Three digital addiction profiles emerge among Turkish middle school students

Researchers have identified three distinct profiles of digital addiction symptoms, attitudes toward physical education, and psychological well-being among middle school students in Turkey, in a study that analyzed data from 736 pupils in grades five through eight. The work, published in BMC Public Health, used a person-centered statistical approach known as latent profile analysis to sort students into groups based on how their scores lined up across three questionnaires rather than treating them as a single homogeneous population.

The study was led by Şinasi Özsaydı, with Özlem Sinem Uslu and Banu Kabak as co-authors, and drew its sample from state middle schools in the provinces of Antalya and Osmaniye. Female students made up 50.7 percent of the sample. Each participant completed the Digital Addiction Scale for Children, the Physical Education Predisposition Scale, and the Stirling Children’s Well-being Scale, and the resulting scores were analyzed in R using latent profile analysis, with between-profile differences examined through analysis of variance, post-hoc tests, and multinomial logistic regression.

The analysis retained a three-profile solution under the EEE covariance specification, with an entropy value of 0.81, a measure indicating reasonably clear separation between the groups. The largest group, labeled Moderate Symptom Severity, contained 597 students, or 81.1 percent of the sample, and was characterized by moderate digital addiction symptom severity, favorable attitudes toward physical education, and positive well-being.

The second profile, labeled Lower Well-Being, included just 39 students, or 5.3 percent of the sample. These students had a numerically higher mean digital addiction symptom score than the first profile, 45.0 against 41.2, and recorded the lowest well-being score in the entire sample, at 22.4. The third profile, High Symptom Severity, comprised 100 students, or 13.6 percent, and showed the highest symptom severity in the sample, with a mean score of 72.5, while its mean well-being score of 45.4 was numerically similar to that of the largest profile, which averaged 46.3.

One of the more striking technical findings concerns the effect sizes for the questionnaire measures. The Digital Addiction Scale showed the largest overall effect size across profiles, with an eta-squared of 0.544, and the Stirling Children’s Well-being Scale showed the second largest, at 0.317. The authors are careful to note, however, that because these indicators were themselves used to generate the profiles, such values describe the retained solution rather than validate it, a distinction that matters when interpreting how strongly the variables appear to differ between groups.

After accounting for classification uncertainty, female gender remained associated with membership in the High Symptom Severity profile, with an odds ratio of 1.68 and a 95 percent confidence interval of 1.04 to 2.73. By contrast, the grade-level association with the Lower Well-Being profile did not remain significant once uncertainty was taken into account, suggesting that the apparent link between a student’s grade and that profile was not robust in this sample.

The researchers also ran sensitivity analyses to test how stable their solution was. Rescaling the indicators left the EEE solution essentially unchanged, with an adjusted Rand index of 0.945, but removing three multivariate outliers reduced the smallest profile below 5 percent of the sample. The retained three-profile partition also depended on the covariance specification and the number of classes specified, meaning that different analytical choices could have produced a different grouping of the same students.

Those caveats shape how the authors frame their conclusions. They write that the retained three-profile solution should be interpreted as one plausible, specification-dependent representation of heterogeneity in this two-province sample, not as evidence of fixed student types, and they note that the smallest profile was the least stable part of the solution. In other words, the groups describe patterns in this particular dataset rather than permanent categories of students waiting to be discovered.

The study is also explicit about what its findings cannot yet support. Whether the observed configurations have implications for psychological-risk assessment requires replication using independent outcomes and more representative samples, according to the authors. The sample came from only two Turkish provinces, and the article was published as an open-access research paper, with the version shared early as a peer-reviewed, accepted manuscript that is citable and carries a permanent DOI but remains subject to further edits before the final version of record replaces it.

By modeling students as members of overlapping profiles instead of averaging across them, the analysis shows that high digital addiction symptom severity and low well-being did not always travel together in this sample: the profile with the most severe symptoms had a mean well-being score close to that of the largest, most favorable group, while the smallest profile combined only moderately elevated symptoms with the lowest well-being of all. Untangling which combinations matter, and whether they predict later outcomes, is the question the authors say future replication studies must answer.

Subject of Research: Latent profiles of digital addiction symptoms, physical education predisposition, and psychological well-being in middle school students

Article Title: An exploratory latent profile analysis of digital addiction symptoms, physical education predisposition, and psychological well-being in middle school students

Article References: Özsaydı, Ş., Uslu, Ö. S., & Kabak, B. (2026). An exploratory latent profile analysis of digital addiction symptoms, physical education predisposition, and psychological well-being in middle school students. BMC Public Health. https://doi.org/10.1186/s12889-026-30009-4

Image Credits: AI Generated

DOI: 10.1186/s12889-026-30009-4

Keywords: digital addiction, middle school students, latent profile analysis, psychological well-being, physical education, adolescent health, Antalya, Osmaniye, screen use, well-being, gender differences, BMC Public Health