% and statistical power of 80%) between the independent variables (IVs) and dependent variable (DV): N ≥ 50 + 8m, where m is the number of IVs in the model⁴⁶.

For this approach, each level of a categorical variable would be counted as one IV⁴⁶. For instance, the continuous PHQ-9 variable is one IV while the categorical variable for change in social disconnectedness with four levels would be counted as four IVs. With 23 IVs in total, the minimum sample size required was 234 (50 + (8 × 23) = 234). This study’s sample of 1,182 participants is therefore adequate.

Statistical analysis

Baseline demographics and covariates were described for the overall study population and by the four patterns of change in social disconnectedness. Categorical variables were presented as frequencies and percentages, while continuous variables were described using means and standard deviations (SD). For categorical variables, between-group differences were assessed using Chi-squared tests. As the continuous variables in this study follow non-normal distributions, both mean [SD] and median [interquartile range (IQR)] were reported, and non-parametric Kruskal–Wallis tests were used for comparison across groups.

Healthcare utilization and costs were analyzed as separately by care setting at two timepoints: 1 and 3 years (cumulative) after the wave 2 survey. Healthcare utilization and cost data typically exhibit excess zeros, with many individuals exhibiting no utilization and cost incurred, and right-skewed distributions among users⁴⁷. A two-step hurdle model was selected for the analyses as it provides better statistical fit compared to single-step models for zero-inflated data and enables separate examination of: 1) the likelihood of any healthcare use (logistic regression), and 2) the extent of utilization among users (count/continuous models)^47,48. This approach also has distinct policy implications, as factors affecting whether someone seeks healthcare may differ from factors affecting how much they utilize healthcare services once care is sought.

For healthcare utilization, the number of visits was dichotomized (0: no visits; 1: ≥ 1 visit), and logistic regressions were conducted in the first step, controlling for baseline utilization (yes, no) and covariates. The coefficients were exponentiated and results were presented as odds ratios (OR) with 95% confidence intervals (CIs). In the second step, participants with no utilization were excluded, and generalized linear model (GLM) analyses were conducted, adjusting for baseline utilization (number of visits) and covariates. Prior to fitting the GLM, overdispersion was assessed using the performance package and check overdispersion function in R Studio⁴⁹. Values greater than 1 would indicate that the variance is larger than the mean, and a p value of < 0.05 would suggest that the overdispersion is significant and the Poisson model assumption is violated. This verification is important as violation of the Poisson model assumption of equal mean–variance relationship could lead to inflated Type 1 error rates in hypothesis testing⁵⁰. For hospitalizations and ED visits in the subsequent 1 year, no overdispersion was detected and analyses were conducted using Poisson GLMs. For the other care settings and at 3-year (cumulative) follow-up, overdispersion was detected and negative binomial GLMs were employed instead. The coefficients were exponentiated, and results were expressed as incidence rate ratios (IRR) with 95% CIs.

Similarly, for healthcare costs, costs were dichotomized (0: no cost incurred; 1: cost incurred) and logistic regressions were conducted in the first step, controlling for baseline cost (yes, no) and covariates. The coefficients were exponentiated and results were presented as ORs with 95% CIs. In the second step, participants with no cost incurred were excluded, and GLM analyses were employed with Gamma distribution and log link to account for the typically right-skewed distribution of cost data among users, controlling for baseline cost (continuous) and covariates. The coefficients were exponentiated, and results were expressed as cost ratios (CR) with 95% CIs.

All analyses were conducted using R Studio version 4.5.0, with statistical significance set at p < 0.05.

Results

Baseline characteristics

The majority of the 1,182 participants were younger adults (61.3%), female (55.0%), of Chinese ethnicity (78.8%), and had attained secondary school education or below (59.1%). Most were employed (60.7%), married (72.3%), did not live alone (89.2%), had perceived money sufficiency (84.3%), and had no multimorbidity (67.2%).

Between the two survey waves, approximately 3 in 10 participants experienced changes in social disconnectedness. A slightly higher proportion became socially connected (15.2%) compared to those who became socially disconnected (11.2%). More than half of the participants (56.7%) remained socially connected, while 16.9% remained socially disconnected. Those who remained socially disconnected were predominantly older adults (62.0%), had attained secondary school education or below (87.0%), were not employed (61.5%), and had multiple chronic conditions (54.5%). On the other hand, those who remained socially connected were mostly younger adults (69.9%), had attained post-secondary school education or above (54.6%), were employed (67.9%), and had no multimorbidity (76.3%). The proportion of participants who had perceived money insufficiency was almost six times higher among the group who remained socially disconnected (39.5%) than the group who remained socially connected (6.7%). Further details on participant characteristics at baseline are presented in Table 1.

Table 1.

Baseline characteristics of study participants by patterns of change in social disconnectedness.

