Balancing Mental Health: Predictive Modeling for Healthcare Workers During Public Health Crises

Balancing Mental Health: Predictive Modeling for Healthcare Workers During Public Health Crises | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Balancing Mental Health: Predictive Modeling for Healthcare Workers During Public Health Crises Jiana Wang, Lin Feng, Nana Meng, Cong Yang, Fanfan Cai, Xin Huang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5228634/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 06 Aug, 2025 Read the published version in Scientific Reports → Version 1 posted 10 You are reading this latest preprint version Abstract Background During public health emergencies such as SARS, Ebola, and COVID-19, healthcare workers (HCWs) are often required to confront these crises, potentially leading to adverse mental health outcomes. Consequently, they are at a heightened risk of experiencing symptoms of depression and anxiety. It is widely recognized that psychological disorders can lead to severe consequences. Despite this, there remains a scarcity of research focused on developing predictive models to forecast the depression and anxiety levels of healthcare workers under these challenging conditions. Methods A total of 349 HCWs were selected from a Class-A tertiary hospital in the city of Shenyang, Liaoning Province in China. Depression and anxiety were assessed using the Patient Health Questionnaire (PHQ-9) and the Generalized Anxiety Disorder (GAD-7) scale, respectively. This study employed a random forest classifier(RFC) to predict the depression and anxiety levels of HCWs from three perspectives: individual, interpersonal, and institutional. Moreover, we employed The Synthetic Minority Over-sampling Technique (SMOTE) to address the issue of imabalanced data distribution. Results The prevalence of depression and anxiety among HCWs were 28.37% and 33.52%, respectively. The prediction model was developed using a training dataset (70%) and a test dataset (30%). The area under the curve (AUC) for depression and anxiety were 0.88 and 0.72, respectively. Additionally, the mean values of the 10-fold cross-validation results were 0.77 for the depression prediction model and 0.79 for the anxiety prediction model. For the depression prediction model, the top ten most significant predictive factors were: burnout, resilience, emotional labor, adaptability, working experience( < = 1year), physician, social support, average work time last week(9–11 hours), age(28–30 years), age(31–35 years old). For the anxiety prediction model, the top ten most significant predictive factors were: burnout, adaptability, emotional labor, age(31–35), average work time last week(9–11 hours), resilience, physician, social support, working experience( < = 1 year), female. Conclusions It is essential to develop multiple interventions that provide support both before and after a public health emergency, aiming at mitigating symptoms of depression and anxiety. SMOTE is a practical method for addressing imbalances in datasets. Mitigating burnout among HCWs, bolstering their resilience and adaptability, and ensuring reasonable work hours are crucial steps to prevent adverse mental health problems. Biological sciences/Psychology/Human behaviour Biological sciences/Psychology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Background According to the report from the Chinese Center for Disease Control and Prevention, the peak of infections was reached on December 22, 2022, with a positive test count of 6.94 million. Since China announced a slight relaxation of its zero-COVID policies on November 11, 2021, followed by a significant easing on December 7, 2022, over 1.41 million people have been infected and lost their lives in the initial wave that ensued(1). The spread of the virus has dramatically changed daily life and the socio-economic environment in affected areas. According to previous studies, these changes are expected to have long-term effects on mental health(2–4). Following the announcement of a major relaxation on December 7, 2022,there has been a notable surge in infections (Fig. 1 ).The rapid spread of COVID-19 throughout China has significantly strained HCWs who are directly or indirectly engaged in combating the pandemic,potentially increasing their risk of adverse mental health outcomes.Notably, HCWs are consistently exposed, both directly and indirectly, to patients and colleagues infected with COVID-19,which elevates their risk of developing psychological disorders(5–10).According to recent research, the prevalence of depression and anxiety among HCWs falls within the ranges of 45.6%-77.6% and 20.7%-60.2%,respectively(11–13).HCWs may exhibit poor performance in both their professional and personal lives if they show symptoms of depression or anxiety, and the effects of these symptoms can be long-lasting.Therefore, early identification of individuals suffering from depression and anxiety is crucial to intervene promptly and minimize the impact on HCWs' health by potentially reducing the progression of the disease.However, the application of machine learning to predict adverse mental health outcomes in HCWs is currently limited.Moreover, our survey occurred at the outset of the routine epidemic prevention and control measures for COVID-19. During this period,HCWs had to confront a substantial influx of infected patients within a condensed time frame,necessitating them to bear heightened social expectations and work pressure. Surprisingly,there is scant research dedicated to investigating the psychological well-being of healthcare personnel during this specific phase. The mental health challenges encountered by HCWs during the COVID-19 pandemic have garnered considerable attention from researchers. Predominantly, depression, anxiety, and post-traumatic stress disorder have been the focal points in studies addressing mental health issues in this context. Notably, depression and anxiety are frequently assessed through cross-sectional surveys, which provide valuable data for statistical analysis.The research approaches in this domain have been significantly influenced by the social-ecological model and ecological models of health behavior. Factors influencing mental health outcomes are categorized within the socio-ecological framework and encompass various aspects: Individual Factors: These encompass age, gender, ethnicity, race, marital status, technical title, residence (urban/rural), and other individual characteristics; Interpersonal Factors: This category includes social support, emotional labor, and other interpersonal dynamics;Institutional Factors: These pertain to work experiences, workload, occupation, burnout, and issues such as shortages of personal protective equipment (PPE).(6,14,15).In summary, we selected the important varibles from these aspects to bulid the prediction models. Early detection and diagnosis is paramount to understand and address mental health illness on a population level. Machine learning is increasingly being utilized in the field of mental health. Research focused on the prediction and analysis of risk factors for depression and anxiety using machine learning techniques has been conducted and displayed a reliable performance(7,16–20).However, according to our research, few studies have developed a machine learning model specifically for predicting depression and anxiety in HCWs during such critical periods.Furthermore, there is a notable lack of research focused on the issue of imbalanced datasets within the realm of mental health. Imbalanced dataset problems are prevalent in real-world scenarios, particularly in the fields of medicine and psychology.The proportion of HCWs diagnosed with depression or anxiety is relatively small in comparison to the total number of HCWs.It’s presumed that the data are drawn from the same distribution as the training dataset,presenting imbalanced dataset to the classifier and producing intolerable results(21).In response to this issue, sampling techniques and the creation of synthetic data are employed to attain a balanced distribution of classes, either by augmenting or reducing samples.Specifically, naive random over-sampling involves generating new samples by randomly selecting and duplicating existing samples from the minority class. In terms of synthetic sampling, researchers have introduced the SMOTE(22,23).SMOTE generates synthetic data for the minority class by identifying some of the nearest neighbors within the minority dataset. It then creates synthetic minority data points along the lines connecting the minority data points to their nearest neighbors(24). Previous researchers have not utilized the SMOTE method to construct a machine learning model for predicting adverse mental health outcomes in HCWs in real-world situations.The typical approach to address imbalanced data involves increasing the sample size or reducing variables. However, these methods are often challenging to implement and may yield inaccurate results. Therefore, there is a need for innovative methods in our field to address this issue.To address this gap, we conducted a survey involving 349 HCWs from a Shenyang's Class-A tertiary hospital.This survey focused on socio-ecological factors associated with depression and anxiety during a critical period. By applying the Random Forest Classifier and incorporating the SMOTE method, we aimed to predict depression and anxiety among HCWs for the purpose to offer valuable insights for developing interventions to support HCWs in outbreak conditions and provide a potential solution for addressing the commonly encountered issue of imbalanced data in real-world scenarios. Methods Study design and participants This study used a single-center, cross-sectional survey design, involving doctors, nurses, and clinical technicians currently working at a Class-A tertiary hospital in the city of Shenyang, Liaoning Province in China.Additionally, individuals in administrative and financial positions were also included.The survey was conducted during the first and second weeks of February 2023, approximately after the government declared the normalization of the COVID-19 pandemic on December 7, 2022.Two researchers entered data into the database double-blindly using Epidata.3.1 to ensure accuracy. A total of 362 HCWs participated in the study, with 349 (96.41%) completing all measures and forming the analytical sample for this research. The final dataset included responses from these 349 HCWs, comprising 73 doctors, 177 nurses, and 32 other staff members.This study was conducted in strict accordance with the principles of the Declaration of Helsinki and received ethical approval from the Ethics Committees of China Medical University and China Medical University Shenyang Yongsen Hospital. Informed consent was obtained from each participant before initiating the survey. Crucially, participants were given the option to withdraw from the study at any time, without any requirement to provide a reason for their withdrawal. Measures Mental health measures :We included measures of key adverse mental health outcomes with prior evidence of strong psychometric properties:The Patient Health Questionnaire-9(PHQ-9) to assess depression symptoms(25–27); Genaralized Anxiety Disorder-7(GAD-7) to assess anxiety symptoms.Based on validation studies of each measure,we defined probable depression as PHQ-9 > = 10,probable GAD as GAD-7 > = 5.(28,29). Socio-ecological factors :We selected socio-ecological factors based on previous study on mental health outcomes of HCWs during pandemics.These factors can be classified to catrgories:individual ,interpersonal,institutional. Individual-level factors Individual-level factors included age,gender(male/female),married-status(married/single/divorce or widowed),minor-child number, education-level(associate/bachelor/master/doctor),whether living with parents(yes/no)personal resilence and adaptability. Personal resilence measured by Chinese version of the Connor-Davidson Resilience Scale. The CD-RISC is a 25-item scale using a 5-point Likert type response scale from not true at all (0) to true nearly all of the time(4), Participants rated each item with reference to the past month. Total scores range from 0 to 100, with higher scores corresponding to higher levels of resilience. The questionnaire has consistently exhibited robust reliability and validity across diverse populations and countries(Xie et al., 2016). The adaptability were measured by The Work-Life Adaptability Scale of Healthcare Workers Scale (WLASH) included 5 dimensions of readiness, work influence, life influence, worry, support, with a total of 21 item. This scale employs a 6-point Likert scale for each item, ranging from “Strongly Disagree” to “Strongly Agree”, with values assigned from 1 to 6 respectively (Items 1–5 and 19–23 are reverse-scored.) .A higher score indicates a greater degree of influence. These questionnaire has demonstrated strong reliability and validity(32). Interpersonal-level factors Interpersonal-level factors mainly included social_support and emotional labor. Social support, as measured by the Chinese version of the Perceived Social Support Scale (PSSS), comprises a 12-item scale with a seven-point rating (ranging from 1 = strongly disagree to 7 = strongly agree). The scale assesses three sources of support: Family, Friends, and Significant Others. This questionnaire has been utilized within the Chinese population and has exhibited high reliability (Chou, 2000). Emotionnal labor measured using the Emotional Labor Scale (ELS), a 14-item scale using a 5-point Likert type response scale from not true at all (1) to true nearly all of the time(5) self-reported questionnaire, Participants rated each item with reference to the past month.This study investigates emotional display in the workplace among hospital employees, focusing on surface acting, deep acting, and the expression of naturally felt emotions. Surface acting is assessed through seven items, while deep acting is measured with four items. Additionally, the expression of naturally felt emotions is gauged using three items. The Chinese version of the Emotional Labor Scale (ELS) has demonstrated robust factorial validity and reliability in China. (Wang et al., 2022). Institutional-level factors institutional-level factors included occupation(physician/nurse/others),working experiences(≤ 1 year/≤3 years/≤5 years/5 days when they were infected),income-level(≤ 4000/≤8000/>8000) and burnout. Burnout measured measured using the Maslach Burnout Inventory - Human Services Survey (MBI-HSS), is a self-report questionnaire comprising 22 items aimed at assessing burnout. This scale is structured into three dimensions: emotional exhaustion (EE), depersonalization (DP), and lack of personal accomplishment (PA). These dimensions are evaluated through specific items, with EE encompassing 9 items, DP including 5 items,and PA involving 8 items, collectively forming a total of 22 items.Each item assesses the frequency of the phenomenon’s occurrence, and responses are scored on a scale from 1 to 7, reflecting the frequency from “never happened” (scored as 1) to “happened every day”(scored as 7). For the EE and DP dimensions, higher scores indicate greater levels of burnout. Conversely, for the PA dimension, which employs a negative scoring method, lower scores signify a higher degree of burnout.The questionnaire has been translated into various languages and widely utilized, demonstrating robust reliability and validity(37). Statistical analysis Analyses were conducted using SPSS version 25,R version 3.4.1 and python 3.8. Categorical data were presented as frequencies and percentages. Data following a normal distribution were reported as means with standard errors (SE). After completing the descriptive analysis, we applied the random forest variable selection model to identify the key features of depression and anxiety. Finally, we utilized the RFC to build the predictive models for depression and anxiety and to gather detailed information about these models. The Random Forest (RF) algorithm is an integrated model designed to address classification and regression problems. It stands out for its ability to apply various models for evaluating responses. Compared to other machine learning algorithms, such as neural networks and support vector machines, RF algorithms can efficiently handle both continuous and categorical data sets. In this paper, we employ the RFC to construct the prediction model. The RFC comprises numerous individual decision trees functioning collectively as an ensemble. Each tree within the random forest contributes to the prediction, with the class receiving the most votes determining the model's overall prediction. One key advantage of the RFC over a single model is its collaborative approach, where each tree classifier, akin to a team member, collectively contributes to the final prediction, often yielding better results