Mapping Occupational Stress and Burnout in the Probation System: A Quantitative Approach

Abstract

This study presents the first nationwide, system-level investigation of occupational stress and professional burnout among probation counselors in Romania, in the context of increasing caseloads, complex job demands and limited institutional support. Building on a comprehensive theoretical analysis, we employ a sociological research design involving a representative sample of 247 probation counselors from all 42 national probation services. Using the Maslach Burnout Inventory—Human Services Questionnaire, along with stress-related factors, we examine the prevalence, most important factors and typologies of burnout. Advanced quantitative techniques—including multiple linear regression, principal component analysis and K-means clustering—allow for a robust identification of key predictors of emotional exhaustion and three distinct psychosocial profiles: stress-resistant seniors, under involved younger staff and overworked, frustrated employees. We also conducted a confirmatory factor analysis (CFA) to check the validity of the MBI-HSS. This typology offers a novel conceptual framework for understanding professional burnout in probation, highlighting systemic vulnerabilities and distinct risk categories. Nevertheless, limitations exist: self-reported data may underestimate stress, and omitting variables like resilience or work meaning constrains explanatory depth. Despite these constraints, this study addresses a significant gap in Romanian probation research and lays the foundation for future longitudinal and qualitative studies. These should incorporate psychological and organizational factors to improve targeted interventions and human resources strategies.

1. Introduction

According to Palmer S., Cooper C. and Thomas K., stress occurs when perceived pressure exceeds one’s coping ability [1]. On the other hand, Wayne J. Pitts mentions that “occupational stress occurs when pressure occurs as a direct result of the tasks and/or conditions experienced during the course of employment and refers to strain, or anxiety that occurs” [2].

The term “staff burnout” was first mentioned in 1969 by Bradley H. B. in the article “Community-based treatment for young adult offenders”, referring to the activity of probation officers [3]. In the specialized literature, we find for the first time the term burnout scientifically defined and described by authors such as Freudenberger [4] and Maslach [5], but Herbert Freudenberger is generally considered to be the founding father of burnout syndrome [6]. The concept of burnout refers to an occupational phenomenon, not to a medical condition, and is a “prolonged occupational stress resulting from the difficult relationships that people have” correlated with their workplace [7]. Maslach’s definition of burnout has remained the same over the years and is seen as a “syndrome of emotional exhaustion, depersonalization, and reduced personal accomplishment that can occur among individuals who work with people in some capacity” [8,9].

According to the International Classification of Diseases—11th Revision, “burnout is a syndrome conceptualized as resulting from chronic workplace stress that has not been successfully managed. It is characterized by three dimensions: feelings of energy depletion or exhaustion; increased mental distance from one’s job, or feelings of negativism or cynicism related to one’s job; and reduced professional efficacy” [10].

In 1981, Maslach and Jackson developed a burnout measurement tool, which they called the Maslach Burnout Inventory (MBI) [8]. After that, new versions of the MBI were developed to match various occupational groups, which have various characteristics. The best-known versions of the MBI are the MBI Human Services Survey (MBI-HSS), MBI Educators Survey (MBI-ES) and MBI General Survey (MBI-GS). But we also find other options, such as the MBI for Medical Personnel (MBI-HSS (MP)) and MBI General Survey for Students (MBI-GS(S)). The MBI is now used worldwide and recognized as one of the most important burnout measurement tools [11].

Over time, other authors have subsequently created new burnout assessment tools, such as the Copenhagen Burnout Inventory (CBI) and Oldenburg Burnout Inventory (OLBI) [12], but our work focuses on the use of the MBI-HSS because, up to this moment, it is the most tested and validated tool for measuring burnout.

There are many studies that validated the Maslach Burnout Inventory—Human Services Survey (MBI-HSS) questionnaire across many countries. The 22-item Maslach burnout inventory has a similar factor structure and is performed similarly across countries [13].

Bria et al. [14] validated the Maslach Burnout Inventory—General Survey among Romanian healthcare professionals, demonstrating acceptable factorial validity (χ²(86) = 432.29, CFI = 0.94, RMSEA = 0.05) and measurement invariance across professional roles, gender, age and organizational tenure.

