AI-Based Precision Prediction of Healthcare Workers’ Antibiotic Use Intentions: Multi-theoretical Psychological and Behavioral Modeling

AI-Based Precision Prediction of Healthcare Workers’ Antibiotic Use Intentions: Multi-theoretical Psychological and Behavioral Modeling | 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 AI-Based Precision Prediction of Healthcare Workers’ Antibiotic Use Intentions: Multi-theoretical Psychological and Behavioral Modeling Le Han, Peng Xian, Xiao Liu, Ying Liu, yue chang, lijun ma, jie ren, and 9 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6768023/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 28 Jan, 2026 Read the published version in Scientific Reports → Version 1 posted 14 You are reading this latest preprint version Abstract Introduction: Antimicrobial resistance (AMR) remains a global health priority, partly driven by irrational antibiotic prescribing among healthcare workers. Existing interventions often overlook the psychological and contextual factors that shape prescribing behaviour. This study aimed to integrate behavioural theory with explainable machine learning to identify psychological predictors of antibiotic use intention among clinicians. Methods: A cross-sectional survey was conducted among 1,135 healthcare workers from four public hospitals in China. Participants completed questionnaires based on constructs from the Theory of Planned Behavior, Health Belief Model, Rational Action Theory, Self-Efficacy Theory, Social Support Theory, and cognitive processing frameworks. LASSO regression and SHAP analysis were applied alongside machine learning classifiers (e.g. XGBoost, LightGBM, CatBoost) to identify key predictors and interactions influencing prescribing intention. Results: Social support, cognitive processing, knowledge and skills, and health beliefs were the most important predictors. SHAP analysis revealed nonlinear interactions, particularly between social support and cognitive engagement. Ensemble models achieved high predictive accuracy (F1-scores >0.94) for high and medium prescribing intention, but classification of low-intention individuals remained more challenging. Conclusion: Combining behavioural theory with explainable AI offers a scalable approach to identifying clinicians at risk of irrational prescribing. These findings support the development of psychologically tailored, real-time interventions that can improve antibiotic stewardship and address AMR in diverse health system settings. Health sciences/Diseases Health sciences/Diseases/Infectious diseases/Bacterial infection antibiotic stewardship prescribing behavior healthcare workers behavioral intention explainable machine learning psychological modeling SHAP analysis LASSO regression Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1. Introduction Antimicrobial resistance (AMR) is among the most urgent global health threats, undermining routine medical care and placing growing strain on healthcare systems. Despite international and national policy efforts, including audit systems, clinical guidelines, and prescription regulations, inappropriate antibiotic use remains widespread. In both high-income and low- and middle-income countries, up to half of all prescriptions are estimated to be irrational.¹⁻³ In China, the overuse of antibiotics—exceeding 60% of prescriptions before 2011—prompted the launch of the National Special Rectification Campaign.⁴ While progress has been made, irrational prescribing persists, particularly in resource-limited and high-pressure clinical environments.⁵ Conventional interventions have focused on improving knowledge and enforcing guidelines. However, growing evidence suggests that prescribing decisions are shaped not only by clinical competence, but also by psychological, social, and institutional pressures. Clinicians’ choices may be influenced by diagnostic uncertainty, time constraints, perceived patient expectations, fear of complications or complaints, and local professional norms. These cognitive and psychosocial factors are deeply embedded in clinical routines but remain insufficiently addressed by current stewardship strategies. Behavioral science provides useful frameworks for understanding these drivers. Models such as the Theory of Planned Behavior, Health Belief Model, Self-Efficacy Theory, and Social Support Theory have identified key constructs—self-efficacy, perceived control, risk perception, and normative beliefs—as significant predictors of prescribing intention.⁶⁻⁹ However, most existing studies adopt narrow theoretical scopes, rely on linear analytical methods, and are based on small or single-center datasets. These limitations restrict their capacity to capture the complex, non-linear, and context-specific nature of clinical decision-making. In a previous study, we used structural equation modeling to examine how social support influences rational antibiotic prescribing among healthcare workers.¹⁰ Our findings revealed several mediating pathways, including cognitive processing and professional expectations. However, the study was limited by its single-center design and the linear assumptions of the model. To address these gaps, the present study integrates multi-theoretical behavioral modeling with explainable machine learning (ML). Using data from four hospitals across different regions of China, we model the psychological and contextual factors influencing healthcare workers’ intentions to use antibiotics appropriately. Explainable ML tools—such as SHAP values and LASSO regression—allow us to identify and interpret the most influential behavioral variables and their interactions, offering both predictive strength and theoretical insight. The aim of this study is to develop an interpretable model to identify healthcare workers at higher risk of inappropriate antibiotic prescribing. Based on individual behavioral profiles, we further aim to design an intelligent system capable of delivering tailored educational and psychological interventions in real time. By combining theory-informed modeling, explainable ML, and adaptive intervention strategies, this work moves antimicrobial stewardship toward greater precision, scalability, and clinical relevance. While this study focuses on antibiotic use, the underlying behavioral mechanisms it examines—such as self-efficacy, perceived norms, and cognitive burden—are not unique to antimicrobial decisions. These factors also influence broader clinical behaviors and healthcare workers’ own health-related practices. ¹¹⁻¹³ As concerns around clinician burnout, diagnostic error, and system-level stress continue to grow, developing scalable, psychologically informed tools to support frontline providers is increasingly recognized as a public health priority.¹⁴ 2. Methods 2.1 Study Design and Participants This study employed a multi-center, cross-sectional survey design conducted between January and December 2024 across public hospitals in China. Participating institutions were located in different regions and represented varying levels of the healthcare system, including both urban and semi-urban settings. Eligible participants were frontline clinical staff engaged in direct patient care and responsible for antimicrobial decision-making. Inclusion criteria were: currently working in clinical departments. aged over 18 years. holding Chinese nationality. native Mandarin speakers. voluntary participation with signed informed consent. Exclusion criteria included: observable cognitive impairment that could affect the ability to complete the questionnaire. serious communication difficulties due to sensory or neurological impairments (e.g., severe hearing or visual impairment). refusal or inability to provide informed consent. 2.2 Scale Development The development of the measurement scale was informed by established scale construction principles and a theory-based conceptual framework. Eight latent constructs were identified through a review of major behavioral and cognitive theories relevant to antimicrobial prescribing among healthcare professionals. These included the Theory of Planned Behavior, the Health Belief Model, Rational Action Theory, Self-Efficacy Theory, Social Support Theory, cognitive processing, and knowledge- and skill-based models.¹⁵⁻¹⁶ Each construct was operationalized through four items designed to reflect core theoretical dimensions and clinical relevance, resulting in a 32-item draft scale. Where possible, items were adapted from previously published instruments.¹⁷ For constructs lacking suitable items, new statements were developed to align with the clinical context of Chinese hospital settings. All items were written in Chinese and employed a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree).The initial item pool was reviewed by a multidisciplinary expert panel, including specialists in infectious diseases, pharmacy, behavioral psychology, and clinical education. Following standard procedures for content validation, experts independently assessed each item’s clarity, relevance, and theoretical alignment.¹⁸⁻²⁰ Revisions were made to improve semantic precision, item distinctiveness, and contextual appropriateness based on their feedback. 2.3 Machine Learning Procedures Overview and Platform All machine learning procedures were implemented in Python 3.10 using the Google Colaboratory (Colab) environment. Commonly used libraries supported the workflow: scikit-learn for model development and evaluation; XGBoost , LightGBM , and CatBoost for gradient boosting algorithms; and SHAP for explainability analyses. Data preprocessing and manipulation were conducted using Pandas and NumPy . Target Variable Definition The primary outcome variable was behavioral intention (BI) to rationally prescribe antibiotics, calculated as the mean of four Likert-scale items capturing future-oriented prescribing intentions. This continuous variable (BI_score) was converted into a three-level categorical outcome (BI_category) based on sample quantiles: Low Intention: BI_score ≤ 33rd percentile Medium Intention: 33rd–67th percentile High Intention: BI_score ≥ 67th percentile This quantile-based discretization ensured balanced class distributions and allowed for meaningful differentiation in the strength of prescribing intentions. The resulting categorical variable served as the dependent variable in all classification models. Feature Standardization and Selection Independent variables comprised composite scores for each psychological construct (e.g., TPB, HBM, knowledge-skill), computed as the mean of four items per construct. Prior to model training, all features were standardized using z-scores via the StandardScaler module in scikit-learn . To reduce dimensionality and enhance interpretability, LASSO regression with cross-validation was employed for feature selection. Only variables with non-zero coefficients were retained for inclusion in downstream models. Model Training and Evaluation Eleven supervised machine learning algorithms were trained to classify BI levels: Logistic Regression Support Vector Machine (SVM) Random Forest XGBoost LightGBM CatBoost K-Nearest Neighbors (KNN) Naive Bayes Gradient Boosting AdaBoost Voting and Stacking Ensembles The dataset was randomly split into training (80%) and testing (20%) subsets using stratified sampling to preserve the distribution of BI categories. Hyperparameter tuning was performed via grid search or internal validation, depending on the algorithm. Model performance was evaluated on the test set using precision, recall, F1-score, and macro-averaged ROC-AUC. 2.5 Model Explainability To enhance interpretability, several explainability techniques were applied: LASSO regression coefficients were examined to assess the relative importance of psychological predictors in a linear model. SHAP (SHapley Additive Explanations) values were calculated to visualize the contribution and directionality of each predictor across models. SHAP interaction values were used to explore potential interactions between psychological constructs. Bootstrapped logistic regression (1,000 resamples) was conducted to estimate 95% confidence intervals for predictor effects in binary classification tasks (High vs. Low intention groups). 