Identification of biological rhythm disorder profiles and associated problematic digital behaviors among university students: a latent profile analysis

Abstract

Objective

This study aimed to identify latent profiles of Biological Rhythm Disorders (BRDs) among university students and to examine their associations with Problematic Digital Behaviors (PDBs).

Methods

A total of 2,660 students from four universities in Hunan Province were surveyed. Biological rhythm disturbances were assessed using the Questionnaire of BRDs for Adolescents, while PDBs were evaluated using the Problematic Short-Form Video Media Use Scale, Problematic Social Network Use Tendency Scale, and Internet Game Addiction Scale. Latent profile analysis (LPA) was conducted with Mplus 8.3 to classify participants into distinct biological rhythm profiles. One-way ANOVA and multinomial logistic regression in SPSS 27.0 were used to examine demographic and behavioral predictors of profile membership.

Results

LPA identified four profiles of BRDs: Stable Rhythm (n = 774, 29.1%), Mild Disturbance (n = 956, 35.9%), Overall Disturbance (n = 500, 18.8%), and Sleep-Activity Disturbance (n = 430, 16.2%). Multinomial logistic regression showed that higher levels of Internet Gaming Addiction were associated with a greater likelihood of belonging to the Overall Disturbance profile (OR = 1.09, 95% CI: 1.07–1.12), whereas Problematic Short-Form Video Use (OR = 1.10, 95% CI: 1.08–1.12) and Problematic Social Network Use (OR = 1.08, 95% CI: 1.05–1.11) was notably associated with the Sleep-Activity Disturbance profile.

Conclusions

BRDs among university students exhibit notable heterogeneity. The influence of PDBs differs across rhythm disturbance profiles. Interventions promoting biological rhythm health should consider students’ sex, academic year, and digital behavior patterns to develop personalized preventive strategies.

Introduction

The university stage represents a critical transitional period from late adolescence to early adulthood, marked by accelerated physiological maturation and increasing psychological independence [1]. During this phase, individuals gradually develop autonomous lifestyles and behavioral patterns. Biological rhythms, also known as the “biological clock,” describe cyclical physiological, psychological, and behavioral changes that occur within a 24-hour period. These rhythms—including sleep, eating, and activity patterns—are regulated by internal “master clocks” (e.g., the suprachiasmatic nucleus) in coordination with external cues such as light exposure, diet, and physical activity [2–4].

The Biological Rhythm Disorders (BRDs) plays a vital role in maintaining both mental and physical health among university students. Empirical evidence shows that rhythm disturbances can disrupt daily routines and academic performance. Prolonged irregularity may further increase the risk of emotional disorders (e.g., depression, anxiety) and maladaptive behaviors (e.g., self-harm, suicidal tendencies) [5, 6]. These findings highlight the need for systematic investigation into the etiology and heterogeneity of BRDs among university students.

Existing research on biological rhythms has largely focused on clinical manifestations or single-dimension disturbances [7, 8]. However, studies identifying latent subtypes of BRDs and their influencing factors among university students remain scarce. Internet-related problematic digital behaviors refers to persistent and uncontrollable engagement in online activities, resulting in significant impairment in social, psychological, and physical functioning [9]. Among university students, common manifestations include Problematic Short-Form Video Use, Problematic Social Network Use, and Internet Gaming Addiction [10, 11]. Problematic Short-Form Video Use is characterized by compulsive viewing and emotional reliance on video content. Problematic Social Network Use reflects excessive online interaction and difficulty in self-regulation. Internet Gaming Addiction involves prolonged immersion in online or offline games, often accompanied by withdrawal symptoms and impaired real-life functioning [12–14].

Previous studies have shown that Internet-related problematic digital behaviors can aggravate biological rhythm disturbances. Specifically, Problematic Short-Form Video Use often delays sleep onset and diminishes sleep quality, thereby disrupting daytime activity rhythms. Problematic Social Network Use not only impairs sleep patterns but also interferes with eating rhythms and daily routines due to excessive engagement and constant feedback stimulation. Likewise, video game addiction affects multiple rhythm dimensions, resulting in irregular sleep schedules, delayed meal times, and reduced physical activity [15, 16]. However, existing research has primarily examined single types of addiction or isolated rhythm dimensions. This fragmented approach limits our understanding of the complex characteristics of BRDs among university students, as well as the combined effects of different forms of internet-related problematic digital behaviors on multiple rhythm domains. To bridge this gap, the present study explores the multidimensional impact of Internet-related problematic digital behaviors on biological rhythm disturbances and examines their potential structural interrelationships, thereby providing theoretical guidance for targeted health interventions.Therefore, the present study focuses on Chinese university students, aiming to identify potential subtypes of BRDs and examine their associations with Internet-related problematic digital behaviors.