Characteristics	Overall n (%) (n = 1182)	Remained socially connected n (%) (n = 670, 56.7%)	Became socially connected n (%) (n = 180, 15.2%)	Remained socially disconnected n (%) (n = 200, 16.9%)	Became socially disconnected n (%) (n = 132, 11.2%)	P value
Age						< 0.001a
Younger adults (21–59 years old)	725 (61.3%)	468 (69.9%)	99 (55.0%)	76 (38.0%)	82 (62.1%)
Older adults (≥ 60 years old)	457 (38.7%)	202 (30.1%)	81 (45.0%)	124 (62.0%)	50 (37.9%)
Gender						< 0.001a
Female	650 (55.0%)	367 (54.8%)	108 (60.0%)	112 (56.0%)	63 (47.7%)
Male	532 (45.0%)	303 (45.2%)	72 (40.0%)	88 (44.0%)	69 (52.3%)
Ethnicity						0.126a
Chinese	932 (78.8%)	530 (79.1%)	137 (76.1%)	158 (79.0%)	107 (81.1%)
Non-Chinese	250 (21.2%)	140 (20.9%)	43 (23.9%)	42 (21.0%)	25 (18.9%)
Highest education attained						< 0.001a
Post-secondary school & above	484 (40.9%)	366 (54.6%)	53 (29.4%)	26 (13.0%)	39 (29.5%)
Secondary school & below	698 (59.1%)	304 (45.4%)	127 (70.6%)	174 (87.0%)	93 (70.5%)
Employment status						< 0.001a
Employed	717 (60.7%)	455 (67.9%)	107 (59.4%)	77 (38.5%)	78 (59.1%)
Not employed	465 (39.3%)	215 (32.1%)	73 (40.6%)	123 (61.5%)	54 (40.9%)
Marital status						< 0.001a
Married	855 (72.3%)	494 (73.7%)	135 (75.0%)	133 (66.5%)	93 (70.5%)
Single/divorced/widowed/separated	80 (27.7%)	176 (26.3%)	45 (25.0%)	67 (33.5%)	39 (29.5%)
Living alone						< 0.001a
Yes	128 (10.8%)	43 (6.4%)	21 (11.7%)	38 (19.0%)	26 (19.7%)
No	1,054 (89.2%)	627 (93.6%)	159 (88.3%)	162 (81.0%)	106 (80.3%)
Perceived money insufficiency						< 0.001a
Yes	186 (15.7%)	45 (6.7%)	40 (22.2%)	79 (39.5%)	22 (16.7%)
No	996 (84.3%)	625 (93.3%)	140 (77.8%)	121 (60.5%)	110 (83.3%)
No. of chronic conditions						< 0.001a
≥ 2	388 (32.8%)	159 (23.7%)	71 (39.4%)	109 (54.5%)	49 (37.1%)
0–1	794 (67.2%)	511 (76.3%)	109 (60.6%)	91 (45.5%)	83 (62.9%)
PHQ-9 score (Mean ± SD Median (IQR))	1.2 ± 2.4 0 (0–1)	0.7 ± 1.4 0 (0–1)	1.4 ± 2.8 0 (0–2)	2.4 ± 3.7 1 (0–3)	1.3 ± 2.2 0 (0–2)	< 0.001b

^aP value derived from Chi-squared test.

^bP value derived from Kruskal–Wallis test.

PHQ-9 Patient Health Questionnaire-9; SD standard deviation; IQR interquartile range.

Change patterns in social disconnectedness and healthcare utilization

The baseline number of visits differed significantly across the four groups at all care settings, with a consistent pattern observed. The group who remained socially disconnected exhibited the highest mean visits, followed by those who became socially disconnected, then those who became socially connected, and finally those who remained socially connected (Supplementary File, Table S2).

Hospitalizations

After adjusting for baseline utilization and covariates (Table 2), participants who experienced persistent social disconnectedness demonstrated significantly higher odds of hospitalization compared to those who remained socially connected. This was observed at both 1-year (OR = 2.66, 95% CI 1.42–4.97, p = 0.002) and 3-year follow-ups (OR = 2.11, 95% CI 1.29–3.43, p = 0.003). Those who became socially connected also demonstrated higher odds and rate of hospitalization at 3-year follow-up (OR = 1.88, 95% CI 1.14–3.10, p = 0.014; IRR = 1.49, 95% CI 1.06–2.08, p = 0.020).

Table 2.

Associations of change patterns in social disconnectedness with healthcare utilization at 1-year and 3-year follow-ups by healthcare settings (in-patient wards, emergency departments, specialist outpatient clinics and polyclinics).

