Associations of social isolation and loneliness with healthcare utilization among older adults: a systematic review and meta-analysis

Abstract

Background and Objectives

Social isolation and loneliness are significant public health concerns associated with increased healthcare utilization among older adults. This review aims to synthesize evidence on the associations between social isolation, loneliness, and healthcare utilization.

Research Design and Methods

Five databases were searched from inception to March 21, 2025, using keyword groups related to social isolation/loneliness, older adults, and healthcare utilization (primary care, emergency visits, inpatient care, and outpatient care). Methodological quality was assessed using the Newcastle-Ottawa Scale. Random-effects models were employed to pool effect sizes (incidence rate ratios [IRRs], odds ratios [ORs]).

Results

A total of 44 studies were included in the systematic review, and 34 were included in the meta-analysis (N = 309,023). Due to insufficient data, meta-analyses for the association between social isolation and primary care or outpatient care utilization were not conducted. Social isolation was statistically associated with increased inpatient care utilization (IRRs = 1.37, 95% CI: 1.24–1.53) but not with emergency department visits. For loneliness, meta-analyses for outpatient care were not feasible due to limited studies. Loneliness was statistically associated with increased emergency department visits (IRRs = 1.15, 95% CI: 1.06–1.24) and inpatient care utilization (OR = 1.13, 95% CI: 1.07–1.20) but not with primary care use.

Discussion and Implications

This is the first meta-analysis to comprehensively synthesize the associations between social isolation, loneliness, and 4 types of healthcare utilization among older adults. The findings highlight the importance of addressing social isolation and loneliness as potential strategies to reduce avoidable healthcare utilization.

Innovation and Translational Significance:

Social isolation and loneliness are growing challenges among older adults, contributing to increased healthcare utilization. This systematic review and meta-analysis reveals significant associations between these social factors and specific types of healthcare use. The findings highlight the need for targeted interventions to reduce isolation and loneliness, which may help lower avoidable healthcare demands. Translating this evidence into practice can inform public health strategies, healthcare planning, and community-based programs to improve aging-related outcomes and reduce strain on healthcare systems.

Background and objectives

Social isolation and loneliness are prevalent public health issues affecting individual well-being and healthcare systems among older adults (Wu, 2020). Social isolation is defined as an objective condition characterized by a lack of social connections, usually measured by the absence of contact with family, friends, and community members (Cacioppo et al., 2011). Loneliness is a distressing subjective feeling of disconnection and dissatisfaction with the quality and/or quantity of social relationships (Bekhet et al., 2008). These two conditions may occur independently and do not always coexist (Sciences et al., 2020). The burden of social isolation and loneliness is substantial, leading to increased healthcare costs and resource utilization (Meisters et al., 2021; Pecanha et al., 2020). Individuals with these conditions often require more frequent medical consultations, straining healthcare systems (Sciences et al., 2020).

Both social isolation and loneliness are associated with increased risk of adverse health outcomes, including heart disease, frailty, dementia, and even increased mortality (Leigh-Hunt et al., 2017). These conditions may significantly influence medical use through shared biological and psychological pathways (Donovan & Blazer, 2020). For example, dysregulated stress responses, such as altered cortisol and chronic systemic inflammation, can accelerate cardiovascular and neurodegenerative diseases, thereby increasing the need for acute medical interventions and overall healthcare utilization (Donovan & Blazer, 2020; Hackett et al., 2012). At the same time, psychological pathways, including increasing stress (Masi et al., 2011), suicidal ideation, depression (Beutel et al., 2017; Cacioppo et al., 2006), and negative beliefs about aging (Hajek & Konig, 2021), may initially reduce healthcare-seeking behaviors or delay timely care. However, these delays can lead to worsening health conditions, reduced individual activation, and increased healthcare needs over time (Gao et al., 2022; Hardcastle et al., 2015). Conversely, some older adults may engage in compensatory healthcare utilization, seeking social interaction through medical visits or relying more heavily on formal healthcare due to limited self-management capacity (Geboers et al., 2016; Kouzis & Eaton, 1998).

Although several mechanisms have been proposed, empirical evidence remains mixed and inconclusive. Some studies report that social isolation is associated with greater inpatient utilization, such as hospitalization (Pomeroy et al., 2024) and re-hospitalization (Giuli et al., 2012), while others show that socially isolated older adults are less likely to use outpatient and primary care services compared to more socially integrated seniors (Xie et al., 2024). Similarly, loneliness has been associated with increased healthcare utilization among older adults in some studies, including primary care use (Burns et al., 2020; Sirois & Owens, 2023), emergency department visits (Burns et al., 2021), inpatient care (e.g., hospitalizations [Molloy et al., 2010], re-hospitalizations [Newall et al., 2015]), and outpatient care (e.g., visiting specialists [Gao et al., 2024]). However, other investigations have shown null associations (Newall et al., 2015). These divergent findings suggest that the impact of social isolation and loneliness on healthcare utilization is complex, potentially varying across different types of services.

