Abstract
The U.S. Surgeon General’s 2023 advisory on America’s loneliness epidemic defines loneliness as a discrepancy in relationship quality linked to higher risk for severe mental illness. It is unclear what is meant by relationship quality and how these qualities would scale large enough to produce a loneliness epidemic.
Keywords
Introduction
In 2023, the U.S. Surgeon General issued an advisory to address serious levels of social isolation or loneliness in up to 40% of U.S. adults, stating, “Social connection is a fundamental human need, as essential to survival as food, water, and shelter” (U.S. Department of Health and Human Services, 2023). Human needs including food, water, shelter, are a distinct type of physiological need. They are considered critical to support the normal functioning of cellular processes (e.g. metabolism, osmotic equilibrium, and body temperature), that are tightly regulated to maintain human survival (McEwen, 1998, 2005). Adding social connection, or human relationships, to this list supports the idea of social homeostasis. Social homeostasis suggests that loneliness is a physiological indicator of deprivation in protective social bonds, generating stress responses that can become chronic until an expected level of exposure to social contact has been met (Cacioppo & Patrick, 2008; Coan & Sbarra, 2015; Fulford & Holt, 2023). Loneliness is thought to be an indicator of poor quality in social connection associated with overall worsening mental health conditions, such as anxiety, depression, suicidal ideation, and severe mental illness (Fulford & Holt, 2023; U.S. Department of Health and Human Services, 2023). Framing social connection as critical to survival, and loneliness as a form of social deprivation and social determinant of mental health, is a paradigm shift that calls for broader policy and institutional participation to intervene (U.S. Department of Health and Human Services, 2023; Wickramaratne et al., 2022). There is a need to better understand what is meant by poor relationship quality to guide loneliness interventions in our communities.
An important obstacle to designing interventions for loneliness is the tension between the objectivity of human need and the subjectivity of loneliness. Loneliness is often distinguished from objective social isolation as a solely subjective reality that individuals feel alone even if they have existing relationships (Lim, Holt-Lunstad, & Badcock, 2020). Loneliness used to be considered a mental health disorder similar to depression (Pettigrew & Roberts, 2008). It can still be treated with psychological interventions to correct dysfunctional or biased perceptions that lonely individuals are thought to have about their social experiences, and discourage behaviors to self-isolate (Cacioppo et al., 2015; Fulford & Holt, 2023; Hickin et al., 2021). Therapeutic interventions for loneliness have shown to be significant in reducing loneliness, but the effect size can be minimal and variable, and seems to depend on the social environment, like living in a group setting in long-term care (Cacioppo et al., 2015; Hickin et al., 2021; Hoang et al., 2022).
Health researchers have also found that objective social experiences such as divorce, death of loved ones, violent assault, and domestic and child abuse, are associated with higher levels of loneliness (DiJulio et al., 2018; Landry et al., 2022; Shepp et al., 2023). An accumulation of these social experiences over the life course increases feelings of loneliness (DiJulio et al., 2018; Liberatore-Maguire et al., 2022). Due to a growing awareness that loneliness is linked to negative social experiences of marginalized populations, loneliness has been repositioned as central to understanding how marginalization of stigmatized communities impacts health outcomes (Shepp et al., 2023). With a historical focus of loneliness and social isolation interventions on older adult experiences, addressing social risk factors like discrimination, domestic abuse, etc., are not being identified as loneliness interventions nor included in systematic reviews to examine their effectiveness on reducing loneliness.
Critical realism may be able to handle the growing tension between objective and subjective understanding of why loneliness occurs to inform interventions. Critical realism is a philosophy about existence that acknowledges an objective world independent of human perception, that humans can only know through the limitations of socially produced beliefs, values, and other forms of perspective (Koopmans & Schiller, 2022). It integrates how knowledge is produced in the natural and social sciences (Hastings, 2021). A critical realist approach in health research assumes that an approximate reality can be produced to explain causal mechanisms well enough to treat health-related conditions (Koopmans & Schiller, 2022). Conclusions from quantitative research to determine patterns and cause and effect relationships are still important in critical realism, but qualitative research can also produce causal explanations that describe the makeup of key factors and complex relationships they have with underlying forces, like social structures (Hastings, 2021; Koopmans & Schiller, 2022; Wiltshire & Ronkainen, 2021).