Timepoint	Healthcare setting	Change patterns in social disconnectedness	Yes, n (%)	Mean ± SD Median (IQR)	Adjusted ORc (95% CI)	Adjusted IRRd (95% CI) Zero values excluded
1-year follow-up	In-patient wards	Remained socially connected (n = 670)	25 (3.7%)	0.1 ± 0.4 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	18 (10.0%)	0.2 ± 0.9 0 (0–0)	1.89 (0.96–3.71)	1.11 (0.69–1.79)
		Remained socially disconnected (n = 200)	38 (19.0%)	0.4 ± 1.0 0 (0–0)	2.66 (1.42–4.97)**	1.05 (0.65–1.70)
		Became socially disconnected (n = 132)	11 (8.3%)	0.1 ± 0.5 0 (0–0)	1.66 (0.76–3.62)	0.84 (0.45–1.57)
		p value	< 0.001a	< 0.001b
	Emergency departments	Remained socially connected (n = 670)	43 (6.4%)	0.1 ± 0.4 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	31 (17.2%)	0.3 ± 0.9 0 (0–0)	1.94 (1.14–3.32)*	1.36 (0.90–2.04)
		Remained socially disconnected (n = 200)	44 (22.0%)	0.5 ± 1.2 0 (0–0)	1.72 (1.01–2.94)*	1.38 (0.90–2.12)
		Became socially disconnected (n = 132)	17 (12.9%)	0.2 ± 0.5 0 (0–0)	1.42 (0.76–2.67)	1.08 (0.65–1.79)
		p value	< 0.001	< 0.001
	Specialist outpatient clinics	Remained socially connected (n = 670)	156 (23.3%)	1.3 ± 3.8 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	57 (31.7%)	2.5 ± 8.7 0 (0–2)	1.13 (0.74–1.72)	1.26 (0.94–1.68)
		Remained socially disconnected (n = 200)	91 (45.5%)	3.1 ± 6.0 0 (0–4)	1.41 (0.93–2.15)	1.05 (0.79–1.38)
		Became socially disconnected (n = 132)	46 (34.8%)	2.4 ± 5.5 0 (0–2)	1.38 (0.87–2.17)	1.22 (0.89–1.66)
		p value	< 0.001	< 0.001
	Polyclinics	Remained socially connected (n = 670)	233 (34.8%)	1.6 ± 3.3 0 (0–2)	1.00	1.00
		Became socially connected (n = 180)	66 (36.7%)	1.8 ± 3.3 0 (0–2)	1.04 (0.66–1.62)	0.99 (0.79–1.24)
		Remained socially disconnected (n = 200)	97 (48.5%)	3.5 ± 9.7 0 (0–5)	1.24 (0.77–1.98)	1.37 (1.12–1.68)**
		Became socially disconnected (n = 132)	53 (40.2%)	2.5 ± 7.4 0 (0–3)	0.95 (0.58–1.55)	1.24 (0.98–1.56)
		p value	0.005	< 0.001
3-year follow-up (cumulative)	In-patient wards	Remained socially connected (n = 670)	21 (3.1%)	0.1 ± 0.6 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	19 (10.6%)	0.5 ± 1.5 0 (0–0)	1.88 (1.14–3.10)*	1.49 (1.06–2.08)*
		Remained socially disconnected (n = 200)	31 (15.5%)	0.8 ± 1.7 0 (0–1)	2.11 (1.29–3.43)**	1.37 (0.97–1.94)
		Became socially disconnected (n = 132)	10 (7.6%)	0.3 ± 1.0 0 (0–0)	1.63 (0.91–2.91)	1.14 (0.75–1.74)
		p value	< 0.001	< 0.001
	Emergency departments	Remained socially connected (n = 670)	38 (5.7%)	0.3 ± 0.9 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	32 (17.8%)	0.8 ± 1.6 0 (0–1)	2.07 (1.37–3.12)**	1.24 (0.96–1.61)
		Remained socially disconnected (n = 200)	43 (21.5%)	1.0 ± 2.0 0 (0–1)	1.77 (1.16–2.70)**	1.11 (0.84–1.46)
		Became socially disconnected (n = 132)	13 (9.8%)	0.5 ± 1.1 0 (0–1)	1.58 (0.99–2.54)	0.98 (0.72–1.35)
		p value	< 0.001	< 0.001
	Specialist outpatient clinics	Remained socially connected (n = 670)	150 (22.4%)	3.6 ± 10.0 0 (0–3)	1.00	1.00
		Became socially connected (n = 180)	63 (35.0%)	6.8 ± 16.1 0 (0–6)	1.13 (0.76–1.67)	1.25 (0.95–1.65)
		Remained socially disconnected (n = 200)	86 (43.0%)	8.2 ± 14.4 1.5 (0–11)	1.36 (0.90–2.05)	1.20 (0.92–1.57)
		Became socially disconnected (n = 132)	40 (30.3%)	6.5 ± 14.0 0 (0–6)	1.33 (0.86–2.04)	1.25 (0.93–1.69)
		p value	< 0.001	< 0.001
	Polyclinics	Remained socially connected (n = 670)	232 (34.6%)	4.7 ± 9.2 0 (0–5)	1.00	1.00
		Became socially connected (n = 180)	70 (38.9%)	5.6 ± 10.3 1 (0–6)	1.21 (0.81–1.80)	0.90 (0.73–1.11)
		Remained socially disconnected (n = 200)	85 (42.5%)	8.6 ± 16.8 2 (0–14)	1.02 (0.65–1.59)	1.25 (1.02–1.54)*
		Became socially disconnected (n = 132)	55 (41.7%)	6.7 ± 11.0 2 (0–9)	1.33 (0.84–2.12)	1.24 (1.00–1.55)
		p value	0.131	< 0.001

*p < 0.05, **p < 0.01.

