Content area
Aim
To identify predictors of academic dropout before graduation and professional attrition after graduation among nursing students, using the theoretical IPOD model and a machine learning approach.
Background
Academic dropout and professional attrition are global challenges.
Design
A retrospective study design.
Method
This study included 878 undergraduate nursing students enrolled between 2007 and 2018. Data were collected from Education Department records and follow-up interviews conducted via phone, email, or social media platforms. To predict academic dropout before graduation, four machine learning models were used: Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), Decision Tree (DT), and Multinomial Logistic Regression (mLR). To predict professional attrition after graduation, Random Forest (RF), SVM, DT, and Binary Logistic Regression (BLR) models were applied.
Outcomes
The academic dropout rate was 2.2 %, while the professional attrition rate was 28.3 %. The XGBoost model, with 91 % accuracy, identified dropout predictors including a higher ratio of failed semesters to total semesters, lower GPA in the first and second semesters, younger age at admission, abnormal early academic status, tuition payment, and male gender. The Random Forest model, with 90 % accuracy, linked professional attrition to higher clinical competency, higher overall GPA, longer waiting time before employment, job burnout, and longer work experience.
Conclusions
Academic performance indicators, particularly during the early semesters, were associated with nursing student dropout, while professional factors such as job burnout and employment delays were linked to post-graduation attrition. These findings may inform targeted interventions to improve retention across both academic and professional stages.
1 Introduction
Turnover in the nursing profession can be categorized into two types: organizational turnover, in which nurses change employers, and professional turnover, in which nurses leave the profession entirely. The latter is commonly referred to as "leaving the profession" or "nursing attrition" ( Halter et al., 2017). One of the global challenges in nursing is the high rate of academic dropout among nursing students during their studies, and professional attrition among nurses after graduation—both of which affect workforce stability, the financial sustainability of health systems, and the quality of care ( Lantz and Fagefors, 2025). The global rate of ( Ren et al., 2024) nursing attrition has been reported to range from 8 % to 36.6 %, with a rate of 19 % in Asia. In Iran, 49.6 % of nurses have expressed an intention to leave the profession ( Maleki et al., 2023). Annually, approximately 15–20 % of nursing students worldwide withdraw from undergraduate nursing programs ( Coakley, 1999). In a study conducted in the United Kingdom, among 534 nursing and midwifery students, 55 had left and returned to their studies, and 281 had considered leaving or discontinuing their education ( Thompson et al., 2025). In a study from Saudi Arabia, more than half of the surveyed nursing students reported an intention to leave the profession in the future ( Kandil et al., 2021). In a study conducted in nursing schools in Tehran, Iran, only 18 % of students had a positive view of the nursing profession, while 69 % expressed a desire to leave it, 63.6 % planned to change their major, and 51.64 % were inclined to drop out ( Joolaee et al., 2006). This situation has led to a shortage of nurses and decreased professional motivation among remaining students ( Coakley, 1999), while the growing population continues to increase the demand for nursing care. According to the WHO, a global shortage of at least 12.9 million nurses is projected by 2035, and the current shortage is estimated to be around 7.2 million ( McCarthy et al., 2020).
Nursing attrition, both professional and academic, is influenced by a range of factors that impact healthcare quality and workforce stability. Professionally, issues such as staff shortages, heavy workloads, burnout, limited opportunities for growth, poor work environments, salary dissatisfaction, and lack of support contribute to nurse turnover—particularly among younger and female nurses ( Lantz and Fagefors, 2025; Mohamed et al., 2024). Additional contributing factors include workplace stress, unfavorable employment conditions, and limited resilience, especially among novice nurses ( Lyu et al., 2024). Academic dropout among nursing students is similarly complex, leading to economic losses and a reduced number of professional entrants. Key reasons include a mismatch with the profession, low confidence, unmet expectations, negative clinical experiences, inadequate support, and abusive supervision ( Canzan et al., 2022; Hong et al., 2024). Despite numerous studies on nurses’ intentions to leave the profession, a significant gap remains in the literature: to date, no study has simultaneously examined the predictors of both nursing student dropout and professional attrition within a coherent theoretical framework ( Lantz and Fagefors, 2025; Bolt et al., 2022). Most previous research has focused on either academic or professional attrition in isolation, lacking the theoretical integration needed for a comprehensive understanding of this multifaceted issue. However, a simultaneous analysis of the academic and professional trajectories of nurses is essential for developing effective strategies to retain the nursing workforce ( Halter et al., 2017; Bolt et al., 2022).
In response to this gap, the present study adopts the IPOD theoretical framework (Input–Process–Output–Development), which provides a comprehensive structure for evaluating educational quality and professional development ( Luo et al., 2018). This framework, with its emphasis on academic and career growth during and after formal education, is particularly well-suited for nursing students and graduates. By encompassing the sequential stages of training and employment, it enables a coherent and targeted analysis of academic and professional pathways and supports the identification of factors influencing academic success, job retention, and career advancement. The IPOD model, through its four key dimensions—Input (e.g., personal characteristics and academic background), Process (educational experiences), Output (academic outcomes), and Development (career progression)—offers a structured and multi-layered platform for analyzing nurses’ professional trajectories ( Luo et al., 2018; Jiabin et al., 2024). Given the aim of this study—to identify predictors of nursing student dropout prior to graduation and professional attrition after clinical entry—this model provides a relevant and practical theoretical foundation for examining the multifaceted nature of attrition in nursing. Its added value lies in its capacity to assess both academic performance and long-term professional development, making it particularly appropriate for investigating the causes of attrition across various stages of the nursing career, especially within the context of Iran’s healthcare system. Utilizing such a framework can yield practical and policy-relevant implications at both the educational and managerial levels of the health system ( Luo et al., 2018). On the other hand, given the large volume and complexity of data related to attrition, recent studies have employed machine learning algorithms for data analysis ( Ganthi et al., 2022; Iparraguirre-Villanueva et al; Mozaffari et al., 2023; Vaarma and Li, 2024; Huo et al., 2023). These methods are particularly effective in uncovering hidden patterns within large and complex datasets that may not be detectable using traditional statistical approaches ( Arian et al., 2025). By simultaneously analyzing individual, educational, and environmental factors, machine learning models can accurately predict risk factors associated with leaving the profession and support managers in making faster, more targeted decisions to improve staff retention. Moreover, these models are both scalable and adaptable, enabling the development of personalized interventions for at-risk personnel ( Alqahtani et al., 2024; Raza et al., 2022).