Traditionally, in applying the MBI-HSS, three components are evaluated through the 22 questions: emotional exhaustion (9 items), depersonalization (5 items) and lack of personal accomplishment (8 items). We find burnout when we encounter high scores on the emotional exhaustion and depersonalization dimensions and low scores on the personal accomplishment dimension. On the other hand, in a new approach to interpreting the results of applying the MBI-HSS instrument, Michael P. Leiter and Christina Maslach, using the same three dimensions, identified five person-centered profiles: “Burnout (high on all three dimensions), Engagement (low on all three), Overextended (high on exhaustion only), Disengaged (high on cynicism only), and Ineffective (high on inefficacy only)”. Also, the authors mentioned that the Disengaged profile is more negative than Overextended and closer to the Burnout profile [15].

1.1. Stress and Burnout in the Probation System

In a 2012 study of the American probation system, K.R. Lewis, L.S. Lewis and T.M. Garby, after an extensive analysis of the research in the field, stated that it is evident that probation officers are impacted by their work with offenders, and this impact is often associated with traumatic stress and burnout; in particular, challenging caseload events, officer victimization and longevity were associated with higher reports [16].

We have already mentioned that “staff burnout” was first used in 1969 by H.B. Bradley to refer to the pressure felt by probation employees in their work with juvenile delinquents, so to some extent we can associate the scientific emergence of this phenomenon with the probation system.

Some studies have indicated that probation officers experience the highest level of stress compared to other actors involved in the criminal justice system, like police or correctional officers [17].

Also, studies published in the specialized literature contain information showing that probation system employees have a higher risk of experiencing situations of intense stress and burnout due to the specificity of the profession (increased risks associated with the profession involving work with criminally convicted persons, the large number of activities, etc.), than other employees in the field of social services [18].

For probation employees, Maslach [19] explained “that negative stress, or distress, occurs as a result of prolonged contact with offenders”, while Chernias [20], Whitehead [18], Stamm [21] and Figley [22] associates stress and distress with factors related to the organizational environment and lack of organization support [2].

Finn and Kuck (2005) said that the main primary sources of stress for probation officers are: high caseloads, large amounts of paperwork and unreasonable deadlines [23].

Job responsibilities of probation officers involve exposure to physical danger, and Wayne J. Pitts notes that if police officers deal with offenders “with appropriate levels of reinforcements, including firearms and back-up support, most probation officers do not have the same levels of training, equipment, or support”, which is a big disadvantage and a reason for additional stress [2]. Moreover, it has been observed that in recent years, with the increasing number of people placed under the supervision of probation services, the rate of supervised people who are at greater risk of being violent has risen. Furthermore, probation services are encountering an increasing number of supervisees with substance use disorders or mental health problems, as well as, more frequently, challenges associated with the intensive migration of probationers [23,24,25,26].

W.J. Pitts observed that probation officers “who feel educationally under-prepared are likely to experience higher levels of occupational stress and more likely to have negative manifestations of stress than those officers who feel well-prepared” [2], education, in this context, referring to the development and training of skills necessary for the profession in the probation system [27].

White W.L, White L.M and collaborators [28,29] identified the main stress-related problems within this profession: the risks associated with working with convicted persons, the high workload, role conflict, bureaucracy, insufficient salary, lack of support from the management team, because as it has been proven, “justice-oriented human resource management related directly to employee burnout and indirectly through employee perceived job control and supervisor social support” [30]. Moreover, in some studies, a series of differences have been noted between experienced and newly hired officers, in the sense that experienced employees have a lower risk of experiencing burnout, as they have developed their own strategies to limit stress, unlike employees with less experience who have not created these strategies and are more at risk of experiencing this disorder [31]. However, we also find other research that does not clearly show the impact of socio-demographic factors such as work experience [32], age [33] and education level [34] on the risk of burnout [35].

So, probation officers, as some studies have established, resort to the coping strategy, which “is a response to excessive job demands and a lack of job resources and can be either adaptive or maladaptive” [27,36] and we find among these adaptive coping strategies examples such as, sense of humor, good family relationships, hobbies, maintaining good relationships with peers, having close friends, physical care, exercise, setting goals, socializing, discussing the problem, religious practice, etc. [28,37,38,39]. Gladfelter A.S. and Haggis W.A. [37] used the Job Demands–Resources (JD-R) model, and they noticed that in probation, job demands and job resources are correlated with burnout, and engagement and resilience significantly predicts every latent variable in the model, as resilience is an important personal characteristic that influences how people react to stressful situations and thus predicts potential burnout.

Another aspect is observed in the literature: often, the clients that probation officers work with have trauma-related problems, and research has shown that working with these people can lead to various reactions, including “vicarious traumatization, compassion fatigue, and burnout” [40].