Previous reviews have provided valuable insights but remain limited in several respects. For example, one review synthesized loneliness and health/social care utilization, but did not distinguish between loneliness and social isolation, and did not conduct a meta-analysis to quantitatively synthesize findings (Smith & Victor, 2022). Another meta-analysis focused narrowly on loneliness and primary care, excluding other healthcare settings such as emergency departments and inpatient care (Sirois & Owens, 2023). Moreover, existing syntheses have not included evidence published beyond 2021, limiting their relevance in rapidly changing healthcare contexts.

To address these gaps, we conducted a systematic review and meta-analysis of studies up to March 2025. We distinguish between social isolation and loneliness to better understand how each relates to healthcare use, quantify these associations across different healthcare settings (primary care, emergency visits, inpatient, and outpatient services). Through this approach, we provide an updated synthesis that clarifies previous inconsistencies and highlights how these social factors are associated with healthcare utilization among older adults.

Research design and methods

This systematic review and meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines. The protocol was registered in the PROSPERO database under record number CRD420251036578.

Study selection and search strategy

Five databases (MEDLINE/PubMed, EMBASE, PsycINFO, CINAHL Plus, Web of Science) were systematically searched from inception to March 21, 2025, using three keyword groups: (1) healthcare utilization, including primary care (e.g., general practitioner or nurse visits), emergency department visits, inpatient care (e.g., hospitalizations, re-hospitalizations), and outpatient care (e.g., specialist consultations); (2) social isolation/loneliness; (3) older adults. The search strategies in the database are shown in Supplementary Tables 1–5.

All retrieved records were managed using EndNote X21. Two researchers independently screened the initial title and abstract, followed by a full-text review. The third researcher resolved disagreements, and all screening outcomes were visualized in the PRISMA flowchart (Figure 1).

Figure 1.

The PRISMA flow diagram.

Eligibility criteria

Studies were selected based on the following criteria: (1) assessment of associations between social isolation/loneliness and healthcare utilization; (2) observational designs including cohort, case–control, and cross-sectional studies; (3) published in the English language; and (4) populations with a mean or median age of ≥60 years. When these statistics were not reported, studies with more than 50% of participants aged ≥60 years were also included, following prior reviews (Laura & George, 2016; Ronzi et al., 2018; Valtorta et al., 2018).

Reviews, meta-analyses, conference abstracts, dissertations, and non-peer-reviewed articles were excluded.

Data extraction

Data extraction was conducted using a pre-designed, standardized form, with one researcher performing the initial extraction and a second researcher verifying the accuracy. The extracted data included the following information: author, year, country/region, age, sample size, male proportion, type of healthcare utilization, measurement tools, and study design.

Risk of bias

The methodological quality of the included studies was independently evaluated by two reviewers using the Newcastle-Ottawa Scale (Wells et al., 2000). For case–control studies, the criteria included selection (0–4 points), comparability (0–2 points), and exposure (0–3 points), with a maximum score of nine. For cohort studies, the criteria comprised: subject selection and exposure assessment (0–4 points), comparability of groups (0–2 points), and measurement validity (0–3 points), resulting in a maximum score of 9. For cross-sectional studies, the criteria included selection criteria (0–5 points), comparability evaluation (0–2 points), and outcome validity (0–3 points), for a maximum score of 10. Studies obtaining scores exceeding seven points were classified as having a low risk of bias. Disagreements between reviewers were resolved through consensus discussions involving a third researcher.

Statistical analysis

Among the 44 included studies, 34 met the criteria for quantitative synthesis in meta-analyses and provided extractable effect estimates. Different types of estimates were included, such as odds ratios (ORs), hazard ratios (HRs), and incidence rate ratios (IRRs). To address heterogeneity in effect estimates, stratified meta-analyses were performed for ORs and HRs/IRRs following Egger et al.’s methodological guidelines (Egger et al., 2008). Pooled ORs were obtained from individual ORs. Pooled IRRs were obtained from individual HRs and IRRs (Hernán, 2010). Pooled effect estimates were calculated using random-effects models. Data on the associations between social isolation and loneliness with each category of healthcare utilization were synthesized separately. For quantitative synthesis, at least three studies per category of healthcare utilization were needed. Heterogeneity was quantified through I² statistics, interpreted as follows: 0%–40% (slight), 30%–60% (moderate), 50%–90% (substantial), and 75%–100% (considerable). The overlapping boundaries acknowledge the need for judgment in interpreting the variability of heterogeneity levels. A narrative synthesis was conducted for studies that were included in the systematic review but excluded from the meta-analysis. All analyses were conducted using Stata MP version 18.0.

Subgroup analysis

To explore potential sources of heterogeneity, three subgroup analyses were conducted based on (1) measurement tools (single-item and multi-item scales), (2) study design (cross-sectional and cohort studies), and (3) confounder adjustment (adjusted and unadjusted effect estimates). In accordance with the Cochrane Handbook, subgroup analyses were only performed when at least 10 studies were available for the outcome (Chandler et al., 2019). Therefore, we only conducted these analyses on the association between social isolation and inpatient care within the OR group.