This study aims to identify key predictors of academic dropout and professional attrition among undergraduate nursing students in Iran, using the IPOD model and machine learning algorithms. Within this framework, the academic and professional trajectories of nurses are analyzed, and the roles of demographic, educational, environmental, and organizational variables in attrition are examined.
2 Methods
2.1 Objective
This retrospective cross-sectional study was conducted to identify key predictors of academic dropout and professional attrition among undergraduate nursing students in Iran, using the IPOD model and machine learning algorithms.
2.2 Study design and setting
The study was conducted at the School of Nursing and Midwifery, Mashhad University of Medical Sciences. Established in 1958, the School of Nursing and Midwifery in Mashhad is one of the oldest, largest, and most reputable nursing schools in Iran. Located in northeastern Iran, the school comprises six academic departments and employs 67 personnel, including 59 faculty members and 8 non-faculty staff with master's degrees. It offers educational programs at the bachelor's, master's, and doctoral levels in various disciplines, including Nursing, Midwifery, Reproductive Health, Critical Care Nursing, Neonatal Intensive Care, Operating Room, Anesthesia, and Prehospital Emergency Services. Currently, 1006 students are enrolled, of whom 217 (21.5 %) are postgraduate students. The school admits undergraduate nursing students twice a year—in the fall and winter semesters—and enrolls an average of 100–120 new undergraduate students annually. The Bachelor of Science in Nursing program has a duration of four years.
2.3 Data collection process
Data collection was conducted in two stages:
2.3.1 Stage one – educational and demographic data (Census)
In this stage, educational and demographic data of undergraduate nursing graduates admitted over the past ten years (from Fall 2007 to Winter 2018) through the national university entrance exam (Konkur) were extracted. Data were obtained from the School’s Education Office between August 17, 2022, and October 18, 2022, and recorded in an Excel file. Data collection was performed as a census, meaning all eligible individuals were included. The Education Office was selected as the data source due to its accuracy and reliability as the official repository of graduates' academic records. Variables such as cumulative GPA, number of credits completed, failed courses, probation status, length of study, and graduation date are recorded in this system and are often more precise than self-reported information collected through phone interviews.
2.3.1.1 Inclusion and exclusion criteria for stage one
Inclusion: All nursing students admitted via the national entrance exam (Konkur) within the specified period were included.
Exclusion: Students who transferred into or out of the school from/to other universities, and those with incomplete or unusable academic records, were excluded.
2.3.2 Stage two – post-graduation data (Follow-up)
In the second stage, data on post-graduation status—including employment, job satisfaction, professional development, intention to leave the profession, and other related retention indicators—were collected directly from individuals, as this information was not available in academic records nor maintained by the Education Office. A trained nursing expert, familiar with data collection principles, contacted all individuals listed in the Excel file by telephone. If a person did not answer after several attempts, a text message was sent, followed by efforts to reach them via email and social media platforms. Upon obtaining verbal consent, data were collected using a standardized form designed based on the IPOD theoretical model (detailed in the Instruments section). Data collection for this stage was completed by January 27, 2024.
2.3.2.1 Inclusion and exclusion criteria for stage two
Inclusion: Individuals whose educational data were recorded in the main file from stage one, and those who responded to emails or social media messages and consented to participate. Exclusion: Individuals unwilling to continue the interview; those who did not respond to repeated phone calls, emails, or messages. (Data management regarding missing data is described in the analysis section.)
2.4 Contextual considerations and limitations in generalizability
In Iran, nursing program admissions and post-graduation employment are influenced by national exams, quota systems, and regional workforce policies. These structural factors shape the composition and distribution of the nursing workforce and may affect the generalizability of study findings.
2.5 Ethical approval
The study was approved by the Ethics Committee of Mashhad University of Medical Sciences (Code: IR.MUMS.REC.1401.147).
2.6 Theoretical framework
The IPOD (Input–Process–Output–Development) theoretical framework is a modern and comprehensive model for analyzing educational quality and individual development (
This framework comprises four main dimensions: Input quality, which assesses individual characteristics and entry conditions at the start of the educational program; Process quality, which involves evaluating academic performance and learning experiences during the program; Output quality, which addresses immediate outcomes following graduation and reflects the effectiveness of the educational program; and Development quality, which focuses on long-term career outcomes and professional sustainability. Initially applied in doctoral education to evaluate academic quality and professional development, the IPOD framework allows for simultaneous analysis of influencing factors throughout and beyond the period of study ( Luo et al., 2018). In the present study, the model is used for the first time at the undergraduate level in nursing education to identify predictors of nursing dropout during education as well as attrition from the nursing profession after graduation. Employing this model enables a systematic and multilayered analysis of this complex phenomenon, emphasizing the interaction of personal, educational, structural, and professional factors.
2.7 Instruments and variables
In this study, the data collection questionnaire was developed based on the IPOD theoretical model ( Luo et al., 2018). The questionnaire comprises four main dimensions: Input Quality, Process Quality, Output Quality, and Development Quality. Each dimension includes specific subcategories consisting of items designed to identify predictors of academic performance and professional development among nursing graduates.
The qualitative face validity of the questionnaire was evaluated by ten experts in nursing education. For quantitative content validity, two indices were used: the Content Validity Ratio (CVR) and the Content Validity Index (CVI). Based on Lawshe’s table, with ten experts participating, the CVR for all items exceeded the minimum threshold of 0.62. The CVI was calculated from expert ratings of simplicity, relevance, and clarity on a 4-point Likert scale, with all items achieving a CVI above 0.79. The reliability of the questionnaire was confirmed through Cronbach’s alpha, which yielded a coefficient of 0.91.