We used a critical realist approach to conduct a qualitative research study that examines causal mechanisms to feeling lonely, as observed by adults anonymously describing loneliness experiences in an online art project. Anonymous narratives were selected for this study because recruiting sufficient participants with serious or chronic loneliness is a known challenge in health research due to stigma and shame (Barreto et al., 2022). Theoretical models that incorporate a causal pathway for social connection as a human need and how this can be related to loneliness through relationship qualities, were used to frame narratives to improve explanation of causal mechanisms rather than exploration of unknown ideas (Fryer, 2022). We conducted thematic analysis to understand qualities within direct social interactions contributing to feeling lonely in adults, ages 18 and older, in 219 loneliness narratives. Results from this study are explanatory, to improve cohesive understanding of relationship quality contributing to adult loneliness that can inform the identification and design of interventions for loneliness.
Materials and Methods
Data Source
The University’s Institutional Review Board approved the study as exempt from human subject research (#23-2083). Secondary data come from The Loneliness Project, a public website developed by a Canadian artist in an effort to destigmatize experiences of loneliness (Korda & Rumball, 2017). The intention of The Loneliness Project was to express how loneliness is a shared and human experience (Kassam, 2018). The project was an open response online survey that provided four prompts to describe loneliness: Last time I felt lonely was; To me, loneliness means; One of the first times I realized I was lonely was; Tell me the story of the time you felt the most lonely. The survey was posted on the website to receive responses from the public between October 2017 to the end of 2018. The survey was publicized internationally, first through a personal Facebook post, and then was picked up by local and global news outlets based in Canada, the United States of America, and the United Kingdom (Korda, 2023). Over 2000 survey responses in English were received (Korda, 2023). The artist selected 300 responses to be posted publicly on The Loneliness Project website, to reflect what she found to be the greatest diversity in ages, lived experiences, and geographic locations (M. Korda, personal communication, January 12, 2023). The study authors were not involved in the Loneliness Project and the artist’s selection process of posting survey responses on the website was not intended for scientific research.
Study Population
There is limited sociodemographic information regarding the study sample available from The Loneliness Project website. Survey respondents were allowed to submit responses anonymously, choosing their own first name. There were not controls in place to prevent respondents from submitting more than once, but the artist selected less than 15% of total responses to post publicly that were intentionally selected to “publish enough variety that anyone could find something that speaks to some aspect of their specific [loneliness] experience” (M. Korda, personal communication, January 12, 2023). The artist did not detect any fraudulent participation and there were no financial incentives provided to respondents.
Distribution of Self-Reported Age of Respondents (N = 219)

Selection of Narratives From the Loneliness Project Used in Thematic Analysis
Theoretical Perspectives
A critical realism approach to thematic analysis uses existing theory as a part of the analysis, so that coding is theory-driven to help produce causal explanations that build on existing knowledge (Fryer, 2022). Two conceptual models were selected to guide thematic analysis of loneliness narratives, because they define a causal pathway from social connection as a human need to health outcomes, and how this causal pathway could specifically apply to loneliness.
The Social Exposome Model
The social exposome is a conceptual model about how the entirety of exposure to a social environment over time impacts health outcomes (Gudi-Mindermann et al., 2023). This model was selected because it identifies social interactions as a necessary component for the broader social world, such as culture, to cross through in order to impact health outcomes. The definition of social interactions guided our thematic analysis to consider three components: the actors or people involved in an interaction, the places where social interactions occur, and the dynamic of a social interaction, such as a relationship quality (Gudi-Mindermann et al., 2023). The model describes embodiment as a biological pathway for how exposure to a social environment can become embedded in human physiology, such as the enforcement of gender norms in social interactions that become learned and influence behavior (Gudi-Mindermann et al., 2023; Krieger, 2024). Neuroscience suggests a physiological dependence on physical exposure to other people, that when there is social deprivation, it can produce stress for humans as a social species (Coan & Sbarra, 2015; Fulford & Holt, 2023).
