Network analysis identifies key factors for screening at-risk adolescents in LOOK@ME project

A large study of adolescents in northern Italy has identified a compact group of psychological warning signs that may help schools recognize young people at risk of serious emotional and behavioral difficulties. The analysis, conducted as part of the LOOK@ME project, suggests that screening programs may not need to ask students dozens of overlapping questions to detect vulnerability. Instead, a smaller set of highly connected features—including emotional problems, conduct difficulties, hyperactivity and inattention, smartphone craving, sleep disruption, weak confidence in emotion-regulation skills and poor impulse control during negative moods—captured much of the information contained in a far broader assessment battery.

The findings come from 702 middle- and high-school students with an average age of 13.44 years; 54 percent were female. Researchers used questionnaires measuring three interconnected areas of adolescent well-being: problematic smartphone use, psychological symptoms and difficulties regulating emotions. The students were drawn randomly from several pre-test phases of LOOK@ME, a school-based research and intervention program in northern Italy. The project first screens students and then offers focus groups tailored to those considered potentially at risk. Its aim is not to label young people, but to identify who might benefit from closer psychological support before difficulties become more severe.

The study reflects a growing concern about how smartphones intersect with adolescent development. Adolescence is already a period of rapid biological, psychological and social change. Brain systems involved in reward, motivation and emotional reactivity develop alongside, but not always at the same pace as, the prefrontal networks responsible for planning, inhibition and self-control. This developmental mismatch can make intense emotions harder to manage, particularly in complicated social environments. Smartphones add another layer to that landscape, providing continuous access to social approval, entertainment, information and rapid rewards. For some teenagers, repeated checking can become difficult to control and may interfere with sleep, relationships, schoolwork and everyday responsibilities.

The researchers focused on problematic smartphone use, or PSU, rather than ordinary or frequent smartphone ownership. PSU describes excessive and difficult-to-control use that produces negative consequences, such as sleep disturbance or interference with social life. One important component is craving: a strong urge to use the device even when doing so creates problems. Smartphone use can also function as an external, though potentially maladaptive, method of regulating emotion. A teenager who feels lonely, bored, anxious or angry may turn to the phone for immediate distraction or reassurance. That short-term relief can reinforce the behavior, making the device increasingly central to coping while leaving the underlying emotion unresolved. The researchers stress, however, that PSU is not currently recognized as a formal diagnosis in the DSM-5 or ICD-11.

To examine how these problems fit together, the team used network analysis. Traditional psychological studies often summarize relationships through broad scores or correlations between total questionnaire results. Network analysis instead treats individual symptoms or traits as nodes in a system and estimates how strongly they are connected to one another. A symptom with many strong links may be especially central to the network, while a bridge variable may connect otherwise separate clusters—for example, linking smartphone-related behavior with emotional dysregulation or behavioral problems. This approach does not prove that one feature causes another. It is a way of identifying the variables most embedded in the observed pattern, which may be particularly useful when researchers are designing efficient screening tools.

The questionnaires included subscales from the Smartphone Addiction Inventory, the Strengths and Difficulties Questionnaire and the short form of the Difficulties in Emotion Regulation Scale. The network revealed several prominent areas. Emotional problems stood out alongside conduct issues, hyperactivity and inattention. Smartphone-related craving and sleep problems were also among the most relevant variables. On the emotion-regulation side, adolescents’ reduced trust in their ability to regulate emotions adaptively and their difficulties controlling impulses during negative emotional states were especially important. Together, the findings depict risk as a connected system rather than a single condition: disturbed sleep may coexist with compulsive use, emotional distress may intensify reliance on a phone, and impulsivity may make it harder to interrupt the cycle.

The team then asked whether the entire questionnaire battery was necessary for classifying students into broad risk profiles. Using the full set of 16 subscales, researchers compared students with a “normative” profile, an at-risk profile or a clinical profile. These categories were created from how far students’ scores deviated from normative averages across multiple domains. Students were considered normative when they had relatively few elevated scores, at risk when several measures were more than one standard deviation above the norm, and clinical when at least two measures exceeded two standard deviations. The categories were designed for screening and research purposes, not for making a psychiatric diagnosis.

The reduced screening set retained the dimensions that repeatedly emerged as central or well-connected in the network, followed by a qualitative review of their clinical relevance. Researchers compared its classification performance with that of the full set using ordinal logistic regression and stratified 10-fold cross-validation. In this procedure, the data are divided into 10 groups; each group is tested after a model has been trained on the other nine, allowing every participant to contribute to an out-of-sample evaluation. The reduced model showed performance close to that of the full battery in distinguishing normative, at-risk and clinical profiles. That result suggests that a carefully selected group of questions could potentially reduce respondent fatigue without sacrificing much of the screening system’s ability to identify elevated risk.

The implications are significant for schools, where time, staffing and students’ willingness to complete lengthy assessments can all limit early-intervention programs. A shorter screening process could make it easier to examine entire school populations and reserve more detailed evaluations for students whose responses indicate concern. Yet the researchers caution that the network reflects associations in one adolescent sample and cannot establish causal pathways. The study also used classifications based on the number and magnitude of elevated questionnaire scores rather than independently validated diagnostic thresholds. The next step will be determining whether these key variables predict later difficulties and whether interventions targeting sleep, emotional coping, impulse control or compulsive smartphone use can alter the broader network. For now, the study offers a practical message: adolescent risk may be easier to detect when researchers look not only at how severe each problem is, but also at how the problems connect.