It is important to note that most previous studies on rhythm disturbances have adopted a variable-centered approach, categorizing individuals into high-, medium-, or low-level groups based on their scores before analyzing correlations with associated factors. While such analyses offer valuable insights, they fail to capture the latent heterogeneity that may exist within populations. Latent Profile Analysis (LPA), a person-centered approach, enables the identification of subgroups characterized by similar response patterns across multiple rhythm indicators. This method enhances between-group differentiation while minimizing within-group variation, thereby providing a more nuanced understanding of multidimensional biological rhythm disturbances. Accordingly, the present study applies LPA to classify subtypes of BRDs among university students, investigate their demographic characteristics, and further examine how Internet-related problematic digital behaviors influence distinct rhythm subtypes.

Participants and methods

Participants

From April to May 2025, a stratified cluster random sampling method was employed. Four universities in Hunan Province were selected as survey sites. Within each university, students were stratified by grade, and two classes from the liberal arts and two from the sciences were randomly selected. All students in the chosen classes were invited to complete an electronic questionnaire. A total of 2,791 students from 54 classes participated. After excluding invalid responses, 2,660 valid questionnaires remained, yielding a response rate of 95.3%. Invalid responses were defined according to clear operational criteria: surveys with more than 20% missing items, responses showing straight-lining or patterned answering in over 80% of items, surveys completed in less than half of the median completion time, and responses with logical inconsistencies across related items. The final sample included 2,660 students, of whom 58.1% were female, with a mean age of 20.02 years (SD = 1.77).

Ethical statement

Ethical approval for this study was obtained from the Biomedical Ethics Committee of Jishou University (Approval No. JSDX-2023-0011). The data were collected exclusively for academic research purposes. Prior to data collection, the research team contacted the administrative offices of the participating schools to provide detailed information regarding the study’s objectives, procedures, and ethical considerations, and obtained written authorization from the relevant authorities. In accordance with ethical guidelines, informed consent was obtained from all participants after they were fully informed about the study’s purpose, procedures, and confidentiality protections.

Procedure

All data were collected through structured, classroom-based assessments. Survey administrators were graduate students majoring in sports science at Jishou University, each trained in research ethics and data collection procedures. Prior to data collection, all administrators received standardized training that covered key terminology, questionnaire content, and academic terminology to ensure consistency.

Before each survey, informed consent was obtained from both instructors and students. Administrators then provided standardized verbal instructions outlining the purpose of the study and the questionnaire completion process. All questionnaires were completed anonymously and collected on-site. Participants were assured that their responses would be used only for research purposes and kept strictly confidential. They were encouraged to answer honestly and to seek clarification if any item was unclear. Any questions raised were answered in accordance with official guidelines to ensure the reliability of responses.

Measures

Biological rhythm disorders

The Questionnaire of Biological Rhythm Disorders for Adolescents (QBRDA), developed by Xie et al. [17], was used to assess the level of biological rhythm disturbances among university students. The instrument consists of 29 items across four dimensions: sleep rhythm, activity rhythm, eating rhythm, and electronic device use rhythm. Each item is rated on a five-point Likert scale (1 = “strongly disagree” to 5 = “strongly agree”), with higher scores indicating greater rhythm disturbance in the corresponding domain. The original scale demonstrated good reliability and validity in adolescent populations, including both middle school and college students, supporting its suitability for the current university student sample.

Given that this study primarily focused on the relationship between students’ daily behavioral rhythms and Internet-related problematic digital behaviors, and to avoid conceptual overlap across measurement dimensions, only three subscales of the QBRDA, sleep rhythm, activity rhythm, and eating rhythm, were included in the analysis, while the “Electronic Device Use Rhythm” subscale was excluded. Scores for each subscale were computed as the mean of their items, consistent with prior research in adolescent populations [18]. The sleep rhythm subscale (9 items, Cronbach’s α = 0.84, McDonald’s ω = 0.85) was used to assess the regularity and quality of students’ sleep patterns. The activity rhythm subscale (6 items, α = 0.87, ω = 0.89) evaluated daily activity patterns. The eating rhythm subscale (7 items, α = 0.89, ω = 0.90) reflected the timing, frequency, and regularity of meals. Confirmatory factor analysis indicated that the three-factor structure was supported in the present sample. The overall questionnaire demonstrated high internal consistency (α = 0.91, ω = 0.91).