than a single decision tree. The RFC is particularly effective for binary classification, capable of managing datasets where the number of variables surpasses the number of observations. It can also handle datasets with a mix of continuous and categorical predictors. Furthermore, the RFC exhibits strong resistance to noise, can process high-dimensional data without feature selection, and is capable of processing various types of data while also ranking the importance of variables(20,38,39). In both the depression and anxiety models, the data were randomly split into two sets: a training set comprising 70% of the sample, and a testing set consisting of the remaining 30%. Given the prevalence of depression and anxiety among HCWs, we employed the SMOTE to balance the 'normal' and 'abnormal' categories. The strategy of under-sampling the majority (normal) class and over-sampling the minority (abnormal) class is recognized as an effective method to enhance the sensitivity of classifiers in cases of imbalanced data. In this study, we focused on over-sampling the minority category (depression/anxiety), creating additional samples by randomly selecting and replicating existing samples from this group. Subsequently, this augmented dataset was utilized for both training and testing the model. In this study, we utilized the GridSearchCV method to optimize the parameters of the RFC, a technique aimed at preventing overfitting by pruning the decision tree and removing its terminal nodes (40). Results Participant characteristics A total of 349 HCWs participated in this study.The findings are reported as frequencies, with the prevalence of depression and anxiety being 28.37% and 33.52%, respectively. Table 1 details the descriptive outcomes of sociodemographic characteristics,including individual,interpersonal and institutional factors, and mental health factors.The average age of participants was 32.55 years (SD = 9.03).A significant majority (80.81%) identified as female, and 45.56% were unmarried. Furthermore, 82.52% of the respondents were frontline HCWs, of whom 28.65% were doctors and a substantial 61.89% were nurses. Regarding work experience, 37.25% of HCWs had more than 10 years in the field.About 57.72% reported working approximately 6–8 hours/day in the past week, while 24.93% did not have adequate rest time.Additionally, nearly half of the HCWs reported occurrences of overtime, including severe cases. Moreover, the scores and standard deviations for burnout, emotional labor, resilience, and adaptability were 63.68 (SD = 15.72), 39.38 (SD = 7.52), 84.49 (SD = 16.47), and 74.94 (SD = 14.47) respectively. Table 1 Descriptive outcomes of sociodemographic characteristics Mean Standard deviation % N Prevalence of probable mental health conditions Depression symptom With depression N/A N/A 28.37 99 Anxiety symptom With anxiety N/A N/A 33.52 117 Individual-level factors Age(years) 32.55 9.03 N/A 349 Genders Male N/A N/A 19.2 67 Female N/A N/A 80.81 349 Maried status Married N/A N/A 45.56 159 unmarried/widowed N/A N/A 54.44 190 Live with parents Yes N/A N/A 21.78 76 No N/A N/A 78.22 273 Minor child number 0 N/A N/A 59.6 208 1 OR 2 N/A N/A 40.4 141 education_level master/doctor N/A N/A 15.76 55 Bachelor N/A N/A 66.19 231 Associate N/A N/A 18.05 63 Whether frontline HCW Yes N/A N/A 82.52 288 No N/A N/A 17.48 61 Occupation Physician N/A N/A 28.65 100 Nurse N/A N/A 61.89 216 Others N/A N/A 9.46 33 Working experience(years) =10 N/A N/A 37.25 150 Average worktime last week(hours/day) [6,8] N/A N/A 57.02 199 [9,11] N/A N/A 32.95 115 ≥ 12 N/A N/A 10.03 35 Income(yuan) ≤ 4000 N/A N/A 32.09 112 [4000,8000] N/A N/A 52.72 184 ≥ 8000 N/A N/A 15.19 53 Have the sufficient rest time Yes N/A N/A 75.07 262 No N/A N/A 24.93 87 Burnout 63.68 15.72 N/A N/A Emotionnallabor 39.38 7.52 N/A N/A Resilence 84.49 16.47 N/A N/A Adaptability 74.94 14.47 N/A N/A Detecting potential predicators We applied the random forest model to identify key variables in the classification of depression and anxiety, treating each issue separately. Based on the selection results and taking into account findings from previous studies, the following variables were incorporated into the random forest models for both depression and anxiety: age, gender, occupation, years of working experience, average work hours in the past week, availability of sufficient rest time, burnout, adaptation, emotional labor, resilience, and social support. Testing prediction accuracy of potential predictors The entire sample was split into a training dataset and a test dataset for statistical analysis using the random forest classification algorithm. Figure 4 presents the confusion matrices for the test set, detailing depression and anxiety classification separately. In line with prior research, we calculated various metrics including Accuracy, Sensitivity, Specificity, Precision, and F1 Score.The accuracy of the random forest model is defined as the percentage of correct predictions made by the method Equations below were used to caculate the sensitivity,specificity,F1 score,presicion,and accuracy in confusion matrix. Sensitivity= \(\:\frac{True\:Positives}{\text{T}\text{r}\text{u}\text{e}\:\text{P}\text{o}\text{s}\text{i}\text{t}\text{i}\text{v}\text{e}\text{s}+\text{F}\text{a}\text{l}\text{s}\text{e}\:\text{N}\text{e}\text{g}\text{a}\text{t}\text{i}\text{v}\text{e}\text{s}}\) Specificity= \(\:\frac{True\:Negatives}{\text{T}\text{r}\text{u}\text{e}\:\text{N}\text{e}\text{g}\text{a}\text{t}\text{i}\text{v}\text{e}\text{s}+\text{F}\text{a}\text{l}\text{s}\text{e}\:\text{P}\text{o}\text{s}\text{i}\text{t}\text{i}\text{v}\text{e}\text{s}}\) F1 Score= \(\:\frac{2\text{*}\left(\text{P}\text{r}\text{e}\text{s}\text{i}\text{c}\text{i}\text{o}\text{n}\text{*}\text{R}\text{e}\text{c}\text{a}\text{l}\text{l}\right)}{Presicion+Recall}\) Presicion= \(\:\frac{True\:Positives}{\text{T}\text{r}\text{u}\text{e}\:\text{P}\text{o}\text{s}\text{i}\text{t}\text{i}\text{v}\text{e}\text{s}+\text{F}\text{a}\text{l}\text{s}\text{e}\:\text{P}\text{o}\text{s}\text{i}\text{t}\text{i}\text{v}\text{e}\text{s}}\) Accuracy= \(\:\frac{\text{T}\text{r}\text{u}\text{e}\:\text{P}\text{o}\text{s}\text{i}\text{t}\text{i}\text{v}\text{e}\text{s}+\text{T}\text{r}\text{u}\text{e}\:\text{N}\text{e}\text{g}\text{a}\text{t}\text{i}\text{v}\text{e}\text{s}}{Total\:Populations}\) Sensitivity, also known as "Recall," is defined as the proportion of true positives accurately identified by the model. In this study, it refers to the rate at which healthcare workers (HCWs) with depression and anxiety are correctly predicted. The sensitivity score for depression was found to be 0.84, while for anxiety, it was 0.88. Modification Table:Specificity is the true negative rate, refers to the propotion of true negatives that are correctly identified by the model.This study refers to the propotion of HCWs without depression and anxiety who are correctly predicted.The specificity score for depression is 0.83 and for anxiety is 0.51. Then, the classifier’s performance was tested by the 10-k cross-validated method, and the result was 0.77 for depression model and 0.79 for anxiety model. The receiver operating characteristic (ROC) curve typically features the sensitivity (true positive rate) on the Y-axis and the specificity (false positive rate) on the X-axis. The ideal point in the upper left corner of the graph indicates a zero false positive rate and a true positive rate of one. A larger area under the curve (AUC) generally denotes higher overall performance. The AUC for the depression model is 0.88 and for the anxiety model is 0.72. Figure 2 and Fig. 3 displays the respective areas under the curve for the depression and anxiety random forest prediction models. However, as indicated in Table 1 , HCWs with depression or anxiety represent approximately 20%-30% of all participants, highlighting clasimbalance. In such scenarios, accuracy alone is not a sufficient measure for the random forest prediction model. Therefore, the F1 score, which considers classification imbalance, becomes a more appropriate metric. This score is effective even with lower accuracy and is calculated as the harmonic mean of precision and recall. The F1 scores for the depression and anxiety models are 0.88 and 0.83, respectively. Feature importance In the development of the RFC model, it is crucial to incorporate local interpretable model-agnostic explanations, such as SHAP (SHapley Additive exPlanations) value calculations and evaluations, to elucidate the model’s data results. SHAP values assign an importance value to each feature for a specific prediction, aiding in the interpretation of complex model predictions(41).Key visualizations include summary plots and beeswarm plots. Summary plots offer a comprehensive view of feature importance by illustrating the average impact of each feature on the model’s output. Beeswarm plots provide an in-depth perspective by showing how the impacts of individual features vary across different predictions. These visualizations play a crucial role in making complex model dynamics understandable, thereby assisting in both validating the model's accuracy and generating hypotheses for further research. Figure 5 -Figure 8 show the importance of the features evaluated by each model in descending order. In this summary plot, the x-axis represents the mean absolute SHAP value (average impact on model output magnitude), which quantifies the average contribution of each feature to the model’s prediction. A higher value indicates a stronger influence on the predictive outcome.In beeswarm plots,red means the characteristic value is relatively high, and blue means that the characteristic value is relatively low. The more right the shap value is, the greater the positive contribution to the prediction of depression. In contrast, the more left, the smaller the shap value is, the greater the negative contribution to the prediction of depression or anxiety. If the shap value can distinguish between red and blue, it can be proved that their high or low values have different effects on the final results. From the Fig. 5 -Figure 8,we can known that the selected variables can predict the depression and anxiety with a high accuracy,while the variables has the different importance in depression and anxiety model seperately. From Fig. 5 and Fig. 6 , we can learn that the top 10 predictaors of depression for HCWs in descending order were:burnout,resilence,emotionallabor,adaptability,working experience( < = 1year),occpation is physician,social support,average worktime last week(9–11 hours/day),age(28–30 years),age(31–35 years old). From Fig. 7 and Fig. 8 , we can learn that the top 10 predictaors of anxiety for HCWs in descending order were:burnout,adaptability,emotional labor,age(31–35 years),average worktime last week(9–11 hours/day),resilence,occupation is physician,social support,working experience( < = 1 year),gender is female. Discussion The objectives of this study were to evaluate the levels of depression and anxiety among HCWs during a public health emergency and to identify which variables significantly contribute to these conditions. Understanding the predictors of such emotional responses is crucial for developing and implementing preventive programs aimed at supporting and enhancing clinical outcomes in this or similar populations. In our research, we employed RFC models to separately predict depression and anxiety in HCWs and utilized the SMOTE to address the common issue of imbalanced datasets in real-world scenarios. The prevalence rates of depression and anxiety among HCWs were found to be 28.97% and 33.52%, respectively.Additionally, the depression model's AUC, Sensitivity (Recall), and Precision were 0.88, 0.84, and 0.93, while the anxiety model's parameters were 0.72, 0.88, and 0.78. These results indicate that the models are effective in screening HCWs vulnerable to depression and anxiety. Key predictive variables identified include age, gender, occupation, working experience, sufficient rest availability,workload(average work hours last week),burnout,adaptability,emotional labor,resilience,and social support.By utilizing these limited variables,we achieved good performance in predicting depression and anxiety, providing an efficient method to screen for adverse mental health outcomes in similar conditions. The prevalence of depression and anxiety among the HCWs during this special period were 28.97%(PHQ ≥ 10),33.52%(GAD ≥ 5),respecitively.The prevalence rates of depression and anxiety observed in this study varied somewhat from those documented in other research studies.When using a cutoff point of PHQ-9 ≥ 5 to determine depression prevalence, the alarming rate among HCWs reaches 90.83%, significantly surpassing rates reported in other studies. Lai et al. reported prevalence of 50.4% (PHQ-9 ≥ 5) for depression, 44.6% (GAD-7 ≥ 5) for anxiety among HCWs during the from January 29, 2020, to February 3, 2020, in China(11),while Zhang et al. reported prevalence of 50.7% (PHQ-9 ≥ 5), 44.7% (GAD-7 ≥ 5),respectively among HCWs from January 29, 2020, to February 3, 2020, in China(13). In contrast to the observed prevalence of depression, the figure of 90.83% appears notably concerning.Although we turn up the cut-off of PHQ to 10,the prevalence is also high to 28.37%.In their study of female HCWs, Li et al. reported a prevalence of 14.2% (PHQ-9 ≥ 10) for depression, 25.2% (GAD-7 ≥ 8) for anxiety(42).And here are some unusual values have been reported by other studies(43–45).This variation is likely attributed to differences in time and social factors.However, in the unique circumstances of this pandemic, HCWs face a daunting challenge as they encounter a surge in patients without adequate protection from PPE, both in their work and social environments.The risk is further heightened by the unfortunate transmission of the virus among colleagues.Under these circumstances,HCWs are mandated to remain at their posts, probably exacerbating the risk of adverse mental health outcomes for them. Generally speaking,it is evident that HCWs are facing widespread mental health issues during public health emergencies.There is a substantial need for psychiatric support for HCWs, both throughout and following such crises.The high incidence of mental health problems identified in this study underscores the immediate and ongoing necessity for psychiatric support for HCWs. Several researchers have employed machine learning methods to predict depression or anxiety in various populations, achieving accuracies ranging from approximately 60–90%.Sau and Bhakta utilized machine learning techniques to screen for depression and anxiety in seafarers and found that these methods outperformed traditional approaches.In their research, they used 14 variables, including age, gender, BMI, marital status, among others, to construct the prediction model.This model was developed using various machine learning algorithms such as CatBoost, Random Forest, Logistic Regression, Naïve Bayes, and Support Vector Machine (SVM).The CatBoost achieve the best accuracy(82.6%),the ROC curve is 0.882,and the precision is 84.1%(46).Kessler et.al,using the baseline patient self-reports of 5877 English-speaking residents of the non-institutionalized civilian US household population ages 15–54 to bulid the predicting model of depression and pointed out that the ML methods has a higher accuracy.They used 913 variables into ML prediction model and 23 variables into logistics model to predicting depression and the AUC is 0.71–0.76.(47). Multiple kernels of support vector machine (SVM) were applied to classify the depression using the social media message and found out the precision,F-measure and the recall scores were the best compare to the other methods.They used features, user microblog text, user profile and user behaviors without the socio-demographic or clinical variabes to bulid the prediction model,and the precision is 75.56%,F1 score is 76.12%,recall is 76.69%.