Publication bias

To assess potential publication bias, we conducted Egger’s test and examined funnel plot asymmetry for each healthcare utilization with at least ten studies in the social isolation and loneliness. For the analysis showing evidence of asymmetry (Egger’s test p < .1), we performed trim-and-fill analysis using a linear estimator with left-side imputation to adjust for potentially missing studies, following the nonparametric approach by Duval and Tweedie (Duval & Tweedie, 2000). Due to the limited number of studies (<10) in most kinds of healthcare utilization, formal testing was only feasible for social isolation, with ORs related to inpatient care.

Sensitivity analyses

For analyses with considerable heterogeneity in each healthcare utilization category (I² ≥75%), we conducted leave-one-out analyses (Steinmann et al., 2024). We also performed additional sensitivity analyses to evaluate the robustness of the age eligibility criterion by excluding studies that met the inclusion threshold solely because more than 50% of participants were aged ≥60 years. This analysis was applied across all relevant pooled estimates to examine whether the operational definition of older adults affected the overall findings.

Results

Search results

A total of 3641 articles were initially identified. After removing duplicates and screening the titles and abstracts, 220 publications remained for the full-text review. Of these, 44 met the inclusion criteria for the systematic review and 34 were included in the meta-analysis. The identification and selection process are shown in Figure 1.

Study characteristics

Table 1 summarizes characteristics of the included studies. Of the 44 studies identified for inclusion, 26 were cohort studies, 17 employed a cross-sectional design, and 1 was a case–control study. Twenty-two studies were conducted in North America, 15 in Europe, 5 in Asia, 1 in Oceania, and 1 in Africa. The total sample size across all included studies was 309,023, with individual study sample sizes ranging from 89 to 76,479 (Dorr et al., 2022; Hand et al., 2014). The male proportion ranged from 20.5% to 100% (Landeiro et al., 2016; Mistry et al., 2001). Only four studies (Barnes et al., 2022; Bu et al., 2020; Chamberlain, Bronskill, et al., 2022; Gao et al., 2024) explored the connections between healthcare utilization and both social isolation and loneliness concurrently.

Table 1.

The characteristics of the included studies (N = 44).

