In March 2024, a convenience sampling method was used to survey 1,219 lower grade medical students (average age 19.4 years, standard deviation 2.08) at a medical college. By removing questionnaires such as invalid answers, contradictory answers, and missing key information, 1,099 valid questionnaires (90.16%) were actually collected. All survey subjects had informed consent for this study, which has been approved by the Ethics Committee of Guangdong Provincial Hospital of Chinese Medicine (Ethics Review Approval No.: ZE2024-054-01).

The Mplus 7.0 software was used for latent category analysis (LCA). The evaluation indicators of the latent category model include AIC: Akaike Information Criterion; BIC: Bayesian Information Criterion; ABIC: BIC for sample calibration; BLRT: Bootstrap Likelihood Ratio Test (K-1 vs. K classes); VLMR: Vuong Lo Mendel Rubin Likelihood Ratio Test (K-1 vs. K classes); Entropy: the smaller the AIC, BIC, and ABIC values, the better the model fit, when the Entropy≥0.8,the classification accuracy exceeds 90%; if the p-values of BLRT and VLMR are less than 0.05, it indicates that the k-category model is superior to the k-1 category model. Referring to the study by Sinha et al. (2021), it is determined that the proportion of people in each category should not be less than 10%.

Since DASS-21 measures three indicators of stress, anxiety, and depression, rather than three dimensions of a single indicator, the LCA method is used to classify the stress, anxiety, and depression status of medical students separately. The DASS-21 entry is assigned a score of 0–3. Considering that in this sample, the response of “Applied to me very much, or most of the time” with a score of 3 for each entry is less than 5%, and considering that LCA is sensitive to outlier, the responses with scores of 2 and 3 will be merged and classified into three categories for latent category analysis.

SPSS 26.0 software was used for data processing and statistical analysis. Categorical variables were represented by frequency and composition ratio, and continuous variables were represented by mean and standard deviation. The inter group differences of categorical variables were compared using contingency table analysis, and logistic regression analysis was used to explore the influencing factors of medical students’ stress, anxiety, and depression.

The fitting results show that when the stress level of medical students is at k = 3, the Entropy value is 0.872, and the p-values corresponding to BLRT and VLMR are both less than 0.01. Moreover, the downward trend of AIC, BIC, and ABIC slows down, and the slope gradually approaches 0. When the anxiety level of medical students is at k = 2, the Entropy value is 0.870, and the p-values corresponding to BLRT and VLMR are both less than 0.01, and the proportion of people in each category is not less than 10%. When the depression level of medical students is at k = 3, the Entropy value is 0.812, and the p-values corresponding to BLRT and VLMR are both less than 0.01. Moreover, the downward trend of AIC, BIC, and ABIC slows down and gradually approaches 0. Taking into account the simplicity of the model and the research objectives, this study retains the 3-category stress and depression latent category model and the 2-category anxiety latent category model, as shown in Tables 1–3.

In this study, the conditional probability was set as the response probability for the option “Did not apply to me at all” in each entry. Based on the conditional probability level, the subgroups of the three latent categories of stress are named as high stress group, low stress group, and no stress group. The stress conditional probability plot shows that the high and low stress groups generally have low probability responses to the stress 7 items in DASS, but have high response probabilities on the Q6 (I tended to over react to situations) and Q18 (I felt I was a rat touch) items. The average scores of Q6 and Q18 were calculated, and the results show that the mean of Q6 was 0.53 and the mean of Q18 was 0.47, reflecting the lower allergic reactions and tactile anger in this sample group under stress conditions (Figure 1).

The anxiety conditional probability plot shows that the high anxiety group exhibits a generally low probability response to the seven items of stress in DASS, but the response probability on Q9 (I was worried about situations in which I might panic and make a fool of myself) is lower, and the low anxiety group has a higher response probability on all six items, but shows a significant low response to Q9, which is similar to the high anxiety group. Therefore, the low anxiety group is named as “low anxiety group afraid of social embarrassment (LAGASE).” The average Q9 of the entire sample is 1.31, reflecting the general anxiety of the sample group toward social embarrassment (Figure 2).