SD Standard deviation; IQR Interquartile range.

^aColumn p values derived from Chi-squared tests.

^bColumn p values derived from Kruskal–Wallis tests.

^cOdds ratio (OR) obtained from logistic regressions, adjusted for the following covariates: baseline visits (yes, no), age group (younger adults, older adults), gender (male, female), ethnicity (Chinese, non-Chinese), education (secondary school & below, post-secondary school & above), employment status (employed, not employed), marital status (married, single/divorced/widowed/separated), living alone (yes, no), perceived money insufficiency for basic living needs (yes, no), number of chronic conditions (0–1, ≥ 2), Patient Health Questionnaire-9 score (integer).

^dIncidence Rate Ratio (IRR) using generalized linear models with negative binomial distribution or Poisson distribution (for hospitalizations and emergency department visits in the subsequent one year), adjusted for the following covariates: baseline visits (integer), age group (younger adults, older adults), gender (male, female), ethnicity (Chinese, non-Chinese), education (secondary school & below, post-secondary school & above), employment status (employed, not employed), marital status (married, single/divorced/widowed/separated), living alone (yes, no), perceived money insufficiency for basic living needs (yes, no), number of chronic conditions (0–1, ≥ 2), Patient Health Questionnaire-9 score (integer).

Emergency department visits

Participants with persistent social disconnectedness demonstrated significantly higher odds of ED visits at both 1-year (OR = 1.72, 95% CI 1.01–2.94, p = 0.048) and 3-year follow-ups (OR = 1.77, 95% CI 1.16–2.70, p = 0.008) compared to those who remained socially connected. Similarly, those who became socially connected also exhibited higher odds of ED visits at both 1-year (OR = 1.94, 95% CI 1.14–3.32, p = 0.015) and 3-year follow-ups (OR = 2.07, 95% CI 1.37–3.12, p = 0.001). No significant differences in IRRs were observed.

Specialist outpatient clinic visits

No significant associations were observed between change patterns in social disconnectedness and subsequent SOC visits at both follow-ups.

Polyclinic visits

Compared to those who remained socially connected, the group with persistent social disconnectedness exhibited higher rates of polyclinic visits at both 1-year (IRR = 1.37, 95% CI 1.12–1.68, p = 0.003) and 3-year follow-ups (IRR = 1.25, 95% CI 1.02–1.54, p = 0.033). No significant differences in ORs were observed.

Change patterns in social disconnectedness and gross healthcare cost

Change patterns in social disconnectedness were significantly associated with unadjusted costs for hospitalizations, polyclinic visits, and SOC visits, but not ED visits (Supplementary File, Table S2). For the three care settings where significant associations were observed, the group who remained socially disconnected exhibited the highest mean cost, followed by those who became socially disconnected, then those who became socially connected, and finally those who remained socially connected (Supplementary File, Table S2).

Hospitalization costs

After adjusting for baseline cost incurred and covariates (Table 3), participants who experienced persistent social disconnectedness demonstrated higher odds of incurring hospitalization costs at both 1-year (OR = 2.30, 95% CI 1.21–4.40, p = 0.012) and 3-year follow-ups (OR = 1.89, 95% CI 1.15–3.13, p = 0.013) compared to those who remained socially connected. No significant associations with hospitalization costs were observed for the other two patterns of change.

Table 3.

Associations of change patterns in social disconnectedness with healthcare cost at 1-year and 3-year follow-ups by healthcare settings (in-patient wards, emergency departments, specialist outpatient clinics and polyclinics).