3. Results 3.1 Demographic Analysis A total of 1,294 questionnaires were distributed across participating sites. Of these, 1,135 valid responses were received, yielding an effective response rate of 87.7%.Among the respondents, the majority were female (81.32%, n = 923), while male participants accounted for 18.68% (n = 212). The age distribution was as follows: 26–35 years (38.85%) comprised the largest group, followed by 36–45 years (27.84%), 18–25 years (22.82%), and 46–55 years (10.22%). Only a small proportion (0.26%) were older than 56 years. In terms of educational background, 45.46% of participants held a bachelor’s degree, 22.03% had education below the bachelor’s level, 20.7% held a master’s degree, 11.54% had completed a doctoral degree, and 0.26% were postdoctoral researchers. Respondents represented a range of clinical departments. Internal medicine was the most common (28.55%), followed by surgery (22.29%), otolaryngology (12.95%), ophthalmology (12.86%), obstetrics and gynecology (7.31%), intensive care (7.75%), emergency medicine (4.41%), dentistry (2.73%), and dermatology (1.15%). Regarding clinical experience, 34.45% had 16–25 years of working experience, 31.81% had more than 25 years, 17.0% had less than 5 years, and 16.65% reported 5–15 years of experience. These demographic characteristics are summarized in Table 1 . Table 1 Demographic Information of Respondents in the Survey Variable Category Count Percent (%) GENDER Male 212 18.68 Female 923 81.32 AGE 18–25 259 22.82 26–35 441 38.85 36–45 316 27.84 46–55 116 10.22 Over 56 3 0.26 EDUCATION Less than Bachelor's degree 250 22.03 Bachelor’s degree 516 45.46 Master’s degree 235 20.7 Doctoral degree (PhD) 131 11.54 Postdoctoral 3 0.26 DEPARTMENT Internal Medicine 324 28.55 Surgery 253 22.29 Ophthalmology 146 12.86 Otolaryngology 147 12.95 Obstetrics and Gynecology 83 7.31 Intensive Care Unit (ICU) 88 7.75 Dentistry 31 2.73 Emergency Medicine 50 4.41 Dermatology 13 1.15 Working-experience 25years 361 31.81 3.2 Feature Selection and Interpretation 3.2.1 Feature Selection and Behavioral Interpretation Using LASSO LASSO regression with five-fold cross-validation was used to identify key psychological predictors of healthcare workers’ intention to rationally prescribe antibiotics. All seven constructs were retained, though their contributions varied in both strength and direction (Fig. 1 ). Social support (β = 0.2564) was the strongest predictor, underscoring the role of institutional culture and peer reinforcement in shaping prescribing behavior. Healthcare workers embedded in supportive, stewardship-oriented environments may be more inclined to follow rational antibiotic practices. This aligns with organizational psychology literature, which emphasizes the behavioral impact of social norms and workplace context. The Health Belief Model (β = 0.1634) was also a significant predictor. Perceived susceptibility, severity, and benefits appear to meaningfully influence prescribing intent—particularly in settings where risk perception and clinical outcomes are closely linked. Cognitive processing (β = 0.1404) and knowledge and skills (β = 0.1023) contributed positively, reflecting the importance of analytical reasoning and clinical competence. These findings support previous work linking reflective decision-making and content knowledge with more appropriate antibiotic use. In contrast, self-efficacy (β = − 0.0383) and TPB constructs (β = − 0.0572) showed weak negative associations. Although these constructs are traditionally associated with behavioral intention, their reduced influence here may reflect the dominant role of institutional policies and norms, which constrain individual agency in tightly regulated clinical environments. Rational Action Theory (β = 0.0456) showed a modest positive effect, suggesting that while deliberate cost-benefit reasoning contributes to decision-making, its role may be limited when other structural or perceptual factors are more salient. 3.2.2 Psychological Feature Contributions and Interactions in SHAP Analysis To further interpret the predictive model, SHapley Additive exPlanations (SHAP) analysis was applied to the top-performing XGBoost classifier. SHAP quantifies the individual contribution of each psychological construct to model predictions, enhancing transparency and interpretability. The SHAP summary plot (Fig. 2 a) illustrates the magnitude, direction, and distribution of each feature’s influence. Cognitive processing (CG_score), social support (SS_score), knowledge and skills (KS_score), and health belief model (HBM_score) constructs showed the highest contributions. Among these, cognitive processing had the strongest positive effect, particularly at higher values, underscoring the role of analytical reasoning in rational prescribing. In contrast, TPB, RAT, and SET constructs were excluded from visualization due to minimal importance, reflecting their limited influence in the final model. The mean absolute SHAP values (Fig. 2 b) confirmed this ranking, with cognitive processing emerging as the most influential factor, followed closely by social support and knowledge and skills. This highlights a hierarchy of predictors in which cognitive and contextual variables outweigh traditional volitional models in this behavioral context. SHAP dependence plots revealed non-linear relationships and potential interactions. In Fig. 2 c, social support displayed a positive but nonlinear association with rational prescribing intentions, with stronger effects at higher scores. Color gradients indicating cognitive processing scores suggest an interaction, where the effect of social support may be modulated by analytical engagement. Similarly, Fig. 2 d shows that the influence of knowledge and skills may also be shaped by social context. Figures 2 e and 2 f explore HBM and cognitive processing effects respectively, both demonstrating threshold-based increases in predicted intention at moderate to high levels. 3.3 Bootstrapped Logistic Regression Analysis of Psychological Predictors Figure 3 presents the results of bootstrapped logistic regression, displaying the median coefficients and 95% confidence intervals (CIs) for seven psychological constructs in relation to healthcare workers’ intentions to prescribe antibiotics rationally. Four constructs showed statistically significant positive associations, with CIs that did not cross zero. Social support (SS_score) emerged as the strongest predictor (Median = 1.72), underscoring the influence of institutional context and peer reinforcement in promoting stewardship-aligned behavior. Cognitive processing (CG_score, Median = 1.42) and knowledge and skills (KS_score, Median = 1.40) followed closely, highlighting the importance of analytical engagement and clinical competence in prescribing decisions. The health belief model (HBM_score, Median = 1.03) also showed a significant positive association, indicating that perceptions of risk and severity surrounding antimicrobial resistance shape prescribing intent. In contrast, constructs derived from the Theory of Planned Behavior (TPB_score, Median = − 0.20) and Self-Efficacy Theory (SET_score, Median = − 0.15) had confidence intervals that included zero, suggesting limited and statistically uncertain effects. While Rational Action Theory (RAT_score, Median = 0.20) had a weak positive coefficient, its wide CI overlapping zero indicates low robustness. 3.4 Comparative Performance of Machine Learning Models by Intention Level Figure 4 presents a comparative evaluation of eleven machine learning models in classifying healthcare workers’ behavioral intentions—categorized as High, Medium, and Low—regarding rational antibiotic use. Model performance was assessed using precision (Fig. 4 a), recall (Fig. 4 b), and F1-score (Fig. 4 c). Ensemble models, including Voting, Stacking, and XGBoost, demonstrated consistently high performance in predicting High and Medium intention groups, with F1-scores exceeding 0.94. The Voting classifier achieved the highest overall balance, with F1-scores of 0.96 and 0.95 for High and Medium groups, respectively. Logistic Regression and Support Vector Machine (SVM) also performed well for High and Medium categories, although performance declined in classifying Low intention individuals. For example, SVM reached an F1-score of 0.97 for High intention but dropped to 0.76 for Low intention, suggesting reduced sensitivity to less distinct behavioral profiles. AdaBoost and Gradient Boosting showed similar performance patterns, with lower F1-scores for the Low category (0.76 and 0.79, respectively). Across all models, classification of the Low intention group proved more challenging. Even top-performing algorithms struggled to balance precision and recall. For instance, Naive Bayes achieved a high recall (0.92) but low precision (0.69) for the Low group, resulting in a moderate F1-score of 0.79—indicating a tendency to over-predict without sufficient specificity. Similarly, LightGBM and Multilayer Perceptron (MLP), despite solid overall performance, recorded F1-scores near 0.80 for the Low group, reflecting consistent difficulty in accurately identifying individuals with weaker behavioral intentions. 3.5 Discriminative Performance of Models Based on Multi-Class ROC Curves Figure 5 illustrates the multi-class Receiver Operating Characteristic (ROC) curves for eleven machine learning models across three levels of behavioral intention: High, Medium, and Low. The Area Under the Curve (AUC) values are reported for each class, providing a quantitative assessment of each model’s ability to discriminate between intention categories. Ensemble models—particularly Voting, Stacking, and XGBoost—demonstrated consistently strong discriminative performance, each achieving AUC values of 0.99 across all three intention levels. These results highlight their robustness in capturing complex decision boundaries and maintaining balanced accuracy across class distributions. LightGBM also performed exceptionally well, with AUCs of 0.98 for the High and Medium groups, and a perfect 1.00 for the Low intention category—traditionally the most challenging to classify due to class imbalance. This suggests LightGBM’s capacity to detect nuanced distinctions in underrepresented subpopulations. AdaBoost and Gradient Boosting followed closely, with AUC values ≥ 0.98 across all categories, further validating their effectiveness as competitive ensemble classifiers. Among traditional algorithms, Logistic Regression, SVM, and MLP also exhibited high AUC values (≥ 0.98), though with slightly more variation across intention levels. For example, K-Nearest Neighbors (KNN) recorded a lower AUC of 0.95 for the Low group, indicating potential limitations in handling imbalanced data or diffuse class boundaries. Naive Bayes, despite its relatively lower F1-scores in earlier analyses, achieved surprisingly strong AUCs—0.97 for High and Medium, and 0.96 for Low—suggesting that it effectively distinguishes between intention levels on a probabilistic level, though its calibration and precision-recall trade-offs may require further refinement. 