Loneliness and Social Isolation as Risk Factors for Mortality a Meta-analytic Review
Cardiac risk factors and prevention
Loneliness and social isolation as risk factors for coronary middle disease and stroke: systematic review and meta-assay of longitudinal observational studies
Abstract
Background The influence of social relationships on morbidity is widely accepted, only the size of the risk to cardiovascular health is unclear.
Objective Nosotros undertook a systematic review and meta-analysis to investigate the association between loneliness or social isolation and incident coronary heart illness (CHD) and stroke.
Methods Sixteen electronic databases were systematically searched for longitudinal studies prepare in high-income countries and published up until May 2015. Two independent reviewers screened studies for inclusion and extracted data. We assessed quality using a component approach and pooled data for analysis using random furnishings models.
Results Of the 35 925 records retrieved, 23 papers met inclusion criteria for the narrative review. They reported data from 16 longitudinal datasets, for a total of 4628 CHD and 3002 stroke events recorded over follow-up periods ranging from 3 to 21 years. Reports of eleven CHD studies and eight stroke studies provided data suitable for meta-assay. Poor social relationships were associated with a 29% increment in risk of incident CHD (pooled relative risk: 1.29, 95% CI 1.04 to 1.59) and a 32% increase in risk of stroke (pooled relative adventure: 1.32, 95% CI i.04 to ane.68). Subgroup analyses did not place whatsoever differences by gender.
Conclusions Our findings suggest that deficiencies in social relationships are associated with an increased risk of developing CHD and stroke. Hereafter studies are needed to investigate whether interventions targeting loneliness and social isolation tin help to prevent two of the leading causes of expiry and disability in high-income countries.
Study registration number CRD42014010225.
This is an Open Admission article distributed in accordance with the terms of the Creative Commons Attribution (CC By 4.0) license, which permits others to distribute, remix, conform and build upon this work, for commercial utilise, provided the original work is properly cited. See: http://creativecommons.org/licenses/past/iv.0/
Statistics from Altmetric.com
Introduction
Adults who have few social contacts (ie, who are socially isolated) or experience unhappy about their social relationships (ie, who are lonely) are at increased run a risk of premature mortality.ane The influence of social relationships on mortality is comparable with well-established chance factors, including physical activity and obesity.2 Yet, compared with our understanding of these hazard factors, we know much less near the implications of loneliness and social isolation for affliction aetiology.
Researchers have identified iii primary pathways through which social relationships may affect health: behavioural, psychological and physiological mechanisms.3 ,four Health-risk behaviours associated with loneliness and social isolation include concrete inactivity and smoking.5 Loneliness is linked to lower self-esteem and limited employ of active coping methods,6 while social isolation predicts decline in self-efficacy.7 Feeling lonely or beingness socially isolated is associated with lacking immune operation and higher claret pressure level.8 ,9 This testify suggests that loneliness and social isolation may be important run a risk factors for developing illness, and that addressing them would benefit public health and well-being.
The aim of this study was to investigate the size of the association between deficiencies in social relationships and incident coronary heart disease (CHD) or stroke, the two greatest causes of burden of disease in high-income countries.ten We conducted a systematic review to answer the following chief question: are deficiencies in social relationships associated with developing CHD and stroke in loftier-income countries? Our secondary objectives included investigating whether loneliness or social isolation was differentially associated with incident heart disease and stroke, and whether the association between social relationships and disease incidence varied according to age, gender, marital status, socioeconomic position, ethnicity and health.
Methods
This written report followed the Centre for Reviews and Dissemination's Guidance for undertaking reviews in healthcare.11 A protocol was registered with the International Prospective Register of Systematic Reviews (registration number: CRD42014010225).12
Study eligibility criteria
To see inclusion criteria, studies had to investigate new CHD and/or stroke diagnosis at the individual level as a office of loneliness and/or social isolation. We divers CHD equally encompassing the diagnoses listed under codes l20–l25 of the 10th revision of the International Statistical Classification of Diseases and Related Health Bug (ICD-10), and stroke as ICD-ten codes I60–69. Nosotros excluded studies where CHD or stroke diagnosis was not the first instance of diagnosis amid participants, except where analyses controlled for previous events. We practical no other exclusion criteria regarding study population. Measures of social relationships met inclusion criteria for loneliness if they were consistent with its definition as a subjective negative feeling associated with someone's perception that their relationships with others are deficient.thirteen Measures of social isolation had to be consequent with its definition equally a more objective measure out of the absence of relationships, ties or contact with others.14 Nosotros focused on longitudinal studies in order to investigate the temporal relationships between loneliness or isolation and subsequent disease. Our purpose was to clarify the public wellness challenge posed by deficiencies in social relationships in high-income countries,15 so nosotros excluded all other settings. We applied no language, publication type or date restrictions to inclusion.
