From habit to high risk: The influence of multidimensional lifestyle changes on internet addiction risk among junior high school students and its predictive utility.

The detected rate of Internet addiction risk increased among junior high school students over a six-month follow-up period.

• Internet addiction risk prevalence rose from 10.62% at baseline (T1, September 2023) to 13.35% at follow-up (T2, April 2024).
• Of 10,535 participants enrolled at baseline, 9,750 were successfully followed up, yielding a retention rate of 92.55%.
• Among those followed up, 7,853 provided complete and valid questionnaire data, corresponding to a T2 effective data rate of 80.54%.
• The study used a longitudinal cohort design with approximately six months between assessments.

Delayed sleep onset or going to bed after 10:00 pm was the strongest behavioral risk factor for Internet addiction.

• Going to bed after 10:00 pm or having delayed sleep onset timing was associated with OR = 2.859 (95% CI: 2.319–3.525).
• This was the highest odds ratio among all identified behavioral risk factors in the multivariate analysis.
• This finding was identified through multivariate logistic regression analysis.

Developing a late-night eating habit was a significant risk factor for Internet addiction.

• Developing a late-night eating habit was associated with OR = 1.932 (95% CI: 1.494–2.499).
• This dietary behavior change was identified as a significant predictor in multivariate analysis.
• Notably, this variable was excluded from the final six-variable predictive ROC model despite being a significant risk factor.

Becoming a smart device owner was associated with increased risk of Internet addiction.

• Becoming a smart device owner was associated with OR = 1.773 (95% CI: 1.307–2.405).
• Smart device ownership was one of four lifestyle dimensions analyzed in the study.
• This finding was identified through multivariate logistic regression analysis.

Being a non-habitual napper was associated with increased risk of Internet addiction.

• Being a non-habitual napper was associated with OR = 1.699 (95% CI: 1.408–2.049).
• This sleep-wake pattern variable was part of the sleep dimension analyzed alongside bedtime timing.
• This factor was included in the six-variable predictive model.

Decreased pursuit of dietary balance was associated with increased risk of Internet addiction.

• Decreased pursuit of dietary balance was associated with OR = 1.654 (95% CI: 1.300–2.104).
• This dietary change was part of the diet dimension, one of four lifestyle dimensions studied.
• This variable was included in the final six-variable predictive model.

Decreases in active exercise and exercise duration were associated with increased risk of Internet addiction.

• Decreased active exercise was associated with OR = 1.575 (95% CI: 1.222–2.031).
• Decrease in exercise duration per session was associated with OR = 1.436 (95% CI: 1.117–1.846).
• Both exercise-related variables were included in the final six-variable predictive model.
• Exercise was one of four lifestyle dimensions analyzed in the study.

A six-variable predictive model for Internet addiction risk demonstrated acceptable discriminative performance.

• The model incorporated six key variables, excluding change in habitual late-night eating.
• The area under the ROC curve (AUC) was 0.721 (95% CI: 0.701–0.741).
• The authors described this performance as 'acceptable.'
• Predictive performance was evaluated using Receiver Operating Characteristic (ROC) curves.