Study	Country/ Region	Sample	% Male	Agea	Type of Healthcare	Measurement Selectionb	SI Measurement	Loneliness Measurement	Design	Risk of Bias	Overall Quality
Barnes et al., 2022	USA	6,994	45	65–85+	ED visits Inpatient care	Both	SNI	UCLA LS-3	Cross-sectional	9/10	High
Bartley et al., 2024	USA	2,320	45.4	55–90+	ED visits Inpatient care	SI	Validated index	—	Cohort	9/10	High
Belachew et al., 2025	Ethiopia	545	42.4	60–80+	Inpatient care	Loneliness	—	—	Cross-sectional	7/10	High
Boehlen et al., 2023	Germany	2,525	47.0	55–84	Outpatient care	Loneliness	—	GFI	Cohort	9/9	High
Bu et al., 2020	UK	4,478	46.3	50–80+	Inpatient care	Both	—	UCLA LS-3	Cohort	7/9	High
Burns et al., 2020	UK	8,175	45.5	Lonely: 63.8 (9.85) Not lonely: 63.4 (9.33)	Primary care ED visits	Loneliness	Three dimensions of social relationships	UCLA LS-5	Cohort	9/9	High
Cantarero-Prieto et al., 2021	Europe	31,536	42.7	68.11	Primary care	SI	Two isolation proxies	—	Cohort	9/9	High
Cayenne et al., 2021	USA	642	46.7	71.75 (8.31)	Outpatient care	SI	PROMIS Social Isolation Short Form	—	Cohort	9/9	High
Chamberlain, Savage, et al., 2022	Canada	18,191	34.2	80.6 (12.8)	ED visits	Both	RAI-HC	RAI-HC	Cohort	9/9	High
Chamberlain, Bronskill, et al., 2022	Canada	44,413	48.8	—	ED visits	Loneliness	—	UCLA LS-3	Cross-sectional	8/10	High
Chamberlain et al., 2023	Canada	44,423	48.8	≥65 = 52.4%	Uncategorized	Loneliness	—	UCLA LS-3	Cross-sectional	9/10	High
Dorr et al., 2022	USA	76,479	45.0	66–91+	Inpatient care	SI	—	—	Cohort	7/9	High
Ellaway et al., 1999	UK	690	44.8	60 years	Primary care	Loneliness	—	One question	Cross-sectional	8/10	High
Gao et al., 2024	USA	6,832	44.5	63.7 (8.0)	Inpatient care Outpatient care	Both	6-item social isolation index	UCLA LS-3	Cohort	9/9	High
Gerst-Emerson & Jayawardhana, 2015	USA	—	Lonely: 38.5 Not lonely: 43.4	Lonely: 71.2 (7.1) Not lonely: 71.2 (6.7)	Primary care	Loneliness	—	UCLA LS-3	Cohort	7/9	High
Giuli et al., 2012	Italy	580	54.1	No SI: 81.4 (5.7) SI: 82.9 (6.6)	Inpatient care	SI	LSNS	—	Cohort	9/9	High
Greysen et al., 2013	USA	1,836	99.0	61	Inpatient care	SI	SIS	—	Cohort	8/9	High
Hand et al., 2014	Canada	89	45.0	81.3 (5.9)	Primary care	SI	Multidimensional definition	—	Cross-sectional	9/10	High
Holaday et al., 2022	USA	8125	43.5	65–75+	Primary care	Loneliness	—	Two questions	Cross-sectional	8/10	High
Humphreys et al., 2018	USA	158	93.0	60.3 (4.8)	ED visits	Loneliness	—	UCLA LS-3	Cohort	9/9	High
Jordan et al., 2008	UK	796	44.1	77	Inpatient care	SI	—	—	Case control	9/9	High
Joseph et al., 2025	UK	11,631	43.8	80.4 (8.0)	ED visits Inpatient care	Loneliness	—	Natural language processing algorithm	Cohort	8/9	High
Kotwal et al., 2023	USA	2,380	47.0	81.2 (9.9)	ED visits Inpatient care	SI	Three dimensions of social relationships	—	Cohort	9/9	High
Landeiro et al., 2016	UK	278	20.5	85.5 (5.8)	Outpatient care	SI	LSNS	—	Cohort	9/9	High
Lim & Chan, 2017	Singapore	2,738	46.9	73.1 (7.2)	Primary care	Loneliness	—	Revised UCLA Loneliness Scale	Cohort	9/9	High
Lofvenmark et al., 2009	Sweden	149	52.3	76 (10.3)	Inpatient care	Loneliness	—	One question	Cross-sectional	8/10	High
Longman et al., 2012	Australia	102	62.0	77.1	Inpatient care	SI	DSSI	—	Cross-sectional	9/10	High
Mira et al., 2024	Spain	932	28.0	77.8 (8.3)	Inpatient care Primary care	Loneliness	—	DeJong score	Cross-sectional	9/10	High
Mistry et al., 2001	USA	123	100	69.9 (7.1)	Inpatient care	SI	LSNS	—	Cross-sectional	10/10	High
Mitsutake et al., 2021	Japan	1,386	40.3	65–75+	Inpatient care	SI	Four questions	—	Cohort	8/9	High
Molloy et al., 2010	Ireland	2,033	43.0	74.1 (6.8)	ED visits Inpatient care	Loneliness	—	One question	Cross-sectional	8/10	High
Mosen et al., 2021	USA	18,557	43.5	73.4 (6.6)	ED visits Inpatient care	SI	One question	—	Cohort	9/9	High
Mosen et al., 2023	USA	9,649	43.4	65–85+	ED visits Inpatient care	SI	Five individual binary social need items	—	Cohort	8/9	High
Newall et al., 2015	Canada	954	46.0	63.5 (10.4)	Primary care Inpatient care	Loneliness	—	Self-conceptualized	Cohort	9/9	High
Pomeroy et al., 2023	USA	11,517	43.0	65–95+	Inpatient care	SI	Self-conceptualized	—	Cohort	8/9	High
Pomeroy et al., 2024	USA	5,533	45.4	75.1 (0.1)	Inpatient care	SI	Self-conceptualized	—	Cohort	8/9	High
Ronneikko et al., 2018	Finland	6,812	32.3	63–90+	Inpatient care	Loneliness	—	RAI-HC	Cohort	8/9	High
Ryan et al., 2023	USA	21,528	44.6	71.0 (8.3)	ED visits Inpatient care	Loneliness	—	Self-reported survey	Cross-sectional	9/10	High
Savitz et al., 2023	USA	3,142	54.8	73.4 (12.0)	ED visits Inpatient care	SI	PROMIS Social Isolation Short Form	—	Cohort	9/9	High
Sterling et al., 2022	UK	690	55.7	76.0 [71.0–82.0]	Inpatient care	SI	—	—	Cohort	7/9	High
Taube et al., 2015	Sweden	153	33.0	81.5	ED visits	Loneliness	—	Four single-item questions	Cross-sectional	8/10	High
Wee et al., 2019	Singapore	928	59.5	60+	ED visits	Loneliness	—	UCLA LS-3	Cross-sectional	9/10	High
Xie et al., 2024	China	8,343	50.3	71.3 (7.3)	Primary care	SI	LSNS	—	Cohort	9/9	High
Zhang et al., 2018	China	5,514	42.9	69.7 (6.5)	Primary care	Loneliness	—	One question	Cross-sectional	8/10	High

Note. DSSI = Duke Social Support Index; ED = emergency department; GFI = Groningen Frailty Indicator; LSNS = Lubben Social Network Scale; PROMIS = The Patient-Reported Outcomes Measurement Information System; RAI-HC = The Resident Assessment Instrument-Home Care; SI = social isolation; SIS = Social Isolation Score; SNI = Social Network Index; UCLA LS-3 = UCLA 3-Item Loneliness Scale; UCLA LS-5 = UCLA 5-Item Loneliness Scale.