(48). In summary, machine learning (ML) methods offer greater flexibility and adaptability compared to conventional methods, and their performance tends to excel, particularly in large datasets. These predictive models often incorporate socio-demographic and other variables to identify the most influential factors. However, our research took a different approach by using the RFC to select the most critical variables for inclusion in the model, and it performed admirably as well. This approach may provide a novel method for constructing prediction models when working with limited datasets. In the real world, imbalanced datasets are a prevalent occurrence. As data availability continues to grow in various large-scale, complex, and networked systems such as surveillance, security, the Internet, finance, and healthcare, it becomes increasingly important to enhance our fundamental understanding of knowledge discovery and analysis, starting from raw data to facilitate decision-making processes. Technically, any dataset that displays an unequal distribution among its classes can be categorized as imbalanced. Nevertheless, the prevailing understanding within the community is that imbalanced data typically involve datasets characterized by significant, and in some instances, extreme imbalances(49). ML method also used to solve the problem,but it’s prone to over predict the majority class.This leads to a larger misclassification rate for the minority class, which in many real-world applications is the class of interest(50).And some researcher take the deep learning method to solve the problem(51). In the medical field, imbalanced data are exceptionally common, with the prevalence of some diseases reaching as low as 1 in 10,000 or even 1 in 100,000 cases. When attempting to create predictive models for such scenarios, traditional methods prove to be inadequate. Consequently, machine learning (ML) techniques coupled with the SMOTE emerge as reliable approaches to tackle these challenges. Despite the widespread occurrence of imbalanced data in the medical domain, it has not received the level of attention it deserves. Our study uncovered an intriguing finding worth discussing regarding both the depression and anxiety models.We utilized individual factors (adaptability/resilience), interpersonal factors (social support/emotional labor), and institutional factors (burnout/occupation) to construct the prediction model.Interestingly,while employing these identical features to forecast depression and anxiety symptoms,they exhibit distinct levels of significance. In both the depression and anxiety models, the paramount feature pertains to institutional factors, specifically burnout. Burnout manifests as a protracted reaction to chronic emotional and interpersonal stressors within the workplace, persistently prevalent among HCWs during the pandemic and prevalent globally(52), and burnout significantly impairs the personal and social functioning of HCWs, indicating its substantial correlation with their mental health.This finding aligns with prior research that has established a correlation between burnout and adverse mental health outcomes such as depression and anxiety(53–55).As burnout is the most important predicors of depression and anxiety,there is a need to create interventions that alleviate HCWs burnout within and outside the hospital,which warrants attention and support from policy makers.Our prediction model highlights the significance of institutional factors in influencing the mental health of HCWs.Institutions play a crucial role in both work and daily life, exerting a mixed effect on HCWs.On one hand, institutions often provide resources and support to alleviate work pressure and enhance job satisfaction.On the other hand,HCWs are frequently affected by institutional factors such as workload, unreasonable work arrangements, and inadequate rest.Our prediction model corroborates these observations.Given these findings, policymakers must not only provide robust support but also prioritize efforts to mitigate the negative impact of institutional factors in the workplace. Individual factors,including age, gender, education, and personal adaptability and resilience, also play significant roles in our prediction models.For the depression model,the second important variable is resilience,while for the anxiety model,the variable is adaptability.The researchers found that a high resilience score consistently correlates with lower levels of depression, anxiety, and other mental health issues, regardless of whether the individuals are adolescents or adults(56,57).Adaptability serves as a crucial mental resource, encompassing an individual's cognitive, behavioral, and emotional regulation or adjustment in situations characterized by change, novelty, and uncertainty(58). Adaptability has been linked to various mental health outcomes. Dyson and Renk reported a negative association between adaptability and levels of stress and depressive symptoms in students(59). Moreover, Martin demonstrated that adaptability served as a predictor of both positive and negative mental health outcomes among adolescents(60).In summary,it is crucial to support HCWs in improving their personal psychological resilience and adaptability.Additionally, individual factors such as age and gender also demonstrate their significance in the prediction model. The interpersonal factors considered in this study include emotional labor and social support. HCWs are frequently influenced by interactions with others and by broader societal dynamics(5,61–63).Hou and Zhang demonstrated that healthcare workers with high levels of social support are more likely to exhibit better mental health outcomes (64).On the other hand, Chinese scholars found that emotional labor significantly impacts the mental health of hotel employees in a longitudinal study (65).In a word,HCWs require ample support from society, both in material resources and emotional sustenance. However, there are some noteworthy observations to consider. In previous studies, some researchers have highlighted that frontline HCWs exhibit a stronger association with depression and anxiety compared to those who are not on the frontline(6,66), However, it's important to note that when using the random forest classifier to select important features, the factor of being on the frontline was not deemed significant and was not included in the models.This observation could be attributed to the notion that depression and anxiety related to public health emergencies extend beyond one's specific role or area.It underscores the significance of making support systems flexible and accessible to all HCWs, not just those on the frontlines.Two other noteworthy factors that emerged as significant for depression and anxiety were the average work time in the past week and whether HCWs enjoyed sufficient rest. HCWs often work longer hours compared to many other professions worldwide, which can have an adverse impact on their mental health. The absence of sufficient rest time also had a similar impact. This highlights the importance of administrators making reasonable adjustments to support the health and well-being of HCWs. Our research is subject to several acknowledged limitations.To begin with, the study adopts a cross-sectional design, restricting our ability to establish clear causality between variables.This limitation underscores the need for longitudinal research to provide a more nuanced understanding of the prevalence of these mental states within the unique context under examination.Secondly, the assessment of depression, anxiety, and other variables relies exclusively on self-reported questionnaires, lacking the inclusion of psychiatric interviews.While self-reported measures offer valuable insights, the absence of clinical interviews may limit the depth of our findings.Thirdly, our study's participants are exclusively drawn from a Shenyang's Class-A tertiary hospital ,the generalizability of our findings may be limited.To address this constraint, it is imperative to conduct further investigations involving a more diverse and representative sample of participants. Expanding the scope of participants will enhance the external validity of our findings and contribute to a more comprehensive understanding of the phenomena under study. Conclusion The AUC values of 0.86 for depression and 0.83 for anxiety demonstrate our models’ effectiveness in identifying these conditions among healthcare workers (HCWs). These findings highlight the complex interplay of individual, interpersonal, and institutional factors affecting HCWs’ adverse outcomes. The study underscores the necessity for policymakers and hospital administrators to prioritize HCWs’ well-being during public health emergencies, such as the COVID-19 pandemic, where their role is pivotal in protecting public health. Employing the RFC model alongside SMOTE techniques has yielded insights into predicting depression and anxiety in HCWs during such crises. Nonetheless, further refinement in data processing techniques is needed to enhance predictive accuracy. Our approach addresses the challenges of imbalanced datasets, contributing significantly to medical research. In sum, our research provides a robust framework for predicting mental health outcomes in HCWs amidst public health emergencies. By integrating diverse factors and applying innovative methodologies, we lay the groundwork for future research focused on improving HCWs’ welfare and reducing the burden of mental health issues in similar scenarios. Declarations Ethics approval and consent to participate Ethical Approval Statement: This study was approved by the Ethics Committee of Medical Ethics Committee of Shengjing Hospital, China Medical University, Shenyang Yongsen Hospital (Approval Number: 20230103). All participants provided written informed consent before participating in the study, and the research design was conducted in accordance with the principles of the Declaration of Helsinki and relevant ethical guidelines. Informed Consent: All participants were fully informed of the purpose, procedure, and potential risks of the study and signed written consent forms before participation. During the study, all data were anonymized to ensure the privacy of the participants. Additional Information: This study did not involve the collection or analysis of data from children or minors and posed no physical or psychological harm to the participants. Consent for publication Not applicable Data availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Conflict of interest The authors declare that they have no competing interests. Funding No funding support Author contribution JW: Conceptualization of this study, Methodology, Software. LF: Data curation, Writing - Original draft preparation. NM: Visualization, Investigation. CY: Software, Validation. FC: Writing - Review \& Editing. XH: Supervision. YS: Project administration. LZ: Conceptualization of this study, Methodology, Software. PY: Conceptualization of this study, Methodology, Software. KS: Conceptualization of this study, Methodology, Software. References Du Z, Wang Y, Bai Y, Wang L, Cowling BJ, Meyers LA. Estimate of COVID-19 Deaths, China, December 2022–February 2023. Emerg Infect Dis. 2023 Oct;29(10):2121–4. Hsieh KY, Kao WT, Li DJ, Lu WC, Tsai KY, Chen WJ, et al. Mental health in biological disasters: From SARS to COVID-19. Int J Soc Psychiatry. 2021 Aug 1;67(5):576–86. Maunder RG. Was SARS a mental health catastrophe? Gen Hosp Psychiatry. 2009;31(4):316–7. Maunder RG, Lancee WJ, Balderson KE, Bennett JP, Borgundvaag B, Evans S, et al. Long-term Psychological and Occupational Effects of Providing Hospital Healthcare during SARS Outbreak. Emerg Infect Dis. 2006 Dec;12(12):1924–32. E Alnazly, OM Khraisat, AM Al-Bashaireh, CL Bryant. Anxiety, depression, stress, fear and social support during COVID-19 pandemic among Jordanian healthcare workers. PLOS ONE. 2021 Mar 12;16(3). J Alonso, G Vilagut, P Mortier, M Ferrer, I Alayo, A Aragon-Pena, et al. Mental health impact of the first wave of COVID-19 pandemic on Spanish healthcare workers: A large cross-sectional survey. REVISTA DE PSIQUIATRIA Y SALUD MENTAL. 2021 Apr;14(2):90–105. Rachel Hennein, Emma J. Mew, Sarah R. Lowe. Socio-ecological predictors of mental health outcomes among healthcare workers during the COVID-19 pandemic in the United States. Joel Msafiri Francis, editor. PLoS ONE. 2021 Feb 5;16(2):e0246602. Yeen Huang, Ning Zhao. Generalized anxiety disorder, depressive symptoms and sleep quality during COVID-19 outbreak in China: a web-based cross-sectional survey. Psychiatry Research. 2020 Jun;288:112954. Yunus Hacimusalar, Aybeniz Civan Kahve, Alisan Burak Yasar, Mehmet Sinan Aydin. Anxiety and hopelessness levels in COVID-19 pandemic: A comparative study of healthcare professionals and other community sample in Turkey. Journal of Psychiatric Research. 2020 Oct;129:181–8. Zhou Zhu, Shabei Xu, Hui Wang, Zheng Liu, Jianhong Wu, Guo Li, et al. COVID-19 in Wuhan: Sociodemographic characteristics and hospital support measures associated with the immediate psychological impact on healthcare workers. eClinicalMedicine. 2020 Jul;24:100443. Lai J, Ma S, Wang Y, Cai Z, Hu J, Wei N, et al. Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019. JAMA NETWORK OPEN. 2020 Mar 23;3(3). Mustafa Kürşat Şahin, Servet Aker, Gülay Şahin, Aytül Karabekiroğlu. Prevalence of Depression, Anxiety, Distress and Insomnia and Related Factors in Healthcare Workers During COVID-19 Pandemic in Turkey. J Community Health. 2020 Dec;45(6):1168–77. Zhang C, Yang L, Liu S, Ma S, Wang Y, Cai Z, et al. Survey of Insomnia and Related Social Psychological Factors Among Medical Staff Involved in the 2019 Novel Coronavirus Disease Outbreak. Frontiers in Psychiatry [Internet]. 2020 [cited 2023 Nov 21];11. Available from: https://www.frontiersin.org/articles/10.3389/fpsyt.2020.00306 An Y, Yang Y, Wang A, Li Y, Zhang Q, Cheung T, et al. Prevalence of depression and its impact on quality of life among frontline nurses in emergency departments during the COVID-19 outbreak. JOURNAL OF AFFECTIVE DISORDERS. 2020 Nov 1;276:312–5. Arafa A, Mohammed Z, Mahmoud O, Elshazley M, Ewis A. Depressed, anxious, and stressed: What have healthcare workers on the frontlines in Egypt and Saudi Arabia experienced during the COVID-19 pandemic? JOURNAL OF AFFECTIVE DISORDERS. 2021 Jan 1;278:365–71. Allahyari E. Predicting Elderly Depression: An Artificial Neural Network Model. Iranian Journal of Psychiatry and Behavioral Sciences [Internet]. 2020 Jan 5 [cited 2023 Nov 29];13(4). Available from: https://brieflands.com/articles/ijpbs-98497.html Chang YS, Hung WC, Juang TY. Depression Diagnosis Based on Ontologies and Bayesian Networks. In: 2013 IEEE International Conference on Systems, Man, and Cybernetics [Internet]. 2013 [cited 2023 Nov 28]. p. 3452–7. Available from: https://ieeexplore.ieee.org/abstract/document/6722342 Priya A, Garg S, Tigga NP. Predicting Anxiety, Depression and Stress in Modern Life using Machine Learning Algorithms. Procedia Computer Science. 2020 Jan 1;167:1258–67. Thanathamathee P. Boosting with feature selection technique for screening and predicting adolescents depression. In: 2014 Fourth International Conference on Digital Information and Communication Technology and its Applications (DICTAP) [Internet]. 2014 [cited 2023 Nov 28]. p. 23–7. Available from: https://ieeexplore.ieee.org/abstract/document/6821650 Xin Y, Ren X. Predicting depression among rural and urban disabled elderly in China using a random forest classifier. BMC Psychiatry. 2022 Feb 15;22(1):118. Khalilia M, Chakraborty S, Popescu M. Predicting disease risks from highly imbalanced data using random forest. BMC Med Inform Decis Mak. 2011 Jul 29;11(1):51. Chawla NV, Bowyer KW, Hall LO, Kegelmeyer WP. SMOTE: Synthetic Minority Over-sampling Technique. Journal of Artificial Intelligence Research. 2002 Jun 1;16:321–57. Fernandez A, Garcia S, Herrera F, Chawla NV. SMOTE for Learning from Imbalanced Data: Progress and Challenges, Marking the 15-year Anniversary. Journal of Artificial Intelligence Research. 2018 Apr 20;61:863–905. Thanathamathee P, Lursinsap C. Handling imbalanced data sets with synthetic boundary data generation using bootstrap re-sampling and AdaBoost techniques. Pattern Recognition Letters. 2013 Sep 1;34(12):1339–47. Chen S, Chiu H, Xu B, Ma Y, Jin T, Wu M, et al. Reliability and validity of the PHQ-9 for screening late-life depression in Chinese primary care. Int J Geriat Psychiatry. 2010 Nov;25(11):1127–33. Kroenke K, Spitzer RL, Williams JBW. The PHQ-9: Validity of a brief depression severity measure. J Gen Intern Med. 2001 Sep;16(9):606–13. Mata DA, Ramos MA, Bansal N, Khan R, Guille C, Di Angelantonio E, et al. Prevalence of Depression and Depressive Symptoms Among Resident Physicians: A Systematic Review and Meta-analysis. Jama. 2015 Dec 8;314(22):2373–83. Spitzer RL, Kroenke K, Williams JBW, Löwe B. A Brief Measure for Assessing Generalized Anxiety Disorder: The GAD-7. Archives of Internal Medicine. 2006 May 22;166(10):1092–7. Zhang C, Wang T, Zeng P, Zhao M, Zhang G, Zhai S, et al. Reliability, Validity, and Measurement Invariance of the General Anxiety Disorder Scale Among Chinese Medical University Students. Front Psychiatry [Internet]. 2021 [cited 2023 Dec 13];12. Available from: https://www.frontiersin.org/articles/10.3389/fpsyt.2021.648755 Connor KM, Davidson JRT. Development of a new resilience scale: The Connor-Davidson Resilience Scale (CD-RISC). Depress Anxiety. 2003;18(2):76–82. Xie Y, Peng L, Zuo X, Li M. The Psychometric Evaluation of the Connor-Davidson Resilience Scale Using a Chinese Military Sample. PLOS ONE. 2016 Feb 9;11(2):e0148843. Lin GL, Guo LH, Lin K, Zhuang JB, Zou C, Chen CW, et al. Development and validation of the Work-Life Adaptability Scale for healthcare workers during the COVID-19 pandemic. J Shantou Univ Med Coll. 2022 Jun;35(2):106–11. Blumenthal JA, Burg MM, Barefoot J, Williams RB, Haney T, Zimet G. Social support, Type A behavior, and coronary artery disease. Psychosomatic Medicine. 