4. Discussion antibiotic prescribing intentions by integrating psychological theory with explainable machine learning, addressing persistent gaps in antimicrobial stewardship research. Traditional models such as the Theory of Planned Behavior (TPB) and Self-Efficacy Theory often conceptualize prescribing behavior as a linear outcome of individual attitudes or beliefs. However, our findings challenge this assumption by demonstrating that, in regulated and high-pressure clinical environments, prescribing intentions are more strongly shaped by contextual and cognitive factors than by individual volitional control [21]. LASSO regression identified social support, cognitive processing, health beliefs, and knowledge and skills as the most robust predictors of rational prescribing intent. This suggests that institutional norms, analytical reasoning, and perceived clinical risk play a more critical role than previously assumed. These findings are consistent with evidence that external drivers such as peer behavior and patient expectations often override personal attitudes in clinical decisions [22][23]. Moreover, TPB-related constructs and self-efficacy showed weak or negative associations, especially in environments where autonomy is constrained by protocols or peer pressure. SHAP analysis revealed nonlinear relationships and interactions among predictors, offering deeper insight into how psychological dimensions operate in combination. For example, the positive effect of social support was significantly amplified in individuals with higher cognitive engagement—a dynamic often missed by conventional statistical models [24]. This underscores the added theoretical value of explainable AI tools not only for prediction but also for uncovering latent structures in behavioral data. Methodologically, the study demonstrates that machine learning can maintain high predictive accuracy while enhancing interpretability. Ensemble models such as Voting, Stacking, and XGBoost consistently outperformed traditional classifiers, especially in identifying subtle patterns among individuals with moderate or high intention. However, classifying healthcare workers with low prescribing intention remained challenging, possibly due to the diffuse nature of weak behavioral intent and overlaps with external constraints such as workload or administrative pressure [25]. Importantly, psychological constructs—especially those reflecting cognitive and contextual dimensions—proved significantly more predictive than demographic or usage-based data. This aligns with research showing that clinicians’ decision-making is often shaped more by environmental and cognitive stressors than by personal attributes [26][27]. These findings reinforce the importance of embedding behavioral science into digital health tools to enable more precise and personalized interventions. By identifying clinicians who are psychologically at risk of irrational prescribing, health systems can move beyond one-size-fits-all training and instead design targeted strategies aligned with individual motivational and cognitive profiles. Beyond antimicrobial use, these insights have broader implications for modeling healthcare behavior. Constructs such as risk perception, cognitive workload, and peer norms influence a wide range of clinical decisions, from diagnostics to treatment adherence. As AI becomes more integrated into health systems, ensuring that behavioral models are both empirically grounded and interpretable will be essential for building trust and achieving sustainable behavioral change [28][29]. This study contributes to that agenda by offering a scalable, explainable, and psychologically informed framework for understanding and influencing healthcare worker behavior in real-world clinical contexts [30]. Conclusion This study presents a theoretically integrated and methodologically innovative framework for predicting healthcare workers’ antibiotic prescribing intentions. By moving beyond single-theory models and incorporating constructs from cognitive, motivational, and environmental domains, we provide a more nuanced understanding of the behavioral drivers of clinical decision-making. The combined use of LASSO regression and SHAP analysis enables both high predictive performance and interpretability, bridging the gap between machine learning and behavioral science. Clinically, our findings lay the foundation for real-time, tailored interventions that identify and engage healthcare workers at risk of irrational prescribing. The interpretability of SHAP further enhances trust and usability, making AI-based decision support more transparent and actionable. Moreover, the framework we propose is adaptable to other domains of healthcare behavior, supporting the broader goal of precision behavioral health. Together, these contributions mark a meaningful step toward developing intelligent, context-sensitive, and psychologically informed tools for antimicrobial stewardship and beyond. Limitations While this study offers valuable insights into psychological and behavioral drivers of antibiotic prescribing, several limitations should be noted to guide interpretation and future research. First, although data were collected from multiple hospitals across diverse regions, all participating institutions were public and based in China. As such, caution is warranted when generalizing findings to other healthcare systems or private settings. However, the inclusion of geographically and professionally diverse participants enhances the relevance of our results within similar institutional contexts. Second, the cross-sectional design limits our ability to draw causal conclusions. While machine learning can reveal complex associations, future longitudinal or experimental studies would be beneficial to verify causal pathways and test intervention effects over time. Third, the use of self-reported behavioral intention—as opposed to actual prescribing behavior—may introduce response bias. Nonetheless, intention remains a widely accepted predictor of behavior in health psychology, and its use here provides a meaningful and feasible indicator in the absence of direct prescription data. Finally, while the psychological constructs were selected based on established theories and validated through expert review, some latent overlap may exist. Future studies could further refine measurement instruments or incorporate qualitative data to deepen contextual understanding. Despite these limitations, the study introduces a scalable and interpretable modeling framework with strong predictive performance and practical utility, offering a solid foundation for future refinement and clinical application. Declarations Human Ethics and Consent to Participate Declarations This study was approved by the Ethics Committee of Beijing Tongren Hospital. It involved an anonymous, minimal-risk survey and did not collect any personally identifiable information, including but not limited to national identification numbers, names, contact details, addresses, or social media accounts. Participation was voluntary, and informed consent was implied through the completion and submission of the questionnaire. The study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. Clinical trial number: not applicable Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Data availability statement Data is provided within the manuscript or supplementary information files" in the submission system Author Contribution Han Le conceptualized and designed the study, conducted data analysis, performed literature review, and drafted the manuscript. Peng Xian (corresponding author) contributed to study conception and design, scale development, literature review, manuscript revision, and provided critical guidance throughout the revision process. Xiao Liu participated in literature review, manuscript review, and revision. Chang Yue, Liu Ying, Ren Jie, Ma Lijun, Long Yunjuan, Wang Xiaosong, Liu Yunshuang, Ren Li, Jing Fang ,Liang ling ,Tang hui , Zhou Yuan and Liu Qi were responsible for data collection. All authors read and approved the final manuscript. References O'Neill, J. (2016). Tackling drug-resistant infections globally: Final report and recommendations . Review on Antimicrobial Resistance. https://amr-review.org/sites/default/files/160518_Final%20paper_with%20cover.pdf World Health Organization. (2014). Antimicrobial resistance: Global report on surveillance 2014 . WHO Press. https://www.who.int/publications/i/item/9789241564748 Klein, E. Y., Van Boeckel, T. P., Martinez, E. M., Pant, S., Gandra, S., Levin, S. A., ... & Laxminarayan, R. (2018). Global increase and geographic convergence in antibiotic consumption between 2000 and 2015. Proceedings of the National Academy of Sciences, 115 (15), E3463–E3470. https://doi.org/10.1073/pnas.1717295115 Xiao, Y., & Li, L. (2013). Legislation of clinical antibiotic use in China. The Lancet Infectious Diseases, 13 (3), 189–191. https://doi.org/10.1016/S1473-3099(12)70318-2 Yin, X., Song, F., Gong, Y., Tu, X., Wang, Y., & Liu, F. (2013). A systematic review of antibiotic utilization in China. Journal of Antimicrobial Chemotherapy, 68 (11), 2445–2452. https://doi.org/10.1093/jac/dkt129 Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50 (2), 179–211.https://doi.org/10.1016/0749-5978(91)90020-T Rosenstock, I. M. (1974). Historical origins of the health belief model. Health Education Monographs, 2 (4), 328–335. https://doi.org/10.1177/109019817400200403 Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84 (2), 191–215. https://doi.org/10.1037/0033-295X.84.2.191 Park, K. (2007). Social support for stress prevention in hospital settings. The Journal of the Royal Society for the Promotion of Health, 127 (6), 260-264. https://doi.org/10.1177/1466424007084066 Han, L., Liu, X., Liu, Y., Zhang, Y., & Zhou, J. (2025). Assessing social support’s mediating effects on rational antimicrobial prescribing: A structural equation modeling study of healthcare professionals’ behavior. Scientific Reports, 15 , 8242. https://doi.org/10.1038/s41598-025-92357-2 West, C. P., Dyrbye, L. N., & Shanafelt, T. D. (2018). Physician burnout: Contributors, consequences and solutions. Journal of Internal Medicine, 283 (6), 516–529. https://doi.org/10.1111/joim.12752 Tannenbaum, S. I., & Cerasoli, C. P. (2013). Do team and individual debriefs enhance performance? A meta-analysis. Human Factors, 55 (1), 231–245. https://doi.org/10.1177/0018720812448394 Hall, L. H., Johnson, J., Watt, I., Tsipa, A., & O’Connor, D. B. (2016). Healthcare staff wellbeing, burnout, and patient safety: A systematic review. BMJ Open, 6 (2), e009714. https://doi.org/10.1136/bmjopen-2015-009714 Shanafelt, T. D., & Noseworthy, J. H. (2017). Executive leadership and physician well-being: Nine organizational strategies to promote engagement and reduce burnout. Mayo Clinic Proceedings, 92 (1), 129–146.https://doi.org/10.1016/j.mayocp.2016.10.004 Glanz, K., Rimer, B. K., & Viswanath, K. (Eds.). (2015). Health behavior: Theory, research, and practice (5th ed.). Jossey-Bass. https://www.wiley.com/en-us/Health+Behavior%3A+Theory%2C+Research%2C+and+Practice%2C+5th+Edition-p-9781118628980 Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50 (2), 179–211. https://doi.org/10.1016/0749-5978(91)90020-T (Duplicate of #6, keep one in final list) DeVellis, R. F. (2016). Scale development: Theory and applications (4th ed.). Sage Publications. https://us.sagepub.com/en-us/nam/scale-development/book233401 Lynn, M. R. (1986). Determination and quantification of content validity. Nursing Research, 35 (6), 382–385. https://doi.org/10.1097/00006199-198611000-00017 Boateng, G. O., Neilands, T. B., Frongillo, E. A., Melgar-Quiñonez, H. R., & Young, S. L. (2018). Best practices for developing and validating scales for health, social, and behavioral research: A primer. Frontiers in Public Health, 6 , 149. https://doi.org/10.3389/fpubh.2018.00149 Hinkin, T. R. (1998). A brief tutorial on the development of measures for use in survey questionnaires. Organizational Research Methods, 1 (1), 104–121. https://doi.org/10.1177/109442819800100106 Lambert, B. L., Salmon, J. W., Stubbings, J., Gilomen-Study, G., Valuck, R. J., & Kezlarian, K. (1997). Factors associated with antibiotic prescribing in a managed care setting: an exploratory investigation. Social Science & Medicine , 45(12), 1767–1779. https://doi.org/10.1016/S0277-9536(97)00108-1 Basu, S., Santra, S., Bhatnagar, N., & Laul, A. (2022). Outpatient antibiotic prescribing behavior and their psychosocial predictors among early-career clinicians in Delhi, India. International Journal of Academic Medicine , 8, 11–15. https://doi.org/10.4103/ijam.ijam_156_20 Liu, C., Liu, C., Wang, D., & Zhang, X. (2019). Intrinsic and external determinants of antibiotic prescribing: a multi-level path analysis of primary care prescriptions in Hubei, China. Antimicrobial Resistance and Infection Control , 8(1). https://doi.org/10.1186/s13756-019-0592-5 Laka, M., & Milazzo, A. (2021). Inappropriate antibiotic prescribing: understanding clinicians. Australian Health Review . https://doi.org/10.1071/AH21197 Chukwu, E. E., Oladele, D. A., Enwuru, C. A., Gogwan, P. L., Abuh, D., Audu, R. A., & Ogunsola, F. T. (2021). Antimicrobial resistance awareness and antibiotic prescribing behavior