According to Ersayan A. E et al. [35], a solution to reduce the risk of burnout among probation officers is to carry out interventions aimed at improving their attitudes toward probationers and these could help in enhancing their mental well-being. The authors propose actions such as increasing resources for training and supervision for probation officers. “These trainings could aim to increase mental health awareness and to increase their understanding of the risk of burnout among officers, to increase their positive attitudes toward probationers that could help ensure a more positive work environment”, professional educational training to improve communication and interpersonal skills with probationers [41], training for increasing coping skills or improve their own personal coping resources, and cognitive–behavioral treatments or tailored psychological interventions for probation officers like Adlerian therapy [29,41,42,43,44].

Since most research in the field has not identified a correlation between burnout and certain specific socio-demographic characteristics, all probation officers may be at risk of being affected by burnout and “could theoretically benefit from burnout prevention and intervention efforts” [29,45,46]. In this context, Salyers M. P. et al. [47] proposed activities to limit potential burnout, as a preventative measure to possibly improve probation practices and reduce staff turnover. Even more, Wirkus Ł. et al. [48], conducting a study to the probation system in Poland and making the correlation between styles of coping with stress, concluded that “effective and rational coping aimed at removing or minimizing stressors should be promoted, instead of emotion-focused coping style that is fantasizing”. That is why the authors propose making changes in the organizational structure, from modifying management strategies to help limit the factors that actually create stress (such as high workload, reducing bureaucracy, salary, control), to engaging in finding strategies that lead to increased innovation in the system, increasing the skills of probation officers, and motivating and rewarding staff [48,49].

Even research conducted in other professional fields has highlighted the fact that promoting a positive professional environment may offer many opportunities to reduce burnout, promote engagement and improve the quality of services provided, increase job satisfaction, improve professional relationships and help with staff retention [50]. Also, new studies reflect the need for intervention programs including strategies to motivate employees, improve the work environment, and enhance stress coping mechanisms [51,52,53,54,55,56].

For Dir A.L. et al. [57], “organizational-level interventions might help to reduce mental health stigma and combat negative burnout effects” for probation officers, through elements such as improving the workplace atmosphere and creating a participatory atmosphere for decision-making, which creates opportunities for the employees of being involved in workplace decision-making and ability to share ideas, aspects that have been shown to decrease “levels of emotional, physical and job-related stress” and increase “job satisfaction and job performance among probation officers” and other employees in the criminal justice system [57,58,59].

1.2. The Romanian Probation System: A High-Stress Environment

Beyond all the potential sources of stress and burnout identified in the specialized literature as affecting employees in the probation system, in this study, we also analyze the situation of the activity in the Romanian system, in order to be able to outline the premises of a research on employees in the field. We considered this analysis opportune to evaluate the level of stress and its impact of burnout among probation counselors in Romania, especially since a national study, covering employees from all 42 probation services, has not yet been conducted.

It should be noted that the probation system in Romania is a relatively recent one: between 1997 and 2000, experimental probation centers were established, and it was only in 2000 that Government Ordinance no. 92/2000 on the organization and functioning of victim protection services and social reintegration of offenders was developed, ensuring the legal foundation for the organization and functioning of future probation services. Subsequently, starting with 2001, 42 probation services were gradually established, based on successive orders of the Minister of Justice (28 services in 2001, 13 services in 2022 and 1 service, namely the Ilfov probation service, in 2012). Over the last 24 years (2001–2025), numerous legislative changes have been recorded with a major impact on the probation system. Particularly significant were the entry into force of the new Criminal Code and the new Criminal Procedure Code in 2014, the enactment of Law no. 252/2013 on the organization and functioning of the probation system and Law no. 253/2013 on the execution of sentences, educational measures and other non-custodial measures ordered by judicial bodies during the criminal process. These reforms have been followed by an expansion of the role of the probation service in criminal execution area and a constant and significant increase in the number of probationers registered in the probation system, making it necessary to permanently supplement the staffing scheme, as a result of overburdening the existing human resources [60,61].

In today’s Romania, the probation counselor contributes the reintegration of individuals convicted for committing offences, reducing the risk of committing new crimes, and ensuring community safety as an alternative to the prison environment [62,63,64], all this through a series of duties that were assigned to them, among which we mention the following: 1. supporting the court in the process of individualizing sentences and educational measures by preparing assessment reports for all minor defendants and some adult defendants in the pre-sentencing stage; 2. coordinating the process of assistance and the monitoring of adult or minor persons who have been convicted for offences and for whom, in the case of adults, penalty such as the deferral of sentence application, suspension of sentence under supervision, parole, or for minors, non-custodial educational measures such as civic training internship, supervision, weekend confinement, daily assistance, or release from an educational or detention center have been ordered. The supervision process for adults and minors is based on the principles and stages of case management, so probation counselors are providing social reintegration programs with probationers, they are collaborating with community institutions to fulfill the measures and obligations imposed by the court on convicted persons; 3. coordinating the execution of criminal fines by performing unpaid community work; 4. carrying out specific activities related to individuals deprived of liberty, participating in parole commissions or social reintegration programs conducted in penitentiaries, etc. [65].

