A Short-Form Measure of Loneliness to Predict Depression Symptoms Among Adolescents

Depression is a common and often devastating psychological disorder that is one of the top five leading causes of disability and disease burden worldwide [1, 2]. Early adolescence is a critical developmental period for the emergence of depression [3], with research suggesting that overall levels of depressive symptoms increase markedly over the course of this period [4, 5]. According to aggregated published data, the probability of depressive disorders increases from 5% in early adolescence to 20% by the end of this age period [6]. Moreover, onset of depression in adolescence confers a specific high risk for chronic recurrence and poor functioning throughout the lifespan [7, 8], including poor psychosocial functioning [9], cognitive impairment [10], and even suicidality [11].

Depression typically leads individuals to become isolated from family and friends and although loneliness has often been identified as an important risk and predictive factor of adolescent depression [12], the impact of loneliness on the health and development of young people is only now being understood [13]. Predicted to reach epidemic proportions by 2030 [14], loneliness is defined as a distressing emotional state people experience when they notice a discrepancy between the desired and perceived quality or quantity of their social relations [15, 16]. The research is unequivocal that loneliness is linked with a constellation of mental health problems (e.g., depression, anxiety: [17‐24]) and has negative impacts on physical wellbeing (e.g., somatic complaints, sleep problems: [25, 26]). A trajectory of increasing loneliness during adolescence predicts future depression, self-harming, and suicide ideation [27, 28].

Recent research [29] has examined risk factors for suicidal behaviours using a network modelling approach (i.e., where mental health problems are viewed as a system where symptoms form interacting networks among themselves and give rise to each other as epiphenomena as the diseases progress) [30]. In this research, items from the Children’s Depression Inventory (CDI) [31] completed by a community sample of approximately 1400 13–19 year olds, identified loneliness as a key central node in the network of symptoms. In a replication study with 5888 12–16 year olds [32], loneliness and sadness were the most central symptoms in a network of adolescent depression symptoms, with loneliness explaining the most variance of suicide ideation.

There is strong evidence that the developmental course of loneliness is heterogeneous [33], and is best represented by a multidimensional model, varying in intensity and across causes and circumstances, and where different social relationships give rise to different forms of loneliness [16, 34‐37]. It is also known that loneliness can fluctuate in daily life, depending on, among others, the context people are in [38, 39]. Only one study seems to have examined the temporal dynamics of adolescent loneliness, however [40]. Although the authors reported not using “stringent statistical tests” (p. 926) the data gathered revealed that levels of adolescent’s loneliness varied over time according to the contexts, situations, and company they were in at that time.

Like loneliness, depressive symptoms can also vary [41] and fluctuations in affective states can signify issues with emotional regulation and exhibit proximal relationships with short-term spikes in suicidal ideation and self-harm [42]. Identifying adolescents at-risk of experiencing psychological distress is inhibited by such fluctuations in affective states and associated risk/protective factors over short-term periods. Although temporal stability may be evident over longer periods there are periods of imbalance over short-term periods that signify risk for adverse outcomes (e.g., psychological distress, substance use, and self-injury). Long time periods between assessments may therefore fail to accurately identify symptoms and obscure associations with predictors.

Aided by advances in mobile technology recent research has focused on the intensive assessment of psychological risk and protective factors to enhance understanding and detection of adverse states. For instance, clinical patients were assessed several times daily using mobile ecological momentary assessment (EMA) software, and nearly all were found to have dramatic changes in suicidal ideation and loneliness over a single day [43]. Compared with healthy controls, depressed individuals tend to report greater instability and reactivity to positive events in prior research [44]. Negative fluctuations in affective states over short-term periods (i.e., hours, days) have been linked to an increased propensity to desire suicide and engage in self-injurious behaviours [45, 46]. A recent meta-analysis [47] examined the associations between daily interactions and negative affect, finding perceptions of social interactions to be associated with within-person fluctuations in positive and negative affect. Other research has found adverse interpersonal states, such as thwarted belongingness (i.e., low perceived support, isolation), to be associated with suicidal ideation [48].

Given the strong associations between loneliness, broader interpersonal states and aspects of mood, clinicians and researchers would benefit from a brief, multi-dimensional measure of loneliness with predictive capabilities. This is not to say that current measures to evaluate loneliness do not already exist. A recent review [49] identified the most commonly used instruments as the UCLA Loneliness Scale (UCLA) [50] and its shortened 11‐item [51], 8‐item [52, 53], 6‐item [54, 55] and 3‐item versions [56]. Theoretically, while these successive versions have conceptualized loneliness as a single dimension, the factor structure has varied between one and three factor solutions. Other measures identified in the review included the Children’s Loneliness and Social Dissatisfaction Scale [57] and the Loneliness and Aloneness Scale for Children and Adolescents ([58]. However, all these measures were developed some time ago [49] and “each omitted key processes, including interviews with children and adolescents” [49 p. 11] suggesting that their views of the loneliness experience did not inform the measures.

A promising measure of adolescent loneliness that did include the voices of adolescents during its development [59] is the Perth A-loneness scale (PALs, 37). The PALs comprises four distinct factors: friendship-related loneliness, isolation, negative attitude to being alone, and positive attitude to being alone. A series of studies [37, 60, 61] has consistently reported satisfactory fit statistics of a four-factor model and internal reliability. Test–retest reliability (9 months apart) also supports a degree of temporal stability (rs = 0.59–0.67) and a recent Rasch analysis of the PALs [62] supported the interval scale measurement properties of the PALs. Of the PALs four factors, friendship-related loneliness (i.e., having reliable, trustworthy supportive friends e.g., “I can turn to my friends for help when I need it”) and isolation (i.e., having few friends or believing that there was no one around offering support e.g., “I feel like I do not have a friend in the world”) have exhibited particularly strong cross-sectional associations with positive mental wellbeing [60] and depression symptoms [61].