among healthcare workers in Nigeria: a national survey. BMC Infectious Diseases , 21(1). https://doi.org/10.1186/s12879-020-05689-x Mbuthia, O. W., Ndonga, E. N., Odiwour, S. O., & Muraguri, M. (2020). Factors associated with antibiotic prescription among healthcare workers at tertiary hospitals in Nairobi County, Kenya. Annals of Health Research . https://doi.org/10.30442/ahr.0602-05-78 Rakhshani, N. S., Kaljee, L., Khan, M. I., Prentiss, T., Turab, A., Mustafa, A., Khalid, M., & Zervos, M. (2022). A formative assessment of antibiotic dispensing/prescribing practices and knowledge and perceptions of antimicrobial resistance among healthcare workers in Lahore, Pakistan. Antibiotics , 11(10), 1418. https://doi.org/10.3390/antibiotics11101418 Ashiru-Oredope, D., Hopkins, S., Vasandani, S., Umoh, E., Oloyede, O., Nilsson, A., Kinsman, J., Elsert, L., & Monnet, D. (2021). Healthcare workers’ knowledge, attitudes and behaviours with respect to antibiotics, antibiotic use and antibiotic resistance across 30 EU/EEA countries in 2019. Eurosurveillance , 26(12). https://doi.org/10.2807/1560-7917.ES.2021.26.12.1900633 Wang, S. Y., Cantarelli, P., Groene, O., Stargardt, T., & Bellé, N. (2022). Patient expectations do matter: experimental evidence on antibiotic prescribing decisions. The European Journal of Public Health , 32(Supplement_3). https://doi.org/10.1093/eurpub/ckac131.188 Kaawa-Mafigiri, D., Ekusai-Sebatta, D., Rutebemberwa, E., Sserwanga, A., Kitutu, F., Kapisi, J., Hopkins, H., Salami, O., Nkeramahame, J., Olliaro, P., & Horgan, P. (2023). A qualitative assessment of a training and communication intervention on antibiotic prescription practices among health workers and outpatients at public health facilities in Uganda. Clinical Infectious Diseases , 77(Supplement_4), S191–S198. https://doi.org/10.1093/cid/ciad329 Additional Declarations No competing interests reported. Supplementary Files forscientific.xlsx Cite Share Download PDF Status: Published Journal Publication published 28 Jan, 2026 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 16 Oct, 2025 Reviews received at journal 15 Oct, 2025 Reviewers agreed at journal 14 Oct, 2025 Reviewers agreed at journal 18 Sep, 2025 Reviewers agreed at journal 31 Jul, 2025 Reviews received at journal 22 Jul, 2025 Reviewers agreed at journal 07 Jul, 2025 Reviewers agreed at journal 18 Jun, 2025 Reviewers agreed at journal 17 Jun, 2025 Reviewers invited by journal 16 Jun, 2025 Editor assigned by journal 16 Jun, 2025 Editor invited by journal 09 Jun, 2025 Submission checks completed at journal 05 Jun, 2025 First submitted to journal 28 May, 2025 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-6768023","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":472652847,"identity":"c8b1e0a4-1b48-4667-8df7-c2253b3c3600","order_by":0,"name":"Le Han","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"Le","middleName":"","lastName":"Han","suffix":""},{"id":472652848,"identity":"fe9e202a-0b21-4d7f-829c-91a01c040fc0","order_by":1,"name":"Peng Xian","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/ElEQVRIiWNgGAWjYDACZiQ2YwODDQ8/fwOxWtjAWtJkJGccINY6iJbDNgYNCfgVGhxnfva4oOKOXYN87+OXM2rO8xgwHGD88DEHtxbJZjZz4xlnniU3sLGbWW44dpvHnLmBWXLmNtxa+JkZzKR52w4nM7CxsRk+bLjNY9lwgI2ZF48WNmb2b9K8/+BazvEYHEjAr4WfmQdoS8NhO6AW5ocbGw4Q1iLZzFMmPePY4QQGtjQ2xhnHknkkZxxsxusXg/PHt0kX1By2Z2A+xvyxp8bOnp+/+eCHj3i0gAAoNhP3H2Bgk4DwgdFDCIC02IPoDwSVjoJRMApGwYgEAFK4S0v8bZ8TAAAAAElFTkSuQmCC","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":true,"prefix":"","firstName":"Peng","middleName":"","lastName":"Xian","suffix":""},{"id":472652849,"identity":"6d2d30ba-9b4d-4cf3-9261-f04888d01dd5","order_by":2,"name":"Xiao Liu","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiao","middleName":"","lastName":"Liu","suffix":""},{"id":472652850,"identity":"04c0880c-f151-498b-b6fb-396382aa3833","order_by":3,"name":"Ying Liu","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Liu","suffix":""},{"id":472652851,"identity":"24a7ca06-791a-4632-8e5a-78bc6915de60","order_by":4,"name":"yue chang","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"yue","middleName":"","lastName":"chang","suffix":""},{"id":472652852,"identity":"37ec4494-4207-4f23-aa7b-ebc8010314e4","order_by":5,"name":"lijun ma","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"lijun","middleName":"","lastName":"ma","suffix":""},{"id":472652853,"identity":"7ee80356-9b5e-45ca-a6c5-7cc24c53c53e","order_by":6,"name":"jie ren","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"jie","middleName":"","lastName":"ren","suffix":""},{"id":472652854,"identity":"bc5dbd84-7ccc-4d31-b140-2c6e3b5e96cc","order_by":7,"name":"yunshuang liu","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"yunshuang","middleName":"","lastName":"liu","suffix":""},{"id":472652855,"identity":"bd3804c2-1909-414a-926a-af33b86efa0e","order_by":8,"name":"xiaosong wang","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"xiaosong","middleName":"","lastName":"wang","suffix":""},{"id":472652856,"identity":"8d6cd479-564a-4c5a-86dc-1f1d20702cda","order_by":9,"name":"yujuan long","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"yujuan","middleName":"","lastName":"long","suffix":""},{"id":472652857,"identity":"d336579c-05ee-4423-a0b1-17634db0444c","order_by":10,"name":"yuan zhou","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"yuan","middleName":"","lastName":"zhou","suffix":""},{"id":472652858,"identity":"a0eb45fa-9776-4c62-859f-561ce8a779d9","order_by":11,"name":"li ren","email":"","orcid":"","institution":"Beijing Tongren Hospital","correspondingAuthor":false,"prefix":"","firstName":"li","middleName":"","lastName":"ren","suffix":""},{"id":472652859,"identity":"614a70ba-fae5-4c2d-9bfb-86a1d30a1c4f","order_by":12,"name":"Fang jin","email":"","orcid":"","institution":"zhangjiakou fourth hospital","correspondingAuthor":false,"prefix":"","firstName":"Fang","middleName":"","lastName":"jin","suffix":""},{"id":472652860,"identity":"5ba37a92-eeeb-4c07-868f-c91fe5faee1b","order_by":13,"name":"Ling liang","email":"","orcid":"","institution":"zhangjiakou fourth hospital","correspondingAuthor":false,"prefix":"","firstName":"Ling","middleName":"","lastName":"liang","suffix":""},{"id":472652861,"identity":"1fd451b9-37ac-48d4-ad4c-3dd7996480f7","order_by":14,"name":"Hui Tang","email":"","orcid":"","institution":"Beijing Aiyuhua Hospital for Children and Women","correspondingAuthor":false,"prefix":"","firstName":"Hui","middleName":"","lastName":"Tang","suffix":""},{"id":472652862,"identity":"9b2834a4-8434-4e4f-b348-8df379043980","order_by":15,"name":"Qi Liu","email":"","orcid":"","institution":"Rehabilitation hospital affiliated to National ResearchCenter for Rehabiitation Technical Aids","correspondingAuthor":false,"prefix":"","firstName":"Qi","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2025-05-28 12:23:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6768023/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6768023/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-026-37495-x","type":"published","date":"2026-01-28T15:58:02+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":85186315,"identity":"6e5c87b7-799d-44ab-bf28-09a111d371ad","added_by":"auto","created_at":"2025-06-23 08:11:27","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":77940,"visible":true,"origin":"","legend":"\u003cp\u003eLasso Feature Importance(Coefficients)\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6768023/v1/4d7f24a5a9aaa0ba5a2f1d93.png"},{"id":85185037,"identity":"ef4d5f75-9da6-4eab-b219-d2f1b0fbb477","added_by":"auto","created_at":"2025-06-23 08:03:27","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":147960,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP-Based Feature Importance and Nonlinear Relationships in Predicting Rational Antibiotic Use Intention\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6768023/v1/523fb4b5be8f8bb696c1a1f5.png"},{"id":85185040,"identity":"4bf914da-658a-4627-bb01-fc5aaf34fdd9","added_by":"auto","created_at":"2025-06-23 08:03:27","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":113498,"visible":true,"origin":"","legend":"\u003cp\u003eBootstrapped Logistic Regression Results for Psychological Predictors of Rational Antibiotic Prescribing Intention\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6768023/v1/2c7f592b831ff9a8533376f1.png"},{"id":85185047,"identity":"ddc277ae-5dee-476d-ac4b-8589914c7dc7","added_by":"auto","created_at":"2025-06-23 08:03:27","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":164715,"visible":true,"origin":"","legend":"\u003cp\u003eClassification Performance of Eleven Machine Learning Models Across Behavioral Intention Levels\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6768023/v1/ed2daf5466f5700d7415b6e7.png"},{"id":85186316,"identity":"c8922545-5939-4afc-9e7a-6dad06f79b08","added_by":"auto","created_at":"2025-06-23 08:11:27","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":275150,"visible":true,"origin":"","legend":"\u003cp\u003eROC Curves for 11 Machine Learning Models Across Intention Level\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6768023/v1/d254d96e2d14af4d5d7a6c47.png"},{"id":101690418,"identity":"843cdf24-dc43-440d-910a-13d23d11e169","added_by":"auto","created_at":"2026-02-02 16:01:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1432039,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6768023/v1/ef54fa93-ac31-4d15-940f-be1219cc9874.pdf"},{"id":85185045,"identity":"70dd026c-ab46-4dce-aa04-e2d23d3a302a","added_by":"auto","created_at":"2025-06-23 08:03:27","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":288762,"visible":true,"origin":"","legend":"","description":"","filename":"forscientific.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6768023/v1/99151c438b4da07f09c61230.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"AI-Based Precision Prediction of Healthcare Workers’ Antibiotic Use Intentions: Multi-theoretical Psychological and Behavioral Modeling","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eAntimicrobial resistance (AMR) is among the most urgent global health threats, undermining routine medical care and placing growing strain on healthcare systems. Despite international and national policy efforts, including audit systems, clinical guidelines, and prescription regulations, inappropriate antibiotic use remains widespread. In both high-income and low- and middle-income countries, up to half of all prescriptions are estimated to be irrational.\u0026sup1;⁻\u0026sup3; In China, the overuse of antibiotics\u0026mdash;exceeding 60% of prescriptions before 2011\u0026mdash;prompted the launch of the National Special Rectification Campaign.⁴ While progress has been made, irrational prescribing persists, particularly in resource-limited and high-pressure clinical environments.⁵\u003c/p\u003e \u003cp\u003e Conventional interventions have focused on improving knowledge and enforcing guidelines. However, growing evidence suggests that prescribing decisions are shaped not only by clinical competence, but also by psychological, social, and institutional pressures. Clinicians\u0026rsquo; choices may be influenced by diagnostic uncertainty, time constraints, perceived patient expectations, fear of complications or complaints, and local professional norms. These cognitive and psychosocial factors are deeply embedded in clinical routines but remain insufficiently addressed by current stewardship strategies.\u003c/p\u003e \u003cp\u003eBehavioral science provides useful frameworks for understanding these drivers. Models such as the Theory of Planned Behavior, Health Belief Model, Self-Efficacy Theory, and Social Support Theory have identified key constructs\u0026mdash;self-efficacy, perceived control, risk perception, and normative beliefs\u0026mdash;as significant predictors of prescribing intention.⁶⁻⁹ However, most existing studies adopt narrow theoretical scopes, rely on linear analytical methods, and are based on small or single-center datasets. These limitations restrict their capacity to capture the complex, non-linear, and context-specific nature of clinical decision-making. In a previous study, we used structural equation modeling to examine how social support influences rational antibiotic prescribing among healthcare workers.