This is a cross-sectional study with a sample of 5,400 employees residing in Spain, surveyed through the CAWI (Computer-Assisted Web Interviewing) method during August and September 2024. Respondents were selected from a panel managed by IPSOS, which includes 57,093 individuals. The IPSOS panel recruitment and quality process focuses on ensuring accurate population representation rather than simply increasing panel size. Ipsos recruits 7,000 to 10,000 new members each month through vetted vendors and direct partnerships, ensuring panel stability. Strict quality controls are applied at every stage, from panelist registration to survey completion, including multi-factor authentication, bot detection, and fraud prevention measures. For the study sample, we established categories for gender, age, autonomous community, and municipality size to ensure that the sample accurately reflects the distribution of employees in Spain, according to data from the National Statistics Institute [8].

The two-item Patient Health Questionnaire (PHQ-2) [23] and the two-item Generalized Anxiety Disorders (GAD-2) [24] scales were employed to measure depression and anxiety, respectively, with both scales referring to the past two weeks. Each scale consists of two items: the PHQ-2 includes “Little interest or pleasure in doing things” and “Feeling down, depressed, or hopeless,” while the GAD-2 comprises “Feeling nervous, anxious, or on edge” and “Not being able to stop or control worrying.” Participants respond using the following categories: “Not at all,” “Several days,” “More than half the days,” and “Nearly every day.” The results from both scales range from 0 to 6, with scores of 3 or higher considered compatible with diagnoses of major depressive disorder (MDD) or generalized anxiety disorder (GAD), respectively [23–24].

The CAGE Questionnaire Adapted to Include Drugs (CAGE-AID) [25] was used to assess substance use disorder during the previous month. It consists of four questions with ‘yes’ or ‘no’ answers: (1) “Have you ever felt you ought to cut down on your drinking or drug use?”; (2) “Have people annoyed you by criticizing your drinking or drug use?”; (3) “Have you ever felt bad or guilty about your drinking or drug use?”; and (4) “Have you ever had a drink or used drugs first thing in the morning to steady your nerves or get rid of a hangover?” Answering ‘yes’ to two or more questions determines substance use disorder [25].

Loneliness was evaluated through a direct question, “How frequently do you experience loneliness?” as well as through the three-item University of California, Los Angeles (UCLA) Loneliness Scale, which is recognized for its satisfactory reliability and demonstrated concurrent and discriminant validity [12]. This scale includes the following inquiries: “How frequently do you experience a lack of companionship?“; “How often do you feel left out?“; and “How frequently do you sense isolation from others?” Participants rated each question on a 3-point scale (1 = rarely; 2 = sometimes; 3 = frequently). The cumulative score can range from 3 to 9, with higher scores indicating greater feelings of loneliness. In this study, loneliness was defined as reporting “sometimes” or “frequently” feeling lonely according to the direct question or achieving a cutoff score of ≥ 6 [13] on the UCLA loneliness scale [12]. In both cases, loneliness was assessed, separately, in workplace and non-workplace contexts. Primary analyses were conducted using the direct question. In the supplementary material, analyses were replicated using the cutoff score of ≥ 6 from the UCLA scale.

Sociodemographic information included gender (male, female, and other), age groups (18–29 years, 30–39 years, 40–49 years, 50–59 years, and 60–65 years), marital status (married, including civil partnerships; never married; separated, including divorced and widowed), nationality (Spanish or non-Spanish), sexual orientation (heterosexual or non-heterosexual), and occupation level (directors, including managers; professionals, including scientists and intellectuals; other non-manual roles, comprising technicians, middle-level professionals, and administrative staff; skilled manual workers, including commerce workers, tradespeople, artisans, and qualified workers; unskilled manual workers, including machine operators and elementary occupations; and military personnel) [26].

Working conditions-related variables included manager status (yes, no), which indicates whether the individual has employees under their supervision; length of service (less than 1 year, 1 to 3 years, more than 3 years); teleworking days (none, less than half, half or more); working under pressure (no, more or less, yes); sick leave days in the last year (5 or fewer days / more than 5 days); and frequent communication with colleagues (yes, more or less, no), which aims to reflect social interactions in the workplace, assessing to what extent work is carried out in isolation or through collaboration.

The Employment Precariousness Scale for Europe (EPRES-E) was utilized to assess labor precariousness. This scale comprises six domains: temporariness (contract duration), disempowerment (the level of negotiation regarding employment conditions), vulnerability (defenselessness against workplace authoritarianism), salary, workplace rights (reconciliation with daily needs), and unpredictability (the predictability of work schedules). Each dimension was measured using ordinal scales, which were recoded into subscales ranging from 0 to 100, where higher values indicate more precarious situations. The total score was derived as a simple average of the six subscales, resulting in a final score ranging from 0 to 100 [27–28].

Descriptive analyses were reported overall and for subpopulations with workplace loneliness and non-workplace loneliness. Proportions and frequencies for categorical variables and means and standard deviation of continuous variables were reported. Differences between population with and without workplace loneliness as well as with and without non-workplace loneliness were measured through Chi-squared test for categorical variables and t-test for continuous variables. Additionally, a Kappa statistic was calculated to assess the level of agreement between workplace loneliness and non-workplace loneliness.

Two separate adjusted logistic regression models were developed to investigate sociodemographic risk factors for both workplace and non-workplace loneliness. Both models incorporated all sociodemographic variables simultaneously (gender, age groups, nationality, marital status, sexual orientation and occupation level). Odds ratio with 95% confidence interval were reported.

Similarly, two separate adjusted logistic regression models were constructed to investigate working conditions-related risk factors for both workplace and non-workplace loneliness. Both models incorporated all working condition-related variables (manager/supervisor, length of service, teleworking days, working under pressure, frequent communication, and labor precariousness) as well gender and age groups.

Unlike categorical variables, the interpretation of odds ratios for continuous variables is more complex. Since precariousness was the only continuous variable, to clarify the association between precariousness and workplace and non-workplace loneliness, estimated probabilities with 95% confidence interval for both contexts of loneliness were calculated. These probabilities were determined using the “margins” command in Stata [29], based on the adjusted logistic regression models for workplace loneliness and non-workplace loneliness. Control variables were centered to their means, considering the sample proportions.

To evaluate the impact of workplace and non-workplace loneliness on absenteeism (more than five days off in the last year), depression, anxiety and substance use disorder, additional logistic regression models were constructed. Firstly, we included as independent variable workplace and non-workplace loneliness separately, and, finally, we included both variables simultaneously. In all cases, gender and age groups were included as covariates. Odds ratio with 95% confidence interval were reported.

While these preliminary analyses were conducted using loneliness defined by a direct question as the dependent variable, sensitivity analyses were also performed replicating all analyses using the definition of loneliness based on a cutoff of ≥ 6 on the three-item UCLA loneliness scale. All reported p-values were based on a two-sided test, where the level of statistical significance was set at p < 0.05. Stata version 13 [30] was used to analyze the survey data.