The depression conditional probability plot shows that the three latent categories have different response patterns to the seven items of depression. The high depression group shows a low response probability on all items, while the non-depression group has a lower response probability in Q5 (I found it difficult to work up the initiative to do things) compared to the high response probability on the other six items. Between the high depression and non-depression groups, an intermediate group is classified, which shows a high response probability to Q17 (I felt that I was worth much as a person) and Q21 (I felt that life was meaningless), and a low response probability to Q5 (I found it difficult to work up the initiative to do things), Q10 (I felt that I had nothing to look forward to) and Q13 (I felt down hearted and blue). The statistical results show that Q5 scoring 0.81, Q10 scoring 0.61, Q13 scoring 0.63, Q17 scoring 0.27, and Q21 scoring 0.3. It indicates that the middle group generally reflects a lack of motivation, depression, and frustration, but rarely experiences self-deprecation and a sense of meaninglessness in life, hence the name “depression group with insufficient motivation (DGIM; Figure 3).

The stress detection rate, anxiety detection rate, and depression detection rate in the sample of this study were 21.7, 46.4, and 31.7%, respectively. According to the norm classification standard, the scores of stress, anxiety, and depression in the sample were classified into four groups: normal, mild, moderate, severe and above. Corresponding to the stress latent category subgroup, anxiety latent category subgroup, and depression latent category subgroup, cross comparison was conducted in a contingency table. The results showed that there was a considerable degree of consistency between the latent category subgroup and the severity subgroup based on the norm (the contingency coefficients were all greater than 0.6), but the differences between the two were also significant. Please refer to Tables 4–6 for details.

The latent category of no pressure group is all distributed in the normal pressure group, the vast majority of the latent category of high pressure group is distributed in the mild, moderate, severe and above pressure level groups, mainly in the moderate, severe and above levels, and the latent category of low pressure is distributed in the normal, mild and moderate pressure groups, mainly in the normal and mild pressure groups. The LAGASE included all normal anxiety samples, and also included mild and moderate anxiety samples based on response patterns, with mild anxiety samples being the main group. The latent high anxiety group is distributed in the mild, moderate, severe and above anxiety levels, with the majority of samples being moderate, severe and above anxiety levels. The latent high depression group is distributed in the mild, moderate, severe and above degree depression groups, with moderate, severe and above degree depression samples being the main ones. The vast majority of the latent non depression group is distributed in the normal depression level group, while the DGIM is mainly distributed in the normal depression group and the mild depression group.

Based on the DASS-21 norm, the anxiety detection rate of medical students in this study was 46.4%, and the depression detection rate was 31.7%, which is similar to previous research results and significantly higher than the average level of Chinese university students.

Among the three subgroups of stress potential categories, the high stress potential category group and the non-stress potential category group have a strong correspondence with the severity grouping based on the norm. At the same time, a low stress group was identified in the main population with normal levels. This subgroup exhibited a series of tense behavioral patterns, but had low allergic reactions and low feelings of anger. On the seven items of anxiety, both high and low anxiety subgroups showed concerns and fears about making a fool of themselves in social situations. Among the three subgroups of depression, the high depression group and the non-depression group have a strong correspondence with the severity grouping based on the norm. At the same time, a depression group with insufficient motivation was identified in the population with normal levels. This subgroup exhibited behavioral patterns such as insufficient motivation, depression, and frustration, but rarely had self-deprecation and a sense of meaninglessness in life. Compared with the severity grouping of the norm, latent category analysis captures some special behavioral patterns of stress, anxiety, and depression in the medical student population, and these patterns have cross group characteristics between groups grouped by severity. It can be considered that latent category analysis based on response patterns and conditional probability has its unique contribution to distinguishing the heterogeneity of stress, anxiety, and depression in the medical student population, and can be combined with norm classification to provide more targeted intervention guidelines.