\u0026sup1;⁰ Our findings revealed several mediating pathways, including cognitive processing and professional expectations. However, the study was limited by its single-center design and the linear assumptions of the model.\u003c/p\u003e \u003cp\u003eTo address these gaps, the present study integrates multi-theoretical behavioral modeling with explainable machine learning (ML). Using data from four hospitals across different regions of China, we model the psychological and contextual factors influencing healthcare workers\u0026rsquo; intentions to use antibiotics appropriately. Explainable ML tools\u0026mdash;such as SHAP values and LASSO regression\u0026mdash;allow us to identify and interpret the most influential behavioral variables and their interactions, offering both predictive strength and theoretical insight. The aim of this study is to develop an interpretable model to identify healthcare workers at higher risk of inappropriate antibiotic prescribing. Based on individual behavioral profiles, we further aim to design an intelligent system capable of delivering tailored educational and psychological interventions in real time. By combining theory-informed modeling, explainable ML, and adaptive intervention strategies, this work moves antimicrobial stewardship toward greater precision, scalability, and clinical relevance.\u003c/p\u003e \u003cp\u003eWhile this study focuses on antibiotic use, the underlying behavioral mechanisms it examines\u0026mdash;such as self-efficacy, perceived norms, and cognitive burden\u0026mdash;are not unique to antimicrobial decisions. These factors also influence broader clinical behaviors and healthcare workers\u0026rsquo; own health-related practices. \u0026sup1;\u0026sup1;⁻\u0026sup1;\u0026sup3; As concerns around clinician burnout, diagnostic error, and system-level stress continue to grow, developing scalable, psychologically informed tools to support frontline providers is increasingly recognized as a public health priority.\u0026sup1;⁴\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cp\u003e2.1 Study Design and Participants\u003c/p\u003e\n\u003cp\u003eThis study employed a multi-center, cross-sectional survey design conducted between January and December 2024 across public hospitals in China. Participating institutions were located in different regions and represented varying levels of the healthcare system, including both urban and semi-urban settings. Eligible participants were frontline clinical staff engaged in direct patient care and responsible for antimicrobial decision-making.\u003c/p\u003e\n\u003cp\u003eInclusion criteria were:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003ecurrently working in clinical departments.\u003c/li\u003e\n \u003cli\u003eaged over 18 years.\u003c/li\u003e\n \u003cli\u003eholding Chinese nationality.\u003c/li\u003e\n \u003cli\u003enative Mandarin speakers.\u003c/li\u003e\n \u003cli\u003evoluntary participation with signed informed consent.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eExclusion criteria included:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eobservable cognitive impairment that could affect the ability to complete the questionnaire.\u003c/li\u003e\n \u003cli\u003eserious communication difficulties due to sensory or neurological impairments (e.g., severe hearing or visual impairment).\u003c/li\u003e\n \u003cli\u003erefusal or inability to provide informed consent.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e2.2 Scale Development\u003c/p\u003e\n\u003cp\u003eThe development of the measurement scale was informed by established scale construction principles and a theory-based conceptual framework. Eight latent constructs were identified through a review of major behavioral and cognitive theories relevant to antimicrobial prescribing among healthcare professionals. These included the Theory of Planned Behavior, the Health Belief Model, Rational Action Theory, Self-Efficacy Theory, Social Support Theory, cognitive processing, and knowledge- and skill-based models.\u0026sup1;⁵⁻\u0026sup1;⁶ Each construct was operationalized through four items designed to reflect core theoretical dimensions and clinical relevance, resulting in a 32-item draft scale. Where possible, items were adapted from previously published instruments.\u0026sup1;⁷ For constructs lacking suitable items, new statements were developed to align with the clinical context of Chinese hospital settings. All items were written in Chinese and employed a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree).The initial item pool was reviewed by a multidisciplinary expert panel, including specialists in infectious diseases, pharmacy, behavioral psychology, and clinical education. Following standard procedures for content validation, experts independently assessed each item\u0026rsquo;s clarity, relevance, and theoretical alignment.\u0026sup1;⁸⁻\u0026sup2;⁰ Revisions were made to improve semantic precision, item distinctiveness, and contextual appropriateness based on their feedback.\u003c/p\u003e\n\u003cp\u003e2.3 Machine Learning Procedures\u003c/p\u003e\n\u003cp\u003eOverview and Platform\u003c/p\u003e\n\u003cp\u003eAll machine learning procedures were implemented in Python 3.10 using the Google Colaboratory (Colab) environment. Commonly used libraries supported the workflow: \u003cem\u003escikit-learn\u003c/em\u003e for model development and evaluation; \u003cem\u003eXGBoost\u003c/em\u003e, \u003cem\u003eLightGBM\u003c/em\u003e, and \u003cem\u003eCatBoost\u003c/em\u003e for gradient boosting algorithms; and \u003cem\u003eSHAP\u003c/em\u003e for explainability analyses. Data preprocessing and manipulation were conducted using \u003cem\u003ePandas\u003c/em\u003e and \u003cem\u003eNumPy\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eTarget Variable Definition\u003c/p\u003e\n\u003cp\u003eThe primary outcome variable was behavioral intention (BI) to rationally prescribe antibiotics, calculated as the mean of four Likert-scale items capturing future-oriented prescribing intentions. This continuous variable (BI_score) was converted into a three-level categorical outcome (BI_category) based on sample quantiles:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eLow Intention: BI_score \u0026le; 33rd percentile\u003c/li\u003e\n \u003cli\u003eMedium Intention: 33rd\u0026ndash;67th percentile\u003c/li\u003e\n \u003cli\u003eHigh Intention: BI_score \u0026ge; 67th percentile\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThis quantile-based discretization ensured balanced class distributions and allowed for meaningful differentiation in the strength of prescribing intentions. The resulting categorical variable served as the dependent variable in all classification models.\u003c/p\u003e\n\u003cp\u003eFeature Standardization and Selection\u003c/p\u003e\n\u003cp\u003eIndependent variables comprised composite scores for each psychological construct (e.g., TPB, HBM, knowledge-skill), computed as the mean of four items per construct. Prior to model training, all features were standardized using z-scores via the \u003cem\u003eStandardScaler\u003c/em\u003e module in \u003cem\u003escikit-learn\u003c/em\u003e. To reduce dimensionality and enhance interpretability, LASSO regression with cross-validation was employed for feature selection. Only variables with non-zero coefficients were retained for inclusion in downstream models.\u003c/p\u003e\n\u003cp\u003eModel Training and Evaluation\u003c/p\u003e\n\u003cp\u003eEleven supervised machine learning algorithms were trained to classify BI levels:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eLogistic Regression\u003c/li\u003e\n \u003cli\u003eSupport Vector Machine (SVM)\u003c/li\u003e\n \u003cli\u003eRandom Forest\u003c/li\u003e\n \u003cli\u003eXGBoost\u003c/li\u003e\n \u003cli\u003eLightGBM\u003c/li\u003e\n \u003cli\u003eCatBoost\u003c/li\u003e\n \u003cli\u003eK-Nearest Neighbors (KNN)\u003c/li\u003e\n \u003cli\u003eNaive Bayes\u003c/li\u003e\n \u003cli\u003eGradient Boosting\u003c/li\u003e\n \u003cli\u003eAdaBoost\u003c/li\u003e\n \u003cli\u003eVoting and Stacking Ensembles\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe dataset was randomly split into training (80%) and testing (20%) subsets using stratified sampling to preserve the distribution of BI categories. Hyperparameter tuning was performed via grid search or internal validation, depending on the algorithm. Model performance was evaluated on the test set using precision, recall, F1-score, and macro-averaged ROC-AUC.\u003c/p\u003e\n\u003cp\u003e2.5 Model Explainability\u003c/p\u003e\n\u003cp\u003eTo enhance interpretability, several explainability techniques were applied:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eLASSO regression coefficients were examined to assess the relative importance of psychological predictors in a linear model.\u003c/li\u003e\n \u003cli\u003eSHAP (SHapley Additive Explanations) values were calculated to visualize the contribution and directionality of each predictor across models.\u003c/li\u003e\n \u003cli\u003eSHAP interaction values were used to explore potential interactions between psychological constructs.\u003c/li\u003e\n \u003cli\u003eBootstrapped logistic regression (1,000 resamples) was conducted to estimate 95% confidence intervals for predictor effects in binary classification tasks (High vs. Low intention groups).\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Demographic Analysis\u003c/h2\u003e \u003cp\u003eA total of 1,294 questionnaires were distributed across participating sites. Of these, 1,135 valid responses were received, yielding an effective response rate of 87.7%.Among the respondents, the majority were female (81.32%, n = 923), while male participants accounted for 18.68% (n = 212). The age distribution was as follows: 26–35 years (38.85%) comprised the largest group, followed by 36–45 years (27.84%), 18–25 years (22.82%), and 46–55 years (10.22%). Only a small proportion (0.26%) were older than 56 years.\u003c/p\u003e \u003cp\u003eIn terms of educational background, 45.46% of participants held a bachelor’s degree, 22.03% had education below the bachelor’s level, 20.7% held a master’s degree, 11.54% had completed a doctoral degree, and 0.26% were postdoctoral researchers. Respondents represented a range of clinical departments. Internal medicine was the most common (28.55%), followed by surgery (22.29%), otolaryngology (12.95%), ophthalmology (12.86%), obstetrics and gynecology (7.31%), intensive care (7.75%), emergency medicine (4.41%), dentistry (2.73%), and dermatology (1.15%). Regarding clinical experience, 34.45% had 16–25 years of working experience, 31.81% had more than 25 years, 17.0% had less than 5 years, and 16.65% reported 5–15 years of experience. These demographic characteristics are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\u003cdiv class=\"gridtable\"\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=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\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\u003eDemographic Information of Respondents in the Survey\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCategory\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCount\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePercent (%)\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eGENDER\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e212\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e18.68\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e923\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e81.32\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003eAGE\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18–25\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e259\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22.82\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26–35\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e441\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e38.85\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36–45\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e316\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e27.84\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e46–55\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e116\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10.22\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOver 56\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.26\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003eEDUCATION\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLess than Bachelor's degree\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e250\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22.03\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBachelor’s degree\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e516\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e45.46\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMaster’s degree\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e235\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20.7\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDoctoral degree (PhD)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e131\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e11.54\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePostdoctoral\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.26\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"8\" rowspan=\"9\"\u003e \u003cp\u003eDEPARTMENT\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInternal Medicine\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e324\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28.55\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSurgery\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e253\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22.29\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOphthalmology\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e146\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12.86\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOtolaryngology\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e147\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12.95\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eObstetrics and Gynecology\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e83\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.31\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIntensive Care Unit (ICU)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e88\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.75\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDentistry\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e31\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.73\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEmergency Medicine\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e50\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.41\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDermatology\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.15\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eWorking-experience\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt; 5 years\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e194\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5-15years\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e189\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16.65\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e16-25years\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e391\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e34.45\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026gt; 25years\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e361\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e31.81\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e \u003cp\u003e\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Feature Selection and Interpretation\u003c/h2\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003e3.2.1 Feature Selection and Behavioral Interpretation Using LASSO\u003c/h2\u003e \u003cp\u003eLASSO regression with five-fold cross-validation was used to identify key psychological predictors of healthcare workers’ intention to rationally prescribe antibiotics. All seven constructs were retained, though their contributions varied in both strength and direction (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Social support (β = 0.2564) was the strongest predictor, underscoring the role of institutional culture and peer reinforcement in shaping prescribing behavior. Healthcare workers embedded in supportive, stewardship-oriented environments may be more inclined to follow rational antibiotic practices. This aligns with organizational psychology literature, which emphasizes the behavioral impact of social norms and workplace context. The Health Belief Model (β = 0.1634) was also a significant predictor. Perceived susceptibility, severity, and benefits appear to meaningfully influence prescribing intent—particularly in settings where risk perception and clinical outcomes are closely linked. Cognitive processing (β = 0.1404) and knowledge and skills (β = 0.1023) contributed positively, reflecting the importance of analytical reasoning and clinical competence. These findings support previous work linking reflective decision-making and content knowledge with more appropriate antibiotic use.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn contrast, self-efficacy (β = − 0.0383) and TPB constructs (β = − 0.0572) showed weak negative associations. Although these constructs are traditionally associated with behavioral intention, their reduced influence here may reflect the dominant role of institutional policies and norms, which constrain individual agency in tightly regulated clinical environments. Rational Action Theory (β = 0.0456) showed a modest positive effect, suggesting that while deliberate cost-benefit reasoning contributes to decision-making, its role may be limited when other structural or perceptual factors are more salient.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section3\"\u003e \u003ch2\u003e3.2.2 Psychological Feature Contributions and Interactions in SHAP Analysis\u003c/h2\u003e \u003cp\u003eTo further interpret the predictive model, SHapley Additive exPlanations (SHAP) analysis was applied to the top-performing XGBoost classifier. SHAP quantifies the individual contribution of each psychological construct to model predictions, enhancing transparency and interpretability. The SHAP summary plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea) illustrates the magnitude, direction, and distribution of each feature’s influence. Cognitive processing (CG_score), social support (SS_score), knowledge and skills (KS_score), and health belief model (HBM_score) constructs showed the highest contributions. Among these, cognitive processing had the strongest positive effect, particularly at higher values, underscoring the role of analytical reasoning in rational prescribing. In contrast, TPB, RAT, and SET constructs were excluded from visualization due to minimal importance, reflecting their limited influence in the final model. The mean absolute SHAP values (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb) confirmed this ranking, with cognitive processing emerging as the most influential factor, followed closely by social support and knowledge and skills. This highlights a hierarchy of predictors in which cognitive and contextual variables outweigh traditional volitional models in this behavioral context.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSHAP dependence plots revealed non-linear relationships and potential interactions. In Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec, social support displayed a positive but nonlinear association with rational prescribing intentions, with stronger effects at higher scores. Color gradients indicating cognitive processing scores suggest an interaction, where the effect of social support may be modulated by analytical engagement. Similarly, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed shows that the influence of knowledge and skills may also be shaped by social context. Figures\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ee and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ef explore HBM and cognitive processing effects respectively, both demonstrating threshold-based increases in predicted intention at moderate to high levels.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Bootstrapped Logistic Regression Analysis of Psychological Predictors\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents the results of bootstrapped logistic regression, displaying the median coefficients and 95% confidence intervals (CIs) for seven psychological constructs in relation to healthcare workers’ intentions to prescribe antibiotics rationally. Four constructs showed statistically significant positive associations, with CIs that did not cross zero. Social support (SS_score) emerged as the strongest predictor (Median = 1.72), underscoring the influence of institutional context and peer reinforcement in promoting stewardship-aligned behavior. Cognitive processing (CG_score, Median = 1.42) and knowledge and skills (KS_score, Median = 1.40) followed closely, highlighting the importance of analytical engagement and clinical competence in prescribing decisions. The health belief model (HBM_score, Median = 1.03) also showed a significant positive association, indicating that perceptions of risk and severity surrounding antimicrobial resistance shape prescribing intent. In contrast, constructs derived from the Theory of Planned Behavior (TPB_score, Median = − 0.20) and Self-Efficacy Theory (SET_score, Median = − 0.15) had confidence intervals that included zero, suggesting limited and statistically uncertain effects. While Rational Action Theory (RAT_score, Median = 0.20) had a weak positive coefficient, its wide CI overlapping zero indicates low robustness.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Comparative Performance of Machine Learning Models by Intention Level\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents a comparative evaluation of eleven machine learning models in classifying healthcare workers’ behavioral intentions—categorized as High, Medium, and Low—regarding rational antibiotic use. Model performance was assessed using precision (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea), recall (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb), and F1-score (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec). Ensemble models, including Voting, Stacking, and XGBoost, demonstrated consistently high performance in predicting High and Medium intention groups, with F1-scores exceeding 0.94. The Voting classifier achieved the highest overall balance, with F1-scores of 0.96 and 0.95 for High and Medium groups, respectively. Logistic Regression and Support Vector Machine (SVM) also performed well for High and Medium categories, although performance declined in classifying Low intention individuals. For example, SVM reached an F1-score of 0.97 for High intention but dropped to 0.76 for Low intention, suggesting reduced sensitivity to less distinct behavioral profiles. AdaBoost and Gradient Boosting showed similar performance patterns, with lower F1-scores for the Low category (0.76 and 0.79, respectively). Across all models, classification of the Low intention group proved more challenging. Even top-performing algorithms struggled to balance precision and recall. For instance, Naive Bayes achieved a high recall (0.92) but low precision (0.69) for the Low group, resulting in a moderate F1-score of 0.79—indicating a tendency to over-predict without sufficient specificity. Similarly, LightGBM and Multilayer Perceptron (MLP), despite solid overall performance, recorded F1-scores near 0.80 for the Low group, reflecting consistent difficulty in accurately identifying individuals with weaker behavioral intentions.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Discriminative Performance of Models Based on Multi-Class ROC Curves\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e illustrates the multi-class Receiver Operating Characteristic (ROC) curves for eleven machine learning models across three levels of behavioral intention: High, Medium, and Low. The Area Under the Curve (AUC) values are reported for each class, providing a quantitative assessment of each model’s ability to discriminate between intention categories.