Timepoint	Healthcare setting	Change patterns in social disconnectedness	Mean ± SD Median (IQR) (SGD)	Adjusted ORb (95% CI)	Adjusted CRc (95% CI) Zero values excluded
1-year follow-up	In-patient wards	Remained socially connected (n = 670)	518 ± 4,555 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	1,360 ± 6,389 0 (0–0)	1.66 (0.83–3.29)	0.84 (0.39–1.78)
		Remained socially disconnected (n = 200)	2,716 ± 9,598 0 (0–0)	2.30 (1.21–4.40)*	0.63 (0.29–1.34)
		Became socially disconnected (n = 132)	750 ± 3,338 0 (0–0)	1.28 (0.56–2.90)	0.49 (0.20–1.19)
		p value	< 0.001 a
	Emergency departments	Remained socially connected (n = 670)	29.70 ± 157 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	132 ± 397 0 (0–0)	1.88 (1.10–3.22)*	1.57 (1.10–2.24)*
		Remained socially disconnected (n = 200)	192 ± 495 0 (0–0)	1.65 (0.96–2.85)	1.50 (1.03–2.20)*
		Became socially disconnected (n = 132)	71.6 ± 222 0 (0–0)	1.36 (0.72–2.56)	1.15 (0.77–1.73)
		p value	< 0.001
	Specialist outpatient clinics	Remained socially connected (n = 670)	233 ± 846 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	640 ± 2,770 0 (0–214)	1.16 (0.71–1.90)	1.77 (1.08–2.91)*
		Remained socially disconnected (n = 200)	637 ± 1,583 0 (0–533)	1.63 (1.00–2.64)*	1.07 (0.67–1.69)
		Became socially disconnected (n = 132)	522 ± 1,459 0 (0–229)	1.67 (0.98–2.83)	0.98 (0.58–1.65)
		p value	< 0.001
	Polyclinics	Remained socially connected (n = 670)	181 ± 450 0 (0–115)	1.00	1.00
		Became socially connected (n = 180)	258 ± 611 0 (0–172)	1.03 (0.66–1.60)	0.89 (0.68–1.15)
		Remained socially disconnected (n = 200)	405 ± 687 0 (0–581)	1.33 (0.83–2.11)	1.13 (0.88–1.44)
		Became socially disconnected (n = 132)	313 ± 850 0 (0–223)	0.93 (0.57–1.52)	1.18 (0.89–1.56)
		p value	< 0.001
3-year follow-up (cumulative)	In-patient wards	Remained socially connected (n = 670)	1,036 ± 5,450 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	3,413 ± 10,541 0 (0–0)	1.64 (0.90–2.73)	1.32 (0.78–2.21)
		Remained socially disconnected (n = 200)	6,352 ± 19,399 0 (0–2,094)	1.89 (1.15–3.13)*	1.58 (0.93–2.70)
		Became socially disconnected (n = 132)	1,940 ± 6,272 0 (0–0)	1.37 (0.75–2.51)	0.80 (0.43–1.49)
		p value	< 0.001
	Emergency departments	Remained socially connected (n = 670)	99.70 ± 340 0 (0–0)	1.00	1.00
		Became socially connected (n = 180)	343 ± 707 0 (0–422)	1.99 (1.32–3.00)**	1.34 (1.03–1.74)*
		Remained socially disconnected (n = 200)	413 ± 824 0 (0–515)	1.72 (1.13–2.64)*	1.14 (0.86–1.52)
		Became socially disconnected (n = 132)	203 ± 454 0 (0–282)	1.56 (0.97–2.51)	0.99 (0.73–1.34)
		p value	< 0.001
	Specialist outpatient clinics	Remained socially connected (n = 670)	717 ± 2,469 0 (0–302)	1.00	1.00
		Became socially connected (n = 180)	1,486 ± 4,162 0 (0–989)	1.12 (0.73–1.73)	1.51 (0.97–2.36)
		Remained socially disconnected (n = 200)	1,978 ± 6,945 153 (0–1,580)	1.45 (0.94–2.26)	1.19 (0.77–1.83)
		Became socially disconnected (n = 132)	1,986 ± 8,416 0 (0–566)	1.54 (0.97–2.46)	1.23 (0.76–1.99)
		p value	< 0.001
	Polyclinics	Remained socially connected (n = 670)	554 ± 1,363 0 (0–437)	1.00	1.00
		Became socially connected (n = 180)	784 ± 1,877 49.20 (0–538)	1.24 (0.83–1.86)	0.82 (0.63–1.06)
		Remained socially disconnected (n = 200)	1,144 ± 1,953 115 (0–1,535)	1.07 (0.69–1.67)	1.13 (0.87–1.48)
		Became socially disconnected (n = 132)	850 ± 1,974 103 (0–808)	1.34 (0.84–2.12)	1.07 (0.80–1.42)
		p value	< 0.001

*p < 0.05, **p < 0.01.

^aColumn p values derived from Kruskal–Wallis tests.

^bOdds Ratio (OR) obtained from logistic regressions, adjusted for the following covariates: baseline cost (yes, no), age group (younger adults, older adults), gender (male, female), ethnicity (Chinese, non-Chinese), education (secondary school & below, post-secondary school & above), employment status (employed, not employed), marital status (married, single/divorced/widowed/separated), living alone (yes, no), perceived money insufficiency for basic living needs (yes, no), number of chronic conditions (0–1, ≥ 2), Patient Health Questionnaire-9 score (integer).