^aUnless indicated otherwise, range, mean (standard deviation), or mean [interquartile range] are reported.

^b“Both” indicates both SI and loneliness were measured in the study.

Healthcare utilization and social isolation

Twenty-four studies examined the association between social isolation and healthcare utilization. The majority focused on inpatient care (Barnes et al., 2022; Mitsutake et al., 2021), including hospitalization (Bartley et al., 2024; Bu et al., 2020; Dorr et al., 2022; Greysen et al., 2013; Jordan et al., 2008; Kotwal et al., 2023; Longman et al., 2012; Mosen et al., 2023; Mosen et al., 2021; Pomeroy et al., 2023; Pomeroy et al., 2024; Savitz et al., 2023), re-hospitalization (Giuli et al., 2012; Mistry et al., 2001; Sterling et al., 2022), and the length of hospital stay (Gao et al., 2024). Seven studies assessed emergency department visits (Barnes et al., 2022; Bartley et al., 2024; Chamberlain, Bronskill, et al., 2022; Kotwal et al., 2023; Mosen et al., 2023; Mosen et al., 2021; Savitz et al., 2023), whereas three reported primary care utilization outcomes (Cantarero-Prieto et al., 2021; Hand et al., 2014; Xie et al., 2024). Three studies (Cayenne et al., 2021; Gao et al., 2024; Landeiro et al., 2016) reported outpatient care utilization. Social isolation was measured differently across studies, with instruments including the Social Network Index (Barnes et al., 2022; Gao et al., 2024), Lubben Social Network Scale (Giuli et al., 2012; Landeiro et al., 2016; Mistry et al., 2001; Xie et al., 2024), and Duke Social Support Index (Longman et al., 2012).

Healthcare utilization and loneliness

Twenty-four studies investigated the relationship between loneliness and healthcare utilization. Eight studies examined primary care utilization (Burns et al., 2020; Ellaway et al., 1999; Holaday et al., 2022; Mira et al., 2024), including general practitioner visits and physician consultations (Gerst-Emerson & Jayawardhana, 2015; Lim & Chan, 2017; Newall et al., 2015; Zhang et al., 2018). Ten studies assessed emergency department visits (Barnes et al., 2022; Burns et al., 2020; Chamberlain, Bronskill, et al., 2022; Chamberlain, Savage, et al., 2022; Humphreys et al., 2018; Joseph et al., 2025; Molloy et al., 2010; Ryan et al., 2023; Taube et al., 2015; Wee et al., 2019), while 11 focused on inpatient care (Barnes et al., 2022), encompassing hospitalization (Bu et al., 2020; Joseph et al., 2025; Lofvenmark et al., 2009; Mira et al., 2024; Molloy et al., 2010; Newall et al., 2015; Ronneikko et al., 2018; Ryan et al., 2023; Zhang et al., 2018), re-hospitalizations (Newall et al., 2015; Ryan et al., 2023), and the length of hospital stay (Gao et al., 2024). Two studies reported outpatient care (Boehlen et al., 2023; Gao et al., 2024). Two studies did not specify the types of healthcare utilization (Belachew et al., 2025; Chamberlain et al., 2023). The different versions of the UCLA Loneliness Scale were the most frequently employed measures (Bu et al., 2020; Burns et al., 2020; Chamberlain et al., 2023; Chamberlain, Savage, et al., 2022; Gerst-Emerson & Jayawardhana, 2015; Humphreys et al., 2018; Lim & Chan, 2017; Wee et al., 2019).

Quality assessment

Among the 44 included studies, the evaluation of bias risk revealed that the mean quality score based on the Newcastle-Ottawa Scale was 8.4 for the 26 cohort studies, with scores ranging from seven to nine. For the 17 cross-sectional studies, the mean quality score was 8.5, with scores ranging from 7 to 10. The single case–control study had a score of nine. Overall, the quality of the studies was high. Table 1 summarizes the study quality.

Narrative synthesis of studies excluded from the meta-analysis

Ten studies met the inclusion criteria for the systematic review but were excluded from the meta-analysis due to missing effect sizes or non-relevant healthcare utilization outcomes.

For social isolation, Cantarero-Prieto et al. (Cantarero-Prieto et al., 2019) reported that socially isolated individuals had more frequent primary care visits across European countries. Cayenne et al. (Cayenne et al., 2021) found that socially isolated older adults had lower rates of in-person follow-up visits after emergency department discharge. Hand et al. (Hand et al., 2014) reported no significant association between social isolation and primary care utilization among frequent healthcare users.