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eEnsemble models—particularly Voting, Stacking, and XGBoost—demonstrated consistently strong discriminative performance, each achieving AUC values of 0.99 across all three intention levels. These results highlight their robustness in capturing complex decision boundaries and maintaining balanced accuracy across class distributions. LightGBM also performed exceptionally well, with AUCs of 0.98 for the High and Medium groups, and a perfect 1.00 for the Low intention category—traditionally the most challenging to classify due to class imbalance. This suggests LightGBM’s capacity to detect nuanced distinctions in underrepresented subpopulations. AdaBoost and Gradient Boosting followed closely, with AUC values ≥ 0.98 across all categories, further validating their effectiveness as competitive ensemble classifiers. Among traditional algorithms, Logistic Regression, SVM, and MLP also exhibited high AUC values (≥ 0.98), though with slightly more variation across intention levels. For example, K-Nearest Neighbors (KNN) recorded a lower AUC of 0.95 for the Low group, indicating potential limitations in handling imbalanced data or diffuse class boundaries. Naive Bayes, despite its relatively lower F1-scores in earlier analyses, achieved surprisingly strong AUCs—0.97 for High and Medium, and 0.96 for Low—suggesting that it effectively distinguishes between intention levels on a probabilistic level, though its calibration and precision-recall trade-offs may require further refinement.\u003c/p\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eantibiotic prescribing intentions by integrating psychological theory with explainable machine learning, addressing persistent gaps in antimicrobial stewardship research. Traditional models such as the Theory of Planned Behavior (TPB) and Self-Efficacy Theory often conceptualize prescribing behavior as a linear outcome of individual attitudes or beliefs. However, our findings challenge this assumption by demonstrating that, in regulated and high-pressure clinical environments, prescribing intentions are more strongly shaped by contextual and cognitive factors than by individual volitional control [21].\u003c/p\u003e\u003cp\u003eLASSO regression identified social support, cognitive processing, health beliefs, and knowledge and skills as the most robust predictors of rational prescribing intent. This suggests that institutional norms, analytical reasoning, and perceived clinical risk play a more critical role than previously assumed. These findings are consistent with evidence that external drivers such as peer behavior and patient expectations often override personal attitudes in clinical decisions [22][23]. Moreover, TPB-related constructs and self-efficacy showed weak or negative associations, especially in environments where autonomy is constrained by protocols or peer pressure.\u003c/p\u003e\u003cp\u003eSHAP analysis revealed nonlinear relationships and interactions among predictors, offering deeper insight into how psychological dimensions operate in combination. For example, the positive effect of social support was significantly amplified in individuals with higher cognitive engagement—a dynamic often missed by conventional statistical models [24]. This underscores the added theoretical value of explainable AI tools not only for prediction but also for uncovering latent structures in behavioral data.\u003c/p\u003e\u003cp\u003eMethodologically, the study demonstrates that machine learning can maintain high predictive accuracy while enhancing interpretability. Ensemble models such as Voting, Stacking, and XGBoost consistently outperformed traditional classifiers, especially in identifying subtle patterns among individuals with moderate or high intention. However, classifying healthcare workers with low prescribing intention remained challenging, possibly due to the diffuse nature of weak behavioral intent and overlaps with external constraints such as workload or administrative pressure [25].\u003c/p\u003e\u003cp\u003eImportantly, psychological constructs—especially those reflecting cognitive and contextual dimensions—proved significantly more predictive than demographic or usage-based data. This aligns with research showing that clinicians’ decision-making is often shaped more by environmental and cognitive stressors than by personal attributes [26][27]. These findings reinforce the importance of embedding behavioral science into digital health tools to enable more precise and personalized interventions. By identifying clinicians who are psychologically at risk of irrational prescribing, health systems can move beyond one-size-fits-all training and instead design targeted strategies aligned with individual motivational and cognitive profiles.\u003c/p\u003e\u003cp\u003eBeyond antimicrobial use, these insights have broader implications for modeling healthcare behavior. Constructs such as risk perception, cognitive workload, and peer norms influence a wide range of clinical decisions, from diagnostics to treatment adherence. As AI becomes more integrated into health systems, ensuring that behavioral models are both empirically grounded and interpretable will be essential for building trust and achieving sustainable behavioral change [28][29]. This study contributes to that agenda by offering a scalable, explainable, and psychologically informed framework for understanding and influencing healthcare worker behavior in real-world clinical contexts [30].\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study presents a theoretically integrated and methodologically innovative framework for predicting healthcare workers’ antibiotic prescribing intentions. By moving beyond single-theory models and incorporating constructs from cognitive, motivational, and environmental domains, we provide a more nuanced understanding of the behavioral drivers of clinical decision-making. The combined use of LASSO regression and SHAP analysis enables both high predictive performance and interpretability, bridging the gap between machine learning and behavioral science.\u003c/p\u003e\u003cp\u003eClinically, our findings lay the foundation for real-time, tailored interventions that identify and engage healthcare workers at risk of irrational prescribing. The interpretability of SHAP further enhances trust and usability, making AI-based decision support more transparent and actionable. Moreover, the framework we propose is adaptable to other domains of healthcare behavior, supporting the broader goal of precision behavioral health. Together, these contributions mark a meaningful step toward developing intelligent, context-sensitive, and psychologically informed tools for antimicrobial stewardship and beyond.\u003c/p\u003e\u003cp\u003e \u003cb\u003eLimitations\u003c/b\u003e \u003c/p\u003e\u003cp\u003eWhile this study offers valuable insights into psychological and behavioral drivers of antibiotic prescribing, several limitations should be noted to guide interpretation and future research. First, although data were collected from multiple hospitals across diverse regions, all participating institutions were public and based in China. As such, caution is warranted when generalizing findings to other healthcare systems or private settings. However, the inclusion of geographically and professionally diverse participants enhances the relevance of our results within similar institutional contexts. Second, the cross-sectional design limits our ability to draw causal conclusions. While machine learning can reveal complex associations, future longitudinal or experimental studies would be beneficial to verify causal pathways and test intervention effects over time. Third, the use of self-reported behavioral intention—as opposed to actual prescribing behavior—may introduce response bias. Nonetheless, intention remains a widely accepted predictor of behavior in health psychology, and its use here provides a meaningful and feasible indicator in the absence of direct prescription data. Finally, while the psychological constructs were selected based on established theories and validated through expert review, some latent overlap may exist. Future studies could further refine measurement instruments or incorporate qualitative data to deepen contextual understanding.\u003c/p\u003e\u003cp\u003eDespite these limitations, the study introduces a scalable and interpretable modeling framework with strong predictive performance and practical utility, offering a solid foundation for future refinement and clinical application.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eHuman Ethics and Consent to Participate Declarations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Ethics Committee of Beijing Tongren Hospital. It involved an anonymous, minimal-risk survey and did not collect any personally identifiable information, including but not limited to national identification numbers, names, contact details, addresses, or social media accounts. Participation was voluntary, and informed consent was implied through the completion and submission of the questionnaire. The study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number:\u003c/strong\u003e not applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData is provided within the manuscript or supplementary information files\u0026quot; in the submission system\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eHan Le conceptualized and designed the study, conducted data analysis, performed literature review, and drafted the manuscript. Peng Xian (corresponding author) contributed to study conception and design, scale development, literature review, manuscript revision, and provided critical guidance throughout the revision process. Xiao Liu participated in literature review, manuscript review, and revision. Chang Yue, Liu Ying, Ren Jie, Ma Lijun, Long Yunjuan, Wang Xiaosong, Liu Yunshuang, Ren Li, Jing Fang ,Liang ling ,Tang hui , Zhou Yuan and Liu Qi were responsible for data collection. All authors read and approved the final manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eO\u0026apos;Neill, J. (2016). \u003cem\u003eTackling drug-resistant infections globally: Final report and recommendations\u003c/em\u003e. Review on Antimicrobial Resistance.\u003cbr\u003e https://amr-review.org/sites/default/files/160518_Final%20paper_with%20cover.pdf\u003c/li\u003e\n\u003cli\u003eWorld Health Organization. (2014). \u003cem\u003eAntimicrobial resistance: Global report on surveillance 2014\u003c/em\u003e. WHO Press.\u003cbr\u003e https://www.who.int/publications/i/item/9789241564748\u003c/li\u003e\n\u003cli\u003eKlein, E. Y., Van Boeckel, T. P., Martinez, E. M., Pant, S., Gandra, S., Levin, S. A., ... \u0026amp; Laxminarayan, R. (2018). Global increase and geographic convergence in antibiotic consumption between 2000 and 2015. \u003cem\u003eProceedings of the National Academy of Sciences, 115\u003c/em\u003e(15), E3463\u0026ndash;E3470.