^cCost Ratio (CR) obtained from Gamma regressions, adjusted for the following covariates: baseline cost (numeric), age group (younger adults, older adults), gender (male, female), ethnicity (Chinese, non-Chinese), education (secondary school & below, post-secondary school & above), employment status (employed, not employed), marital status (married, single/divorced/widowed/separated), living alone (yes, no), perceived money insufficiency for basic living needs (yes, no), number of chronic conditions (0–1, ≥ 2), Patient Health Questionnaire-9 score (integer).

Emergency department costs

Compared to those who remained socially connected, the group who remained socially disconnected exhibited higher cost ratio at 1-year (CR = 1.50, 95% CI 1.03–2.20, p = 0.039) and higher odds ratio at 3-year follow-ups (OR = 1.72, 95% CI 1.13–2.64, p = 0.012). Participants who became socially connected demonstrated higher ED visit cost incurred at both 1-year (OR = 1.88, 95% CI 1.10–3.22, p = 0.021; CR = 1.57, 95% CI 1.10–2.24, p = 0.014) and 3-year follow-ups (OR = 1.99, 95% CI 1.32–3.00, p = 0.001; CR = 1.34, 95% CI 1.03–1.74, p = 0.032).

Specialist outpatient clinic costs

At 1-year follow-up, participants who remained socially disconnected had higher odds of incurring SOC visit costs (OR = 1.63, 95% CI 1.00–2.64, p = 0.049) while individuals who became socially connected had higher cost ratio (CR = 1.77, 95% CI 1.08–2.91, p = 0.025) compared to those who remained socially connected. At 3-year follow-up, no significant associations with cost incurred were observed for all patterns of change in social disconnectedness.

Polyclinic costs

No significant associations were observed between change patterns in social disconnectedness and subsequent costs incurred for polyclinic visits at both follow-ups.

Discussion

Main findings

This study examined the associations between four patterns of change in social disconnectedness and healthcare utilization and gross costs incurred at 1- and 3-year (cumulative) follow-ups for four public healthcare settings. In the following sections, we discuss our findings organized according to the four patterns of change, with those who remained socially connected as the reference group, followed by each care setting.

Persistent social disconnectedness and increased healthcare burden

Persistent social disconnectedness was associated with higher odds of hospitalizations and ED visits, as well as associated costs, at both follow-ups. Among polyclinic users, those with persistent social disconnectedness exhibited higher rates of visits at both follow-ups. Additionally, higher odds were observed for SOC visit costs at 1-year follow-up. These observations of higher utilization and costs could be explained by various pathways through which social disconnectedness could influence health outcomes. For instance, social disconnection may directly increase healthcare needs through experiences of pain that are disruptive to daily activities⁵¹ or the development of acute health conditions triggered by prolonged stress⁵².

Becoming socially connected: mixed healthcare utilization patterns

Encouragingly, unlike those who experienced persistent social disconnectedness, individuals who became socially connected did not have elevated odds of hospitalizations and cost at 1-year follow-up, nor elevated odds of polyclinic visits at both follow-ups, compared to those who remained socially connected. While these results would require cautious interpretation given the exploratory nature of the analyses, they suggest potential benefits of interventions targeting social disconnectedness. Nevertheless, we observed higher odds of ED visits and costs for this group, as well as hospitalizations at 3-year follow-up, which suggests that prior social disconnection could have lasting health effects even after social connections improve. This warrants further investigation, integrated with qualitative approaches, to better understand the nuances of these observed patterns and inform future interventions.

Becoming socially disconnected: no significant associations

Participants who became socially disconnected did not exhibit differential subsequent healthcare utilization and costs compared to those who remained socially connected. We posit that this could be due to several factors. First, there could be a lag between becoming socially disconnected and observable changes in healthcare outcomes, which may not yet be detectable within our follow-up period. Second, these individuals might have engaged in healthcare avoidant behaviors such as the preference to self-medicate⁵³ or tendency to delay seeking medical care²¹. It is also plausible that the relatively small sample size of this group (n = 132, 11.2%) had limited statistical power to detect associations.

In-patient wards

Participants who remained socially disconnected exhibited higher odds of hospitalizations compared to those who remained socially connected. One possible explanation for this observation is the engagement in healthcare-avoidant behavior due to contributing factors such as low health literacy^20,54 or financial constraints⁵⁵. Consequently, such tendencies may lead to further deterioration of their health conditions, ultimately necessitating emergency or in-patient interventions. These findings are also aligned with existing literature demonstrating a dose–response relationship, where prolonged social disconnectedness is associated with worse health outcomes than temporary disconnection^56,57.

Participants who became socially connected showed no elevated hospitalization odds at 1-year follow-up compared to those who remained socially connected. This suggests that interventions aimed at improving social connectedness may help reduce avoidable hospitalizations, the most resource-intensive form of care, in the short term. Potential mechanisms may include reduced stress-induced adverse health outcomes⁵², improved medication adherence⁵³, and engagement in healthier lifestyle choices⁵⁸. However, the elevated odds observed at 3-year follow-up highlight the possibility that sustained social connectedness may be necessary to maintain these benefits over time.