The study variables were categorized according to the four dimensions of the IPOD framework, with the variables for each dimension presented separately in
In this study, alongside presenting results within the IPOD framework, the researchers aimed to identify variables associated with nursing student dropout and nurse attrition before and after graduation from the undergraduate nursing program. Nursing student dropout was assessed using the variable Retention, which measured academic progression through to graduation. This variable was categorized into three classes: ideal (graduation within the standard 8-semester curriculum), continuous (graduation after 9 or more semesters), and dropout (failure to graduate and withdrawal from the program).
Nurse attrition was evaluated using the binary variable Nursing Attrition, with two classes: yes (left the profession) and no (remained in the profession).
To identify predictors relevant to both pre- and post-graduation stages, two separate datasets were constructed. The pre-graduation dataset included only student-related variables, while the post-graduation dataset included both student-related and post-graduation variables. In both datasets, irrelevant features and highly correlated variables were either transformed or excluded to improve model performance and interpretability. For example, instead of using the absolute number of semesters in which a student was not accepted, the proportion of unaccepted semesters relative to the total number of semesters was used. Similarly, the proportion of passed credits to total credits was used in place of the raw number of passed credits.
2.8 Data analysis
2.8.1 Data preprocessing and analysis
Data preprocessing and analysis were conducted using the Python programming language (version 3.10.12) ( Rossum and Drake, 2009). Key packages included scikit-learn (version 1.6.1), XGBoost (version 2.0.0), SHAP (version 0.47.1), pandas (version 2.1.0), numpy (version 1.26.4), and matplotlib (version 3.7.1). Descriptive and inferential statistics were applied for analyses within the IPOD model, while machine learning models were used to identify predictors of nursing student dropout and professional attrition.
The following steps were taken to prepare the data features for analysis:
Label encoding
To improve model performance, nominal variables were transformed using One-Hot Encoding. In this method, each level of a nominal variable was converted into a binary variable. These variables were coded as 1 (Yes) if the data matched the respective level and 0 (No) otherwise, with one level designated as the reference ( Potdar et al., 2017).
Missing data management
In the first phase, data extracted from the Education Department were largely complete. However, where grades for the seventh or eighth semesters were missing, imputation was performed using estimation methods based on available information, particularly first-semester grades and overall GPA. This approach aimed to preserve data integrity and reduce bias from excluding incomplete cases. For the variable "Number of Failed Credits," imputation was based on the difference between total credits taken and credits passed, assuming a predictable pattern in academic performance and using adjacent data for estimation. For students who had dropped out, imputation was not applied to subsequent semesters since no academic activity existed, and these missing values were classified as "Missing Not Applicable" rather than "Missing at Random."
In the second phase, which combined academic records with data from follow-up phone interviews, complete data were available only for individuals who responded. No post-graduation information was collected from those who declined or did not respond. Given that most variables in this phase were qualitative and based on subjective assessments (e.g., intention to leave the profession), imputation was neither statistically justified nor scientifically appropriate. Preliminary analyses showed no significant differences in academic variables (such as GPA and academic status) between respondents and non-respondents. Therefore, missing data in this phase were considered "Missing at Random," minimizing the risk of selection bias. Nonetheless, the exclusion of incomplete cases reduced the sample size, potentially limiting the statistical power of the predictive models.
Feature Scaling
Features with wide numerical ranges (e.g., continuous variables) can disproportionately influence model training or cause computational instability. To standardize feature impact, Min–Max scaling was applied, rescaling values to a 0–1 range ( Jain and Bhandare, 2011).
2.8.2 Classification models
To identify factors associated with academic dropout, the target variable "Academic Dropout" (i.e., retention status through graduation) was classified into three categories: Ideal, Continuous, and Dropout. Accordingly, multi-class classification algorithms—Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), Multinomial Logistic Regression (MLR) ( Otieno et al., 2024), and Decision Tree (DT) ( Potdar et al., 2017)—were applied, consistent with previous studies.
In contrast, to examine post-graduation professional attrition, the binary variable "Professional Attrition" (Yes/No) indicated whether individuals left or remained in the nursing profession. Binary classification models—Random Forest, SVM, Decision Tree, and Binary Logistic Regression (BLR)—were used based on their proven effectiveness in medical research ( Lee et al., 2020). Detailed model descriptions are provided in Appendix A.
2.8.3 Model training process
Data were randomly split into a training set (75 %) and a test set (25 %) for model training. The models were implemented using the scikit-learn library. Some hyperparameters (e.g., tree depth, kernel type, hidden layer size) were manually initialized, while others were kept at their default values, either due to their general suitability (e.g., booster type, learning rate) or insensitivity to tuning (e.g., kernel cache size, max iterations) ( JNv and Hutter, 2018)Hyperparameter optimization was conducted via Grid Search with 5-fold cross-validation, using the weighted F1 score (f1_weighted) as the performance metric. This process helped identify the optimal parameter set, improve prediction accuracy, and reduce overfitting risk ( Belete and Huchaiah, 2021). A fixed random seed (random_state = 42) was consistently applied to ensure reproducibility.
2.8.4 Model performance metrics
Model performance was evaluated using Precision, Recall, Accuracy, Cohen’s Kappa, Matthews Correlation Coefficient (MCC), F1 Score, and the Area Under the Receiver Operating Characteristic Curve (AUC–ROC) ( Grandini et al., 2020). All metrics were calculated from the confusion matrix, and estimates were reported along with their 95 % confidence intervals.
2.8.5 Model explainability
Modern machine learning models often lack interpretability, a limitation in clinical and educational settings compared to models like Logistic Regression (
Watson et al., 2019). To enhance model explainability, SHAP (SHapley Additive exPlanations) values were calculated using the SHAP Python library (
Lundberg and Lee, 2017). SHAP values quantified the contribution of each feature to the model’s output. To interpret the analysis of educational outcomes (i.e., the
Retention variable), SHAP values were visualized using Force plots (individual level) and Beeswarm plots (overall patterns).