For loneliness, Taube et al. (2015) and Lofvenmark et al. (2009) reported higher frequencies of emergency visits and longer inpatient stays among lonely older adults. Mira et al. (Mira et al., 2024) observed increased healthcare resource use among lonely older adults in Spain. Ellaway et al. (Ellaway et al., 1999) reported that lonely individuals had more frequent general practitioner consultations. Humphreys et al. found that loneliness was associated with emergency department use among older adults released from jail (Humphreys et al., 2018). Boehlen et al. (Boehlen et al., 2023) reported that loneliness was linked to greater use of outpatient and mental health services, particularly among women. Chamberlain et al. (Chamberlain et al., 2023) found that lonely individuals had higher rates of unmet healthcare needs.

Meta-analysis

Social isolation and healthcare utilization

For primary care and outpatient care utilization, meta-analyses were not feasible due to insufficient data.

Social isolation and emergency department visit.

Findings for emergency department visits were inconsistent across studies. In the IRR group, some studies reported an increased risk of visits among socially isolated individuals, while others found a protective effect. The pooled analysis of four studies, including 37,580 participants, showed no statistically significant association (IRR = 1.17, 95% CI: 0.78, 1.76), with substantial heterogeneity (I² = 96.8%, p < .001). In the OR group, results also varied, with one study indicating higher odds and the others suggesting reduced or null associations. Across three studies including 23,653 participants, the pooled effect was not statistically significant (OR = 1.07, 95% CI: 0.87, 1.33), with considerable heterogeneity (I² = 87.5%, p < .001) (Figure 2).

Figure 2.

Forest plot for the association between social isolation and emergency department visits: (A) incidence rate ratios (IRRs) and (B) odds ratios (ORs).

The leave-one-out sensitivity analysis showed that excluding Chamberlain et al. (2022) significantly strengthened the association (IRR = 1.40, 95% CI: 1.08, 1.81, p = .011), while the omission of other studies resulted in a non-significant association in the IRR group (Supplementary Figure 1). In the OR group, excluding Barnes et al. (2021), the effect size increased (OR = 1.23, 95% CI: 1.14, 1.33, p < .001) (Supplementary Figure 2).

Social isolation and inpatient care.

In the IRR group, most studies indicated a higher risk of inpatient care utilization among socially isolated older adults, although one study reported a non-significant effect. The pooled analysis of six studies, including 13,046 participants, demonstrated a 37.4% increased risk (IRR = 1.37, 95% CI: 1.24, 1.53), with moderate heterogeneity (I² = 57.9%, p = .026). In the OR group, some studies suggested strong positive associations, while others showed null or even protective effects. When combined with 13 studies involving 140,626 participants, the pooled effect was not statistically significant (OR = 1.21, 95% CI: 0.99, 1.47), and heterogeneity was considerable (I² = 94.8%, p < .001) (Figure 3).

Figure 3.

Forest plot for the association between social isolation and inpatient care: (A) incidence rate ratios (IRRs) and (B) odds ratios (ORs).

The leave-one-out sensitivity analysis showed that the pooled OR values ranged from 1.12 to 1.26. Excluding the study by Kotwal (2024) resulted in a statistically significant pooled OR of 1.26 (95% CI: 1.01, 1.56, p = .038). Similarly, excluding the studies by Longman (2012) and Mitsutake (2020) also yielded significant results (Supplementary Figure 3).

Loneliness and healthcare utilization

For outpatient care utilization, meta-analyses were not feasible due to insufficient data.

Loneliness and primary care.

Based on five studies including 17,331 participants, loneliness showed no statistically significant association with primary care utilization (OR = 0.98, 95% CI: 0.78, 1.22, I² = 93.3%, p < .001) (Figure 4).

Figure 4.

Forest plot for the association between loneliness and primary care.

The leave-one-out sensitivity analysis did not reveal any significant change in the reported effect estimates (Supplementary Figure 4).

Loneliness and emergency department visits.

In the IRR group, all three studies indicated a positive association, and the pooled analysis, including 37,997 participants, showed a 14.6% higher risk of emergency department visits (IRR = 1.15, 95% CI: 1.06, 1.24), with substantial heterogeneity (I² = 86.1%, p < .001). In the OR group, the pooled analysis of five studies with 54,698 participants demonstrated a 34.2% increased risk (OR = 1.34, 95% CI: 1.20, 1.50), with moderate heterogeneity (I² = 40.5%, p = .136) (Figure 5).

Figure 5.

Forest plot for the association between loneliness and emergency department visits: (A) incidence rate ratios (IRRs) and (B) odds ratios (ORs).

After excluding one study at a time, the leave-one-out sensitivity analysis showed that the association between loneliness and emergency department visits among older adults remained consistent with the overall findings (Supplementary Figure 5).

Loneliness and inpatient care.