Problematic short-form video use

The Problematic Short-Form Video Media Use Scale (PSVMUS), developed by Mao et al. [19], was employed to assess the level of Problematic Short-Form Video Use among university students. The scale consists of 13 items across three dimensions—cognitive-behavioral changes (e.g., difficulty controlling use, preoccupation), physiological impairment (e.g., fatigue, sleep disturbance), and social viscosity (e.g., reduced engagement in offline activities, social interference)—each rated on a five-point Likert scale (1 = “strongly disagree” to 5 = “strongly agree”), with higher scores indicating greater Problematic Short-Form Video Use. The original scale demonstrated good reliability and validity in adolescent populations. In the present study, the scale demonstrated good internal consistency (Cronbach’s α = 0.88, McDonald’s ω = 0.89).

Problematic Social Network Use

The Social Network Addiction Tendency Scale (SNATS), developed by Wang et al. [20], was used to assess the level of Problematic Social Network Use among university students. The scale consists of 8 items rated on a five-point Likert scale (1 = “strongly disagree” to 5 = “strongly agree”), with higher scores indicating a greater tendency toward Problematic Social Network Use. The original scale demonstrated good reliability and validity in adolescent populations.In the present study, the scale demonstrated acceptable internal consistency (Cronbach’s α = 0.83, McDonald’s ω = 0.84).

Internet Game Addiction

The Internet Game Addiction Scale (IGAS), a subscale of the College Students’ Internet Addiction Scale developed by Zhou et al. [21], was used to assess the level of Internet Gaming Addiction among university students. The subscale consists of 8 items rated on a five-point Likert scale (1 = “strongly disagree” to 5 = “strongly agree”), with higher scores indicating greater levels of Internet Gaming Addiction. The original scale demonstrated good reliability and validity in adolescent populations. In the present study, the subscale demonstrated good internal consistency (Cronbach’s α = 0.88, McDonald’s ω = 0.89).

Statistical analysis

All valid questionnaires were coded uniformly and double-entered by two researchers using EpiData 3.0 to ensure data accuracy [22]. Descriptive analyses were performed using SPSS 27.0. Continuous variables are presented as mean ± standard deviation (M ± SD) and were compared across latent profiles or demographic groups using one-way ANOVA, with effect sizes reported as partial eta-squared (ηp²). Post hoc comparisons were conducted using Tukey’s HSD test to adjust for multiple comparisons. Categorical variables are presented as counts and percentages, and group differences were assessed using chi-square tests [23].

LPA was conducted in Mplus 8.3 to identify subgroups of BRDs among university students. The analysis started with a single-class model and incrementally increased the number of classes. To ensure solution stability and avoid local maxima, 200 random starts were used in the initial stage and 50 final-stage optimizations. The best log-likelihood value was successfully replicated, confirming model convergence and stability. Multiple random start sets were tested to avoid local optima. The maximum number of M-step iterations was set to 100, and the model estimation terminated normally. Missing data were handled using Full Information Maximum Likelihood (FIML). Model fit was evaluated using AIC, BIC, sample-size-adjusted BIC (aBIC), entropy, the Lo-Mendell-Rubin adjusted likelihood ratio test (LMRT), and the parametric bootstrapped likelihood ratio test (BLRT). Lower AIC, BIC, and aBIC values indicate better model fit. Significant LMRT or BLRT results (p < 0.05) indicate that the K-class model fits the data significantly better than the K-1 class model. Entropy was reported to indicate classification quality.

To examine potential predictors of latent profile membership, univariate analyses were first conducted to explore differences across demographic variables (sex, grade) and measures of Problematic Digital Behaviors. Prior to multinomial logistic regression, multicollinearity among independent variables was assessed using variance inflation factors (VIFs), which ranged from 1.45 to 1.58, indicating no serious multicollinearity. Continuous predictors were entered using their original scale scores, so the reported odds ratios represent the change in odds associated with a one-unit increase in each measure. Overall model fit was evaluated using − 2 log likelihood, likelihood ratio chi-square tests, and pseudo R² indices (Cox & Snell, Nagelkerke, and McFadden) to assess model adequacy and explanatory power. Subsequently, multinomial logistic regression was conducted with latent profile membership as the dependent variable, and odds ratios (ORs) with 95% confidence intervals were reported to assess the independent predictive effects of each factor. All tests were two-tailed, and statistical significance was set at α = 0.05.

Common method bias

To control for common method bias, Harman’s single-factor test was conducted. The variance explained by the first common factor was 27.34%, well below the critical threshold of 40%, indicating that common method bias had a limited impact.

Results

Descriptive statistics

Table 1 presents the descriptive statistics of the sample, including age, grade distribution, biological rhythm disturbances, and problematic digital behaviors. These results provide a general overview of the study population prior to latent profile analysis.

Table 1.