3 Results
The results based on the IPOD model are presented in
The following section presents the results of the machine learning analyses, focusing on the identification of predictors for nursing student dropout and professional attrition. The first dataset, comprising only academic variables collected during the course of study, was used to identify factors associated with student dropout. The second dataset, which included both academic and post-graduation professional variables, was employed to determine predictors of attrition from the nursing profession. After removing irrelevant and highly correlated variables to enhance model performance and interpretability, the final sets of variables used in each dataset are illustrated in
According to the results presented in
According to the results presented in
4 Discussion
Nursing student dropout and professional attrition before and after graduation have become pressing global concerns, with serious implications for both healthcare quality and economic efficiency. This study applied the IPOD model in conjunction with machine learning techniques to identify predictive factors associated with academic success, retention during nursing education, and continued professional engagement after graduation. The dropout rate among undergraduate nursing students at the School of Nursing and Midwifery in Mashhad, Iran, from 2007 to 2018 was 2.2 %. In contrast, a study at a university in southern Turkey reported an average dropout rate of 7.6 % ( Kantek, 2010), and other international studies have documented rates as high as 60 % ( Coakley, 1997; Glossop, 2002). These differences likely reflect variations in educational environments across countries. The relatively low dropout rate in Mashhad may be attributed to the academic rigor of the nursing program and the sociocultural context of the city itself. As the largest religious center in Iran and the country’s spiritual capital, Mashhad is home to a highly reputable medical university, which may contribute to greater student commitment and retention throughout the four-year program. The predictive profile for pre-graduation dropout included a higher proportion of failed semesters, lower GPAs in the first and second semesters, younger age at entry, abnormal academic status in the first semester (such as probation or academic leave), tuition-paying status, and male gender. These factors can serve as early indicators of risk, enabling the implementation of more focused and timely interventions to support at-risk students.
Previous research highlights the early stages of nursing education as a particularly vulnerable period for student attrition. Kantek’s study emphasized that the initial years are critical for dropout ( Kantek, 2010) risk, while Last and Fulbrook noted that students require heightened support and guidance during their first year more than at any other stage ( Last and Fulbrook, 2003). Academic failure and declining grades have been identified as the most common reasons for dropout by Richardson (1996) and White et al. (1999). Specifically, nursing students with lower GPAs are more likely to discontinue their studies ( Kantek, 2010). Glossop reported that 56 % of nursing students drop out during their educational program, particularly in the first year ( Glossop, 2002). The findings of the present study align with these previous observations ( Hannaford et al., 2021). Jackson and Sandiford, in their review titled "Predictors of First-Year Nursing Student Dropout," reported that the dropout rate during the first semester among nursing students at Dallas College, Texas, ranged between 20 % and 30 %. They categorized the factors contributing to student dropout into three groups: academic/university-related, socio-economic, and motivational. They emphasized the importance of addressing the causes of nursing student dropout and called for further research in this area ( Sandiford and Jackson, 2003).
A study by Joolaee in Iran found that male students tend to have a more negative attitude toward the nursing field compared to female students and report a higher likelihood of dropping out of the nursing program ( Joolaee et al., 2006). Similarly, research by Parker and Lindop highlights the significant impact of factors such as negative societal attitudes and perceptions of nurses and other healthcare team members, especially on men ( Lindop, 1991; Parker, 1996). During their student years, core perceptions about nursing are still developing, and gender stereotypes often persist. Many students view nursing as a “practical” profession and caregiving as a “female” trait, which may contribute to higher dropout rates among male nursing students ( van der Cingel and Brouwer, 2021). Kandil’s study showed that the prevailing image of the nursing profession strongly influenced Saudi female students’ preference for nursing education. However, more than half of the students expressed a desire to leave the profession after graduation. Among those intending to leave, over half cited a lack of interest in nursing as the primary reason. Family opposition and the social image of nursing were also commonly reported reasons for leaving the profession ( Kandil et al., 2021).
In the present study, 28.3 % of nurses had left the nursing profession, and among those still practicing, 25.4 % expressed an intention to leave. A study conducted in Iran reported that 49.6 % of nurses intended to leave the profession, with the average intention score exceeding a moderate level ( Maleki et al., 2023). Another Iranian study found that nearly one-third of nurses stated they would leave the profession if given another job opportunity ( Sokhanvar et al., 2018). Research by Chegini and colleagues and Al Momani in Iran showed that 64 % and 60.9 % of nurses, respectively, intended to leave the profession ( Al Momani, 2017; Chegini et al., 2019), whereas rates of intention to leave were considerably lower in Canada (13 %) and Australia (15 %) ( Fernet et al., 2017; Guo et al., 2019). In this study, the predictive factors for continuing in the nursing profession included higher salary, job satisfaction, experience in home care, interest in nursing, older age at admission, and prior student work experience. These factors highlight the link between individual motivation and favorable job conditions. Conversely, nursing attrition after graduation was more strongly associated with high clinical competence, high overall GPA, long waiting time before starting work, job burnout, and extensive work experience.
Maleki’s study found no statistically significant differences between nurses who intended to leave their jobs and those who did not, in terms of age, marital status, gender, type of employment, work shift, and work experience. However, a significant association was observed between both workplace and job title and the intention to leave the profession ( Maleki et al., 2023). The results of the present study revealed nearly twice the intention to leave the profession compared to studies conducted outside Iran, but lower than those within the country—possibly because participants in this study had less work experience. Work experience is an important and influential factor in professional attrition. One likely reason for nurses’ intention to leave is excessive mandatory overtime caused by workforce shortages. Another contributing factor may be the work environment; evidence suggests that clinical settings characterized by respect, autonomy, freedom, and independence serve as important deterrents to leaving the profession ( Maleki et al., 2023).
In the present study, individuals who entered nursing at an older age were less likely to leave the profession, both before and after graduation. These individuals likely entered the field with greater awareness and without the pressure of entrance exams. Based on the researchers’ previous observations, these nurses often had prior hospital work experience and were admitted through practical nursing quotas, viewing the transition as a significant career advancement. Conversely, nurses with more work experience showed higher rates of attrition. Most studies have reported a positive correlation between length of service and intention to leave—meaning that as years of service increase, so does the likelihood of leaving the profession ( Javanmardi et al., 2022). In contrast, Bumeister and colleagues found that younger age and less experience were key factors contributing to absenteeism and the intention to leave nursing ( Burmeister et al., 2019).