1987;49(4):331–40. Chou KL. Assessing Chinese adolescents’ social support: the multidimensional scale of perceived social support. Pers Indiv Differ. 2000 Feb;28(2):299–307. Diefendorff JM, Croyle MH, Gosserand RH. The dimensionality and antecedents of emotional labor strategies. J Vocat Behav. 2005 Apr;66(2):339–57. Wang M, Wu F, Wu P, Wang N. Validation of Chinese Version Emotional Labor Scale for Social Workers in Mainland China. Research on Social Work Practice. 2022 Sep 1;32(6):678–86. Lin CY, Alimoradi Z, Griffiths MD, Pakpour AH. Psychometric properties of the Maslach Burnout Inventory for Medical Personnel (MBI-HSS-MP). Heliyon. 2022 Feb;8(2):e08868. Gokten ES, Uyulan C. Prediction of the development of depression and post-traumatic stress disorder in sexually abused children using a random forest classifier. J Affect Disord. 2021 Jan 15;279:256–65. Rigatti SJ. Random Forest. J Insur Med. 2017 Jan 1;47(1):31–9. Probst P, Wright MN, Boulesteix AL. Hyperparameters and tuning strategies for random forest. Wires Data Min Knowl Discovery. 2019;9(3):e1301. Lundberg SM, Lee SI. A Unified Approach to Interpreting Model Predictions. In: Advances in Neural Information Processing Systems [Internet]. Curran Associates, Inc.; 2017 [cited 2023 Dec 28]. Available from: https://proceedings.neurips.cc/paper/2017/hash/8a20a8621978632d76c43dfd28b67767-Abstract.html Li G, Miao J, Wang H, Xu S, Sun W, Fan Y, et al. Psychological impact on women health workers involved in COVID-19 outbreak in Wuhan: a cross-sectional study. Journal of Neurology, Neurosurgery & Psychiatry. 2020 Aug;91(8):895–7. Yin Q, Sun Z, Liu T, Ni X, Deng X, Jia Y, et al. Posttraumatic stress symptoms of health care workers during the corona virus disease 2019. CLINICAL PSYCHOLOGY & PSYCHOTHERAPY. 2020 May;27(3):384–95. Zhang W rui, Wang K, Yin L, Zhao W feng, Xue Q, Peng M, et al. Mental Health and Psychosocial Problems of Medical Health Workers during the COVID-19 Epidemic in China. Psychother Psychosom. 2020;89(4):242–50. Zhou Y, Zhou Y, Song Y, Ren L, Ng CH, Xiang YT, et al. Tackling the mental health burden of frontline healthcare staff in the COVID-19 pandemic: China’s experiences. Psychol Med. 2021 Aug;51(11):1955–6. Sau A, Bhakta I. Screening of anxiety and depression among seafarers using machine learning technology. Informatics in Medicine Unlocked. 2019;16:100228. Kessler RC, Van Loo HM, Wardenaar KJ, Bossarte RM, Brenner LA, Cai T, et al. Testing a machine-learning algorithm to predict the persistence and severity of major depressive disorder from baseline self-reports. Mol Psychiatry. 2016 Oct;21(10):1366–71. Peng Z, Hu Q, Dang J. Multi-kernel SVM based depression recognition using social media data. International Journal of Machine Learning and Cybernetics. 2019 Jan 1;10(1):43–57. Haibo He, Garcia EA. Learning from Imbalanced Data. IEEE Trans Knowl Data Eng. 2009 Sep;21(9):1263–84. Esposito C, Landrum GA, Schneider N, Stiefl N, Riniker S. GHOST: Adjusting the Decision Threshold to Handle Imbalanced Data in Machine Learning. J Chem Inf Model. 2021 Jun 28;61(6):2623–40. Lee J, Park K. GAN-based imbalanced data intrusion detection system. Pers Ubiquit Comput. 2021 Feb;25(1):121–8. Zheng Y, Tang PK, Lin G, Liu J, Hu H, Wu AMS, et al. Burnout among healthcare providers: Its prevalence and association with anxiety and depression during the COVID-19 pandemic in Macao, China. PLOS ONE. 2023 Mar 16;18(3):e0283239. Barello S, Palamenghi L, Graffigna G. Burnout and somatic symptoms among frontline healthcare professionals at the peak of the Italian COVID-19 pandemic. PSYCHIATRY RESEARCH. 2020 Aug;290. Dobson H, Malpas CB, Burrell AJC, Gurvich C, Chen L, Kulkarni J, et al. Burnout and psychological distress amongst Australian healthcare workers during the COVID-19 pandemic. AUSTRALASIAN PSYCHIATRY. 2021 Feb;29(1):26–30. Duarte I, Teixeira A, Castro L, Marina S, Ribeiro C, Jacome C, et al. Burnout among Portuguese healthcare workers during the COVID-19 pandemic. BMC PUBLIC HEALTH. 2020 Dec 7;20(1). Hjemdal O, Vogel PA, Solem S, Hagen K, Stiles TC. The relationship between resilience and levels of anxiety, depression, and obsessive–compulsive symptoms in adolescents. Clin Psychol Psychot. 2011 Jul;18(4):314–21. Ran L, Wang W, Ai M, Kong Y, Chen J, Kuang L. Psychological resilience, depression, anxiety, and somatization symptoms in response to COVID-19: A study of the general population in China at the peak of its epidemic. Soc Sci Med. 2020 Oct;262:113261. Waldeck D, Pancani L, Holliman A, Karekla M, Tyndall I. Adaptability and psychological flexibility: Overlapping constructs? Journal of Contextual Behavioral Science. 2021 Jan 1;19:72–8. Dyson R, Renk K. Freshmen adaptation to university life: Depressive symptoms, stress, and coping. J Clin Psychol. 2006 Oct;62(10):1231–44. Martin AJ, Nejad HG, Colmar S, Liem GAD. Adaptability: How students’ responses to uncertainty and novelty predict their academic and non-academic outcomes. J Educ Psychol. 2013 Aug;105(3):728–46. Ohrnberger J, Fichera E, Sutton M. The relationship between physical and mental health: A mediation analysis. Soc Sci Med. 2017 Dec;195:42–9. Yoon SL, Kim JH. Job-Related Stress, Emotional Labor, and Depressive Symptoms Among Korean Nurses. J Nurs Scholarsh. 2013;45(2):169–76. Zhang Y, ElGhaziri M, Siddique S, Gore R, Kurowski A, Nobrega S, et al. Emotional Labor and Depressive Symptoms Among Healthcare Workers: The Role of Sleep. Workplace Health & Safety. 2021 Aug;69(8):383–93. Hou T, Zhang T, Cai W, Song X, Chen A, Deng G, et al. Social support and mental health among health care workers during Coronavirus Disease 2019 outbreak: A moderated mediation model. PLOS ONE. 2020 May 29;15(5). Xiong W, Huang M, Okumus B, Leung XY, Cai X, Fan F. How emotional labor affect hotel employees’ mental health: A longitudinal study. Tourism Management. 2023 Feb 1;94:104631. James Gilleen, Aida Santaolalla, Lorena Valdearenas, Clara Salice, Montserrat Fusté. Impact of the COVID-19 pandemic on the mental health and well-being of UK healthcare workers. BJPsych open. 2021 May;7(3):e88. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 06 Aug, 2025 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 26 Jun, 2025 Reviews received at journal 25 Jun, 2025 Reviews received at journal 25 Jun, 2025 Reviewers agreed at journal 16 Jun, 2025 Reviewers agreed at journal 16 Jun, 2025 Reviewers invited by journal 05 Nov, 2024 Editor assigned by journal 05 Nov, 2024 Editor invited by journal 04 Nov, 2024 Submission checks completed at journal 01 Nov, 2024 First submitted to journal 08 Oct, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5228634","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":375462453,"identity":"1e4b65f7-e43b-46c4-bd28-0d2bee2952a4","order_by":0,"name":"Jiana Wang","email":"","orcid":"","institution":"Ningbo University","correspondingAuthor":false,"prefix":"","firstName":"Jiana","middleName":"","lastName":"Wang","suffix":""},{"id":375462454,"identity":"33aa3687-abdf-446b-8177-db05e89200c6","order_by":1,"name":"Lin Feng","email":"","orcid":"","institution":"China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lin","middleName":"","lastName":"Feng","suffix":""},{"id":375462455,"identity":"71aa0b63-0f7f-40e4-bc2c-4cbaaafb8870","order_by":2,"name":"Nana Meng","email":"","orcid":"","institution":"China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Nana","middleName":"","lastName":"Meng","suffix":""},{"id":375462456,"identity":"31956fff-fa21-4f77-a09a-064a21cb79a1","order_by":3,"name":"Cong Yang","email":"","orcid":"","institution":"China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Cong","middleName":"","lastName":"Yang","suffix":""},{"id":375462457,"identity":"7dfa0f7c-7b7e-4767-aae7-0a15a7a0f3a5","order_by":4,"name":"Fanfan Cai","email":"","orcid":"","institution":"China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Fanfan","middleName":"","lastName":"Cai","suffix":""},{"id":375462458,"identity":"1e677f72-475c-4092-8f2d-362e51d2a116","order_by":5,"name":"Xin Huang","email":"","orcid":"","institution":"China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xin","middleName":"","lastName":"Huang","suffix":""},{"id":375462459,"identity":"7bf5c180-adac-4ebe-96a2-feb69df2c0b3","order_by":6,"name":"Yihang Sun","email":"","orcid":"","institution":"China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yihang","middleName":"","lastName":"Sun","suffix":""},{"id":375462460,"identity":"b923c558-c6be-4883-ad5d-3469eb0b3d57","order_by":7,"name":"Lu Zhang","email":"","orcid":"","institution":"China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lu","middleName":"","lastName":"Zhang","suffix":""},{"id":375462461,"identity":"a160401c-ce46-4c3c-a106-c1882ea1b0af","order_by":8,"name":"Pin Yao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAElEQVRIiWNgGAWjYLCCBAYbBsYGHhBTgoGBnTgtaXAtEgzMxNlzGIjBWhgIazE4fvjYg4c7ztszN/AefPhzh0UdfzMD4+OKX3i0nElLN0g8czuxsYEv2Zj3jISExGEGZsOzfXi0HMgxk0hsu53AOP+NmTRjG9AvhxnYJBt78Gg5/wak5Zw90Ptmkj+BWuQJarkBtuUAYyNQiwQvUIsBSEvDD9xaJG88A/qlLRnoFx5jkF8kNx5mbDZsbMCthe988rGHP9vs7A0beAyBIVbHL3e8+eDDhj+4tSgcYGADMwxB5jKCDQeSjG24tcg3QLXIM8C1gAAeW0bBKBgFo2DEAQDo5VB5h2gitQAAAABJRU5ErkJggg==","orcid":"","institution":"China Medical University","correspondingAuthor":true,"prefix":"","firstName":"Pin","middleName":"","lastName":"Yao","suffix":""},{"id":375462462,"identity":"6e26a248-c788-4373-9368-628ca24538a7","order_by":9,"name":"Kristin K Sznajder","email":"","orcid":"","institution":"Pennsylvania State University","correspondingAuthor":false,"prefix":"","firstName":"Kristin","middleName":"K","lastName":"Sznajder","suffix":""}],"badges":[],"createdAt":"2024-10-09 02:38:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5228634/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5228634/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-025-14403-3","type":"published","date":"2025-08-06T15:57:37+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":70039429,"identity":"7e954299-10d0-4785-a797-c8335fbbd1d2","added_by":"auto","created_at":"2024-11-27 17:46:18","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":119987,"visible":true,"origin":"","legend":"\u003cp\u003eThe infected numbers\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5228634/v1/30e046e0001c44893866fa53.png"},{"id":70039427,"identity":"c68977c8-af5b-4f79-a36f-226f390371ef","added_by":"auto","created_at":"2024-11-27 17:46:17","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":60716,"visible":true,"origin":"","legend":"\u003cp\u003eDepression ROC curve\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5228634/v1/50f09ed11ea1de6746a938ad.png"},{"id":70040313,"identity":"a9886bf9-647a-4d24-86f3-9123dfdce4ab","added_by":"auto","created_at":"2024-11-27 17:54:17","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":58886,"visible":true,"origin":"","legend":"\u003cp\u003eAnxiety ROC curve\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5228634/v1/c3b5c16f2f43bdc01b3e8d79.png"},{"id":70039433,"identity":"dc328793-a9b2-4f05-add5-5f253847c248","added_by":"auto","created_at":"2024-11-27 17:46:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":27432,"visible":true,"origin":"","legend":"\u003cp\u003edepression and aniety confusion matrix\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5228634/v1/ab5a4d5024c56000dad511db.png"},{"id":70039435,"identity":"29c5d991-65bc-4f64-8c90-ed8c3de72b42","added_by":"auto","created_at":"2024-11-27 17:46:18","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":36249,"visible":true,"origin":"","legend":"\u003cp\u003eMean SHAP value of depression\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-5228634/v1/502103574639868ad51d2ae3.png"},{"id":70039434,"identity":"06af42d2-362e-49e7-aa3c-04322ae5aa0f","added_by":"auto","created_at":"2024-11-27 17:46:18","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":65886,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP value of depression\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-5228634/v1/5313ed7a21b525d89ce4ee7b.png"},{"id":70040314,"identity":"200252a6-2a3e-485e-a9e3-ff5bcbe88191","added_by":"auto","created_at":"2024-11-27 17:54:18","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":38877,"visible":true,"origin":"","legend":"\u003cp\u003eMean SHAP value of anxiety\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-5228634/v1/7a343ec183848dda057e0438.png"},{"id":70039431,"identity":"e57163b3-57af-4e20-b3a0-653be0e0e78f","added_by":"auto","created_at":"2024-11-27 17:46:18","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":65195,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP value of anxity\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-5228634/v1/a3e590e673e75916048b7567.png"},{"id":88814811,"identity":"bed81225-beb0-468d-9d2e-1b8bba889973","added_by":"auto","created_at":"2025-08-11 16:10:03","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1321833,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5228634/v1/2fab4f2a-9171-48bd-933a-73d675d90ca7.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Balancing Mental Health: Predictive Modeling for Healthcare Workers During Public Health Crises","fulltext":[{"header":"Background","content":"\u003cp\u003eAccording to the report from the Chinese Center for Disease Control and Prevention, the peak of infections was reached on December 22, 2022, with a positive test count of 6.94\u0026nbsp;million. Since China announced a slight relaxation of its zero-COVID policies on November 11, 2021, followed by a significant easing on December 7, 2022, over 1.41\u0026nbsp;million people have been infected and lost their lives in the initial wave that ensued(1). The spread of the virus has dramatically changed daily life and the socio-economic environment in affected areas. According to previous studies, these changes are expected to have long-term effects on mental health(2\u0026ndash;4).\u003c/p\u003e \u003cp\u003eFollowing the announcement of a major relaxation on December 7, 2022,there has been a notable surge in infections (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).The rapid spread of COVID-19 throughout China has significantly strained HCWs who are directly or indirectly engaged in combating the pandemic,potentially increasing their risk of adverse mental health outcomes.Notably, HCWs are consistently exposed, both directly and indirectly, to patients and colleagues infected with COVID-19,which elevates their risk of developing psychological disorders(5\u0026ndash;10).According to recent research, the prevalence of depression and anxiety among HCWs falls within the ranges of 45.6%-77.6% and 20.7%-60.2%,respectively(11\u0026ndash;13).HCWs may exhibit poor performance in both their professional and personal lives if they show symptoms of depression or anxiety, and the effects of these symptoms can be long-lasting.Therefore, early identification of individuals suffering from depression and anxiety is crucial to intervene promptly and minimize the impact on HCWs' health by potentially reducing the progression of the disease.However, the application of machine learning to predict adverse mental health outcomes in HCWs is currently limited.Moreover, our survey occurred at the outset of the routine epidemic prevention and control measures for COVID-19. During this period,HCWs had to confront a substantial influx of infected patients within a condensed time frame,necessitating them to bear heightened social expectations and work pressure. Surprisingly,there is scant research dedicated to investigating the psychological well-being of healthcare personnel during this specific phase.\u003c/p\u003e \u003cp\u003eThe mental health challenges encountered by HCWs during the COVID-19 pandemic have garnered considerable attention from researchers. Predominantly, depression, anxiety, and post-traumatic stress disorder have been the focal points in studies addressing mental health issues in this context. Notably, depression and anxiety are frequently assessed through cross-sectional surveys, which provide valuable data for statistical analysis.The research approaches in this domain have been significantly influenced by the social-ecological model and ecological models of health behavior. Factors influencing mental health outcomes are categorized within the socio-ecological framework and encompass various aspects: Individual Factors: These encompass age, gender, ethnicity, race, marital status, technical title, residence (urban/rural), and other individual characteristics; Interpersonal Factors: This category includes social support, emotional labor, and other interpersonal dynamics;Institutional Factors: These pertain to work experiences, workload, occupation, burnout, and issues such as shortages of personal protective equipment (PPE).(6,14,15).In summary, we selected the important varibles from these aspects to bulid the prediction models.