The Cognitive Discrepancy Model
In addition, the cognitive discrepancy model is commonly referenced in loneliness research to suggest that loneliness occurs when the quality of relationships one has is different than the quality of relationships that one wants (U.S. Department of Health and Human Services, 2023). For the purposes of this study, the definition of social interactions was used to focus on relationship quality in loneliness experiences to better explain this discrepancy.
Analysis
Secondary Data Considerations
Using online written material that was unintended for research requires additional thought about conducting secondary qualitative data analysis (Chatfield, 2020; Keim-Malpass et al., 2014). Validation of results is improved when research questions are aligned with the original intention of the data (Chatfield, 2020). Data came from a single website focused on sharing stories about loneliness. In news interviews, the artist describes The Loneliness Project as a reaction to social media’s attention on human happiness that does not allow for sharing more difficult parts of human life that are also normal (Kassam, 2018).
After reading respondents’ stories, the artist reaffirms the cognitive discrepancy model – noticing that there is a disconnect between the kinds of relationships that people want and their reality (Kassam, 2018). The artist, though, rejects the idea that loneliness should be treated like a health concern or medical condition, believing this will further stigmatize those who feel lonely. This is an important contradiction we include in our analysis, to allow for loneliness to be normal without need for intervention. It aligns with our approach to include anonymous contributors in loneliness research who may seek out anonymity because they feel stigmatized for their loneliness (Barreto et al., 2022; Rains, 2014). Online written material was also selected to have a large amount of naturalistic information that reflects a wide range of views on the same health topic, as our study is trying to understand a shared human need (Keim-Malpass et al., 2014).
Thematic Analysis
A critical realist approach to thematic analysis of data was conducted. It followed 5 main steps (Campbell et al., 2021; Fryer, 2022). These steps are iterative, being continuously returned to throughout the analytic process.
The first step is to identify the research question (Fryer, 2022). To reiterate, our study primarily examined the question of what are the underlying causes of loneliness in adults, as related to relationship quality?
Step two is developing familiarity with the data (Fryer, 2022). Data were organized using ATLAS.ti 23. A supervising researcher and the lead author conducted preliminary work from January to May 2023 to become familiar with data. We met each week to reflect on initial thoughts about narratives, our contributing areas of expertise, and to improve on study design (Fryer, 2022). We also discussed key term definitions to build a shared understanding. Reflections were documented in an on-going research journal (Fryer, 2022). We used a random number generator to select only 30 narratives to read and analyze to avoid limiting future analysis. The research team was trained and oriented using the selected 30 narratives.
Step three was the act of developing codes for our total sample size of 219 loneliness narratives (Fryer, 2022). Our coding was theory driven. We incorporated our theoretical models into the qualitative analysis itself. For example, we used the components of social interactions (actor, place, dynamic) to guide coding, with an emphasis on understanding relationship qualities (Figure 2). Example of How Theory Guided Thematic Analysis
Descriptive codes were used to capture the complexity of social interactions across a diversity of social experiences contributing to loneliness, as well as to incorporate the intentions of the artist and knowledge within our research team (Fryer, 2022). Each narrative was double coded independently between the lead author and a research assistant. To improve validity and reliability, the lead author and research assistant regularly met to standardize the naming and definitions of codes to best reflect the data, and consider consolidating them when the volume of code increased (Fryer, 2022; Wiltshire & Ronkainen, 2021). An advising researcher who is an expert in qualitative research also reviewed code names, definitions, and key decisions for consolidation. These key decisions were tracked in an audit trail (Cutcliffe & McKenna, 2004).