According to the information contained in the Report “SPACE II-2022. Council of Europe Annual Penal Statistics: Persons under the supervision of probation agencies”, regarding activities in 2021, Romania was placed among the top three European states in terms of the ratio between the number of supervisions (stock) and the total number of employees in the probation system, after Slovakia and Monaco, with an average number of 84.1 files, well above the European average of 39.1 files per counselor, and the median of 33.5 files [66]. In the latest available Report, SPACE II-2023, we observe for the year 2022 a European average of 43 files for each counselor and a median of 24.5 files [67], and from the Statistical Report of the National Probation Directorate, an average number of 104.47 supervisions files (from the stock of supervisions, without taking into account the flow of exits from supervision) for one probation counselor, on 31 December 2022 [68].

According to the Statistical Report on the Probation System in Romania, from 2024, the total number of supervisions and assistance activities carried out by probation counselors in 2023 (stock on 31 December 2023 + outflow flow from the period 1 January 2023–31 December 2023) remained constant, at a general level, at 102.400 files [64]. At the same time, the average number of supervisions (from the stock of supervisions, without taking into account the outflow flow from supervision) for one probation counselor on 31 December 2023, was approximately 91 [69].

In Romania, in 2022, the number of reports and evaluations prepared was 10,022, to which were added 1460 recommendations/proposals made for the commissions in the penitentiary units. The average of the reports/presentative evaluation reports prepared by a probation counselor in 2022 was 8.8, higher than the European average of 5.82 reports or presentative evaluation reports/probation counselor [69].

Moreover, all employees within the probation system have had and continue to have the status of petitioner–claimer in several legal actions in courts, in the field of labor disputes, concerning their salary rights or other rights of theirs, actions directed against the Ministry of Justice and the National Probation Directorate, all of which are related to the dissatisfaction of employees in the system regarding the method of calculating the salary and the method of assigning other rights.

Research conducted in Romania in 2014 focused on presenting the stress felt among probation counselors, but this material included a study conducted on five probation counselors; thus, the data were not statistically relevant [70].

Another study was conducted in 2021 on 20 (out of 42) probation services in Romania, (of which 10 services with a high caseload), based on the Maslach Burnout Inventory (MBI) burnout assessment questionnaire, to which 101 counselors responded, and it revealed that 40.6% of respondents recorded high levels in the emotional exhaustion dimension. At the same time, low scores were highlighted in the depersonalization dimension, which reveals that probation staff are concerned that in interactions with people in the service’s records, they respect human dignity, show empathy, and establish cooperative relationships with other colleagues. In the dimension of reduced personal achievements, 8.9% of subjects recorded high scores, while low scores were reported in the case of 31.7% of subjects [71]. The authors noted that other research has shown that the feeling of personal fulfillment is more closely related to aspects such as the meaning of work or the existence of a feeling of dissatisfaction with the work performed, compared to other aspects related to the volume of work [72]. As a result of this study, three counselors were identified with chronic burnout, for whom the authors propose an immediate intervention.

We note that the average number of evaluation reports prepared by a probation counselor in Romania, especially the average number of supervision files managed by a Romanian counselor, is much higher than the European average, and to this, the large number of court actions is added, which results in dissatisfaction regarding the salary level and other rights of these individuals, correlated with other causes of stress found in the literature as being specific to employees in the probation system. All these aspects, put together, have outlined the premise of a high level of stress and professional burnout among employees in the Romanian probation system.

1.3. Filling the Empirical Gap in Burnout Research Among Romanian Probation Counselors

Despite the international and national evidence pointing to the stress burden in probation services, no previous nationwide, system-level research has been conducted in Romania to assess burnout using validated measurement tools. This study addresses that gap by offering a comprehensive, quantitative analysis of stress and burnout among probation counselors in all 42 services nationwide. The findings aim to support better institutional policy, targeted intervention, and the long-term well-being of professionals in this essential component of the criminal justice system.