Search strategy and selection criteria
Nosotros searched xvi electronic databases for published and gray literature published up until May 2015: MEDLINE, EMBASE, CINAHL Plus, PsycINFO, ASSIA, Web of Science, Cochrane Library, Social Policy and Practice, National Database of Ageing Research, Open up Grey, HMIC, ETHOS, NDLTD, NHS Evidence, SCIE and National Institute for Health and Care Excellence (Dainty). Thesaurus and complimentary text terms (eg, loneliness, social isolation, social relationships, social support, social network) were combined with filters for observational study designs and tailored to each database. The search strategy included no health terms, as it aimed to capture all disease outcomes, rather than focus on CHD and stroke. For the full electronic strategy used to search MEDLINE, come across online supplementary appendix 1.
Supplementary appendix ane
To complement the electronic search, we screened reference lists, searched for citations in Scopus (the largest database of abstracts and citations) and contacted topic experts identified through the Britain Entrada to Finish Loneliness' Enquiry Hub.
Subsequently removing duplicates, ii researchers independently screened titles and abstracts before assessing full records using a standardised screening sail. Additional information was sought from authors when necessary (3 (60%) responded). When authors did not reply, nosotros searched for information from related publications to inform our decision.
Data extraction and quality assessment
Data were extracted into a standardised form by one researcher, and checked by a second. Study authors were contacted to obtain missing information.
Based on the Bureau for Healthcare Research and Quality framework and taxonomy of threats to validity and precision,sixteen nosotros selected the post-obit domains as relevant for assessing studies: sampling bias, non-response bias, missing information, differential loss to follow-upward, information error with regard to exposure and outcome measure, detection bias, misreckoning and study size. We identified age, gender and socioeconomic condition as potential confounders (ie, factors correlated with exposure, predictive of upshot and non on the causal pathway).17 ,eighteen No studies were excluded due to quality; instead, subgroup and sensitivity analyses were performed, to test the stability of findings according to internal validity.
Quantitative synthesis
We hypothesised that social relationships were associated with disease incidence, and that this association may differ according to the dimension of relationships measured, and individual-level and contextual-level factors. A preliminary synthesis was developed by group study characteristics and results according to their measure of relationships. The bulk of papers reported relative hazards of new diagnosis, comparing people with higher versus lower levels of loneliness or social isolation. Since incidence of disease was low (<10%) in the three studies reporting ORs, these estimates were approximated to relative risks.xix Where the lonely or isolated group was used as the reference, results were transformed to allow comparison across studies.
Patterns identified in the preliminary synthesis were formally investigated. Only papers for which an effect estimate and SE or CI were available (reported in the paper or provided by contacted authors), or could exist calculated, contributed to this stage of the analysis. Where several papers reported results from the same accomplice, we privileged the findings with the longest follow-up time. If a study included multiple measures of exposure and/or effect, we selected the result relating to the most comprehensive mensurate. Where a study used statistical controls to calculate an effect size, we extracted data from the nearly circuitous model to minimise risk of confounding. All consequence sizes were transformed to the natural log for analyses. Using Revman V.five.iii (Review Manager (RevMan) Version 5.3 [plan]. Copenhagen: The Nordic Cochrane Centre, 2014), CHD and stroke effect estimates were plotted in separate forest plots, and heterogeneity between studies was assessed using the Iii statistic.
Potential sources of variation were explored with prespecified subgroup analyses. Since heterogeneity could not exist explained and removed based on these analyses, only nosotros accounted studies sufficiently like to warrant aggregation, we combined results using random effects models. This approach allows for between-report variation, and is consistent with our assumption that the effects estimated in the different studies were not identical, since they investigated unlike dimensions of social relationships and derived from different populations.