\u003c/p\u003e \u003cp\u003eEarly detection and diagnosis is paramount to understand and address mental health illness on a population level. Machine learning is increasingly being utilized in the field of mental health. Research focused on the prediction and analysis of risk factors for depression and anxiety using machine learning techniques has been conducted and displayed a reliable performance(7,16\u0026ndash;20).However, according to our research, few studies have developed a machine learning model specifically for predicting depression and anxiety in HCWs during such critical periods.Furthermore, there is a notable lack of research focused on the issue of imbalanced datasets within the realm of mental health. Imbalanced dataset problems are prevalent in real-world scenarios, particularly in the fields of medicine and psychology.The proportion of HCWs diagnosed with depression or anxiety is relatively small in comparison to the total number of HCWs.It\u0026rsquo;s presumed that the data are drawn from the same distribution as the training dataset,presenting imbalanced dataset to the classifier and producing intolerable results(21).In response to this issue, sampling techniques and the creation of synthetic data are employed to attain a balanced distribution of classes, either by augmenting or reducing samples.Specifically, naive random over-sampling involves generating new samples by randomly selecting and duplicating existing samples from the minority class. In terms of synthetic sampling, researchers have introduced the SMOTE(22,23).SMOTE generates synthetic data for the minority class by identifying some of the nearest neighbors within the minority dataset. It then creates synthetic minority data points along the lines connecting the minority data points to their nearest neighbors(24).\u003c/p\u003e \u003cp\u003ePrevious researchers have not utilized the SMOTE method to construct a machine learning model for predicting adverse mental health outcomes in HCWs in real-world situations.The typical approach to address imbalanced data involves increasing the sample size or reducing variables. However, these methods are often challenging to implement and may yield inaccurate results. Therefore, there is a need for innovative methods in our field to address this issue.To address this gap, we conducted a survey involving 349 HCWs from a Shenyang's Class-A tertiary hospital.This survey focused on socio-ecological factors associated with depression and anxiety during a critical period. By applying the Random Forest Classifier and incorporating the SMOTE method, we aimed to predict depression and anxiety among HCWs for the purpose to offer valuable insights for developing interventions to support HCWs in outbreak conditions and provide a potential solution for addressing the commonly encountered issue of imbalanced data in real-world scenarios.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy design and participants\u003c/h2\u003e \u003cp\u003eThis study used a single-center, cross-sectional survey design, involving doctors, nurses, and clinical technicians currently working at a Class-A tertiary hospital in the city of Shenyang, Liaoning Province in China.Additionally, individuals in administrative and financial positions were also included.The survey was conducted during the first and second weeks of February 2023, approximately after the government declared the normalization of the COVID-19 pandemic on December 7, 2022.Two researchers entered data into the database double-blindly using Epidata.3.1 to ensure accuracy.\u003c/p\u003e \u003cp\u003eA total of 362 HCWs participated in the study, with 349 (96.41%) completing all measures and forming the analytical sample for this research. The final dataset included responses from these 349 HCWs, comprising 73 doctors, 177 nurses, and 32 other staff members.This study was conducted in strict accordance with the principles of the Declaration of Helsinki and received ethical approval from the Ethics Committees of China Medical University and China Medical University Shenyang Yongsen Hospital. Informed consent was obtained from each participant before initiating the survey. Crucially, participants were given the option to withdraw from the study at any time, without any requirement to provide a reason for their withdrawal.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eMeasures\u003c/h3\u003e\n\u003cp\u003e \u003cb\u003eMental health measures\u003c/b\u003e:We included measures of key adverse mental health outcomes with prior evidence of strong psychometric properties:The Patient Health Questionnaire-9(PHQ-9) to assess depression symptoms(25\u0026ndash;27);\u003c/p\u003e \u003cp\u003eGenaralized Anxiety Disorder-7(GAD-7) to assess anxiety symptoms.Based on validation studies of each measure,we defined probable depression as PHQ-9\u0026thinsp;\u0026gt;\u0026thinsp;=\u0026thinsp;10,probable GAD as GAD-7\u0026thinsp;\u0026gt;\u0026thinsp;=\u0026thinsp;5.(28,29).\u003c/p\u003e \u003cp\u003e \u003cb\u003eSocio-ecological factors\u003c/b\u003e:We selected socio-ecological factors based on previous study on mental health outcomes of HCWs during pandemics.These factors can be classified to catrgories:individual ,interpersonal,institutional.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eIndividual-level factors\u003c/strong\u003e \u003cp\u003eIndividual-level factors included age,gender(male/female),married-status(married/single/divorce or widowed),minor-child number, education-level(associate/bachelor/master/doctor),whether living with parents(yes/no)personal resilence and adaptability.\u003c/p\u003e \u003c/p\u003e \u003cp\u003ePersonal resilence measured by Chinese version of the Connor-Davidson Resilience Scale. The CD-RISC is a 25-item scale using a 5-point Likert type response scale from not true at all (0) to true nearly all of the time(4), Participants rated each item with reference to the past month. Total scores range from 0 to 100, with higher scores corresponding to higher levels of resilience. The questionnaire has consistently exhibited robust reliability and validity across diverse populations and countries(Xie et al., 2016).\u003c/p\u003e \u003cp\u003eThe adaptability were measured by The Work-Life Adaptability Scale of Healthcare Workers Scale (WLASH) included 5 dimensions of readiness, work influence, life influence, worry, support, with a total of 21 item. This scale employs a 6-point Likert scale for each item, ranging from \u0026ldquo;Strongly Disagree\u0026rdquo; to \u0026ldquo;Strongly Agree\u0026rdquo;, with values assigned from 1 to 6 respectively (Items 1\u0026ndash;5 and 19\u0026ndash;23 are reverse-scored.) .A higher score indicates a greater degree of influence. These questionnaire has demonstrated strong reliability and validity(32).\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eInterpersonal-level factors\u003c/strong\u003e \u003cp\u003eInterpersonal-level factors mainly included social_support and emotional labor.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eSocial support, as measured by the Chinese version of the Perceived Social Support Scale (PSSS), comprises a 12-item scale with a seven-point rating (ranging from 1\u0026thinsp;=\u0026thinsp;strongly disagree to 7\u0026thinsp;=\u0026thinsp;strongly agree). The scale assesses three sources of support: Family, Friends, and Significant Others. This questionnaire has been utilized within the Chinese population and has exhibited high reliability (Chou, 2000).\u003c/p\u003e \u003cp\u003eEmotionnal labor measured using the Emotional Labor Scale (ELS), a 14-item scale using a 5-point Likert type response scale from not true at all (1) to true nearly all of the time(5) self-reported questionnaire, Participants rated each item with reference to the past month.This study investigates emotional display in the workplace among hospital employees, focusing on surface acting, deep acting, and the expression of naturally felt emotions. Surface acting is assessed through seven items, while deep acting is measured with four items. Additionally, the expression of naturally felt emotions is gauged using three items. The Chinese version of the Emotional Labor Scale (ELS) has demonstrated robust factorial validity and reliability in China. (Wang et al., 2022).\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eInstitutional-level factors\u003c/strong\u003e \u003cp\u003einstitutional-level factors included occupation(physician/nurse/others),working experiences(\u0026le;\u0026thinsp;1 year/\u0026le;3 years/\u0026le;5 years/\u0026lt; 10 years/\u0026ge;10 years),frontline status(yes/no),workload(6\u0026ndash;8 hours/9\u0026ndash;11 hours/\u0026ge;12 hours in the past week),whether enjoying sufficient rest(\u0026le;\u0026thinsp;5 days/\u0026gt;5 days when they were infected),income-level(\u0026le;\u0026thinsp;4000/\u0026le;8000/\u0026gt;8000) and burnout.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eBurnout measured measured using the Maslach Burnout Inventory - Human Services Survey (MBI-HSS), is a self-report questionnaire comprising 22 items aimed at assessing burnout. This scale is structured into three dimensions: emotional exhaustion (EE), depersonalization (DP), and lack of personal accomplishment (PA). These dimensions are evaluated through specific items, with EE encompassing 9 items, DP including 5 items,and PA involving 8 items, collectively forming a total of 22 items.Each item assesses the frequency of the phenomenon\u0026rsquo;s occurrence, and responses are scored on a scale from 1 to 7, reflecting the frequency from \u0026ldquo;never happened\u0026rdquo; (scored as 1) to \u0026ldquo;happened every day\u0026rdquo;(scored as 7). For the EE and DP dimensions, higher scores indicate greater levels of burnout. Conversely, for the PA dimension, which employs a negative scoring method, lower scores signify a higher degree of burnout.The questionnaire has been translated into various languages and widely utilized, demonstrating robust reliability and validity(37).\u003c/p\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eAnalyses were conducted using SPSS version 25,R version 3.4.1 and python 3.8. Categorical data were presented as frequencies and percentages. Data following a normal distribution were reported as means with standard errors (SE). After completing the descriptive analysis, we applied the random forest variable selection model to identify the key features of depression and anxiety. Finally, we utilized the RFC to build the predictive models for depression and anxiety and to gather detailed information about these models.\u003c/p\u003e \u003cp\u003eThe Random Forest (RF) algorithm is an integrated model designed to address classification and regression problems. It stands out for its ability to apply various models for evaluating responses. Compared to other machine learning algorithms, such as neural networks and support vector machines, RF algorithms can efficiently handle both continuous and categorical data sets. In this paper, we employ the RFC to construct the prediction model. The RFC comprises numerous individual decision trees functioning collectively as an ensemble. Each tree within the random forest contributes to the prediction, with the class receiving the most votes determining the model's overall prediction. One key advantage of the RFC over a single model is its collaborative approach, where each tree classifier, akin to a team member, collectively contributes to the final prediction, often yielding better results than a single decision tree. The RFC is particularly effective for binary classification, capable of managing datasets where the number of variables surpasses the number of observations. It can also handle datasets with a mix of continuous and categorical predictors. Furthermore, the RFC exhibits strong resistance to noise, can process high-dimensional data without feature selection, and is capable of processing various types of data while also ranking the importance of variables(20,38,39). In both the depression and anxiety models, the data were randomly split into two sets: a training set comprising 70% of the sample, and a testing set consisting of the remaining 30%.\u003c/p\u003e \u003cp\u003eGiven the prevalence of depression and anxiety among HCWs, we employed the SMOTE to balance the 'normal' and 'abnormal' categories. The strategy of under-sampling the majority (normal) class and over-sampling the minority (abnormal) class is recognized as an effective method to enhance the sensitivity of classifiers in cases of imbalanced data. In this study, we focused on over-sampling the minority category (depression/anxiety), creating additional samples by randomly selecting and replicating existing samples from this group. Subsequently, this augmented dataset was utilized for both training and testing the model.\u003c/p\u003e \u003cp\u003eIn this study, we utilized the GridSearchCV method to optimize the parameters of the RFC, a technique aimed at preventing overfitting by pruning the decision tree and removing its terminal nodes (40).\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eParticipant characteristics\u003c/h2\u003e \u003cp\u003eA total of 349 HCWs participated in this study.The findings are reported as frequencies, with the prevalence of depression and anxiety being 28.37% and 33.52%, respectively. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e details the descriptive outcomes of sociodemographic characteristics,including individual,interpersonal and institutional factors, and mental health factors.The average age of participants was 32.55 years (SD\u0026thinsp;=\u0026thinsp;9.03).A significant majority (80.81%) identified as female, and 45.56% were unmarried. Furthermore, 82.52% of the respondents were frontline HCWs, of whom 28.65% were doctors and a substantial 61.89% were nurses. Regarding work experience, 37.25% of HCWs had more than 10 years in the field.About 57.72% reported working approximately 6\u0026ndash;8 hours/day in the past week, while 24.93% did not have adequate rest time.Additionally, nearly half of the HCWs reported occurrences of overtime, including severe cases. Moreover, the scores and standard deviations for burnout, emotional labor, resilience, and adaptability were 63.68 (SD\u0026thinsp;=\u0026thinsp;15.72), 39.38 (SD\u0026thinsp;=\u0026thinsp;7.52), 84.49 (SD\u0026thinsp;=\u0026thinsp;16.47), and 74.94 (SD\u0026thinsp;=\u0026thinsp;14.47) respectively.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescriptive outcomes of sociodemographic characteristics\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eStandard deviation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e%\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePrevalence of probable mental health conditions\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDepression symptom\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWith depression\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28.37\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e99\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAnxiety symptom\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWith anxiety\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e33.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e117\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003eIndividual-level factors\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge(years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e32.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e349\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGenders\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e19.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e67\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e80.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e349\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMaried status\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarried\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e45.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e159\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eunmarried/widowed\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e54.44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e190\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLive with parents\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e21.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e76\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e78.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e273\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMinor child number\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e59.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e208\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1 OR 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e40.