In Iran, nurses are required to complete a mandatory two-year service period within the national healthcare system after graduation. Only after fulfilling this requirement can they enter the formal employment process. Consequently, most young nurses in this study were either completing their mandatory service or in the early stages of employment and had not yet experienced occupational burnout. In contrast, Lantz’s study found that age was significantly associated with factors contributing to job turnover, with younger nurses more likely to report dissatisfaction with compensation and fairness, as well as higher levels of stress and burnout. Newly graduated nurses often face considerable challenges in adapting to the realities of professional practice ( Lantz and Fagefors, 2025).
In the present study, unlike the pre-graduation period when men were more likely to leave the nursing field, after graduation men were less likely to leave the profession, while women exhibited higher rates of attrition. The main reasons reported for leaving the profession included low salary, marriage and childbirth, and lack of interest. Being male, single, and having less work experience were positively correlated with retention in nursing; however, only work experience emerged as a key variable in the predictive model. According to a review study, age, gender, and marital status showed mixed relationships with nursing attrition—approximately half of the studies reported a negative association, while the other half found a positive one. Therefore, it is not possible to definitively conclude whether these variables have a direct or inverse effect on leaving the profession, highlighting the need for further research to clarify their roles. In addition, findings indicated that the intention to leave was less common among nurses with permanent employment ( Javanmardi et al., 2022). In the present study, a positive correlation was observed between holding nursing managerial positions and retention. Lee and colleagues similarly reported that leadership and management roles can influence nurses’ intention to leave, emphasizing that fostering a supportive work environment may help reduce attrition rates ( Lee et al., 2019).
Three key variables found to significantly contribute to retaining nursing staff in this study were higher salary, job satisfaction, and the absence of burnout. A review study categorized factors influencing nursing attrition into six broad groups: demographic factors, individual and family factors, organizational and environmental factors, organizational behavior factors, job satisfaction factors, and factors related to the nature of the profession. This study showed that in Iran, nurses’ tendency to leave the profession ranges from moderate to high, with burnout, job satisfaction, and income level identified as the most important contributors to attrition within the Iranian healthcare system ( Javanmardi et al., 2022). Similarly, a study conducted in Saudi Arabia found that being single, having a low gross monthly salary, and lower job satisfaction were associated with a higher likelihood of leaving the job ( Albougami et al., 2020). Job dissatisfaction is a common issue among nurses, both in Iran and globally. It is defined as an emotional response to the perception that one’s job fails to fulfill important values and goals ( Meier and Spector, 2015). This issue is especially critical in healthcare, as job dissatisfaction directly affects the quality of patient care and health outcomes. Simply put, nurses who are dissatisfied with their job or workplace treatment face greater challenges in meeting patients’ needs ( Boon Ooi et al., 2007; Arian et al., 2018). Supporting this, another study in Iran found that more than half of nurses reported dissatisfaction with their work ( Mehrdad et al., 2013).
Research in Saudi Arabia further highlighted that job satisfaction and intention to leave are key factors influencing nurses’ motivation. Incentives were shown to positively affect job satisfaction and productivity, ultimately improving the quality of nursing care. Monthly salary showed a significant relationship with the intention to leave; low income negatively impacts job behavior, increasing dissatisfaction, burnout, and turnover tendency. Regression analysis from this study revealed that income level, incentives, hospital type, and intention to leave significantly influenced job satisfaction, while gender and nationality were also significantly associated with the intention to leave ( Alanazi et al., 2023).
Overall, job satisfaction, work environment, and individual factors remain among the strongest determinants of nursing attrition. Nurses satisfied with their jobs are less likely to leave, underscoring the importance for healthcare organizations to enhance job satisfaction through supportive work environments, adequate resources, and career advancement opportunities. Conversely, factors such as heavy workload and lack of managerial support are major contributors to burnout. Addressing these through appropriate staffing, resource allocation, and organizational leadership is essential to reducing attrition and sustaining a committed nursing workforce ( Mohamed et al., 2024; Arian et al., 2023).
According to the findings of the present study conducted in Mashhad, Iran, individuals with higher academic and professional competence were more likely to leave the nursing profession. Although this may seem counterintuitive, it reflects broader professional dynamics. In Mashhad—similar to many other healthcare systems—challenges such as limited opportunities for advancement, modest salaries, and demanding work conditions may prompt high-performing individuals, who often seek greater professional development, to pursue alternative career paths. Many choose to continue their careers abroad or transition to other health-related fields such as medicine, dentistry, or pharmacy through national entrance examinations. This trend has been further facilitated by a recent policy reform in Iran, which allows nursing graduates to apply for medical school under specific conditions and by passing a designated exam. Another factor influencing nurse retention in Mashhad was participation in home care services. This form of employment, more common among male nurses, not only provides additional income but also strengthens professional relationships with physicians and supervisors. These connections can lead to career advancement, increased job satisfaction, and a reduced likelihood of leaving the nursing profession—echoing previous research that highlights the role of workplace engagement in improving retention.
Similarly, the study by Hörberg et al. emphasized that job-related activities fostering better workplace relationships play an important role in reducing nursing attrition ( Hörberg et al., 2023). According to a systematic review, several factors—including job satisfaction, organizational commitment, quality of work life, work environment, leadership style, workplace bullying, family-related reasons, and job security—have been identified as negatively associated with the intention to leave and positively associated with the intention to stay within an organization ( Al Zamel et al., 2020). Job satisfaction has also been linked to factors such as organizational justice, job outlook and stability, relationships with managers and colleagues, and the overall work environment ( Sokhanvar et al., 2018). An inductive qualitative content analysis exploring motivational factors among experienced nurses, their intention to leave (ITL), and the reasons behind nursing attrition identified five major categories: organizational characteristics, job-related features, workplace relationships, recognition and appreciation, and health-related issues. Rarely was a single reason cited for leaving the profession; rather, multiple factors accumulated over time influenced nurses’ motivation to either remain in or leave their profession. Nursing attrition and retention are complex, multifactorial processes that should be understood as interconnected phenomena rather than isolated or one-dimensional events ( Hörberg et al., 2023).