It is generally believed that the incidence rate of depression in women is higher than that in men. The results of this study showed that compared with the non-depression group, men were significant predictors of high depression (OR: 1.77, p: 0.01, 95% CI: 1.18–2.64). Without distinguishing majors, the meta-analysis study by Yu and Huang (2023) showed no statistically significant difference in depression scores between men and women, while a meta-analysis study of medical college students showed that men were more depressed than women (OR: 1.17, 95% CI: 1.10–1.24; Mao et al., 2019). Based on the social factor hypothesis, relevant studies suggest that Chinese society generally has higher expectations for men’s economic ability and family social responsibility, and the pressure of long-term education and uncertainty about future employment faced by medical students may be one of the important reasons for the higher depression detection rate among men than women (Chen et al., 2013; Pan et al., 2016). Considering that gender does not have a significant impact on the depression group with low motivation, unlike the high depression group, which mainly has moderate to severe depression, and the depression group with low motivation, which mainly consists of normal and mildly depressed individuals, the above explanation should also consider the susceptibility factors of individuals with high depression based on the Comprehensive Quality Stress Model (Parker and Brotchie, 2010).

The results of meta-analysis showed that left-behind children exhibited higher rates of depression and anxiety compared to their peers (Cheng and Sun, 2015). For medical college students, this study found that previous experiences of left behind children were a significant predictor of depression with low motivation, reflecting the long-term impact of left-behind experiences on individual psychology. Left-behind (separation from father/mother) means the loss of actual or symbolic attachment objects. From the perspective of attachment theory, left behind is a disturbance and destruction of secure attachment relationships, and insecure attachment is considered an important risk factor for mental health problems and a core element in the occurrence and development trajectory of mental disorders. The meta-analysis results of Herstell et al. (2021) showed that insecure attachment is a cross diagnostic common risk factor for bipolar disorder, depression, and schizophrenia spectrum disorders.

Since middle school, many Chinese students have started to leave their families and move into schools. In this study, 90.6% of medical students had experience living on campus, reflecting the prevalence of living on campus. Compared to the low anxiety group, the experience of living on campus was a significant predictor for the high anxiety group. Living on campus means living and learning with others. From the common concern and fear of social embarrassment exhibited by medical students, it can be considered that in a cultural environment that emphasizes interpersonal harmony with others, living on campus increases an individual’s social chances in the group, combined with the psychological pressure brought by the Chinese expectation of “building good relationships” with others, and jointly promotes anxiety about social embarrassment.

In this study, low subjective socioeconomic status was a common predictor for the high stress group, high anxiety group, and high depression group, similar to the results of Präg et al. (2016) based on data from 29 countries, which confirmed a robust association between subjective socioeconomic status (SSS) and self-rated health (SRH) and psychological well-being outcomes. SSS showed a significant negative correlation with psychological symptoms of SRH and a significant positive correlation with psychological well-being, and this result was universal across all study countries.

Logistic regression revealed the importance of family atmosphere as a factor. Compared to harmonious family atmosphere, disharmony or average in family atmosphere can significantly increase the chances of medical students being classified into high stress group, high anxiety group, and high depression group. According to Olson (2011) understanding of family atmosphere, harmonious family atmosphere is defined as a balance of cohesion, flexibility, and communication quality. A systematic evaluation of the relationship between family atmosphere and adolescent adaptability shows a significant correlation between family atmosphere and adolescent adaptive behavior (Kurock et al., 2022). According to Bronfenbrenner (1979) ecosystem theory, the family as a microsystem has a direct impact on individual development, and supportive family environments can enhance individuals’ ability to cope with stress. Cross cultural studies have shown that family influence is more prominent in collectivist cultures (Asia; He et al., 2023).

The systematic review and meta-analysis conducted by Luo et al. (2025) on the correlation between psychological resilience and mental health in adolescents and youth groups showed that the correlation coefficient between psychological resilience and negative indicators of mental health was −0.391 (95% CI: −0.469, −0.308, p < 0.001), and the correlation coefficient with positive indicators of mental health was 0.499 (95% CI: 0.400, 0.586, p < 0.001). This study refined the relationship between the three factors of psychological resilience (tenacity, strength, and optimism) and stress, anxiety, and depression. The results showed that psychological resilience factors have different effects on the subgroups of medical students in the latent categories of stress, anxiety, and depression.