Finally, sensitivity analyses were performed to test whether our overall results were affected past internal study validity and pocket-sized-study effects. Profile-enhanced funnel plots for asymmetry were drawn using STATA Five.12 (Stata Statistical Software: Release 12 [plan]. Higher Station, TX: StataCorp LP, 2011). The limited number and the heterogeneity of studies did non back up the use of tests for funnel plot asymmetry.20
Results
A total of 23 studies based on 16 cohorts were identified for inclusion in the review, after a two-stage process. Encounter figure 1 for a catamenia diagram of the report pick process. Eleven studies on CHD and viii studies on stroke met inclusion criteria for the quantitative syntheses (ie, studies based on contained samples reporting information from which the natural log of the estimate and its SE could derived).
Table one summarises the descriptive characteristics of the evidence included in our review (meet online supplementary appendix 2 for individual report characteristics).
Supplementary appendix 2
View this table:
- View inline
Tabular array 1
Characteristics of the included show
Assessment of loneliness and social isolation
Prevalence of loneliness or social isolation ranged from 2.viii%forty to 77.ii%.31 Three papers measured loneliness,21 ,30 ,42 xviii measured social isolation22–43 and two papers used a measure out combining both dimensions.34 ,35 The 3 papers on loneliness used different tools: a direct question request nigh loneliness feelings during the day,30 a question on feelings of loneliness in the past week42 and a thirteen-detail tool encompassing the perceived availability, adequacy or accessibility of social relationships.21 Across the 18 studies on social isolation, 11 tools were used: half-dozen studies used the Berkman–Syme Social Network Index,44 two studies used the 10-item Lubben Social Network Scale45 and the remainder used nine different tools on the availability and/or frequency of contacts. 1 cohort written report used a measure combining social isolation and loneliness, the 11-item Knuckles Social Support Index, which asks about frequency of interaction and satisfaction with relationships.46
Loneliness and social isolation were predominantly treated equally a categorical variable; two studies analysed them every bit continuous variables.29 ,42 Just one written report reported results based on measuring social relationships more than in one case.42
Ascertainment of CHD and stroke
A total of 4628 CHD and 3002 stroke events were recorded across the 23 papers. Eighteen studies measured incident CHD and 10 measured stroke (five studies reported on both outcomes). Diagnosis was ascertained from medical records, death certificates or national registers in all but 4 studies. Others used self-written report,34 ,35 or telephone interviews with a nurse or physician.33 Two studies verified self-reported events confronting medical records.29 ,36 ,38 The majority of studies with a measure of CHD focused on myocardial infarction and/or CHD death (11/18). Four studies included angina pectoris within their measure of CHD and two presented results for angina separately. The remit of the CHD measure was unclear in one report.43
Report validity
Figure 2 summarises risk of bias across the studies included in our review (encounter online supplementary appendix 3 for details of criteria). For many of the instruments assessing social relationships, information on reliability and validity was limited (online supplementary appendix 4 displays detailed data on the validity and reliability of tools). Iv cohorts (vi articles) relied on subjects reporting new diagnosis for all or role of the outcomes measured, and were judged to be at greater risk of misclassification (see online supplementary appendix two for details of upshot assessment). Limited data on attrition and blinding of outcome assessment meant that susceptibility to differential loss to follow-upwards and detection bias was unclear. We note that the multiplicity of risk factors investigated and the differential length of follow-up suggest that outcome assessment is unlikely to take been influenced by knowledge of baseline information on social relationships.