4.1 Strengths and limitations of the study
While this study offers valuable insights into nursing student dropout and professional attrition in Mashhad, Iran, its findings should be considered alongside certain limitations. Below is a summary of the study’s key strengths and weaknesses, providing a balanced view of its contributions and constraints. The study has several notable strengths. It employs multiple machine learning models, which enhance predictive accuracy and robustness. Furthermore, the application of the IPOD theoretical framework in nursing education and workforce research offers a novel and structured approach to understanding the complex interactions among academic, institutional, and professional factors. The longitudinal design, spanning over a decade, adds depth and reliability to the findings, providing a strong foundation for future research and informed policy development. Additionally, the use of SHAP-based interpretability improves model transparency and delivers clear, actionable insights for decision-makers. However, it is important to acknowledge some limitations. While SHAP enhances interpretability, its outputs can sometimes be misinterpreted as indicating strictly linear or causal relationships, which may lead to oversimplified conclusions. The research was conducted at a single nursing faculty, potentially limiting generalizability. Follow-up data collected via telephone may introduce selection bias, as respondents might differ meaningfully from non-respondents. Moreover, important factors such as mental health, peer influence, and family expectations were not measured, leaving room for residual confounding. Although longitudinal, data were collected at only two time points per participant, restricting deeper insight into attrition dynamics over time.
5 Conclusion
This study identified predictors of nursing student dropout and professional attrition over a decade among graduates of the School of Nursing and Midwifery in Mashhad, Iran. The dropout rate among nursing students in Mashhad was relatively low, at 2.2 %, which may reflect the favorable educational environment of one of Iran’s leading nursing schools. However, professional attrition was considerably higher, at 28.3 %, largely influenced by organizational structures and the socio-legal context of the Iranian healthcare system. Early academic challenges—particularly declining grades—were the main contributors to student dropout, while factors such as job burnout, dissatisfaction, and income concerns significantly influenced nurses’ decisions to leave the profession. Although these findings are based on data from a single institution and should be generalized with caution, the identified predictors hold substantial practical value. Nursing schools can leverage early academic indicators—such as GPA trends, failed courses, and academic engagement—to identify students at risk of dropping out. This enables the development of targeted intervention strategies, including personalized counseling, tutoring, and academic support programs designed to address specific challenges before they escalate. Meanwhile, healthcare organizations and policymakers can utilize post-graduation predictors—such as burnout levels, time to employment, and professional competencies—to inform workforce retention policies. By improving workplace conditions, ensuring equitable shift scheduling, and introducing performance-based incentives, they can enhance job satisfaction and reduce burnout among nurses. Together, these focused interventions provide a practical, evidence-based roadmap for reducing dropout rates and professional attrition, thereby strengthening nursing education outcomes and sustaining a resilient nursing workforce.
CRediT authorship contribution statement
Mahdieh Arian: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Funding acquisition, Formal analysis, Conceptualization. Azadeh Kamali: Writing – review & editing, Visualization, Resources, Conceptualization. Zahra Dalir: Visualization, Validation, Resources, Data curation, Conceptualization. Fatemeh Hajiabadi: Validation, Methodology, Formal analysis, Conceptualization. Seyed Reza Mazloum: Visualization, Resources, Conceptualization.
Ethics approval
Approval for this study was obtained from Mashhad University of Medical Sciences (Code: IR.MUMS.REC.1401.147). Strict measures were taken to ensure the confidentiality of participants' information.
Consent to participate
Informed consent was obtained from all individual participants included in the study.
Funding
This study was conducted with the financial and spiritual support of