Reflexivity is an important part of critical realism that recognizes it is not possible to remove researcher bias from thematic analysis (Fryer, 2022). As a team we reflected on patterns in the data from the perspective of a social worker trained in clinical interventions to improve systems of relational health, of a research assistant trained in global health to consider cultural differences in health topics, and of the lead author trained in systems biology (e.g. ecology) and public health to address social determinants of health. We found our shared knowledge of systems thinking helped us understand the complexity of social environments found in loneliness experiences (Meadows, 2008). Accepting all loneliness narratives at face value, despite concerns raised about faulty perceptions of survey respondents, was a consistent tension in our research team. This was done to adhere to critical realist philosophy that there is no need to remove or determine subjectivity, as underlying participants’ personal experiences, an approximate understanding of objective reality can still be found (Wiltshire & Ronkainen, 2021). This also aligned with our efforts to include the intentions of the artist and survey respondents in The Loneliness Project to destigmatize loneliness as a normal part of human life (Kassam, 2018).
We found thematic saturation to be an important part of our analysis. Given the diversity of loneliness experiences from anonymous adults around the world, it wasn’t clear that we would be able to consolidate our code to describe what survey respondents had in common. Our team defined thematic saturation as no new ideas were arising from loneliness experiences to understand relationship quality or other causal explanations. Preliminary thematic saturation was reached after coding 83 loneliness narratives (Guest et al., 2020). However, we did not have theoretical reason for stopping and coded the full study sample to ensure saturation (Fryer, 2022; Guest et al., 2020). We divided up remaining narratives to independently code until all 219 narratives had been reviewed, only discussing them as a research team if a new idea was found.
Step four was the development of themes (Fryer, 2022). For a critical realist approach to thematic analysis, the importance of this step is distinguishing among the personal experience of loneliness, identifying shared events contributing to loneliness among a group of narrators, and developing themes to explain a causal mechanism producing the shared events (Fryer, 2022). For example, a survey response in our study may describe the personal experience of loneliness as feeling alone and afraid, the event that led to those feelings as being rejected by a teacher for being gay, and the underlying cause of that event as a cultural message that stigmatizes gay men.
The lead author organized the final codes into shared categories, which guided the discussion in developing themes. The research team discussed themes and reached decisions about thematic definitions. However, we decided to develop all three - experiences, events, and causal explanations - into themes. We found the novel use of the social exposome model to give new insight into each of these considerations. We assessed the validity of our themes that described a causal explanation, based on their explanatory power found within the larger body of loneliness population health research (Fryer, 2022; Hastings, 2021; Wiltshire & Ronkainen, 2021).
The final fifth step was to share and discuss results in the following sections.
Results
Thematic Analysis of Narratives From the Loneliness Project (N = 219)
A critical realist approach may consider these themes about relationship quality to be events, rather than underlying causal explanations, because they describe characteristics that were observable in direct social interactions. However, the theory we used in our analysis already explains the underlying force of human need for social connection, that we found it more accurate to describe themes about relationship qualities as an immediate causal explanation for how social interactions can result in adult loneliness.
The remaining themes were grouped into Underlying Causal Explanations to describe underlying forces contributing to the discrepancy in qualities in social interactions, Experiences of Loneliness to describe the personal or internal experience of loneliness, and Responses to Loneliness to describe how survey respondents ultimately reacted to the experience. Underlying causal explanations do not have to be observable, but need to explain how and why events, like poor quality social interactions, are taking place (Fryer, 2022; Hastings, 2021).
Discussion
Our results further explain the Cognitive Discrepancy Model that suggests loneliness develops when there is a discrepancy in the quality of relationships that someone wants versus what they have (U.S. Department of Health and Human Services, 2023). Our study explains the discrepancy as four poor qualities in social interactions that are different ways of lacking being cared for by others. These qualities could encompass a large range of social experiences. Review studies trying to understand risk factors for loneliness often end up collecting a long list of demographic and social factors that can be difficult to make sense of how to address (Barjaková et al., 2023; Lim, Eres, & Vasan, 2020). By using a critical realist approach to consider the hypothesized biological pathway of social connection as a human need and personal accounts of loneliness experiences, a short list of approximate causes of loneliness emerged in our study as simpler ideas that could contain a great deal of complexity within them.