Descriptive statistics of the sample

Variable	Total	Male	Female
Age[Mean ± SD]	20.02 ± 1.77	20.11 ± 1.80	19.94 ± 1.75
Grade Level[n (%)]
First-year	871 (32.74)	367 (32.94)	504 (32.60)
Second-year	1050 (39.47)	428 (38.42)	622 (40.23)
Third-year	439 (16.51)	222 (19.93)	217 (14.04)
Fourth-year	300 (11.28)	97 (8.71)	203 (13.13)
Biological Rhythm Disorders [Mean ± SD]
Sleep Rhythm	2.67 ± 0.57	2.60 ± 0.60	2.72 ± 0.53
Activity Rhythm	1.85 ± 0.53	1.82 ± 0.52	1.88 ± 0.54
Eating Rhythm	2.28 ± 0.46	2.23 ± 0.48	2.31 ± 0.43
Problematic Digital Behaviors [Mean ± SD]
Short-Form Video Use	32.12 ± 12.13	30.62 ± 11.90	33.16 ± 12.17
Problematic Social Network Use	16.18 ± 5.94	16.02 ± 5.98	16.30 ± 5.92
Internet Gaming Addiction	22.03 ± 7.95	20.70 ± 7.11	22.94 ± 8.35

LPA of BRDs among University Students

LPA was conducted to identify distinct subtypes of BRDs among university students. Model estimation began with a one-profile baseline model, and the number of latent profiles was incrementally increased.

As shown in Table 2, AIC, BIC, and aBIC values decreased with more profiles, indicating improved model fit. The four-profile model exhibited high entropy (0.954), and both LMRT and BLRT tests were statistically significant (p < 0.001), suggesting superior fit compared with the three-profile model. The best log-likelihood value was successfully replicated, confirming model convergence and stability. Although the five-profile model showed further decreases in fit indices, it included a very small class representing less than 1% of the sample with lower average posterior probabilities, indicating unstable classification.

Table 2.

Comparison of Model Fit Indices for Latent Profile Models of BRDs

Model	AIC	BIC	aBIC	Entropy	LMRT(p)	BLRT(p)	Class Probabilities
1	12373.461	12408.777	12389.713	-	-	-	-
2	8795.592	8854.453	8822.680	0.933	< 0.001	< 0.001	0.65/0.35
3	6881.643	6964.048	6919.566	0.954	< 0.001	< 0.001	0.33/0.29/0.38
4	5913.321	6019.270	5962.079	0.955	< 0.001	< 0.001	0.29/0.36/0.19/0.16
5	5648.106	5777.600	5707.700	0.960	< 0.001	< 0.001	0.29/0.16/0.36/0.18/0.01

As shown in Table 3, average posterior probabilities for the most likely class membership ranged from 0.968 to 0.998, supporting reliable classification.

Table 3.

Classification quality for latent profile analysis

Latent Class	Class 1	Class 2	Class 3	Class 4
Class 1	0.993	0.007	0.000	0.000
Class 2	0.000	0.989	0.000	0.011
Class 3	0.000	0.000	0.963	0.037
Class 4	0.000	0.004	0.087	0.908

As shown in Fig. 1; Table 4, the four latent profiles differed significantly across all three dimensions of BRDs (p < 0.001). Profile 1 (“Stable Rhythm”, n = 774, 29.10%) scored lowest on sleep, eating, and activity rhythms, indicating relatively stable biological rhythms. Profile 2 (“Mild Disturbance”, n = 956, 35.94%) exhibited intermediate scores across all dimensions. Profile 3 (“Overall Disturbance”, n = 500, 18.80%) showed high scores on sleep and eating rhythms, with moderately elevated activity rhythm scores relative to the other profiles. Profile 4 (“Sleep-Activity Disturbance”, n = 430, 16.17%) demonstrated the highest activity rhythm scores and significantly higher eating rhythm scores than Profile 2, suggesting disturbances primarily in sleep and activity, with moderate disruption in eating rhythm. Post hoc analyses confirmed that pairwise differences among profiles were significant across most dimensions.

Fig. 1.

Profile Characteristics of Latent Classes of BRDs. Note. Mean scores of sleep, activity, and eating rhythms are shown for four latent profiles: Stable Rhythm, Mild Disturbance, Overall Disturbance, and Sleep-Activity Disturbance. Scores represent raw means (range 1–5). Error bars are not shown

Table 4.