Supplementary appendix 3
Supplementary appendix 4
The results reported in 12 papers were at lower risk of misreckoning, that is, analyses controlled or accounted for age, gender and socioeconomic status.21 ,22 ,27 ,28 ,xxx ,33 ,36 ,37 ,39 ,40 ,42 ,43 4 studies presented results from univariate analyses,31 ,34 ,35 ,41 with a further report adjusting for age only.26 The remaining viii reports did non control for socioeconomic condition, although in the case of the Wellness Professionals Follow-up Study the relative socioeconomic homogeneity of the sample may limit the touch of this omission.22 ,24
Loneliness, social isolation and CHD
Across 11 studies (3794 events; one study did not report numbers) based on independent samples, the boilerplate relative take chances of new CHD when comparing high versus low loneliness or social isolation was 1.29 (95% CI i.04 to 1.59; see figure 3). We found evidence of heterogeneity inside this comparing (I2=66%, χ2=29.16, df=10, p=0.001) and explored whether this could be explained by social relationship domain (loneliness vs social isolation), gender, risk of misreckoning and college risk of bias due to exposure measurement error. We establish no evidence that effects differed according to each subgroup (see online supplementary appendix 5). We were non able to explore other potential sources of heterogeneity due to express information and study numbers.
Supplementary appendix 5
Social isolation and stroke
Across 9 independent study samples (2577 events; ane study did not report numbers), the average relative risk of stroke incidence was 1.32 (95% CI i.04 to one.68; see figure four). Post-obit confirmation of heterogeneity (I2=53%, χ2=17.07, df=eight, P=0.03) we performed subgroup analyses according to chance of misreckoning and chance of bias due to outcome measurement error (there were besides few studies to perform whatsoever other analyses). At that place was no evidence of furnishings differing according to subgroup (encounter online supplementary appendix 6); we had insufficient data to explore other potential sources of heterogeneity.
Supplementary appendix 6
Risk of bias across studies
To test whether our findings were sensitive to internal study validity, we compared results with and without studies at greater hazard of bias. We found no evidence of a departure in the ratio of the relative risks for CHD and stroke according to written report validity (see table 2).
View this table:
- View inline
Table 2
Sensitivity analyses
Visual assessment of contour-enhanced funnel plots suggested that studies might be missing in areas of statistical significance (see figure 5A, B). Comparison fixed-furnishings and random-effects estimates, nosotros constitute the random-effects gauge to exist more than beneficial (CHD: relative risk (RR) random-furnishings: 1.29, 95% CI 1.04 to ane.59, compared with RR fixed-furnishings: 1.xviii, 95% CI 1.06 to i.31; stroke: RR, random-furnishings: i.32, 95% CI 1.04 to ane.68, compared with RR stock-still-effects: 1.xix, 95% CI 1.03 to one.36). This suggests the presence of modest-study furnishings, which could be due to reporting bias. Although we found no evidence that study quality and truthful heterogeneity explained small-study effects in our review, these, forth with risk, remain possible explanations.
Additional studies
7 papers with a mensurate of social isolation were excluded from quantitative synthesis since they either did not written report information in a format suitable for pooling and/or shared information with other studies.23 ,25–27 ,29 ,38 ,41 Of the four papers that did not duplicate data from other studies, two reported results based on the Honolulu Heart Programme: social isolation appeared to predict CHD simply not stroke, in analyses adjusted for historic period, though the association disappeared in multivariate analysis.26 ,27 In a univariate analysis of information from the Atherosclerosis Take chances in Communities Report (USA) the Lubben Social Network score was not significantly associated with incident CHD among people with prehypertension.41 A further study establish no evidence of an clan between social isolation and CHD amongst men in France and Northern Ireland,29 although we note that this report controlled for low, one of the possible pathways through which social isolation might pb to disease.
Discussion
Summary of findings and comparison with other work
Our review establish that poor social relationships were associated with a 29% increase in chance of incident CHD and a 32% increment in risk of stroke. This is the first systematic review to focus on the prospective association between loneliness or social isolation and showtime occurrence of CHD or stroke.
Before reviews reported that cardiovascular disease (CVD) prognosis is worse amidst people with poorer social relationships.one ,2 Narrative reviews on social support and CHD have described an clan with prognosis likewise as incidence, but the strength of evidence was depression.47 ,48 A recent review of seven papers linked loneliness and social isolation to occurrence of CHD,49 but the effect on prognosis and incidence could non be disentangled.