When considering which relationship qualities lonely adults may need, survey respondents in our study desired to be cared for by others (Table 2). Being Cared For encompassed a group of physical actions that helped adults feel visible and known as their authentic selves, such as showing up in person or being engaged in conversation to be understood. Social support that is provided in personal relationships may be a similar idea that has been found in loneliness research to reduce loneliness levels (Zhang & Dong, 2022). While social support can have many definitions, perceived social support scales often measure emotional and functional help provided by friends, family, and partners, such as listening to problems or helping with chores (Zhang & Dong, 2022). These measures overlook exposure to the broader social environment, as adults in our study also described needing to be cared for in any social space, such as at work or in healthcare settings that won’t involve intimate relationships. Science that has studied and researched what “caring” means has been developed in nursing, that could serve as a guide for how to measure and provide care across our communities (Turkel et al., 2017).
Grieving Relational Loss, Rejection of Who I Truly Am, and Excessive Labor (Table 1) are three social interaction qualities we identified in our study that contributed to feeling lonely and have been documented in loneliness research. With regards to loss, loneliness has a long history of being studied in older populations, recognizing that death in significant relationships impacts feeling lonely (O'Rourke et al., 2018). Older adults were also found to lose relationships due to major health problems that can create communication barriers with others and lead to isolation in homes or care facilities (O'Rourke et al., 2018). Our study indentified these same life transitions can result in loneliness at any adult age, in addition to the loss of relationships from moving away. Moving is known as a significant stressful life event, especially for children, due to the impact it has on changes in relationships (McMahon et al., 2020). Loneliness studies on older populations have mentioned that adult children moving away can impact loneliness in parents (Schröders et al., 2021). Our study found moving away was important across adult ages, especially frequent moving described by military families or working adults finding new jobs.
The second relationship quality of Rejection of Who I Truly Am, was described by survey respondents as feeling criticized or disregarded. It resulted in loneliness because rejection pointed out something wrong or unlikeable about the adult. This highlights the importance of incorporating anonymous stories into loneliness research and the need to address stigma. Rejection has been discussed in loneliness research by Dr. John T. Cacioppo, the founder of social neuroscience, who conducted psychology experiments that intentionally rejected individuals who ended up feeling lonely (Cacioppo & Patrick, 2008). However, these experiments were in controlled environments. Rejection described in our study was complex, as it was often being informed by dominant cultural messages to reject an individual based on a social identity or life experience. Known stigmas against sexual and gender minorities or those of certain weights (Gillani et al., 2024; Talumaa et al., 2022), for example, showed up in our study, as well as unrecognized stigmas, such as the social exclusion of single adults. Narratives in our study also described more severe forms of rejection as abuse and bullying. Recent loneliness research has found correlations between perceived discrimination, violence, and childhood and domestic abuse with population loneliness (Landry et al., 2022; Refaeli & Achdut, 2024; Shepp et al., 2023).
The third relationship quality of Excessive Labor, was described by survey respondents as any kind of labor that exceeded their capacity, either at work or in caring for others at home. They felt lonely because they were solely responsible for the labor. Job burnout was found to be related to loneliness levels by Dr. Daniel W. Russel, who developed the UCLA Loneliness Scale as a validated psychometric measure (Russell, 1996). Public school teachers and hospital-based nurses were studied as a population to help develop the scale, and the high job burnout in their professions continues to be correlated with loneliness (Russell, 1996; Wood et al., 2023). Loneliness has also been recognized as a common experience among caregivers for children or adults with serious mental or physical health problems (Vasileiou et al., 2017). Burnout may not fully capture the experience as survey respondents in our study described being in crisis with suicidal thoughts when unable to meet labor demands, from an inability to pay bills as a single parent to working 24-hour shifts at a hospital.
Present But Emotionally Absent (Table 2) is the fourth relationship quality we identified in our study that contributed to feeling lonely but is scant in loneliness research. Survey respondents described emotional absence as lacking in fulfilling interaction, such as an individual sitting in silence while others talked, or a person crying among a crowd of people who walk by. Emotional neglect has been examined in few loneliness studies to describe a lack of “love, support, a sense of belonging, and care”, primarily as it relates to children (Musetti et al., 2021). A sense of belonging is more commonly measured in adult loneliness research, that captures an individual’s feeling of being valued, accepted, and fitting in a community (Mellinger et al., 2024). Fitting in was included in our study under the umbrella of Present But Emotionally Absent, but it should not be confused with rejection. Instead, survey respondents described a misalignment between their authentic self and the people around them as lacking common ground. This quality may require further exploration.