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The authors of this study express their gratitude and appreciate the support of the Vice-Chancellor’s Office for Research at Mashhad University of Medical Sciences.
Appendix A Supporting information
Supplementary data associated with this article can be found in the online version at
Appendix A Supplementary material
Supplementary material
Table 1
| Variables | Female Count | Female Mean | Female Std | Male Count | Male Mean | Male Std | Total Count | Total Mean | Total Std | p-value | |
| Input quality | College entry age | 442 | 19.74 | 2.85 | 384 | 20.18 | 3.45 | 826 | 19.95 | 3.15 | 0.001 |
| Process quality | Number of credits in the first semester | 470 | 17.76 | 2.69 | 407 | 17.75 | 2.71 | 877 | 17.76 | 2.70 | 0.57 |
| Number of failed credits in the first semester | 470 | 0.37 | 1.16 | 407 | 0.87 | 1.80 | 877 | 0.60 | 1.51 | 0.001 | |
| GPA semester 1 | 470 | 15.44 | 1.67 | 408 | 14.66 | 1.80 | 878 | 15.07 | 1.77 | 0.001 | |
| GPA semester 2 | 470 | 16.07 | 1.60 | 408 | 15.20 | 1.80 | 878 | 15.66 | 1.75 | 0.001 | |
| GPA semester 7 | 468 | 17.69 | 1.06 | 399 | 16.64 | 1.98 | 867 | 17.21 | 1.64 | 0.001 | |
| GPA semester 8 | 465 | 17.94 | 0.99 | 391 | 17.22 | 1.28 | 856 | 17.61 | 1.19 | 0.001 | |
| GPA Total | 469 | 16.63 | 1.15 | 404 | 15.65 | 1.51 | 873 | 16.18 | 1.42 | 0.001 | |
| Total semester | 470 | 8.11 | 0.56 | 408 | 8.07 | 1.01 | 878 | 8.10 | 0.80 | 0.34 | |
| Total credits | 470 | 135.69 | 10.41 | 408 | 135.74 | 18.40 | 878 | 135.72 | 14.67 | 0.001 | |
| Rank | 470 | 23.28 | 15.28 | 408 | 24.06 | 16.43 | 878 | 23.64 | 15.82 | 0.72 | |
| Number of failed credits | 466 | 1.40 | 3.73 | 393 | 3.28 | 5.52 | 859 | 2.26 | 4.73 | 0.001 | |
| Number of total credits passed | 470 | 133.31 | 9.16 | 408 | 130.81 | 17.39 | 878 | 132.15 | 13.67 | 0.16 | |
| Number of failed semesters | 470 | 0.06 | 0.36 | 408 | 0.21 | 0.63 | 878 | 0.13 | 0.51 | 0.001 | |
| Credit passed/ Total credits | 470 | 0.98 | 0.05 | 408 | 0.96 | 0.07 | 878 | 0.97 | 0.06 | 0.001 | |
| Failed semesters /Total semester | 470 | 0.01 | 0.06 | 408 | 0.03 | 0.11 | 878 | 0.02 | 0.09 | 0.001 | |
| Output quality | Interest to nursing | 470 | 3.68 | 1.11 | 408 | 3.39 | 1.24 | 878 | 3.55 | 1.18 | 0.001 |
| Waiting time for start work | 202 | 3.37 | 2.13 | 174 | 4.43 | 3.16 | 376 | 3.86 | 2.70 | 0.001 | |
| level of clinical competency | 231 | 2.78 | 1.21 | 186 | 2.69 | 1.23 | 417 | 2.74 | 1.22 | 0.38 | |
| Development quality | Years of work experience | 470 | 5.7 | 2.7 | 408 | 5.4 | 3.01 | 878 | 5.58 | 2.88 | 0.02 |
Table 2
| Variables | Sub_Variables | Total Count | Total % | Female Count | Female % | Male Count | Male % | p-value | |
| Input quality | Start year | 2007 | 31 | 3.5 | 15 | 3.2 | 16 | 3.9 | 0.001 |
| 2008 | 52 | 5.9 | 18 | 3.8 | 34 | 8.3 | |||
| 2009 | 51 | 5.8 | 32 | 6.8 | 19 | 4.7 | |||
| 2010 | 68 | 7.7 | 43 | 9.1 | 25 | 6.1 | |||
| 2011 | 86 | 9.8 | 58 | 12.3 | 28 | 6.9 | |||
| 2012 | 91 | 10.4 | 58 | 12.3 | 33 | 8.1 | |||
| 2013 | 85 | 9.7 | 46 | 9.8 | 39 | 9.6 | |||
| 2014 | 81 | 9.2 | 43 | 9.1 | 38 | 9.3 | |||
| 2015 | 98 | 11.2 | 46 | 9.8 | 52 | 12.7 | |||
| 2016 | 97 | 11 | 48 | 10.2 | 49 | 12 | |||
| 2017 | 87 | 9.9 | 39 | 8.3 | 48 | 11.8 | |||
| 2018 | 51 | 5.8 | 24 | 5.1 | 27 | 6.6 | |||
| Start semester | First | 415 | 55.1 | 218 | 54.1 | 197 | 56.3 | 0.56 | |
| Second | 338 | 44.9 | 185 | 45.9 | 153 | 43.7 | |||
| Gender | Female | 470 | 53.5 | 470 | 100 | 0 | 0 | 0.001 | |
| Male | 408 | 46.5 | 0 | 0 | 408 | 100 | |||
| Marital status | Married | 281 | 64.3 | 160 | 65.3 | 121 | 63 | 0.61 | |
| Single | 132 | 30.2 | 70 | 28.6 | 62 | 32.3 | |||
| Divorced | 24 | 5.5 | 15 | 6.1 | 9 | 4.7 | |||
| Native status | Yes | 580 | 74 | 311 | 74 | 269 | 73.9 | 0.99 | |
| No | 204 | 26 | 109 | 26 | 95 | 26.1 | |||
| Dormitory | Yes | 375 | 43.1 | 176 | 37.8 | 199 | 49 | 0.99 | |
| No | 496 | 56.9 | 289 | 62.2 | 207 | 51 | |||
| Tuition | Yes | 95 | 10.8 | 50 | 10.6 | 45 | 11 | 0.91 | |
| No | 783 | 89.2 | 420 | 89.4 | 363 | 89 | |||
| Process quality | academic status in the first semester | Normal | 836 | 95.2 | 453 | 96.4 | 383 | 93.9 | 0.11 |
| Not_normal | 42 | 4.8 | 17 | 3.6 | 25 | 6.1 | |||
| Retention until graduation | IDEAL | 729 | 83 | 408 | 86.8 | 321 | 78.7 | 0.001 | |
| Continuous | 130 | 14.8 | 58 | 12.3 | 72 | 17.6 | |||
| Drop out | 19 | 2.2 | 4 | 0.9 | 15 | 3.7 | |||
| Work experience as a student | Yes | 266 | 34.1 | 111 | 26.3 | 155 | 43.3 | 0.001 | |
| No | 514 | 65.9 | 311 | 73.7 | 203 | 56.7 | |||
| Output quality | Nursing attrition | Yes | 122 | 28.3 | 82 | 34 | 40 | 21.1 | 0.01 |
| No | 309 | 71.7 | 159 | 66 | 150 | 78.9 | |||
| Reasons for nursing attrition | Lack of interest | 39 | 41.5 | 23 | 34.8 | 16 | 57.1 | 0.07 | |
| Marriage/Children | 33 | 35.1 | 24 | 36.4 | 9 | 32.1 | |||