Scores of three biological rhythm dimensions across four latent profiles

Profile	Sleep Rhythm	Activity Rhythm	Eating Rhythm
1 Stable Rhythm[Mean ± SD]	1.99 ± 0.09	1.42 ± 0.27	1.79 ± 0.22
2 Mild Disturbance[Mean ± SD]	2.62 ± 0.13	1.70 ± 0.28	2.32 ± 0.24
3 Overall Disturbance[Mean ± SD]	3.55 ± 0.16	2.39 ± 0.43	2.80 ± 0.35
4 Sleep-Activity Disturbance[Mean ± SD]	3.10 ± 0.13	2.53 ± 0.36	2.49 ± 0.39
F (3, 2656)	17345.56***	1598.35***	1408.26***
Post Hoc Comparisons	1 < 2<4 < 3	1 < 2<3 < 4	1 < 2<4 < 3
Effect Size	0.95	0.64	0.61

Univariate Analysis of Factors Influencing Latent Profiles of BRDs among University Students

To investigate the factors associated with latent profiles of BRDs among university students, univariate analyses were conducted on selected demographic characteristics and Internet-related problematic digital behavior variables. Variance inflation factors (VIFs) for all predictors ranged from 1.45 to 1.58, indicating no significant multicollinearity. As shown in Table 5, gender, grade level, Problematic Short-Form Video Use, Problematic Social Network Use, and Internet Gaming Addiction emerged as the primary factors associated with membership in different biological rhythm disorder profiles.

Table 5.

Univariate Analysis of Factors Associated with Latent Profiles of BRDs

Variable	Total	Stable Rhythm	Mild Disturbance	Overall Disturbance	Sleep-Activity Disturbance	x2/F	df	p
Gender [n (%)]						19.299a	1	< 0.001
Male	1114(41.9)	371(47.9)	394(41.2)	184(36.8)	165(38.4)
Female	1546(58.1)	403(52.1)	562(58.8)	316(63.2)	265(61.6)
Grade Level [n (%)]						105.887a	3	< 0.001
First-year	871(32.7)	233(30.1)	374(39.1)	153(30.6)	111(25.8)
Second-year	1038(39.0)	265(34.2)	362(37.9)	218(43.6)	193(44.9)
Third-year	451(17.0)	207(26.7)	103(10.8)	68(13.6)	73(17.0)
Fourth-year	300(11.3)	69(8.9)	117(12.2)	61(12.2)	53(12.3)
Residence Status [n (%)]						3.715a	1	0.294
On-campus	2442(91.8)	702(90.7)	880(92.1)	468(93.6)	392(91.2)
Off-campus	218(8.2)	72(9.3)	76(7.9)	32(6.4)	38(8.8)
Registered Residence [n (%)]						6.361a	1	0.095
Urban	1630(61.3)	454(58.7)	615(64.3)	301(60.2)	260(60.5)
Rural	1030(38.7)	320(41.3)	341(35.7)	199(39.8)	170(39.5)
Academic Performance [n (%)]						9.996a	4	0.616
Excellent	403(15.2)	116(15.0)	135(14.1)	82(16.4)	70(16.3)
Above Average	435(16.4)	131(16.9)	164(17.2)	72(14.4)	68(15.8)
Average	915(34.4)	259(33.5)	346(36.2)	157(31.4)	153(35.6)
Below Average	513(19.3)	149(19.3)	182(19.0)	103(20.6)	79(18.4)
Poor	394(14.8)	119(15.4)	129(13.5)	86(17.2)	60(14.0)
Problematic Short-Form Video Use [Score, ±SD]	32.27 ± 12.2	24.62 ± 7.2	34.14 ± 11.7	35.89 ± 12.4	37.67 ± 13.8	181.987b	3, 2656	< 0.001
Problematic Social Network Use [Score, ±SD]	16.25 ± 6.0	13.16 ± 3.8	17.01 ± 6.1	17.94 ± 6.1	18.16 ± 6.8	114.769b	3, 2656	< 0.001
Internet Gaming Addiction [Score, ±SD]	22.11 ± 8.1	18.12 ± 3.1	22.67 ± 7.7	25.29 ± 9.8	24.33 ± 9.8	114.500b	3, 2656	< 0.001

^aχ²: Chi-square test (applied to categorical variables, e.g., gender, grade level)

^bF: One-way analysis of variance (applied to continuous variables, e.g., three types of addiction scores)

Predictive effects of factors on latent profiles of BRDs among university students

To further examine the predictive effects of demographic characteristics and Internet-related problematic digital behaviors on latent profiles of BRDs, a multinomial logistic regression was conducted. The latent profile membership was set as the dependent variable, with the Stable Rhythm profile as the reference category. Gender (male as reference), grade level (senior year as reference), and scores on each dimension of Internet-related problematic digital behaviors were included as independent variables. ORs were calculated to assess predictive effects.