In the IRR group, some studies suggested an elevated risk of hospitalization among lonely older adults, whereas others reported null or opposite associations. The pooled analysis of four studies, including 16,984 participants, showed no statistically significant association (IRR = 1.20, 95% CI: 0.85, 1.70), with considerable heterogeneity (I² = 98.7%, p < .001). In contrast, results were more consistent in the OR group. Most studies reported higher odds of inpatient care among lonely individuals, and the pooled analysis of seven studies with 30,093 participants demonstrated a statistically significant 13% increased risk (OR = 1.13, 95% CI: 1.07, 1.20), with no evidence of heterogeneity (I² = 0.0%, p = .178) (Figure 6).

Figure 6.

Forest plot for the association between loneliness and inpatient care: (A) incidence rate ratios (IRRs) and (B) odds ratios (ORs).

The leave-one-out sensitivity analysis for the IRR group indicated that the overall effect size remained nonsignificant regardless of which study was excluded (Supplementary Figure 6).

Publication bias

The funnel plot is shown in Supplementary Figure 7. Egger’s test revealed significant small-study effects (β = 2.58, p < .001), suggesting potential publication bias. Trim-and-fill analysis imputed three missing studies on the left side of the funnel plot to adjust for asymmetry. The pooled effect size decreased from an observed IRR of 1.21 (95% CI: 0.99, 1.47) to an adjusted IRR of 1.07 (95% CI: 0.78, 1.47) after imputation. The original nonsignificant association between social isolation and inpatient care remained consistent, even accounting for the hypothetical missing studies (Supplementary Figure 8).

Subgroup analysis

We conducted subgroup analyses within the OR group on social isolation and inpatient care based on measurement instruments, study design, and adjustment for confounders. For measurement instruments, studies were categorized into single-item (n = 1) and multi-item (n = 10) scales, with no statistically significant difference observed (Supplementary Figure 9). Study design subgroup analysis showed pooled ORs of 0.596 (95% CI: 0.36, 1.55) for cross-sectional studies (n = 4) and 0.15 (95% CI: 0.01, 0.31) for cohort studies (n = 6), with no statistically significant difference (Supplementary Figure 10). Regarding confounder adjustment, the pooled OR was 0.28 (95% CI: 0.07, 0.64) for adjusted studies (n = 9), and 0.12 (95% CI: 0.07, 0.17) for unadjusted studies (n = 2), with no statistically significant difference between groups (Supplementary Figure 11).

Sensitivity analyses

We conducted sensitivity analyses excluding studies that met the age eligibility criterion solely by including over 50% of participants aged ≥60 years (Bartley et al., 2024; Bu et al., 2020; Chamberlain, Savage, et al., 2022). Across all relevant pooled estimates, the exclusion of each study did not significantly change the results, and the statistical significance or nonsignificance of the findings remained consistent (Supplementary Figures 12-15).

Discussion and implications

This study is the first systematic review and meta-analysis to synthesize the evidence on the associations of social isolation, loneliness, and four types of healthcare utilization across various study designs, including cohort, case–control, and cross-sectional studies. Social isolation was statistically associated with inpatient care utilization but not with emergency department visits. Loneliness was statistically associated with emergency department visits and inpatient care but not primary care utilization.

Our analysis did not reveal a statistically significant association between social isolation and emergency department visits. However, substantial heterogeneity was observed, with nearly half of the included studies reporting significant associations (Bartley et al., 2024; Mosen et al., 2023; Mosen et al., 2021; Savitz et al., 2023). This inconsistency may reflect differences in healthcare access and population characteristics across study settings. For instance, Barnes et al. found reduced emergency department utilization among socially isolated older adults, which may indicate that structural barriers, such as limited mobility or transportation access, simultaneously restrict social engagement, and healthcare-seeking behaviors (Barnes et al., 2022). Meanwhile, Chamberlain et al. found no increased risk of emergency department visits among residents in supportive living facilities (Chamberlain, Bronskill, et al., 2022). In these settings, decisions to transfer residents to the emergency department are influenced by hierarchical reporting structures, clinical knowledge of staff, and communication processes, rather than by the residents themselves (Tate et al., 2020). Regular health monitoring and proactive care in these environments likely reduce both medical emergencies and the perceived need for emergency department visits as social contact, weakening the link between isolation and acute care use (Lemoyne et al., 2019).

The statistically significant association between social isolation and inpatient care may be due to its cumulative effects on physical health. Socially isolated individuals may experience progressive health deterioration due to limited access to preventive care, leading to unmanaged chronic conditions and avoidable hospitalizations (Mistry et al., 2001). This is consistent with our IRR results, which reflect long-term risks, indicating that prolonged isolation worsens health over time. In contrast, ORs from cross-sectional studies capture associations at a single point in time, which may overlook cumulative effects, suggesting that inpatient care is not directly caused by isolation itself but could be influenced by the health issues resulting from prolonged isolation (Alzahrani, 2021).