| Low salary | 22 | 23.4 | 19 | 28.8 | 3 | 10.7 | |||
| Research experience | Yes | 387 | 44.2 | 201 | 42.9 | 186 | 45.8 | 0.41 | |
| No | 488 | 55.8 | 268 | 57.1 | 220 | 54.2 | |||
| Teaching experience | Yes | 261 | 29.9 | 131 | 28.1 | 130 | 32.1 | 0.21 | |
| No | 611 | 70.1 | 336 | 71.9 | 275 | 67.9 | |||
| participation in seminars or scientific event | Yes | 272 | 31.2 | 156 | 33.3 | 116 | 28.6 | 0.14 | |
| No | 601 | 68.8 | 312 | 66.7 | 289 | 71.4 | |||
| Development quality | Home care experience | Yes | 136 | 34 | 43 | 19.4 | 93 | 52.2 | 0.001 |
| No | 264 | 66 | 179 | 80.6 | 85 | 47.8 | |||
| Continuation of education | Yes | 36 | 8.9 | 23 | 10.4 | 13 | 7.2 | 0.22 | |
| No | 367 | 91.1 | 199 | 89.6 | 168 | 92.8 | |||
| Education grading | BS | 367 | 91.1 | 199 | 89.6 | 168 | 92.8 | 0.12 | |
| MS | 31 | 7.7 | 18 | 8.1 | 13 | 7.2 | |||
| PhD | 5 | 1.2 | 5 | 2.3 | 0 | 0 | |||
| Salary | Lower12 | 62 | 15.7 | 47 | 21.1 | 15 | 8.7 | 0.001 | |
| 12–16 | 188 | 47.5 | 108 | 48.4 | 80 | 46.2 | |||
| 16–20 | 105 | 26.5 | 44 | 19.7 | 61 | 35.3 | |||
| More20 | 41 | 10.4 | 24 | 10.8 | 17 | 9.8 | |||
| Job burnout | Yes | 209 | 51.2 | 126 | 55.5 | 83 | 45.9 | 0.06 | |
| No | 199 | 48.8 | 101 | 44.5 | 98 | 54.1 | |||
| Job satisfaction | Yes | 151 | 37 | 71 | 31.3 | 80 | 44.2 | 0.01 | |
| No | 257 | 63 | 156 | 68.7 | 101 | 55.8 | |||
| Current job position | Nurse | 306 | 81.4 | 177 | 85.9 | 129 | 75.9 | 0.01 | |
| Best nurse | 29 | 7.7 | 7 | 3.4 | 22 | 12.9 | |||
| Headnurse | 15 | 4 | 9 | 4.4 | 6 | 3.5 | |||
| Teacher | 13 | 3.5 | 8 | 3.9 | 5 | 2.9 | |||
| Supervisor | 9 | 2.4 | 3 | 1.5 | 6 | 3.5 | |||
| Matron | 4 | 1.1 | 2 | 1 | 2 | 1.2 | |||
| Type of hospital | Government_Educational | 164 | 37.5 | 97 | 39.6 | 67 | 34.9 | 0.03 | |
| Government_Non_Educational | 145 | 33.2 | 73 | 29.8 | 72 | 37.5 | |||
| Private | 103 | 23.6 | 66 | 26.9 | 37 | 19.3 | |||
| Military | 25 | 5.7 | 9 | 3.7 | 16 | 8.3 | |||
| Type of employment | Permanent | 226 | 51.7 | 119 | 48.6 | 107 | 55.7 | 0.15 | |
| Temporary | 211 | 48.3 | 126 | 51.4 | 85 | 44.3 | |||
| Intention to change job | Yes | 94 | 25.4 | 57 | 28.1 | 37 | 22.2 | 0.23 | |
| No | 276 | 74.6 | 146 | 71.9 | 130 | 77.8 | |||
| Intention to migration | Yes | 108 | 29.1 | 56 | 27.9 | 52 | 30.6 | 0.57 | |
| No | 263 | 70.9 | 145 | 72.1 | 118 | 69.4 | |||
| Migration | Yes | 33 | 7.6 | 23 | 9.4 | 10 | 5.2 | 0.14 | |
| No | 404 | 92.4 | 222 | 90.6 | 182 | 94.8 |
Table 3
This symbol indicates that those variables have been identified in the algorithm
| Feature & metrics\ Algorithms | mLR (A) | XGBoost | SVM (A) | Decision Tree (A) |
| Age_start_uni | * | * | * | |
| Dormitory | * | |||
| Gender_Male | * | |||
| Grade_semester 1 | * | * | * | * |
| Grade_semester 2 | * | * | ||
| Semesters_not_accepted / Total_semester | * | * | * | * |
| Situation_term 1_Normal | * | * | ||
| Student_work | * | |||
| Tuition | * | * | ||
| course_pass_Total_course | * | |||
| Number of features imported to model | 3 | 7 | 7 | 3 |
| Accuracy | 0.91 | 0.91 | 0.90 | 0.87 |
| Macro-average-Sensitivity | 0.74 | 0.77 | 0.78 | 0.74 |
| Macro-average-Precision | 0.94 | 0.92 | 0.89 | 0.82 |
| F1-Score: IDEAL | 0.58 | 0.64 | 0.65 | 0.52 |
| F1-Score: Continuous | 0.95 | 0.95 | 0.95 | 0.92 |
| F1-Score: Drop_out | 0.88 | 0.89 | 0.88 | 0.88 |
| Macro-average-F1-Score | 0.80 | 0.83 | 0.82 | 0.78 |
| Weighted-average-F1-Score | 0.89 | 0.91 | 0.90 | 0.86 |
| Cohen’s Kappa | 0.61 | 0.65 | 0.65 | 0.52 |
| Matthew’s correlation coefficient (MCC) | 0.65 | 0.67 | 0.66 | 0.52 |
Table 4
This symbol indicates that those variables have been identified in the algorithm
| Feature & metrics\ Algorithms | BLR | Random Forest | SVM (B) | Decision Tree (B) |
| Situation_term 1_Normal | * | * | ||
| Interest_nursing | * | |||
| Student_work | * | * | * | |
| Grade_Total | * | * | * | |
| Salary | * | |||
| Gender_Male | * | * | ||
| Domain_Manager | * | |||
| Domain_Business | * | |||
| Marital_status_Divorced | * | * | ||
| Tuition | * | * | ||
| Hospital_type_Government | * | |||
| Home care | * | * | * | |
| Age_start_uni | * | |||
| Migration | * | * | ||
| Satisfaction | * | * | * | |
| Domain_Nurse | * | |||
| Competency | * | * | * | * |
| Years_experience | * | * | ||
| Teaching | * | * | ||
| Waiting_time_start_work | * | |||
| Burnout | * | * | ||
| Number of features imported to model | 12 | 11 | 12 | 5 |
| Accuracy | 0.70 | 0.90 | 0.58 | 0.73 |
| Sensitivity | 0.73 | 0.95 | 0.61 | 0.70 |
| Specificity | 0.50 | 0.50 | 0.40 | 0.90 |
| Positive Predictive Value (PPV) | 0.92 | 0.93 | 0.88 | 0.98 |
| Negative Predictive Value (NPV) | 0.20 | 0.56 | 0.12 | 0.29 |
| Area Under the Curve (AUC) | 0.64 | 0.83 | 0.52 | 0.78 |
© 2025 Elsevier Ltd