2. Materials and Methods

2.1. Research Justification

Although the specialized literature has repeatedly emphasized the existence of high levels of occupational stress and burnout among criminal justice system employees, particularly in probation, a number of methodological controversies and limitations persist, which affect the consistency and applicability of the conclusions drawn. First, there is no clear consensus on the impact of socio-demographic factors (such as age, experience or educational level) on the risk of burnout, with some studies indicating a significant correlation, while others do not identify statistically relevant differences. Second, most previous research has used limited samples or predominantly qualitative approaches, which reduces the generalizability of the results. Also, organizational aspects such as institutional culture, professional recognition deficits or dysfunctional collaborations are rarely integrated into a multidimensional analytical framework.

For these reasons, we conducted this study, which is the first nationally representative statistical analysis of the Romanian probation system using the Maslach Burnout Inventory–Human Services Survey (MBI-HSS) and using a representative sample from all 42 probation services in Romania, selected through stratified sampling by county. The study correlates psychological data with demographic and organizational variables and applies advanced statistical methods (multiple regression, principal component analysis, K-means clustering) to describe the phenomenon and identify psychosocial profiles based on stress and burnout levels. This approach offers both a quantitative overview and an in-depth understanding of the mechanisms driving, sustaining, or preventing stress and burnout, supporting tailored organizational interventions and human resource policies in the Romanian context.

Another shortcoming of previous studies is that they have rarely integrated a complex organizational perspective, limiting themselves to individual or immediate context factors. The current research overcomes this limitation by including variables related to institutional structure, professional recognition, difficulties in inter-institutional collaboration and relational climate, thus providing a comprehensive approach necessary for the formulation of effective public policies.

Moreover, this study solves the lack of a differentiated perspective on burnout, previously treated as a unitary phenomenon. Through the construction of distinct psychosocial profiles, the research highlights the existence of multiple typologies of stress reaction (compliant young people, resilient seniors, overworked staff), which allows the development of personalized interventions instead of general and uniform measures.

2.2. Methodology

2.2.1. Research Objectives and Design

Given the theoretical framework and the gaps identified in the specialized literature on occupational stress and burnout among probation staff, we chose to conduct sociological research based on a questionnaire, with the following main objectives:

• To assess the level of occupational stress experienced by probation counselors in Romania, by analyzing individual perceptions of organizational, relational and systemic sources of pressure;
• To determine the degree of occupational burnout, using the Maslach Burnout Inventory—Human Services Survey (MBI-HSS) instrument;
• To identify the main determinants of stress and burnout by applying quantitative statistical models (correlations, multiple regressions);
• To identify psychosocial profiles of employees based on the level of stress and burnout, using factor analysis and clustering methods (PCA and K-means), in order to identify organizational typologies with distinct characteristics (e.g., overworked staff, resilient staff, disengaged staff);
• To provide applied recommendations for human resources policies to support burnout prevention and support professional motivation within the Romanian probation system.

2.2.2. Instruments and Measures

In our research, burnout in the Romanian probation system was measured using the Maslach Burnout Inventory— Human Services Survey (MBI-HSS), with 22 items [9] assessing emotional exhaustion, depersonalization and personal accomplishment. Participants used a seven-point scale (from 0 meaning “never” to 6, meaning “daily”) to indicate the frequency with which they experienced each item.

We realized a confirmatory factor analysis (CFA) to check the validity of MBI-HSS among Romanian probation system professionals and we found acceptable factorial validity (χ²(86) = 176.206, CFI = 0.951, RMSEA = 0.075, p = 0.000). We also conducted internal reliability analysis showing Cronbach’s α values: Emotional Exhaustion (α = 0.924), Depersonalization (α = 0.844), Personal Accomplishment (α = 0.892).

In addition to the MBI items, we also asked 15 other questions that aimed to gain knowledge of elements such as age, gender, education, county in which they work, experience in the profession, the number of files they have worked with and are working with at the time of the evaluation, the level of stress felt at work in general, but also the level of stress related to certain aspects such as the number of files, the difficulty of work tasks, complicated situations in case management, low salary level, low level of benefits, tension in the relationship with colleagues and hierarchical superiors, but also the risk of giving up the current profession.

The items for evaluating stress were developed based on systematic literature review of probation stress research and theoretical grounding in the Job Demands–Resources model [73,74]. We conducted an internal consistency measure (Cronbach’s α = 0.897 for the overall stress scale).

The questionnaire has been designed to include the following:

−

A section for collecting demographic and occupational characteristics (questions 1 to 6, 6 items);

−

A section for assessing stress levels (questions 7–15, 17 items);

−

A section for assessing burnout (question 16, 22 items).