Therefore, the current study explores the effectiveness of a multidimensional measure of loneliness in predicting current and future symptoms of depression. Developing an abbreviated version of the PALs scale, which shows strong connections to depressive symptoms, is beneficial to both comprehensive long-term longitudinal (i.e., monthly, yearly) and intensive research (e.g., daily diary, ecological momentary assessment), and minimises unnecessary burden on participants.

Given the known strong associations between loneliness and depressive symptoms and that both constructs can fluctuate, it is important to develop a brief measure of loneliness capable of predicting symptoms of depression in a way comparable to a more comprehensive, well-validated measure. A brief measure of loneliness would also reduce the burden on young people when repeated measurements over time are necessary. Future advances in this area of research are likely to be dependent, in part, on the development of such a brief loneliness measure. A short form can facilitate both intensive research work (such as that using ecological momentary assessment) and more traditional longitudinal work stretching across months or years.

The outcome of the present research, the reduced PALs—6, is a multi-dimensional measure of loneliness which halves participant load to 6 items as compared to the 12 items of the two subscales of the PALs. PALs—6 preserves cross-sectional relationships with extreme depression, though both the PALs and PALS—6 were less successful in longitudinal prediction. The reduced 6-item measure (PALs—6) preserved the two-factor nature of the original two PALs subscales when capturing friendship related loneliness and isolation. Thus, the new brief scale captures both quality of friendships and aspects of isolation to define adolescents’ experiences of loneliness.

Adolescence is known to be a sensitive developmental phase, a time when insufficient connections to others can lead to profound and lasting negative consequences on physical and mental health, even leading to increased mortality [68]. Conversely, quality friendships provide numerous social and emotional benefits [69]. This applies to young people worldwide. Thus, the abbreviated PALs—6 scale is an important contribution to research aiming to assess an important risk factor for current psychopathology and for use in intensive, short-term research.

Cross-sectional prediction of very elevated depression symptoms in the current research was satisfactory, showing a good balance between sensitivity and specificity. On the other hand, longitudinal prediction was associated with higher misclassification rates. This may in part be due to the limitations of classification over long-term periods, particularly with both dependent and independent variables being prone to change over time. For instance, interpersonal factors and depression symptoms are known to fluctuate over a matter of hours or days, and therefore isolated measurement may not capture “typical” patterns in symptoms and risk factors [42]. Interpreting positive predictive value is largely dependent on the outcome in question (e.g., prediction of suicide completion is typically associated with exceptionally high false positive rates of approximately 99%; [70]). Further, the consequences from false negatives (i.e., failing to detect someone with depression) may outweigh the cost of false positives (i.e., incorrectly targeting resources to adolescents not at short-term risk of depression). Inferences regarding an acceptable PPV can be made through comparisons to similar studies predicting depression symptoms. For example, Seeley et al. [71] assessed the predictive quality of several interpersonal related composite scores, including low peer support (PPV = 0.17), low parental support (PPV = 0.22), and poor school functioning (PPV = 0.24). Another community-based study of older adults [72] found a brief measure of loneliness predicted moderate to severe depression symptoms cross-sectionally with a PPV of 0.26. The PPV of the PALs—6 scale is therefore high relative to the few studies which have provided PPV statistics of individual scales when predicting depression symptoms. Improving PPV may be enhanced through repeated administration of the current scale to identify adolescents with persistent loneliness and by assessing a broader range of correlates for depression.

There are limitations to the current study. First, data were self-report and based on reflection. Young people frequently have difficulty reporting their internal states to other sources such as parents and teachers, and parents and teachers have difficulty perceiving the internal world of their children [73]. Therefore, self-report measures may elicit valid responses from young people. Second, predicting depressive symptoms over short-term periods was beyond the scope of the current study, and should be assessed to determine optimal lags between assessments when predicting future symptoms using the PALs—6. Third, there was a notable proportion of missing data over time, although there was no systematic pattern of missing data detected. That is, adolescents with better or worse mental health were not significantly more likely to have missing data. Fourth, there was a high proportion of adolescents who met the criteria for “very elevated” depression symptoms. Approximately 3% of adolescents should be two standard deviations above the normative average [74]. However, some CDI 2: SR[S] items were endorsed a considerably high amount, and may have inflated the number of adolescents in the very elevated range. For instance, some 25% of adolescents endorsed the items “I have to push myself all the time to do my schoolwork” and “I am tired all the time”. With the original CDI-2 dating back to 1992, there may be important changes in pressures faced by adolescents.

Although we are only now beginning to develop an understanding about adolescent loneliness, the research evidence is unequivocal that it is an important risk factor for current psychopathology. Loneliness is predicted to reach epidemic proportions within the next 10 years and is rightly viewed as a global public health disorder [14]. Following a U-shaped curve over the lifespan, loneliness peaks among adolescents and older adults [75]. For some adolescents, loneliness can create a sense of “paralyzing hopelessness and unutterable futility” [76, p. 7] and can lead to catastrophic outcomes. The research evidence is unequivocal that loneliness is an important risk factor for current psychopathology. Thus, research aimed at using loneliness to predict adverse events in longitudinal research, aided by brief measures that preserve the qualities of a more comprehensive measure, is critical.

Although half the length of the PALs, the PALs—6 was almost identical in its ability to predict adolescents with very elevated levels of depression both cross-sectionally and longitudinally. Given that loneliness can fluctuate over time and has strong associations with adverse mental health, the PALs—6 provides clinicians and researchers with a suitable brief measure for identifying young people at risk.