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e141\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eeducation_level\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emaster/doctor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e15.76\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBachelor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e66.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e231\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAssociate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e18.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e63\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWhether frontline HCW\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e288\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e17.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e61\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOccupation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhysician\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNurse\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e61.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e216\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOthers\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e33\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWorking experience(years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt;=1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[1,3]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e15.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e53\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[3,5]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e78\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[5,10]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e15.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e53\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026gt;=10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e37.25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e150\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAverage worktime last week(hours/day)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[6,8]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e57.02\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e199\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[9,11]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e32.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e115\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026ge;\u0026thinsp;12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIncome(yuan)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026le;\u0026thinsp;4000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e32.09\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e112\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[4000,8000]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e52.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e184\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026ge;\u0026thinsp;8000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e15.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e53\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHave the sufficient rest time\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e75.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e262\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e24.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e87\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBurnout\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e63.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEmotionnallabor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e39.38\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eResilence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e84.49\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAdaptability\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e74.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eDetecting potential predicators\u003c/h2\u003e \u003cp\u003eWe applied the random forest model to identify key variables in the classification of depression and anxiety, treating each issue separately. Based on the selection results and taking into account findings from previous studies, the following variables were incorporated into the random forest models for both depression and anxiety: age, gender, occupation, years of working experience, average work hours in the past week, availability of sufficient rest time, burnout, adaptation, emotional labor, resilience, and social support.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eTesting prediction accuracy of potential predictors\u003c/h3\u003e\n\u003cp\u003eThe entire sample was split into a training dataset and a test dataset for statistical analysis using the random forest classification algorithm. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents the confusion matrices for the test set, detailing depression and anxiety classification separately. In line with prior research, we calculated various metrics including Accuracy, Sensitivity, Specificity, Precision, and F1 Score.The accuracy of the random forest model is defined as the percentage of correct predictions made by the method\u003c/p\u003e \u003cp\u003eEquations below were used to caculate the sensitivity,specificity,F1 score,presicion,and accuracy in confusion matrix.\u003c/p\u003e \u003cp\u003eSensitivity=\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{True\\:Positives}{\\text{T}\\text{r}\\text{u}\\text{e}\\:\\text{P}\\text{o}\\text{s}\\text{i}\\text{t}\\text{i}\\text{v}\\text{e}\\text{s}+\\text{F}\\text{a}\\text{l}\\text{s}\\text{e}\\:\\text{N}\\text{e}\\text{g}\\text{a}\\text{t}\\text{i}\\text{v}\\text{e}\\text{s}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eSpecificity=\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{True\\:Negatives}{\\text{T}\\text{r}\\text{u}\\text{e}\\:\\text{N}\\text{e}\\text{g}\\text{a}\\text{t}\\text{i}\\text{v}\\text{e}\\text{s}+\\text{F}\\text{a}\\text{l}\\text{s}\\text{e}\\:\\text{P}\\text{o}\\text{s}\\text{i}\\text{t}\\text{i}\\text{v}\\text{e}\\text{s}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eF1 Score=\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{2\\text{*}\\left(\\text{P}\\text{r}\\text{e}\\text{s}\\text{i}\\text{c}\\text{i}\\text{o}\\text{n}\\text{*}\\text{R}\\text{e}\\text{c}\\text{a}\\text{l}\\text{l}\\right)}{Presicion+Recall}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003ePresicion=\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{True\\:Positives}{\\text{T}\\text{r}\\text{u}\\text{e}\\:\\text{P}\\text{o}\\text{s}\\text{i}\\text{t}\\text{i}\\text{v}\\text{e}\\text{s}+\\text{F}\\text{a}\\text{l}\\text{s}\\text{e}\\:\\text{P}\\text{o}\\text{s}\\text{i}\\text{t}\\text{i}\\text{v}\\text{e}\\text{s}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eAccuracy=\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{\\text{T}\\text{r}\\text{u}\\text{e}\\:\\text{P}\\text{o}\\text{s}\\text{i}\\text{t}\\text{i}\\text{v}\\text{e}\\text{s}+\\text{T}\\text{r}\\text{u}\\text{e}\\:\\text{N}\\text{e}\\text{g}\\text{a}\\text{t}\\text{i}\\text{v}\\text{e}\\text{s}}{Total\\:Populations}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eSensitivity, also known as \"Recall,\" is defined as the proportion of true positives accurately identified by the model. In this study, it refers to the rate at which healthcare workers (HCWs) with depression and anxiety are correctly predicted. The sensitivity score for depression was found to be 0.84, while for anxiety, it was 0.88.\u003c/p\u003e \u003cp\u003eModification Table:Specificity is the true negative rate, refers to the propotion of true negatives that are correctly identified by the model.This study refers to the propotion of HCWs without depression and anxiety who are correctly predicted.The specificity score for depression is 0.83 and for anxiety is 0.51.\u003c/p\u003e \u003cp\u003eThen, the classifier\u0026rsquo;s performance was tested by the 10-k cross-validated method, and the result was 0.77 for depression model and 0.79 for anxiety model.\u003c/p\u003e \u003cp\u003eThe receiver operating characteristic (ROC) curve typically features the sensitivity (true positive rate) on the Y-axis and the specificity (false positive rate) on the X-axis. The ideal point in the upper left corner of the graph indicates a zero false positive rate and a true positive rate of one. A larger area under the curve (AUC) generally denotes higher overall performance. The AUC for the depression model is 0.88 and for the anxiety model is 0.72. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e displays the respective areas under the curve for the depression and anxiety random forest prediction models. However, as indicated in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, HCWs with depression or anxiety represent approximately 20%-30% of all participants, highlighting clasimbalance. In such scenarios, accuracy alone is not a sufficient measure for the random forest prediction model. Therefore, the F1 score, which considers classification imbalance, becomes a more appropriate metric. This score is effective even with lower accuracy and is calculated as the harmonic mean of precision and recall. The F1 scores for the depression and anxiety models are 0.88 and 0.83, respectively.\u003c/p\u003e\n\u003ch3\u003eFeature importance\u003c/h3\u003e\n\u003cp\u003eIn the development of the RFC model, it is crucial to incorporate local interpretable model-agnostic explanations, such as SHAP (SHapley Additive exPlanations) value calculations and evaluations, to elucidate the model\u0026rsquo;s data results. SHAP values assign an importance value to each feature for a specific prediction, aiding in the interpretation of complex model predictions(41).Key visualizations include summary plots and beeswarm plots. Summary plots offer a comprehensive view of feature importance by illustrating the average impact of each feature on the model\u0026rsquo;s output. Beeswarm plots provide an in-depth perspective by showing how the impacts of individual features vary across different predictions. These visualizations play a crucial role in making complex model dynamics understandable, thereby assisting in both validating the model's accuracy and generating hypotheses for further research.\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e-Figure 8 show the importance of the features evaluated by each model in descending order. In this summary plot, the x-axis represents the mean absolute SHAP value (average impact on model output magnitude), which quantifies the average contribution of each feature to the model\u0026rsquo;s prediction. A higher value indicates a stronger influence on the predictive outcome.In beeswarm plots,red means the characteristic value is relatively high, and blue means that the characteristic value is relatively low. The more right the shap value is, the greater the positive contribution to the prediction of depression. In contrast, the more left, the smaller the shap value is, the greater the negative contribution to the prediction of depression or anxiety. If the shap value can distinguish between red and blue, it can be proved that their high or low values have different effects on the final results.\u003c/p\u003e \u003cp\u003eFrom the Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e-Figure 8,we can known that the selected variables can predict the depression and anxiety with a high accuracy,while the variables has the different importance in depression and anxiety model seperately.\u003c/p\u003e \u003cp\u003eFrom Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e, we can learn that the top 10 predictaors of depression for HCWs in descending order were:burnout,resilence,emotionallabor,adaptability,working experience(\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;1year),occpation is physician,social support,average worktime last week(9\u0026ndash;11 hours/day),age(28\u0026ndash;30 years),age(31\u0026ndash;35 years old).\u003c/p\u003e \u003cp\u003eFrom Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e, we can learn that the top 10 predictaors of anxiety for HCWs in descending order were:burnout,adaptability,emotional labor,age(31\u0026ndash;35 years),average worktime last week(9\u0026ndash;11 hours/day),resilence,occupation is physician,social support,working experience(\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;1 year),gender is female.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe objectives of this study were to evaluate the levels of depression and anxiety among HCWs during a public health emergency and to identify which variables significantly contribute to these conditions. Understanding the predictors of such emotional responses is crucial for developing and implementing preventive programs aimed at supporting and enhancing clinical outcomes in this or similar populations. In our research, we employed RFC models to separately predict depression and anxiety in HCWs and utilized the SMOTE to address the common issue of imbalanced datasets in real-world scenarios. The prevalence rates of depression and anxiety among HCWs were found to be 28.97% and 33.52%, respectively.Additionally, the depression model's AUC, Sensitivity (Recall), and Precision were 0.88, 0.84, and 0.93, while the anxiety model's parameters were 0.72, 0.88, and 0.78. These results indicate that the models are effective in screening HCWs vulnerable to depression and anxiety. Key predictive variables identified include age, gender, occupation, working experience, sufficient rest availability,workload(average work hours last week),burnout,adaptability,emotional labor,resilience,and social support.By utilizing these limited variables,we achieved good performance in predicting depression and anxiety, providing an efficient method to screen for adverse mental health outcomes in similar conditions.\u003c/p\u003e \u003cp\u003eThe prevalence of depression and anxiety among the HCWs during this special period were 28.97%(PHQ\u0026thinsp;\u0026ge;\u0026thinsp;10),33.52%(GAD\u0026thinsp;\u0026ge;\u0026thinsp;5),respecitively.The prevalence rates of depression and anxiety observed in this study varied somewhat from those documented in other research studies.When using a cutoff point of PHQ-9\u0026thinsp;\u0026ge;\u0026thinsp;5 to determine depression prevalence, the alarming rate among HCWs reaches 90.83%, significantly surpassing rates reported in other studies. Lai et al. reported prevalence of 50.4% (PHQ-9\u0026thinsp;\u0026ge;\u0026thinsp;5) for depression, 44.6% (GAD-7\u0026thinsp;\u0026ge;\u0026thinsp;5) for anxiety among HCWs during the from January 29, 2020, to February 3, 2020, in China(11),while Zhang et al. reported prevalence of 50.7% (PHQ-9\u0026thinsp;\u0026ge;\u0026thinsp;5), 44.7% (GAD-7\u0026thinsp;\u0026ge;\u0026thinsp;5),respectively among HCWs from January 29, 2020, to February 3, 2020, in China(13). In contrast to the observed prevalence of depression, the figure of 90.83% appears notably concerning.Although we turn up the cut-off of PHQ to 10,the prevalence is also high to 28.37%.In their study of female HCWs, Li et al. reported a prevalence of 14.2% (PHQ-9\u0026thinsp;\u0026ge;\u0026thinsp;10) for depression, 25.2% (GAD-7\u0026thinsp;\u0026ge;\u0026thinsp;8) for anxiety(42).And here are some unusual values have been reported by other studies(43\u0026ndash;45).This variation is likely attributed to differences in time and social factors.However, in the unique circumstances of this pandemic, HCWs face a daunting challenge as they encounter a surge in patients without adequate protection from PPE, both in their work and social environments.The risk is further heightened by the unfortunate transmission of the virus among colleagues.Under these circumstances,HCWs are mandated to remain at their posts, probably exacerbating the risk of adverse mental health outcomes for them.\u003c/p\u003e \u003cp\u003eGenerally speaking,it is evident that HCWs are facing widespread mental health issues during public health emergencies.There is a substantial need for psychiatric support for HCWs, both throughout and following such crises.The high incidence of mental health problems identified in this study underscores the immediate and ongoing necessity for psychiatric support for HCWs.\u003c/p\u003e \u003cp\u003eSeveral researchers have employed machine learning methods to predict depression or anxiety in various populations, achieving accuracies ranging from approximately 60\u0026ndash;90%.Sau and Bhakta utilized machine learning techniques to screen for depression and anxiety in seafarers and found that these methods outperformed traditional approaches.In their research, they used 14 variables, including age, gender, BMI, marital status, among others, to construct the prediction model.This model was developed using various machine learning algorithms such as CatBoost, Random Forest, Logistic Regression, Na\u0026iuml;ve Bayes, and Support Vector Machine (SVM).The CatBoost achieve the best accuracy(82.6%),the ROC curve is 0.882,and the precision is 84.1%(46).Kessler et.al,using the baseline patient self-reports of 5877 English-speaking residents of the non-institutionalized civilian US household population ages 15\u0026ndash;54 to bulid the predicting model of depression and pointed out that the ML methods has a higher accuracy.They used 913 variables into ML prediction model and 23 variables into logistics model to predicting depression and the AUC is 0.71\u0026ndash;0.76.