Emergency departments

Paradoxically, similar to those with persistent social disconnectedness, participants who became socially connected demonstrated higher odds of ED visits and costs at both follow-ups. We note several potential explanations for this counterintuitive observation. It remains unclear whether individuals with apparent improvements subsequently reverted to being socially disconnected after wave 2, or whether prior exposure had already impacted their health⁵⁹. These individuals may also have experienced extended periods of social disconnection preceding and between data collections, or encountered frequent relapses due to vulnerability from socioeconomic factors⁵⁹. We also recognize that improved social connections may facilitate health-seeking behavior²¹. Finally, our analyses only adjusted for the number of chronic conditions, while disease severity could have been a stronger mediator of healthcare utilization and costs. Overall, these findings underscore the importance of early prevention strategies to mitigate potential lasting effects of social disconnectedness on ED visits and costs.

Specialist outpatient clinics

Our exploratory analyses also endeavored to address an understudied area in the literature regarding the association of social disconnectedness with outpatient specialist consultations. While SOC visit frequency did not differ by groups after adjusting for baseline visits and covariates such as multimorbidity, persistently socially disconnected individuals displayed higher odds of incurring SOC visit costs at 1-year follow-up. Among those who incurred SOC visit costs, individuals who became socially connected demonstrated higher cost ratios at 1-year but not 3-year follow-up. The lack of significant associations with visit frequency, despite differences in costs, suggests that these two groups might have presented with more complex care needs requiring costlier specialist care in the shorter term.

Polyclinics

Compared to those who remained socially connected, the group with persistent social disconnectedness exhibited higher rates of polyclinic visits at both 1- and 3-year follow-ups after controlling for baseline visits and covariates. As polyclinics cater to a sizeable patient population with chronic conditions by providing a comprehensive range of services at subsidized rates in Singapore⁶⁰, the sustained higher utilization among persistently disconnected polyclinic attendees may reflect ongoing management of chronic health conditions that could have been exacerbated by social disconnectedness. Encouragingly, we note from our exploratory analyses that participants who became socially connected did not demonstrate higher polyclinic utilization at both follow-ups. This finding suggests that strengthening social ties may help reduce the burden on primary care services, which is particularly noteworthy for Singapore’s healthcare system where polyclinics are already experiencing significant resource constraints from high patient volumes⁶⁰.

Limitations

We acknowledge several limitations of this study that should be considered when interpreting the findings. First, this study only included participants with prior healthcare service utilization at NHG Health institutions, which may limit generalizability to individuals who avoid healthcare entirely or those who solely use private healthcare providers or services under other clusters. However, NHG Health serves approximately 1.4 million residents in Central Singapore. Additionally, as healthcare utilization and cost data were only available for NHG Health institutions, there may be underestimation of the total healthcare utilization and costs for participants who also sought care from other healthcare providers. Second, we examined social isolation and loneliness under the broader concept of social disconnectedness, which carried the assumption that both constructs have equivalent contributions to healthcare utilization and cost outcomes. However, differentiating their respective effects was not within the scope of this study as population-level interventions in Singapore tend to target social disconnectedness broadly. Third, while holding covariates constant at baseline values enhanced the interpretability of our results, potential changes in time-varying sociodemographic factors that may influence social connectedness could not be accounted for. Finally, we also could not account for potential fluctuations in social connectedness and health conditions between measurement points, and the lag time between changes in social disconnectedness and health outcomes may mean that some implications might not be detectable within our follow-up period.

Implications and future directions

To the best of our knowledge, this is the first longitudinal study to prospectively examine four patterns of change in social disconnectedness and their associations with healthcare utilization and costs using administrative data from multiple public healthcare settings. While our findings yield important insights specific to Singapore’s unique sociocultural context, they could also serve as reference points for other countries, as the factors that link social disconnectedness to poorer health outcomes, such as low health literacy, unhealthy behaviors and chronic stress that could necessitate healthcare interventions, are likely similar based on international literature. However, it should be noted that the magnitude and specific patterns of these associations may vary across countries due to differences in context including healthcare system structures, healthcare accessibility and affordability, and cultural aspects that influence the ease of social connectedness and healthcare-seeking behaviors.

As Singapore works to meet rising healthcare demands from a rapidly aging population and seeks to keep costs contained⁶⁰, our results suggest that social prescribing initiatives and community-based interventions to promote social connectedness could be promising strategies for easing healthcare burden and costs across multiple healthcare settings. Among those who remained socially disconnected, there was a higher proportion of older adults (62.0%), underscoring the vulnerability of this demographic. In Singapore, there has been an active push to promote social engagement and community integration among the older adult population. For instance, the national Age Well SG program, launched in 2023, has various community initiatives to engage, encourage and enable seniors to keep physically and socially active¹⁴. Such community programs that promote social connections through group activities or volunteer opportunities can help to reduce feelings of loneliness and improve health outcomes⁶¹. There are also ongoing national outreach efforts to older adults who may be socially isolated⁶².