Compared to inpatient or primary care, the stronger association between loneliness and emergency department visits may reflect distinct psychosocial pathways. Loneliness likely increases sensitivity to physical symptoms and psychological distress (Mushtaq et al., 2014), leading older adults to perceive non-emergent issues as urgent, thus increasing emergency department utilization. This aligns with evidence that lonely individuals often use emergency departments to address unmet social or mental health needs, especially during acute distress (Chamberlain, Bronskill, et al., 2022; Chamberlain, Savage, et al., 2022). In contrast, inpatient care is mainly driven by medical necessities, such as surgeries or life-threatening conditions, where clinical needs are the primary focus, aligning with our non-significant IRR findings. The lack of association with primary care might be due to barriers such as stigma, transportation issues, or fears of burdening providers, which can prevent lonely older adults from making routine visits (Holaday et al., 2022), redirecting their needs to emergency departments. Therefore, emergency departments serve as safety nets for both medical crises and psychosocial vulnerabilities, while primary and inpatient care settings are less responsive to help-seeking related to loneliness (Lim & Chan, 2017).

Three subgroup analyses did not reveal statistically significant differences in effect estimates, and substantial heterogeneity persisted within all groups (I² >90%). This may suggest that none of these methodological factors alone adequately explain the observed heterogeneity. One likely contributor is the variability in how social isolation is measured across studies. For instance, the Lubben Social Network Scale emphasizes structural aspects of social relationships, such as contact frequency and network size (Siette et al., 2021), whereas the Duke Social Support Index focuses on perceived behavioral or emotional support (Pachana et al., 2008). Additionally, the inclusion of both cross-sectional and cohort studies introduces further complexity. Cross-sectional designs cannot establish temporality, whereas cohort studies offer stronger evidence regarding the direction of associations. However, substantial heterogeneity persisted across both study types, and inconsistencies in the adjustment for confounders further limit comparability. These findings highlight the need for cautious interpretation of pooled estimates and reflect the methodological and contextual complexity involved in synthesizing evidence across diverse study designs and contexts.

Strengths and limitations

This meta-analysis has several strengths. First, while previous research often focused on a single type of healthcare utilization, our study is the first to include four types of healthcare utilization. Second, we searched five major databases to minimize the risk of missing relevant studies. This comprehensive approach enhances the understanding of the relationship between social isolation/loneliness and healthcare utilization. Third, instead of converting different effect measures into a single metric, we performed stratified analyses for IRR and OR estimates to reduce bias and ensure methodological validity.

However, several limitations should be acknowledged. First, we only include English-language articles, which may create language bias. Second, the lack of sufficient data on social isolation and loneliness regarding outpatient care utilization prevented the conduct of meta-analyses, limiting the ability to draw comprehensive conclusions about these specific healthcare settings. Third, the substantial heterogeneity observed in the pooled analyses. Differences in study designs (e.g., inconsistent adjustment for confounders such as socioeconomic status), populations (e.g., community-dwelling seniors, institutionalized older adults), and measurement tools (e.g., varying scales for social isolation and loneliness) may limit the comparability of studies. In addition, unmeasured contextual factors, such as geographic differences in healthcare access, disruptions related to the pandemic, and cultural norms affecting healthcare-seeking behavior, may further contribute to inconsistent findings. These variabilities may limit the generalizability of our findings.

Implications for future research

Our findings highlight several areas for future research to better understand the complex relationships between social isolation, loneliness, and healthcare utilization. First, expanding investigations in primary and outpatient care is important to address the current evidence gap, as limited data in these areas limit the ability to draw robust conclusions. Second, the substantial heterogeneity in effect sizes across studies highlights the need to identify factors that influence these relationships. Measurement discrepancies in social isolation and loneliness may contribute to inconsistent findings, suggesting that standardized measurement tools and reporting practices are essential to improve comparability across studies. Additionally, future studies should address unmeasured confounders, such as socioeconomic disparities and geographic barriers. Addressing these areas can contribute to the development of targeted interventions and policies aimed at reducing social isolation and loneliness, thereby improving healthcare access and outcomes for older adults. For healthcare providers, integrating routine screening for social isolation and loneliness into primary care visits could identify high-risk older adults earlier, enabling referrals to community resources or mental health services.

Conclusion

This systematic review and meta-analysis provides comprehensive insights into the associations between social isolation, loneliness, and healthcare utilization across various settings. The findings indicate that social isolation is statistically associated with increased inpatient care utilization, while loneliness is linked to higher emergency department visits and inpatient care. The study highlights the importance of considering different effects when interpreting results, as well as the influence of environmental factors. Future research should focus on diverse healthcare settings, the impact of external factors, and the development of targeted interventions to address social isolation and loneliness and improve healthcare outcomes for older populations.

Supplementary material

Supplementary data are available at Innovation in Aging online.

Funding

None declared.

Conflict of interest

None declared.

Data Availability

This study was preregistered in the PROSPERO database under record number CRD420251036578. All data analyzed in this study are included in this published article and its supplementary information files.