The overall model fit was acceptable, as indicated by significant likelihood ratio tests (p < 0.001) and pseudo R² values (Cox & Snell = 0.255; Nagelkerke = 0.274; McFadden = 0.110). As shown in Table 6, gender, grade level, Problematic Short-Form Video Use, Problematic Social Network Use, and Internet Gaming Addiction were factors associated with latent profile membership. Specifically, regarding gender, females were more likely than males to belong to the Overall Disturbance profile. Concerning grade level, students in the Mild Disturbance and Overall Disturbance profiles showed significant grade-level differences, with second- and third-year students in the Mild Disturbance profile, and first- and third-year students in the Overall Disturbance profile being less likely to be classified into these profiles compared with seniors. For the Sleep-Activity Disturbance profile, only first-year students showed a significantly lower risk than seniors, while second- and third-year students did not differ significantly, indicating that fourth-year students face higher risk primarily relative to first-year students. Regarding Internet-related problematic digital behaviors, higher scores were associated with an increased likelihood of belonging to the Mild Disturbance, Overall Disturbance, and Sleep-Activity Disturbance profiles. Notably, students with higher levels of Internet Gaming Addiction were associated with the Overall Disturbance profile, whereas those with higher Problematic Short-Form Video Use or Problematic Social Network Use were associated with the Sleep-Activity Disturbance profile.

Table 6.

Multinomial logistic regression analysis of factors influencing latent profiles of BRDs

Variable	β	SE	Wald χ2	p	OR	95%CI
Mild Disturbance
Intercept	-3.617	0.296	148.939	< 0.001	—	—
Female	0.145	0.107	1.839	0.175	1.156	0.937 ~ 1.427
Grade Level
First-year	-0.324	0.187	2.998	0.083	0.724	0.502 ~ 1.044
Second-year	-0.485	0.185	6.910	0.009	0.616	0.429 ~ 0.884
Third-year	-0.999	0.205	23.853	< 0.001	0.368	0.247 ~ 0.550
Problematic Short-Form Video Use	0.075	0.007	116.637	< 0.001	1.078	1.064 ~ 1.093
Problematic Social Network Use	0.070	0.013	31.600	< 0.001	1.073	1.047 ~ 1.100
Internet Gaming Addiction	0.053	0.011	23.180	< 0.001	1.054	1.032 ~ 1.077
Overall Disturbance
Intercept	-5.300	0.341	242.075	< 0.001	—	—
Female	0.270	0.129	4.402	0.036	1.310	1.018 ~ 1.686
Grade Level
First-year	-0.576	0.223	6.697	0.010	0.562	0.364 ~ 0.870
Second-year	-0.338	0.215	2.454	0.117	0.714	0.468 ~ 1.088
Third-year	-0.623	0.241	6.662	0.010	0.537	0.334 ~ 0.861
Problematic Short-Form Video Use	0.078	0.008	104.687	< 0.001	1.081	1.065 ~ 1.097
Problematic Social Network Use	0.071	0.014	25.301	< 0.001	1.074	1.044 ~ 1.104
Internet Gaming Addiction	0.088	0.012	56.995	< 0.001	1.092	1.067 ~ 1.117
Sleep-Activity Disturbance
Intercept	-5.607	0.357	247.371	< 0.001	—	—
Female	0.235	0.135	3.046	0.081	1.265	0.972 ~ 1.647
Grade Level
First-year	-0.754	0.236	10.239	0.001	0.470	0.296 ~ 0.747
Second-year	-0.320	0.225	2.025	0.155	0.726	0.467 ~ 1.128
Third-year	-0.420	0.248	2.866	0.090	0.657	0.404 ~ 1.069
Problematic Short-Form Video Use	0.094	0.008	146.970	< 0.001	1.098	1.082 ~ 1.115
Problematic Social Network Use	0.080	0.015	30.130	< 0.001	1.083	1.053 ~ 1.114
Internet Gaming Addiction	0.066	0.012	29.646	< 0.001	1.068	1.043 ~ 1.093

Discussion

Latent profiles of BRDs among university students

This study revealed the heterogeneous patterns of BRDs among university students through LPA, identifying four distinct profiles: Stable Rhythm, Mild Disturbance, Overall Disturbance, and Sleep-Activity Disturbance. Compared with the Stable Rhythm profile, the other three profiles exhibited varying degrees of rhythm disruption. Specifically, the Mild Disturbance profile displayed moderate scores across all dimensions, indicating that overall rhythms remained relatively stable. The Overall Disturbance profile showed elevated scores in sleep, eating, and activity rhythms, suggesting severe disturbances across all aspects of biological rhythm. In contrast, the Sleep-Activity Disturbance profile was characterized by high scores in sleep and activity rhythms, while eating rhythm scores were relatively preserved compared to the Overall Disturbance profile, reflecting localized imbalances primarily in sleep and activity rhythms. This preserved eating rhythm may reflect the influence of fixed university dining schedules, which serve as social zeitgebers that help maintain relatively regular meal times despite disrupted sleep and activity patterns [24, 25]. Notably, over one-third of students belonged to the Overall Disturbance or Sleep-Activity Disturbance profiles, highlighting that the prevalence and severity of biological rhythm disruptions among university students warrant concern. These findings suggest that educators and parents should monitor students’ biological rhythms and, based on specific patterns in sleep, eating, and activity rhythms, classify students accordingly to implement targeted interventions and guidance.