The optimistic physical and mental benefits have been supported by a wealth of evidence in previous studies (Scheier and Carver, 2018; Rasmussen et al., 2009), and in this study, the optimistic factor has a significant predictive effect on stress groups. Bonanno et al. (2024) defines optimism as a motivation and belief that accepting challenges is valuable, believing that they can and ultimately will effectively move toward a positive future. Based on this, optimists provide the necessary motivational push, handle cognitive and emotional events in adversity, and initiate behavioral adjustments they may need. Carver and Scheier (2017) used a healthy biological behavior model to explain the mechanism of the relationship between optimism and individual physical and mental health. They believe that there are three important pathways of action. The first is the behavioral pathway mediated by differential coping strategies. The experimental research results of Wang and Wang (2023) showed that individuals facing different levels of negative situations are more inclined to choose positive reappraisal strategies compared to pessimists, and can flexibly choose multiple cognitive reappraisal strategies to regulate negative emotions. There is considerable evidence to suggest that optimism is associated with the use of positive, method oriented coping strategies, which directly address the problems faced or change perceptions of stressors in a positive way, thereby exerting adaptive functions (Carver and Scheier, 2017). The second is the physiological pathway of stress response. Optimism is related to the immune response to stress response. In the face of stressful situations, optimists tend to persist in overcoming stressors, which is reflected at the physiological level, that is, sustained physiological stress response and cell-mediated immune reduction. Researchers believe that although this phenomenon may lead to an increase in stress levels or even disorder in the short term, this strategy may have long-term benefits in ultimately solving stressors (Segerstrom, 2005). The third is social support and social integration pathways. Researches have shown that optimism promotes more beneficial social relationships and networks (Smith et al., 2013; Andersson, 2012). The three approaches reflect the important impact of optimism on individuals’ perception of stress and stress management in the face of adversity or stressors from different perspectives.

Yu and Zhang (2007) have argued that tenacity described a person’s perseverance, calmness, agility, and sense of control in the face of difficulties and challenges, while strength focused on an individual’s ability to recover and become stronger after experiencing setbacks.

The results showed that tenacity and strength were significant predictive factors for the high depression group, while strength was a significant predictive factor for the high anxiety group. The resource conservation theory proposed by Hobfoll (1989, 2002) provided a reference explanatory model for this impact. Hobfoll (1989, 2002) resource conservation theory suggested that in stressful situations, a person’s willingness to act and subsequent behavioral activities were regulated by two orientations: resource gain orientation and resource loss orientation. Hobfoll defines resources as personal characteristics, objects, conditions, or energy that were valued by individuals, or as means, personal characteristics, conditions, or energy to achieve these objects. The study by Bouckenooghe et al. (2019) considered tenacity and strength as loss oriented resources. Based on the framework of resource conservation theory, anxiety and depression can be conceptualized as individuals’ physical and mental responses to actual or imagined resource loss. As effective defense mechanisms against resource loss, tenacity and strength (loss oriented resources) enable people to recover from negative experiences and threats (Block and Kremen, 1996). Further research showed that when loss oriented individuals had fewer resources, their coping abilities were compromised, making them unable to effectively face challenging stressors (Demerouti et al., 2004) and exhibiting more physical and mental symptoms. In this study, high strength can significantly reduce the odds of medical students being classified as high anxiety group, and high tenacity and strength can significantly reduce the odds of medical students being classified as high depression group, further verifying the important role of tenacity and strength in resisting resource loss and maintaining physical and mental health. As the saying goes, “there are unpredictable storms in the sky, and misfortunes and blessings in a single day,” in the uncertain risks of modern society, encountering setbacks and difficulties, and bearing possible high pressure, may be a high probability event for every individual, especially for medical students. In this context, improving physical and mental reactions such as anxiety and depression under pressure, strengthening tenacity and strength and other psychological capital, is not only supported by research evidence, but also in line with the value guidance of traditional Chinese culture. For example, emphasizing the cultivation of resilience character such as “good as water” and “resilient as bamboo,” emphasizing the spirit of self-improvement and tolerance by learning from heaven and earth, such as “heaven’s actions are healthy, gentlemen strive for self-improvement; the terrain is vast, and gentlemen carry things with kindness.”

The limitation of this study was that cross-sectional studies could only provide data at a certain time point. Future studies should comprehensively consider longitudinal variation data at different time points to derive more clinically significant symptom subtypes. Secondly, non-random sampling limited the extrapolation of research results, and the universality and representativeness of the results need further verification.