In light of these efforts, future studies could not only evaluate the effectiveness of existing initiatives in addressing social disconnectedness but also examine the individual and collective impacts of these initiatives on healthcare utilization and associated costs using population-representative and administrative data. Building upon our findings on the associations between persistent social disconnectedness and subsequent hospitalizations, future research may also examine more granular outcomes such as the length of stay and other relevant healthcare services, including mental health services, in Singapore. International evidence supports the value of such detailed analyses, with Christiansen et al. (2023) observing that loneliness was linked to longer hospital stays in the Danish general population²³, and Molloy et al. (2010) noting that loneliness was associated with unplanned hospitalizations but not planned admissions among community-dwelling adults in Ireland⁶³. Future longitudinal studies may also consider planning for more frequent assessment intervals to better capture the temporal dynamics of changes in social disconnectedness, as our study was limited to two annual measurement points with potential changes occurring between assessments that we could not account for. Where sample sizes permit, studies could investigate the associations of social isolation and loneliness with healthcare utilization separately, and conduct sub-group analyses by demographics such as age groups or socioeconomic status, to ascertain whether there are differential patterns of effects⁶⁴. Evidence from these studies would provide valuable insights to inform the refinement of interventions as well as healthcare policy decisions.

Finally, these findings highlight the importance of addressing social disconnectedness as a modifiable determinant of healthcare utilization and costs. Given that screening tools for social isolation and loneliness require minimal cost and time to implement, policymakers and stakeholders may consider integrating social disconnectedness screening and targeted interventions into Singapore’s broader health strategies, leveraging existing avenues such as Healthier SG and Age Well SG to identify and support at-risk individuals. This could include enhanced integration of social support services and therapeutic interventions within healthcare delivery systems. For instance, individuals may require referral to structured interventions such as cognitive-behavioral therapy to alleviate feelings of loneliness if community activities are less effective for their needs⁶⁵. However, it should be noted that these strategies might shift healthcare utilization patterns rather than decrease them overall; for example, if primary care professionals were to be the first line of screening and intervention for social isolation and loneliness, utilization and cost incurred at this setting may rise⁶⁶. Insights from qualitative studies could therefore illuminate the acceptability and feasibility of addressing social disconnectedness in clinical and community settings within Singapore’s context, as older people may view loneliness as a personal problem and demonstrate resistance towards addressing it in primary care settings⁶⁷. This is particularly relevant considering that the majority of Singapore’s population is of Chinese ethnicity (73.9%), where seeking professional help for psychological or social concerns could be perceived as “losing face”—reflecting personal failure and potentially bringing shame to their family⁶⁸. To overcome such barriers, there may be a need for public education campaigns that frame interventions as approaches to boost personal wellness, which could be more culturally acceptable, or peer-led community initiatives to encourage discussions about social disconnectedness as one ages.

Conclusion

This longitudinal study provides the first examination of four patterns of change in social disconnectedness and their associations with subsequent healthcare utilization and costs at four regional public healthcare settings in Singapore. Key findings include higher odds of hospitalizations and ED visits, as well as higher rates of polyclinic visits and higher healthcare costs across multiple settings, among those with persistent social disconnectedness compared to those who remained socially connected. Notably, individuals who became socially connected continued to show elevated ED utilization, suggesting potential lasting health effects of social disconnection. These findings support targeted interventions to promote social connectedness as a strategy to reduce healthcare burden, while highlighting the need for sustained engagement. Future research should examine mechanisms underlying these associations and evaluate the cost-effectiveness of social connectedness interventions within healthcare systems.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

G.H. analyzed and interpreted the data, drafted the initial manuscript, and made subsequent revisions. C.W.Y. interpreted the data and reviewed the article. L.G. conceptualized the study, interpreted the data, and reviewed and revised the article. All authors read and approved the final version of the manuscript.

Funding

This work did not receive any external funding support. All authors were employees of the National Healthcare Group Pte Ltd. However, the employer had no role in/influence on study design, data collection and analysis, result interpretation, decision to publish, or preparation of the manuscript.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Ethics approval

The PHI study received approval from the National Healthcare Group Domain Specific Review Board Ethics Review Committee (Reference Number: 2015/00269). Written informed consent was obtained from all participants following comprehensive briefings about the study objectives and procedures. To ensure participant confidentiality, all collected data were de-identified prior to analysis and stored separately from personally identifiable information. Identifiable information is password-protected and maintained in secure storage accessible only to designated survey team members. In accordance with institutional data retention policies, all study data will be retained for a minimum of 6 years following study completion.