Demographic characteristics of latent profiles of BRDs among university students

The results of the multinomial logistic regression indicated that gender and grade level significantly predicted latent profile membership of BRDs among university students. Compared with males, females were more likely to belong to the Overall Disturbance profile, suggesting a higher risk of rhythm disruption, consistent with previous research [21]. On one hand, females experience more frequent physiological fluctuations and hormonal changes (e.g., menstrual cycles, variations in estrogen and progesterone), which may interfere with sleep quality, energy levels, and the stability of daily rhythms [26]. On the other hand, females generally exhibit greater sensitivity in emotional regulation and cognitive processing, tending to focus on negative information and engage in extensive rumination. This sustained psychological load may affect stress regulation and daily scheduling, thereby disrupting established life rhythm balance [27, 28]. The combined effects of physiological changes and cognitive-emotional traits may further reduce female students’ adaptability to external rhythm variations, influencing daily behavioral patterns and psychological regulation, and thus exacerbating and prolonging biological rhythm disturbances.

Students beyond the first year exhibited higher probabilities of belonging to the Mild Disturbance, Overall Disturbance, or Sleep-Activity Disturbance profiles, with differences plateauing after the first year. While employment-related stress may contribute to elevated risk among seniors, other factors such as academic workload, adaptation to university life, and lifestyle changes across grades may also influence biological rhythm disturbances. Under these pressures, seniors often have to adjust their schedules to meet escalating academic and career demands. Additionally, greater autonomy over time with reduced institutional supervision can challenge self-regulatory capacities, potentially contributing to physiological and psychological dysregulation, which may be associated with further disruptions in biological rhythms [29, 30].

Associations between BRDs and Internet-related problematic digital behaviors among University Students

Univariate analyses indicated that Problematic Short-Form Video Use, Problematic Social Network Use, and Internet Gaming Addiction were significant factors influencing latent profiles of BRDs among university students. According to self-control theory, individuals possess limited self-regulatory resources, and excessive expenditure on a particular behavior can weaken the regulation of other behaviors, leading to behavioral dysregulation [31, 32]. As a digital behavioral pattern, Internet-related problematic digital behavior can continuously deplete self-control resources, impairing the planning and execution of daily routines [33, 34]. University students face multiple demands, including academic pressures, social expansion, and the development of self-management skills, which already place high demands on their self-control resources [35]. The intrusion of Internet-related problematic digital behaviors may further contribute to the imbalance of resource allocation. When self-control resources are excessively consumed by short-video use, social networking, or gaming, students’ ability to maintain regular sleep, eating, and activity rhythms is compromised, thereby disrupting the stability and continuity of biological rhythms [36]. These findings support the potential disruptive effects of Internet-related problematic digital behaviors on students’ biological rhythm stability.

Multinomial logistic regression further revealed significant differences in the effects of different types of Internet-related problematic digital behaviors on latent profile membership. Specifically, Problematic Short-Form Video Use and Problematic Social Network Use were notably associated with localized disturbances in sleep and activity rhythms, whereas Internet Gaming Addiction was associated with disruptions across all dimensions of biological rhythms. The pronounced effects of short-video and social network use on sleep and activity rhythms may be attributed to the fragmented nature of these platforms and their intrusion into flexible time periods [37, 38]. According to social comparison theory, individuals often evaluate their abilities, emotions, and social status by comparing themselves to others in the absence of objective standards [38]. On short-video platforms, users are frequently exposed to content depicting “others’ curated lives” or “peers’ achievements,” while the continuous updates and feedback mechanisms on social networks promote upward or downward social comparisons [39]. These comparison behaviors typically occur during flexible time periods, such as before sleep or during study breaks [40]. Pre-sleep comparisons may be associated with anxiety or arousal, potentially prolonging sleep onset latency, while comparisons during activity intervals may be linked to interruptions in studying or exercising. In contrast, eating rhythms are primarily regulated by intrinsic physiological drives, with hunger often taking precedence over emotionally related effects from social comparisons [41, 42]. This may partially explain why the Sleep-Activity Disturbance profile shows relatively preserved eating rhythms compared to the Overall Disturbance profile, while eating rhythms are clearly disrupted in the Overall Disturbance profile [43, 44].

Internet Gaming Addiction, however, was associated with pervasive disruptions in sleep, activity, and eating rhythms, consistent with previous findings [45, 46]. The sustained and immersive nature of gaming may disrupt normal physiological cycles, which may contribute to deviations in sleep, exercise, and dietary behaviors. This deviation not only directly disturbs daily routines but may also generate dual physiological and psychological stress, continually undermining the self-regulatory capacity of biological rhythms and creating a “gaming addiction–physiological and psychological imbalance–rhythm disruption” vicious cycle [47, 48].