We plant an association between poor social relationships and incident CVD comparable in size to other recognised psychosocial risk factors, such equally anxiety50 and chore strain.51 Our findings point that efforts to reduce the risk of CHD and stroke could benefit from taking both loneliness and social isolation into account, as we establish no show to suggest that one was more strongly related to disease incidence than the other. This is in line with other research linking subjective and objective isolation to hypertension, a risk gene for both stroke and CHD.8 ,nine
Strengths and limitations
Our focus on longitudinal studies allowed us to comment on the direction of the relationship between social relationships and wellness, and avert the problem of reverse causation. Pooling results from studies of CHD that measured loneliness and isolation immune us to answer the broader question of whether deficiencies in social relationships are associated with disease incidence. We anticipated and explored heterogeneity where possible only found no statistical evidence that components of internal validity were associated with issue estimates.
Subgroup analyses specified a priori showed no difference between the clan of loneliness or social isolation with CHD incidence, and we institute no evidence beyond studies of differences betwixt men and women. We establish insufficient data to explore the relative effects of the quantity and quality of relationships, or study effect modifiers in depth. Seven of the estimates included in our meta-analyses (five CHD, two stroke) were extracted from studies where participants were of college socioeconomic status and in better health than the target population. The function of deficiencies in social relationships may be greater among individuals under stress,52 and our results may underestimate the health-damaging implications of loneliness and social isolation amongst disadvantaged groups. Our review included some data nerveless from 1965; more contempo strategies for CHD prevention may take modified the influence of loneliness and social isolation on disease incidence.
In mutual with other reviews of observational studies, we cannot infer causality from our findings, nor can nosotros exclude confounding by unmeasured mutual causes, or reverse causation if deficiencies in social relationships are the outcome of subclinical disease. Publication bias is a concern in every review, and may lead us to overestimate the 'true' effect of poor social relationships. Conversely, our pooled effects could be a conservative estimate: most of the studies in this review statistically adjusted for factors that are probable to be on the causal pathway, such every bit depression or health-related behaviour.
Implications
The main finding of our review, that isolated individuals are at increased run a risk of developing CHD and stroke, supports public health concerns over the implications of social relationships for health and well-being. Our work suggests that addressing loneliness and social isolation may have an important part in the prevention of 2 of the leading causes of morbidity in high-income countries.
A variety of interventions directed at loneliness and social isolation have been developed, ranging from grouping initiatives such every bit educational programmes and social activities, to one-to-one approaches including befriending and cognitive-behavioural therapy. These have primarily focused on secondary prevention, targeting people identified as isolated or lonely, but their effectiveness is unclear. Evaluative research is needed to investigate their touch on a range of wellness outcomes. Addressing health-damaging behaviours is as well likely to exist important, with solitary and isolated people more likely to smoke and be physically inactive, for example5 main prevention strategies, such as promoting social networks or developing resilience, have received limited attending to date. Adventure factors for loneliness and social isolation such equally gender, socioeconomic position, bereavement and health condition are well established14 ,18 and concur the central to identifying people who may do good from intervention.
Our findings advise that tackling loneliness and isolation may be a valuable addition to CHD and stroke prevention strategies. Health practitioners have an of import role to play in acknowledging the importance of social relations to their patients.53 54
Key messages
What is already known on this subject area?
-
People with poorer social relationships are at increased take a chance of premature death. The implications of social relationships for disease onset are unclear.
What might this written report add together?
-
This systematic review of prospective longitudinal studies constitute that deficiencies in social relationships are associated with an increased hazard of developing coronary center disease and stroke of around 30%. This association is comparable in size to other recognised psychosocial risk factors, such as feet and job strain.
How might this touch on on clinical exercise?
-
Efforts to reduce cardiovascular disease incidence demand to consider loneliness and social isolation.
Acknowledgments
We thank Rocio Rodriguez-Lopez and Melissa Harden for carrying out the electronic literature searches, and Martin Bland and Dan Pope for their advice on meta-analysis software and data assay.
References
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
View Abstract
Request Permissions
If you wish to reuse whatsoever or all of this article please use the link below which volition take you to the Copyright Clearance Center'southward RightsLink service. You will be able to get a quick cost and instant permission to reuse the content in many unlike ways.
Copyright information:
Linked Articles
- Editorial
- Heartbeat
Source: https://heart.bmj.com/content/102/13/1009
0 Response to "Loneliness and Social Isolation as Risk Factors for Mortality a Meta-analytic Review"
Post a Comment