Beyond the Cognitive Discrepancy Model, our study identified approximate underlying causes to to the quality of social interactions adults experienced. For example, Breaking Expectations of Dominant Social Narratives seemed to be an underlying force contributing to poor qualities in social interactions. One survey respondent described stigmatized stories about homosexuality in their culture as a way to explain why they were rejected as a gay man in their social interactions with others. The Social Exposome Model describes this as the process of embodying the social environment (Gudi-Mindermann et al., 2023). Cultural norms as stories can take on a physical form when stored in our brains, acted out in our behavior with others, and can be enforced through our social interactions (Gudi-Mindermann et al., 2023).
However, the four underlying causes we identified seemed to have a complex relationship with each other, that could not be easily described through thematic analysis alone (Fryer, 2022). They may reflect what critical realist approaches in healthcare call an open system (Koopmans & Schiller, 2022). Systems science is used to describe a collection of interacting factors that ultimately contribute to a single outcome in a dynamic way (Meadows, 2008). Systems methods are used to identify key causes in complex health issues, that can have bidirectional relationships and multiple pathways to an outcome (de Pinho & Larsen, 2015; Diez Roux, 2011). Our study found the qualities of social interactions, dominant social narratives, significant life transitions, physical isolation, and social isolation, could interact with each other to ultimately contribute to loneliness experiences in adults.
Limitations & Strengths
Secondary data for this study is anonymous and 300 survey responses out of the 2000 submitted were selected by one individual to be publicly posted. Those who submitted survey responses may be different than lonely persons who chose not to submit a survey response. As our study aim is to understand causes of loneliness as a health-related issue, health researchers will be concerned about generalizability and selection bias to understand which populations these results may apply to. We can only confirm that these results apply to our study sample of adults; however, we incorporated theory in our analysis to understand a human need that will have shared characteristics across all adult humans. We found that after analyzing 83 loneliness survey responses from anonymous adults, there were no new ideas in the remaining 136 responses to develop causal explanations for loneliness, indicating our results might be broadly shared among adults.
Understanding of loneliness from our study sample may emphasize a Western perspective that would be improved by examining loneliness experiences in other languages than English and across cultures. However, survey responses in our study sample described a rich diversity in loneliness experiences that could be positive or negative, and range in severity. Some survey respondents also explicitly described being from other countries or cultures that are not Western. We intentionally included contradictions in our analysis and final results, that maintained this diversity, to allow for loneliness to be a normal part of any adult life.
Conclusions
In interventions to reduce loneliness at any adult age, there are at least 5 key relationship qualities to consider: Being Cared For, Grieving Relational Loss, Rejection of Who I Truly Am, Excessive Labor, and Present But Emotionally Absent. These qualities overall reflect a wide range of well understood social concerns, that it may be more useful to identify health interventions already addressing these issues, and reframe or readjust them as loneliness interventions. Our results support the paradigm shift of understanding the need for social connection as a broad need across all relationship types in a community. Meaning, loneliness interventions should not limit themselves to fostering new or personal relationships, but should consider existing social structures at work, in healthcare, in religion, in neighborhoods, etc., in which poor quality social connection can be embedded. Our future research will improve upon these results to understand the system of underlying causes contributing to poor quality connection in social environments.
Footnotes
Ethical Considerations
The Colorado Multiple Institutional Review Board approved the study as exempt from human subject research (#23-2083).
Author Contributions
Conceptualization – Original Design, KLB, RAJK; Conceptualization – Review & Editing, GT, JAL, MPF, BMM; Methodology, KLB, SE, RAJK; Formal Analysis, KLB, SE, RAJK; Resources, All Authors; Writing – Original Draft Preparation, KLB; Writing – Review & Editing, All Authors; Supervision, RAJK.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This publication is supported in part by the NIH/NCATS Colorado CTSA Grant Number UL1 TR001082.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
Disclaimer
Content are the authors’ sole responsibility and do not necessarily represent official NIH views.