In summary, BRDs among university students exhibit significant heterogeneity and can be classified into four profiles: Stable Rhythm, Mild Disturbance, Overall Disturbance, and Sleep-Activity Disturbance. Under the influence of Internet-related problematic digital behaviors, these profiles demonstrate differentiated patterns: Problematic Short-Form Video Use and Problematic Social Network Use primarily are associated with localized disturbances in sleep and activity rhythms, while Internet Gaming Addiction is more likely linked to widespread rhythm dysregulation. Based on these findings, educators and parents should closely monitor students’ biological rhythms and develop personalized interventions tailored to the specific type of rhythm disturbance and associated addictive behaviors.

Strengths and limitations

Strengths

This study adopted a person-centered perspective, employing LPA to identify heterogeneous subtypes of BRDs among university students. The large sample size (N > 2,600) enhances the statistical power and robustness of the findings. Notably, the identification of a distinct “Sleep–Activity Disturbance” profile represents a novel contribution to the literature on biological rhythm disturbances in university populations.Additionally, it explored the differential impact of various Internet-related problematic digital behaviors on these rhythm disorder profiles. The findings provide both theoretical and empirical support for understanding the complex interplay between university students’ lifestyle patterns and digital behaviors. Furthermore, the study has important practical implications for promoting healthy biological rhythms among university students and for developing effective prevention and intervention strategies targeting Internet-related problematic digital behaviors.

Limitations

Nevertheless, several limitations should be acknowledged. First, the sample was drawn from four public universities in Hunan Province, China, which may limit the generalizability of the findings; future research should consider expanding the sample to other regions and cultural contexts. Second, all data were collected via self-report questionnaires, which may be subject to social desirability and recall biases, potentially affecting the accuracy of measurements. Third, the cross-sectional design of this study precludes causal inference regarding the relationships between biological rhythm disturbances and Internet-related problematic digital behaviors. Future studies are encouraged to incorporate more objective data sources, such as physiological assessments or behavioral observations, to complement self-reported information.

Conclusion

This study investigated the heterogeneous patterns of BRDs among university students and examined the differential impact of Internet-related problematic digital behaviors on these profiles. Several key findings emerged: (1) university students’ BRDs can be classified into four distinct profiles—Stable Rhythm, Mild Disturbance, Overall Disturbance, and Sleep-Activity Disturbance—highlighting substantial heterogeneity in rhythm disturbances; (2) demographic factors, including gender and grade level, significantly predicted profile membership, with female and senior-year students more likely to experience severe disruptions; (3) Internet-related problematic digital behaviors exert differentiated effects on biological rhythms: Problematic Short-Form Video Use and Problematic Social Network Use primarily contributed to localized disturbances in sleep and activity rhythms, whereas Internet Gaming Addiction was associated with pervasive disruptions across sleep, activity, and eating rhythms. These findings underscore the need for personalized interventions that consider both students’ demographic characteristics and digital behavior patterns. Specifically, targeted strategies addressing the type of rhythm disorder and associated problematic digital behaviors may help restore biological rhythm stability, promote overall well-being, and reduce potential adverse physical and psychological outcomes among university students.

Abbreviations

BRDs

Biological Rhythm Disorders

LPA

Latent Profile Analysis

AIC

Akaike Information Criterion

BIC

Bayesian Information Criterion

aBIC

Adjusted Bayesian Information Criterion

LMRT

Lo-Mendell-Rubin Likelihood Ratio Test

BLRT

Bootstrap Likelihood Ratio Test

M

Mean

SD

Standard Deviation

OR

Odds Ratio

Authors’ contributions

Ziyi Chen, Chengjun Chai, and Zeng Zhou conceptualized and designed the study. Ziyi Chen and Shunchi Xue collected and analyzed the data. Ziyi Chen drafted the manuscript. Hongli Wang and Chengjun Chai provided critical revisions and supervision. All authors reviewed and approved the final manuscript.

Funding

No Funding.

Data availability

The datasets of this study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval for this cross-sectional study was obtained from the Biomedical Ethics Committee of Jishou University (Approval No. JSDX-2023-0011), which covered the 2025 data collection wave. The data were collected exclusively for academic research purposes. Prior to data collection, the research team contacted the administrative offices of the participating schools to provide detailed information regarding the study’s objectives, procedures, and ethical considerations, and obtained written authorization from the relevant authorities. In accordance with ethical guidelines, written informed consent was obtained from all participants, who were fully informed about the study’s purpose, procedures, and confidentiality protections.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.