(47). Multiple kernels of support vector machine (SVM) were applied to classify the depression using the social media message and found out the precision,F-measure and the recall scores were the best compare to the other methods.They used features, user microblog text, user profile and user behaviors without the socio-demographic or clinical variabes to bulid the prediction model,and the precision is 75.56%,F1 score is 76.12%,recall is 76.69%.(48). In summary, machine learning (ML) methods offer greater flexibility and adaptability compared to conventional methods, and their performance tends to excel, particularly in large datasets. These predictive models often incorporate socio-demographic and other variables to identify the most influential factors. However, our research took a different approach by using the RFC to select the most critical variables for inclusion in the model, and it performed admirably as well. This approach may provide a novel method for constructing prediction models when working with limited datasets.\u003c/p\u003e \u003cp\u003eIn the real world, imbalanced datasets are a prevalent occurrence. As data availability continues to grow in various large-scale, complex, and networked systems such as surveillance, security, the Internet, finance, and healthcare, it becomes increasingly important to enhance our fundamental understanding of knowledge discovery and analysis, starting from raw data to facilitate decision-making processes. Technically, any dataset that displays an unequal distribution among its classes can be categorized as imbalanced. Nevertheless, the prevailing understanding within the community is that imbalanced data typically involve datasets characterized by significant, and in some instances, extreme imbalances(49). ML method also used to solve the problem,but it\u0026rsquo;s prone to over predict the majority class.This leads to a larger misclassification rate for the minority class, which in many real-world applications is the class of interest(50).And some researcher take the deep learning method to solve the problem(51). In the medical field, imbalanced data are exceptionally common, with the prevalence of some diseases reaching as low as 1 in 10,000 or even 1 in 100,000 cases. When attempting to create predictive models for such scenarios, traditional methods prove to be inadequate. Consequently, machine learning (ML) techniques coupled with the SMOTE emerge as reliable approaches to tackle these challenges. Despite the widespread occurrence of imbalanced data in the medical domain, it has not received the level of attention it deserves.\u003c/p\u003e \u003cp\u003eOur study uncovered an intriguing finding worth discussing regarding both the depression and anxiety models.We utilized individual factors (adaptability/resilience), interpersonal factors (social support/emotional labor), and institutional factors (burnout/occupation) to construct the prediction model.Interestingly,while employing these identical features to forecast depression and anxiety symptoms,they exhibit distinct levels of significance. In both the depression and anxiety models, the paramount feature pertains to institutional factors, specifically burnout. Burnout manifests as a protracted reaction to chronic emotional and interpersonal stressors within the workplace, persistently prevalent among HCWs during the pandemic and prevalent globally(52), and burnout significantly impairs the personal and social functioning of HCWs, indicating its substantial correlation with their mental health.This finding aligns with prior research that has established a correlation between burnout and adverse mental health outcomes such as depression and anxiety(53\u0026ndash;55).As burnout is the most important predicors of depression and anxiety,there is a need to create interventions that alleviate HCWs burnout within and outside the hospital,which warrants attention and support from policy makers.Our prediction model highlights the significance of institutional factors in influencing the mental health of HCWs.Institutions play a crucial role in both work and daily life, exerting a mixed effect on HCWs.On one hand, institutions often provide resources and support to alleviate work pressure and enhance job satisfaction.On the other hand,HCWs are frequently affected by institutional factors such as workload, unreasonable work arrangements, and inadequate rest.Our prediction model corroborates these observations.Given these findings, policymakers must not only provide robust support but also prioritize efforts to mitigate the negative impact of institutional factors in the workplace.\u003c/p\u003e \u003cp\u003eIndividual factors,including age, gender, education, and personal adaptability and resilience, also play significant roles in our prediction models.For the depression model,the second important variable is resilience,while for the anxiety model,the variable is adaptability.The researchers found that a high resilience score consistently correlates with lower levels of depression, anxiety, and other mental health issues, regardless of whether the individuals are adolescents or adults(56,57).Adaptability serves as a crucial mental resource, encompassing an individual's cognitive, behavioral, and emotional regulation or adjustment in situations characterized by change, novelty, and uncertainty(58). Adaptability has been linked to various mental health outcomes. Dyson and Renk reported a negative association between adaptability and levels of stress and depressive symptoms in students(59). Moreover, Martin demonstrated that adaptability served as a predictor of both positive and negative mental health outcomes among adolescents(60).In summary,it is crucial to support HCWs in improving their personal psychological resilience and adaptability.Additionally, individual factors such as age and gender also demonstrate their significance in the prediction model.\u003c/p\u003e \u003cp\u003eThe interpersonal factors considered in this study include emotional labor and social support. HCWs are frequently influenced by interactions with others and by broader societal dynamics(5,61\u0026ndash;63).Hou and Zhang demonstrated that healthcare workers with high levels of social support are more likely to exhibit better mental health outcomes (64).On the other hand, Chinese scholars found that emotional labor significantly impacts the mental health of hotel employees in a longitudinal study (65).In a word,HCWs require ample support from society, both in material resources and emotional sustenance.\u003c/p\u003e \u003cp\u003eHowever, there are some noteworthy observations to consider. In previous studies, some researchers have highlighted that frontline HCWs exhibit a stronger association with depression and anxiety compared to those who are not on the frontline(6,66), However, it's important to note that when using the random forest classifier to select important features, the factor of being on the frontline was not deemed significant and was not included in the models.This observation could be attributed to the notion that depression and anxiety related to public health emergencies extend beyond one's specific role or area.It underscores the significance of making support systems flexible and accessible to all HCWs, not just those on the frontlines.Two other noteworthy factors that emerged as significant for depression and anxiety were the average work time in the past week and whether HCWs enjoyed sufficient rest. HCWs often work longer hours compared to many other professions worldwide, which can have an adverse impact on their mental health. The absence of sufficient rest time also had a similar impact. This highlights the importance of administrators making reasonable adjustments to support the health and well-being of HCWs.\u003c/p\u003e \u003cp\u003eOur research is subject to several acknowledged limitations.To begin with, the study adopts a cross-sectional design, restricting our ability to establish clear causality between variables.This limitation underscores the need for longitudinal research to provide a more nuanced understanding of the prevalence of these mental states within the unique context under examination.Secondly, the assessment of depression, anxiety, and other variables relies exclusively on self-reported questionnaires, lacking the inclusion of psychiatric interviews.While self-reported measures offer valuable insights, the absence of clinical interviews may limit the depth of our findings.Thirdly, our study's participants are exclusively drawn from a \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eShenyang's Class-A tertiary hospital\u003c/span\u003e,the generalizability of our findings may be limited.To address this constraint, it is imperative to conduct further investigations involving a more diverse and representative sample of participants. Expanding the scope of participants will enhance the external validity of our findings and contribute to a more comprehensive understanding of the phenomena under study.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe AUC values of 0.86 for depression and 0.83 for anxiety demonstrate our models\u0026rsquo; effectiveness in identifying these conditions among healthcare workers (HCWs). These findings highlight the complex interplay of individual, interpersonal, and institutional factors affecting HCWs\u0026rsquo; adverse outcomes. The study underscores the necessity for policymakers and hospital administrators to prioritize HCWs\u0026rsquo; well-being during public health emergencies, such as the COVID-19 pandemic, where their role is pivotal in protecting public health.\u003c/p\u003e \u003cp\u003eEmploying the RFC model alongside SMOTE techniques has yielded insights into predicting depression and anxiety in HCWs during such crises. Nonetheless, further refinement in data processing techniques is needed to enhance predictive accuracy. Our approach addresses the challenges of imbalanced datasets, contributing significantly to medical research.\u003c/p\u003e \u003cp\u003eIn sum, our research provides a robust framework for predicting mental health outcomes in HCWs amidst public health emergencies. By integrating diverse factors and applying innovative methodologies, we lay the groundwork for future research focused on improving HCWs\u0026rsquo; welfare and reducing the burden of mental health issues in similar scenarios.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch3\u003eEthics approval and consent to participate\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003eEthical Approval Statement:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Ethics Committee of Medical Ethics Committee of Shengjing Hospital, China Medical University, Shenyang Yongsen Hospital (Approval Number: 20230103). All participants provided written informed consent before participating in the study, and the research design was conducted in accordance with the principles of the Declaration of Helsinki and relevant ethical guidelines.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInformed Consent:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll participants were fully informed of the purpose, procedure, and potential risks of the study and signed written consent forms before participation. During the study, all data were anonymized to ensure the privacy of the participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdditional Information:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study did not involve the collection or analysis of data from children or minors and posed no physical or psychological harm to the participants.\u003c/p\u003e\n\u003ch3\u003eConsent for publication\u003c/h3\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003ch3\u003eData availability\u003c/h3\u003e\n\u003cp\u003eThe datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003ch3\u003eConflict of interest\u003c/h3\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003ch3\u003eFunding\u003c/h3\u003e\n\u003cp\u003eNo funding support\u003c/p\u003e\n\u003ch3\u003eAuthor contribution\u003c/h3\u003e\n\u003cp\u003eJW: Conceptualization of this study, Methodology, Software.\u003c/p\u003e\n\u003cp\u003eLF: Data curation, Writing - Original draft preparation.\u003c/p\u003e\n\u003cp\u003eNM: Visualization, Investigation.\u003c/p\u003e\n\u003cp\u003eCY: Software, Validation.\u003c/p\u003e\n\u003cp\u003eFC: Writing - Review \\\u0026amp; Editing.\u003c/p\u003e\n\u003cp\u003eXH: Supervision.\u003c/p\u003e\n\u003cp\u003eYS: Project administration.\u003c/p\u003e\n\u003cp\u003eLZ: Conceptualization of this study, Methodology, Software.\u003c/p\u003e\n\u003cp\u003ePY: Conceptualization of this study, Methodology, Software.\u003c/p\u003e\n\u003cp\u003eKS: Conceptualization of this study, Methodology, Software.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eDu Z, Wang Y, Bai Y, Wang L, Cowling BJ, Meyers LA. Estimate of COVID-19 Deaths, China, December 2022\u0026ndash;February 2023. Emerg Infect Dis. 2023 Oct;29(10):2121\u0026ndash;4.\u003c/li\u003e\n\u003cli\u003eHsieh KY, Kao WT, Li DJ, Lu WC, Tsai KY, Chen WJ, et al. Mental health in biological disasters: From SARS to COVID-19. Int J Soc Psychiatry. 2021 Aug 1;67(5):576\u0026ndash;86.\u003c/li\u003e\n\u003cli\u003eMaunder RG. Was SARS a mental health catastrophe? Gen Hosp Psychiatry. 2009;31(4):316\u0026ndash;7.\u003c/li\u003e\n\u003cli\u003eMaunder RG, Lancee WJ, Balderson KE, Bennett JP, Borgundvaag B, Evans S, et al. Long-term Psychological and Occupational Effects of Providing Hospital Healthcare during SARS Outbreak. Emerg Infect Dis. 2006 Dec;12(12):1924\u0026ndash;32.\u003c/li\u003e\n\u003cli\u003eE Alnazly, OM Khraisat, AM Al-Bashaireh, CL Bryant. Anxiety, depression, stress, fear and social support during COVID-19 pandemic among Jordanian healthcare workers. PLOS ONE. 2021 Mar 12;16(3).\u003c/li\u003e\n\u003cli\u003eJ Alonso, G Vilagut, P Mortier, M Ferrer, I Alayo, A Aragon-Pena, et al. Mental health impact of the first wave of COVID-19 pandemic on Spanish healthcare workers: A large cross-sectional survey. REVISTA DE PSIQUIATRIA Y SALUD MENTAL. 2021 Apr;14(2):90\u0026ndash;105.\u003c/li\u003e\n\u003cli\u003eRachel Hennein, Emma J. Mew, Sarah R. Lowe. Socio-ecological predictors of mental health outcomes among healthcare workers during the COVID-19 pandemic in the United States. Joel Msafiri Francis, editor. PLoS ONE. 2021 Feb 5;16(2):e0246602.\u003c/li\u003e\n\u003cli\u003eYeen Huang, Ning Zhao. Generalized anxiety disorder, depressive symptoms and sleep quality during COVID-19 outbreak in China: a web-based cross-sectional survey. Psychiatry Research. 2020 Jun;288:112954.\u003c/li\u003e\n\u003cli\u003eYunus Hacimusalar, Aybeniz Civan Kahve, Alisan Burak Yasar, Mehmet Sinan Aydin. Anxiety and hopelessness levels in COVID-19 pandemic: A comparative study of healthcare professionals and other community sample in Turkey. Journal of Psychiatric Research. 2020 Oct;129:181\u0026ndash;8.\u003c/li\u003e\n\u003cli\u003eZhou Zhu, Shabei Xu, Hui Wang, Zheng Liu, Jianhong Wu, Guo Li, et al. COVID-19 in Wuhan: Sociodemographic characteristics and hospital support measures associated with the immediate psychological impact on healthcare workers. eClinicalMedicine. 2020 Jul;24:100443.\u003c/li\u003e\n\u003cli\u003eLai J, Ma S, Wang Y, Cai Z, Hu J, Wei N, et al. Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019. JAMA NETWORK OPEN. 2020 Mar 23;3(3).\u003c/li\u003e\n\u003cli\u003eMustafa K\u0026uuml;rşat Şahin, Servet Aker, G\u0026uuml;lay Şahin, Ayt\u0026uuml;l Karabekiroğlu. Prevalence of Depression, Anxiety, Distress and Insomnia and Related Factors in Healthcare Workers During COVID-19 Pandemic in Turkey. J Community Health. 2020 Dec;45(6):1168\u0026ndash;77.\u003c/li\u003e\n\u003cli\u003eZhang C, Yang L, Liu S, Ma S, Wang Y, Cai Z, et al. Survey of Insomnia and Related Social Psychological Factors Among Medical Staff Involved in the 2019 Novel Coronavirus Disease Outbreak. Frontiers in Psychiatry [Internet]. 2020 [cited 2023 Nov 21];11. Available from: https://www.frontiersin.org/articles/10.3389/fpsyt.2020.00306\u003c/li\u003e\n\u003cli\u003eAn Y, Yang Y, Wang A, Li Y, Zhang Q, Cheung T, et al. Prevalence of depression and its impact on quality of life among frontline nurses in emergency departments during the COVID-19 outbreak. JOURNAL OF AFFECTIVE DISORDERS. 2020 Nov 1;276:312\u0026ndash;5.