\u003cbr\u003e https://doi.org/10.1073/pnas.1717295115\u003c/li\u003e\n\u003cli\u003eXiao, Y., \u0026amp; Li, L. (2013). Legislation of clinical antibiotic use in China. \u003cem\u003eThe Lancet Infectious Diseases, 13\u003c/em\u003e(3), 189\u0026ndash;191. https://doi.org/10.1016/S1473-3099(12)70318-2\u003c/li\u003e\n\u003cli\u003eYin, X., Song, F., Gong, Y., Tu, X., Wang, Y., \u0026amp; Liu, F. (2013). A systematic review of antibiotic utilization in China. \u003cem\u003eJournal of Antimicrobial Chemotherapy, 68\u003c/em\u003e(11), 2445\u0026ndash;2452. https://doi.org/10.1093/jac/dkt129\u003c/li\u003e\n\u003cli\u003eAjzen, I. (1991). The theory of planned behavior. \u003cem\u003eOrganizational Behavior and Human Decision Processes, 50\u003c/em\u003e(2), 179\u0026ndash;211.https://doi.org/10.1016/0749-5978(91)90020-T\u003c/li\u003e\n\u003cli\u003eRosenstock, I. M. (1974). Historical origins of the health belief model. \u003cem\u003eHealth Education Monographs, 2\u003c/em\u003e(4), 328\u0026ndash;335. https://doi.org/10.1177/109019817400200403\u003c/li\u003e\n\u003cli\u003eBandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. \u003cem\u003ePsychological Review, 84\u003c/em\u003e(2), 191\u0026ndash;215. https://doi.org/10.1037/0033-295X.84.2.191\u003c/li\u003e\n\u003cli\u003ePark, K. (2007). Social support for stress prevention in hospital settings. \u003cem\u003eThe Journal of the Royal Society for the Promotion of Health, 127\u003c/em\u003e(6), 260-264. https://doi.org/10.1177/1466424007084066\u003c/li\u003e\n\u003cli\u003eHan, L., Liu, X., Liu, Y., Zhang, Y., \u0026amp; Zhou, J. (2025). Assessing social support\u0026rsquo;s mediating effects on rational antimicrobial prescribing: A structural equation modeling study of healthcare professionals\u0026rsquo; behavior. \u003cem\u003eScientific Reports, 15\u003c/em\u003e, 8242. https://doi.org/10.1038/s41598-025-92357-2\u003c/li\u003e\n\u003cli\u003eWest, C. P., Dyrbye, L. N., \u0026amp; Shanafelt, T. D. (2018). Physician burnout: Contributors, consequences and solutions. \u003cem\u003eJournal of Internal Medicine, 283\u003c/em\u003e(6), 516\u0026ndash;529. https://doi.org/10.1111/joim.12752\u003c/li\u003e\n\u003cli\u003eTannenbaum, S. I., \u0026amp; Cerasoli, C. P. (2013). Do team and individual debriefs enhance performance? A meta-analysis. \u003cem\u003eHuman Factors, 55\u003c/em\u003e(1), 231\u0026ndash;245.\u003cbr\u003e https://doi.org/10.1177/0018720812448394\u003c/li\u003e\n\u003cli\u003eHall, L. H., Johnson, J., Watt, I., Tsipa, A., \u0026amp; O\u0026rsquo;Connor, D. B. (2016). Healthcare staff wellbeing, burnout, and patient safety: A systematic review. \u003cem\u003eBMJ Open, 6\u003c/em\u003e(2), e009714. https://doi.org/10.1136/bmjopen-2015-009714\u003c/li\u003e\n\u003cli\u003eShanafelt, T. D., \u0026amp; Noseworthy, J. H. (2017). Executive leadership and physician well-being: Nine organizational strategies to promote engagement and reduce burnout. \u003cem\u003eMayo Clinic Proceedings, 92\u003c/em\u003e(1), 129\u0026ndash;146.https://doi.org/10.1016/j.mayocp.2016.10.004\u003c/li\u003e\n\u003cli\u003eGlanz, K., Rimer, B. K., \u0026amp; Viswanath, K. (Eds.). (2015). \u003cem\u003eHealth behavior: Theory, research, and practice\u003c/em\u003e (5th ed.). Jossey-Bass. https://www.wiley.com/en-us/Health+Behavior%3A+Theory%2C+Research%2C+and+Practice%2C+5th+Edition-p-9781118628980\u003c/li\u003e\n\u003cli\u003eAjzen, I. (1991). The theory of planned behavior. \u003cem\u003eOrganizational Behavior and Human Decision Processes, 50\u003c/em\u003e(2), 179\u0026ndash;211. https://doi.org/10.1016/0749-5978(91)90020-T \u003cem\u003e(Duplicate of #6, keep one in final list)\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eDeVellis, R. F. (2016). \u003cem\u003eScale development: Theory and applications\u003c/em\u003e (4th ed.). Sage Publications. https://us.sagepub.com/en-us/nam/scale-development/book233401\u003c/li\u003e\n\u003cli\u003eLynn, M. R. (1986). Determination and quantification of content validity. \u003cem\u003eNursing Research, 35\u003c/em\u003e(6), 382\u0026ndash;385. https://doi.org/10.1097/00006199-198611000-00017\u003c/li\u003e\n\u003cli\u003eBoateng, G. O., Neilands, T. B., Frongillo, E. A., Melgar-Qui\u0026ntilde;onez, H. R., \u0026amp; Young, S. L. (2018). Best practices for developing and validating scales for health, social, and behavioral research: A primer. \u003cem\u003eFrontiers in Public Health, 6\u003c/em\u003e, 149.\u003cbr\u003e https://doi.org/10.3389/fpubh.2018.00149\u003c/li\u003e\n\u003cli\u003eHinkin, T. R. (1998). A brief tutorial on the development of measures for use in survey questionnaires. \u003cem\u003eOrganizational Research Methods, 1\u003c/em\u003e(1), 104\u0026ndash;121.\u003cbr\u003e https://doi.org/10.1177/109442819800100106\u003c/li\u003e\n\u003cli\u003eLambert, B. L., Salmon, J. W., Stubbings, J., Gilomen-Study, G., Valuck, R. J., \u0026amp; Kezlarian, K. (1997). Factors associated with antibiotic prescribing in a managed care setting: an exploratory investigation. \u003cem\u003eSocial Science \u0026amp; Medicine\u003c/em\u003e, 45(12), 1767\u0026ndash;1779. https://doi.org/10.1016/S0277-9536(97)00108-1\u003c/li\u003e\n\u003cli\u003eBasu, S., Santra, S., Bhatnagar, N., \u0026amp; Laul, A. (2022). Outpatient antibiotic prescribing behavior and their psychosocial predictors among early-career clinicians in Delhi, India. \u003cem\u003eInternational Journal of Academic Medicine\u003c/em\u003e, 8, 11\u0026ndash;15. https://doi.org/10.4103/ijam.ijam_156_20\u003c/li\u003e\n\u003cli\u003eLiu, C., Liu, C., Wang, D., \u0026amp; Zhang, X. (2019). Intrinsic and external determinants of antibiotic prescribing: a multi-level path analysis of primary care prescriptions in Hubei, China. \u003cem\u003eAntimicrobial Resistance and Infection Control\u003c/em\u003e, 8(1). https://doi.org/10.1186/s13756-019-0592-5\u003c/li\u003e\n\u003cli\u003eLaka, M., \u0026amp; Milazzo, A. (2021). Inappropriate antibiotic prescribing: understanding clinicians. \u003cem\u003eAustralian Health Review\u003c/em\u003e. https://doi.org/10.1071/AH21197\u003c/li\u003e\n\u003cli\u003eChukwu, E. E., Oladele, D. A., Enwuru, C. A., Gogwan, P. L., Abuh, D., Audu, R. A., \u0026amp; Ogunsola, F. T. (2021). Antimicrobial resistance awareness and antibiotic prescribing behavior among healthcare workers in Nigeria: a national survey. \u003cem\u003eBMC Infectious Diseases\u003c/em\u003e, 21(1). https://doi.org/10.1186/s12879-020-05689-x\u003c/li\u003e\n\u003cli\u003eMbuthia, O. W., Ndonga, E. N., Odiwour, S. O., \u0026amp; Muraguri, M. (2020). Factors associated with antibiotic prescription among healthcare workers at tertiary hospitals in Nairobi County, Kenya. \u003cem\u003eAnnals of Health Research\u003c/em\u003e. https://doi.org/10.30442/ahr.0602-05-78\u003c/li\u003e\n\u003cli\u003eRakhshani, N. S., Kaljee, L., Khan, M. I., Prentiss, T., Turab, A., Mustafa, A., Khalid, M., \u0026amp; Zervos, M. (2022). A formative assessment of antibiotic dispensing/prescribing practices and knowledge and perceptions of antimicrobial resistance among healthcare workers in Lahore, Pakistan. \u003cem\u003eAntibiotics\u003c/em\u003e, 11(10), 1418. https://doi.org/10.3390/antibiotics11101418\u003c/li\u003e\n\u003cli\u003eAshiru-Oredope, D., Hopkins, S., Vasandani, S., Umoh, E., Oloyede, O., Nilsson, A., Kinsman, J., Elsert, L., \u0026amp; Monnet, D. (2021). Healthcare workers\u0026rsquo; knowledge, attitudes and behaviours with respect to antibiotics, antibiotic use and antibiotic resistance across 30 EU/EEA countries in 2019. \u003cem\u003eEurosurveillance\u003c/em\u003e, 26(12). https://doi.org/10.2807/1560-7917.ES.2021.26.12.1900633\u003c/li\u003e\n\u003cli\u003eWang, S. Y., Cantarelli, P., Groene, O., Stargardt, T., \u0026amp; Bell\u0026eacute;, N. (2022). Patient expectations do matter: experimental evidence on antibiotic prescribing decisions. \u003cem\u003eThe European Journal of Public Health\u003c/em\u003e, 32(Supplement_3). https://doi.org/10.1093/eurpub/ckac131.188\u003c/li\u003e\n\u003cli\u003eKaawa-Mafigiri, D., Ekusai-Sebatta, D., Rutebemberwa, E., Sserwanga, A., Kitutu, F., Kapisi, J., Hopkins, H., Salami, O., Nkeramahame, J., Olliaro, P., \u0026amp; Horgan, P. (2023). A qualitative assessment of a training and communication intervention on antibiotic prescription practices among health workers and outpatients at public health facilities in Uganda. \u003cem\u003eClinical Infectious Diseases\u003c/em\u003e, 77(Supplement_4), S191\u0026ndash;S198. https://doi.org/10.1093/cid/ciad329\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":"antibiotic stewardship, prescribing behavior, healthcare workers, behavioral intention, explainable machine learning, psychological modeling, SHAP analysis, LASSO regression","lastPublishedDoi":"10.21203/rs.3.rs-6768023/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6768023/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIntroduction: Antimicrobial resistance (AMR) remains a global health priority, partly driven by irrational antibiotic prescribing among healthcare workers. Existing interventions often overlook the psychological and contextual factors that shape prescribing behaviour. This study aimed to integrate behavioural theory with explainable machine learning to identify psychological predictors of antibiotic use intention among clinicians.\u003c/p\u003e\n\u003cp\u003eMethods: A cross-sectional survey was conducted among 1,135 healthcare workers from four public hospitals in China. Participants completed questionnaires based on constructs from the Theory of Planned Behavior, Health Belief Model, Rational Action Theory, Self-Efficacy Theory, Social Support Theory, and cognitive processing frameworks. LASSO regression and SHAP analysis were applied alongside machine learning classifiers (e.g. XGBoost, LightGBM, CatBoost) to identify key predictors and interactions influencing prescribing intention.\u003c/p\u003e\n\u003cp\u003eResults: Social support, cognitive processing, knowledge and skills, and health beliefs were the most important predictors. SHAP analysis revealed nonlinear interactions, particularly between social support and cognitive engagement. Ensemble models achieved high predictive accuracy (F1-scores \u0026gt;0.94) for high and medium prescribing intention, but classification of low-intention individuals remained more challenging.\u003c/p\u003e\n\u003cp\u003eConclusion: Combining behavioural theory with explainable AI offers a scalable approach to identifying clinicians at risk of irrational prescribing. These findings support the development of psychologically tailored, real-time interventions that can improve antibiotic stewardship and address AMR in diverse health system settings.\u003c/p\u003e","manuscriptTitle":"AI-Based Precision Prediction of Healthcare Workers’ Antibiotic Use Intentions: Multi-theoretical Psychological and Behavioral Modeling","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-23 08:03:22","doi":"10.21203/rs.3.rs-6768023/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-10-16T05:37:11+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-15T18:57:43+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"161112305085974016057648139835280572166","date":"2025-10-14T14:17:52+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"323119437153638795436372753733571514059","date":"2025-09-18T14:00:54+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"28266486954444944325844589941953708140","date":"2025-07-31T21:08:53+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-22T17:09:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"106512198146544528459977144119733213359","date":"2025-07-07T13:00:12+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"183160740828874743355417748730805498226","date":"2025-06-19T00:24:30+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"325266791896590452787591138758212684687","date":"2025-06-17T07:06:22+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-06-16T23:43:05+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-06-16T23:41:10+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-06-09T05:34:22+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-06-05T13:03:49+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-05-28T12:16:09+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":"f3310bf6-0a5b-484f-827e-bfe043bd3282","owner":[],"postedDate":"June 23rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":50196032,"name":"Health sciences/Diseases"},{"id":50196033,"name":"Health sciences/Diseases/Infectious diseases/Bacterial infection"}],"tags":[],"updatedAt":"2026-02-02T16:00:11+00:00","versionOfRecord":{"articleIdentity":"rs-6768023","link":"https://doi.org/10.1038/s41598-026-37495-x","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2026-01-28 15:58:02","publishedOnDateReadable":"January 28th, 2026"},"versionCreatedAt":"2025-06-23 08:03:22","video":"","vorDoi":"10.1038/s41598-026-37495-x","vorDoiUrl":"https://doi.org/10.1038/s41598-026-37495-x","workflowStages":[]},"version":"v1","identity":"rs-6768023","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6768023","identity":"rs-6768023","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}