\u003c/li\u003e\n\u003cli\u003eArafa A, Mohammed Z, Mahmoud O, Elshazley M, Ewis A. Depressed, anxious, and stressed: What have healthcare workers on the frontlines in Egypt and Saudi Arabia experienced during the COVID-19 pandemic? JOURNAL OF AFFECTIVE DISORDERS. 2021 Jan 1;278:365\u0026ndash;71.\u003c/li\u003e\n\u003cli\u003eAllahyari E. Predicting Elderly Depression: An Artificial Neural Network Model. Iranian Journal of Psychiatry and Behavioral Sciences [Internet]. 2020 Jan 5 [cited 2023 Nov 29];13(4). Available from: https://brieflands.com/articles/ijpbs-98497.html\u003c/li\u003e\n\u003cli\u003eChang YS, Hung WC, Juang TY. Depression Diagnosis Based on Ontologies and Bayesian Networks. In: 2013 IEEE International Conference on Systems, Man, and Cybernetics [Internet]. 2013 [cited 2023 Nov 28]. p. 3452\u0026ndash;7. Available from: https://ieeexplore.ieee.org/abstract/document/6722342\u003c/li\u003e\n\u003cli\u003ePriya A, Garg S, Tigga NP. Predicting Anxiety, Depression and Stress in Modern Life using Machine Learning Algorithms. Procedia Computer Science. 2020 Jan 1;167:1258\u0026ndash;67.\u003c/li\u003e\n\u003cli\u003eThanathamathee P. Boosting with feature selection technique for screening and predicting adolescents depression. In: 2014 Fourth International Conference on Digital Information and Communication Technology and its Applications (DICTAP) [Internet]. 2014 [cited 2023 Nov 28]. p. 23\u0026ndash;7. Available from: https://ieeexplore.ieee.org/abstract/document/6821650\u003c/li\u003e\n\u003cli\u003eXin Y, Ren X. Predicting depression among rural and urban disabled elderly in China using a random forest classifier. BMC Psychiatry. 2022 Feb 15;22(1):118.\u003c/li\u003e\n\u003cli\u003eKhalilia M, Chakraborty S, Popescu M. Predicting disease risks from highly imbalanced data using random forest. BMC Med Inform Decis Mak. 2011 Jul 29;11(1):51.\u003c/li\u003e\n\u003cli\u003eChawla NV, Bowyer KW, Hall LO, Kegelmeyer WP. SMOTE: Synthetic Minority Over-sampling Technique. Journal of Artificial Intelligence Research. 2002 Jun 1;16:321\u0026ndash;57.\u003c/li\u003e\n\u003cli\u003eFernandez A, Garcia S, Herrera F, Chawla NV. SMOTE for Learning from Imbalanced Data: Progress and Challenges, Marking the 15-year Anniversary. Journal of Artificial Intelligence Research. 2018 Apr 20;61:863\u0026ndash;905.\u003c/li\u003e\n\u003cli\u003eThanathamathee P, Lursinsap C. Handling imbalanced data sets with synthetic boundary data generation using bootstrap re-sampling and AdaBoost techniques. Pattern Recognition Letters. 2013 Sep 1;34(12):1339\u0026ndash;47.\u003c/li\u003e\n\u003cli\u003eChen S, Chiu H, Xu B, Ma Y, Jin T, Wu M, et al. Reliability and validity of the PHQ-9 for screening late-life depression in Chinese primary care. Int J Geriat Psychiatry. 2010 Nov;25(11):1127\u0026ndash;33.\u003c/li\u003e\n\u003cli\u003eKroenke K, Spitzer RL, Williams JBW. The PHQ-9: Validity of a brief depression severity measure. J Gen Intern Med. 2001 Sep;16(9):606\u0026ndash;13.\u003c/li\u003e\n\u003cli\u003eMata DA, Ramos MA, Bansal N, Khan R, Guille C, Di Angelantonio E, et al. Prevalence of Depression and Depressive Symptoms Among Resident Physicians: A Systematic Review and Meta-analysis. Jama. 2015 Dec 8;314(22):2373\u0026ndash;83.\u003c/li\u003e\n\u003cli\u003eSpitzer RL, Kroenke K, Williams JBW, L\u0026ouml;we B. A Brief Measure for Assessing Generalized Anxiety Disorder: The GAD-7. Archives of Internal Medicine. 2006 May 22;166(10):1092\u0026ndash;7.\u003c/li\u003e\n\u003cli\u003eZhang C, Wang T, Zeng P, Zhao M, Zhang G, Zhai S, et al. Reliability, Validity, and Measurement Invariance of the General Anxiety Disorder Scale Among Chinese Medical University Students. Front Psychiatry [Internet]. 2021 [cited 2023 Dec 13];12. Available from: https://www.frontiersin.org/articles/10.3389/fpsyt.2021.648755\u003c/li\u003e\n\u003cli\u003eConnor KM, Davidson JRT. Development of a new resilience scale: The Connor-Davidson Resilience Scale (CD-RISC). Depress Anxiety. 2003;18(2):76\u0026ndash;82.\u003c/li\u003e\n\u003cli\u003eXie Y, Peng L, Zuo X, Li M. The Psychometric Evaluation of the Connor-Davidson Resilience Scale Using a Chinese Military Sample. PLOS ONE. 2016 Feb 9;11(2):e0148843.\u003c/li\u003e\n\u003cli\u003eLin GL, Guo LH, Lin K, Zhuang JB, Zou C, Chen CW, et al. Development and validation of the Work-Life Adaptability Scale for healthcare workers during the COVID-19 pandemic. J Shantou Univ Med Coll. 2022 Jun;35(2):106\u0026ndash;11.\u003c/li\u003e\n\u003cli\u003eBlumenthal JA, Burg MM, Barefoot J, Williams RB, Haney T, Zimet G. Social support, Type A behavior, and coronary artery disease. Psychosomatic Medicine. 1987;49(4):331\u0026ndash;40.\u003c/li\u003e\n\u003cli\u003eChou KL. Assessing Chinese adolescents\u0026rsquo; social support: the multidimensional scale of perceived social support. Pers Indiv Differ. 2000 Feb;28(2):299\u0026ndash;307.\u003c/li\u003e\n\u003cli\u003eDiefendorff JM, Croyle MH, Gosserand RH. The dimensionality and antecedents of emotional labor strategies. J Vocat Behav. 2005 Apr;66(2):339\u0026ndash;57.\u003c/li\u003e\n\u003cli\u003eWang M, Wu F, Wu P, Wang N. Validation of Chinese Version Emotional Labor Scale for Social Workers in Mainland China. Research on Social Work Practice. 2022 Sep 1;32(6):678\u0026ndash;86.\u003c/li\u003e\n\u003cli\u003eLin CY, Alimoradi Z, Griffiths MD, Pakpour AH. Psychometric properties of the Maslach Burnout Inventory for Medical Personnel (MBI-HSS-MP). Heliyon. 2022 Feb;8(2):e08868.\u003c/li\u003e\n\u003cli\u003eGokten ES, Uyulan C. Prediction of the development of depression and post-traumatic stress disorder in sexually abused children using a random forest classifier. J Affect Disord. 2021 Jan 15;279:256\u0026ndash;65.\u003c/li\u003e\n\u003cli\u003eRigatti SJ. Random Forest. J Insur Med. 2017 Jan 1;47(1):31\u0026ndash;9.\u003c/li\u003e\n\u003cli\u003eProbst P, Wright MN, Boulesteix AL. Hyperparameters and tuning strategies for random forest. Wires Data Min Knowl Discovery. 2019;9(3):e1301.\u003c/li\u003e\n\u003cli\u003eLundberg SM, Lee SI. A Unified Approach to Interpreting Model Predictions. In: Advances in Neural Information Processing Systems [Internet]. Curran Associates, Inc.; 2017 [cited 2023 Dec 28]. Available from: https://proceedings.neurips.cc/paper/2017/hash/8a20a8621978632d76c43dfd28b67767-Abstract.html\u003c/li\u003e\n\u003cli\u003eLi G, Miao J, Wang H, Xu S, Sun W, Fan Y, et al. Psychological impact on women health workers involved in COVID-19 outbreak in Wuhan: a cross-sectional study. Journal of Neurology, Neurosurgery \u0026amp; Psychiatry. 2020 Aug;91(8):895\u0026ndash;7.\u003c/li\u003e\n\u003cli\u003eYin Q, Sun Z, Liu T, Ni X, Deng X, Jia Y, et al. Posttraumatic stress symptoms of health care workers during the corona virus disease 2019. CLINICAL PSYCHOLOGY \u0026amp; PSYCHOTHERAPY. 2020 May;27(3):384\u0026ndash;95.\u003c/li\u003e\n\u003cli\u003eZhang W rui, Wang K, Yin L, Zhao W feng, Xue Q, Peng M, et al. Mental Health and Psychosocial Problems of Medical Health Workers during the COVID-19 Epidemic in China. Psychother Psychosom. 2020;89(4):242\u0026ndash;50.\u003c/li\u003e\n\u003cli\u003eZhou Y, Zhou Y, Song Y, Ren L, Ng CH, Xiang YT, et al. Tackling the mental health burden of frontline healthcare staff in the COVID-19 pandemic: China\u0026rsquo;s experiences. Psychol Med. 2021 Aug;51(11):1955\u0026ndash;6.\u003c/li\u003e\n\u003cli\u003eSau A, Bhakta I. Screening of anxiety and depression among seafarers using machine learning technology. Informatics in Medicine Unlocked. 2019;16:100228.\u003c/li\u003e\n\u003cli\u003eKessler RC, Van Loo HM, Wardenaar KJ, Bossarte RM, Brenner LA, Cai T, et al. Testing a machine-learning algorithm to predict the persistence and severity of major depressive disorder from baseline self-reports. Mol Psychiatry. 2016 Oct;21(10):1366\u0026ndash;71.\u003c/li\u003e\n\u003cli\u003ePeng Z, Hu Q, Dang J. Multi-kernel SVM based depression recognition using social media data. International Journal of Machine Learning and Cybernetics. 2019 Jan 1;10(1):43\u0026ndash;57.\u003c/li\u003e\n\u003cli\u003eHaibo He, Garcia EA. Learning from Imbalanced Data. IEEE Trans Knowl Data Eng. 2009 Sep;21(9):1263\u0026ndash;84.\u003c/li\u003e\n\u003cli\u003eEsposito C, Landrum GA, Schneider N, Stiefl N, Riniker S. GHOST: Adjusting the Decision Threshold to Handle Imbalanced Data in Machine Learning. J Chem Inf Model. 2021 Jun 28;61(6):2623\u0026ndash;40.\u003c/li\u003e\n\u003cli\u003eLee J, Park K. GAN-based imbalanced data intrusion detection system. Pers Ubiquit Comput. 2021 Feb;25(1):121\u0026ndash;8.\u003c/li\u003e\n\u003cli\u003eZheng Y, Tang PK, Lin G, Liu J, Hu H, Wu AMS, et al. Burnout among healthcare providers: Its prevalence and association with anxiety and depression during the COVID-19 pandemic in Macao, China. PLOS ONE. 2023 Mar 16;18(3):e0283239.\u003c/li\u003e\n\u003cli\u003eBarello S, Palamenghi L, Graffigna G. Burnout and somatic symptoms among frontline healthcare professionals at the peak of the Italian COVID-19 pandemic. PSYCHIATRY RESEARCH. 2020 Aug;290.\u003c/li\u003e\n\u003cli\u003eDobson H, Malpas CB, Burrell AJC, Gurvich C, Chen L, Kulkarni J, et al. Burnout and psychological distress amongst Australian healthcare workers during the COVID-19 pandemic. AUSTRALASIAN PSYCHIATRY. 2021 Feb;29(1):26\u0026ndash;30.\u003c/li\u003e\n\u003cli\u003eDuarte I, Teixeira A, Castro L, Marina S, Ribeiro C, Jacome C, et al. Burnout among Portuguese healthcare workers during the COVID-19 pandemic. BMC PUBLIC HEALTH. 2020 Dec 7;20(1).\u003c/li\u003e\n\u003cli\u003eHjemdal O, Vogel PA, Solem S, Hagen K, Stiles TC. The relationship between resilience and levels of anxiety, depression, and obsessive\u0026ndash;compulsive symptoms in adolescents. Clin Psychol Psychot. 2011 Jul;18(4):314\u0026ndash;21.\u003c/li\u003e\n\u003cli\u003eRan L, Wang W, Ai M, Kong Y, Chen J, Kuang L. Psychological resilience, depression, anxiety, and somatization symptoms in response to COVID-19: A study of the general population in China at the peak of its epidemic. Soc Sci Med. 2020 Oct;262:113261.\u003c/li\u003e\n\u003cli\u003eWaldeck D, Pancani L, Holliman A, Karekla M, Tyndall I. Adaptability and psychological flexibility: Overlapping constructs? Journal of Contextual Behavioral Science. 2021 Jan 1;19:72\u0026ndash;8.\u003c/li\u003e\n\u003cli\u003eDyson R, Renk K. Freshmen adaptation to university life: Depressive symptoms, stress, and coping. J Clin Psychol. 2006 Oct;62(10):1231\u0026ndash;44.\u003c/li\u003e\n\u003cli\u003eMartin AJ, Nejad HG, Colmar S, Liem GAD. Adaptability: How students\u0026rsquo; responses to uncertainty and novelty predict their academic and non-academic outcomes. J Educ Psychol. 2013 Aug;105(3):728\u0026ndash;46.\u003c/li\u003e\n\u003cli\u003eOhrnberger J, Fichera E, Sutton M. The relationship between physical and mental health: A mediation analysis. Soc Sci Med. 2017 Dec;195:42\u0026ndash;9.\u003c/li\u003e\n\u003cli\u003eYoon SL, Kim JH. Job-Related Stress, Emotional Labor, and Depressive Symptoms Among Korean Nurses. J Nurs Scholarsh. 2013;45(2):169\u0026ndash;76.\u003c/li\u003e\n\u003cli\u003eZhang Y, ElGhaziri M, Siddique S, Gore R, Kurowski A, Nobrega S, et al. Emotional Labor and Depressive Symptoms Among Healthcare Workers: The Role of Sleep. Workplace Health \u0026amp; Safety. 2021 Aug;69(8):383\u0026ndash;93.\u003c/li\u003e\n\u003cli\u003eHou T, Zhang T, Cai W, Song X, Chen A, Deng G, et al. Social support and mental health among health care workers during Coronavirus Disease 2019 outbreak: A moderated mediation model. PLOS ONE. 2020 May 29;15(5).\u003c/li\u003e\n\u003cli\u003eXiong W, Huang M, Okumus B, Leung XY, Cai X, Fan F. How emotional labor affect hotel employees\u0026rsquo; mental health: A longitudinal study. Tourism Management. 2023 Feb 1;94:104631.\u003c/li\u003e\n\u003cli\u003eJames Gilleen, Aida Santaolalla, Lorena Valdearenas, Clara Salice, Montserrat Fust\u0026eacute;. Impact of the COVID-19 pandemic on the mental health and well-being of UK healthcare workers. BJPsych open. 2021 May;7(3):e88.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-5228634/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5228634/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eDuring public health emergencies such as SARS, Ebola, and COVID-19, healthcare workers (HCWs) are often required to confront these crises, potentially leading to adverse mental health outcomes. Consequently, they are at a heightened risk of experiencing symptoms of depression and anxiety. It is widely recognized that psychological disorders can lead to severe consequences. Despite this, there remains a scarcity of research focused on developing predictive models to forecast the depression and anxiety levels of healthcare workers under these challenging conditions.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eA total of 349 HCWs were selected from a Class-A tertiary hospital in the city of Shenyang, Liaoning Province in China. Depression and anxiety were assessed using the Patient Health Questionnaire (PHQ-9) and the Generalized Anxiety Disorder (GAD-7) scale, respectively. This study employed a random forest classifier(RFC) to predict the depression and anxiety levels of HCWs from three perspectives: individual, interpersonal, and institutional. Moreover, we employed The Synthetic Minority Over-sampling Technique (SMOTE) to address the issue of imabalanced data distribution.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe prevalence of depression and anxiety among HCWs were 28.37% and 33.52%, respectively. The prediction model was developed using a training dataset (70%) and a test dataset (30%). The area under the curve (AUC) for depression and anxiety were 0.88 and 0.72, respectively. Additionally, the mean values of the 10-fold cross-validation results were 0.77 for the depression prediction model and 0.79 for the anxiety prediction model. For the depression prediction model, the top ten most significant predictive factors were: burnout, resilience, emotional labor, adaptability, working experience(\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;1year), physician, social support, average work time last week(9\u0026ndash;11 hours), age(28\u0026ndash;30 years), age(31\u0026ndash;35 years old). For the anxiety prediction model, the top ten most significant predictive factors were: burnout, adaptability, emotional labor, age(31\u0026ndash;35), average work time last week(9\u0026ndash;11 hours), resilience, physician, social support, working experience(\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;1 year), female.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eIt is essential to develop multiple interventions that provide support both before and after a public health emergency, aiming at mitigating symptoms of depression and anxiety. SMOTE is a practical method for addressing imbalances in datasets. Mitigating burnout among HCWs, bolstering their resilience and adaptability, and ensuring reasonable work hours are crucial steps to prevent adverse mental health problems.\u003c/p\u003e","manuscriptTitle":"Balancing Mental Health: Predictive Modeling for Healthcare Workers During Public Health Crises","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-11-27 17:46:13","doi":"10.21203/rs.3.rs-5228634/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-06-26T11:04:05+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-25T18:38:14+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-25T14:49:15+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"155824341362433324557740985128405353103","date":"2025-06-16T17:21:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"309311469214579755541816736159531016422","date":"2025-06-16T14:13:38+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-11-05T08:07:24+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-11-05T08:03:48+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-11-04T13:57:54+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-11-01T09:49:09+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2024-10-09T02:25:07+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"6d14bd51-3bbe-4c3d-92b5-e2445f8997a0","owner":[],"postedDate":"November 27th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":39972239,"name":"Biological sciences/Psychology/Human behaviour"},{"id":39972240,"name":"Biological sciences/Psychology"}],"tags":[],"updatedAt":"2025-08-11T16:08:56+00:00","versionOfRecord":{"articleIdentity":"rs-5228634","link":"https://doi.org/10.1038/s41598-025-14403-3","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2025-08-06 15:57:37","publishedOnDateReadable":"August 6th, 2025"},"versionCreatedAt":"2024-11-27 17:46:13","video":"","vorDoi":"10.1038/s41598-025-14403-3","vorDoiUrl":"https://doi.org/10.1038/s41598-025-14403-3","workflowStages":[]},"version":"v1","identity":"rs-5228634","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5228634","identity":"rs-5228634","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}