2.2.3. Participants and Sampling

Using the Google forms platform, we distributed the questionnaire to probation counselors in Romania, and the application period was January–March 2025. Out of the 732 counselors employed in Romania at the time of the research, 688 were active, while the rest were on unpaid or parental leave, and of these, 247 responded. The sample used is representative of the Romanian probation system and is calculated with a probability of 95% and an error of +/−5%.

A stratified random sampling method was used to ensure the representativeness of the sample across all 42 national probation services. Stratification was based on geographic distribution and institutional size (on county-level distribution and service size), which allowed us to account for regional variations in workload and organizational structure. We took into account the county distribution of probation counselors (we note that there are 42 probation services in Romania, one in each county of the country and one for the municipality of Bucharest—

Table 1. Sample used within the Romanian probation system.

County	Total Employees	Employees Who Were Not Working at the Time of the Questionnaire Application	Employed in the Activity, at the Time of Applying the Questionnaire	Sample
Alba	13	1	12	4
Arad	19	3	16	6
Argeș	20	1	19	7
Bacau	22	1	21	7
Bihor	20	1	19	7
Bistrița-Năsăud	12	1	11	4
Botoșani	14		14	5
Brașov	25	1	24	9
Brăila	8		8	3
București	54		54	19
Buzău	17	2	15	5
Caraș-Severin	9	1	8	3
Călărași	7		7	3
Cluj	26	3	23	8
Constanța	31	3	28	10
Covasna	7		7	2
Dâmbovița	14		14	5
Dolj	32		32	11
Galați	20	3	17	6
Giurgiu	8		8	3
Gorj	10	1	9	3
Harghita	5	1	4	1
Hunedoara	11		11	4
Ialomița	6	1	5	2
Iași	31	3	28	10
Ilfov	18	2	16	6
Maramureș	17	3	14	5
Mehedinți	15	2	13	5
Mureș	17	1	16	6
Neamț	23	1	22	8
Olt	16		16	6
Prahova	20	3	17	6
Satu-Mare	12		12	4
Sălaj	12	1	11	4
Sibiu	16		16	6
Suceava	31	2	29	10
Teleorman	10		10	4
Timiș	34	1	33	12
Tulcea	10		10	4
Vaslui	14		14	5
Valcea	15	1	14	5
Vrancea	11		11	4
Total	732	44	688	247

). Only active counselors at the time of the survey (January–March 2025) were eligible, excluding those on unpaid or parental leave. The sample represents 35.901% of the active population ($n$ = 688). Participants were predominantly female (86.191%), with balanced age groups and two main experience categories—less than 10 years and over 15 years in the profession. Most respondents had degrees in law or social work. This structure reflects the national profile of the probation workforce, supporting the representativeness of the findings.

In order to check the representativity of the sample, we applied chi-square goodness-of-fit tests comparing our sample distribution to population parameters across counties, and the results (χ² = 1.786, p = 1.000) indicated no significant difference. We conducted a power analysis confirming adequate sample size ($n$ = 247), and the results indicated a value of 97.5% power to detecting medium effect sizes (Cohen’s f = 0.25).

2.3. Statistical Analysis

2.3.1. Descriptive Analysis

The perceived stress level was assessed using descriptive statistical indicators at the sample level, including mean, standard deviation, and frequency distribution across the response categories. Stress was measured using a 6-point ordinal Likert-type scale ranging from 0 (“no stress”) to 5 (“very high stress”), applied in Question 9 (“On a scale from 0 (zero) to 5 (five) how would you estimate the level of stress you are currently experiencing at work?”), 10 (“On a scale from 0 (zero) to 5 (five) how do you estimate the level of stress you felt when working with the maximum number of files mentioned by you above?”) and 12 of the questionnaire (“On a scale from 0 (zero) to 5 (five) please indicate the level of stress you feel in relation to different situations you may encounter in your work). The specific situations were listed, as can be seen in the questionnaire and in the subsequent fragment. For analytical purposes, high stress was operationalized as responses with values of 4 (“high stress”) or 5 (“very high stress”). The percentage of respondents falling into this category was calculated and tested for statistical significance using the Z test for proportions.

2.3.2. Correlation and Regression Analysis

The determinants of stress and burnout experienced by the respondents were analyzed. To identify the determinants of stress levels, we analyzed the correlation matrix between the variables recorded in Question 9 (Current Stress Level) and those related to Question 12 (Large Number of Files, Difficult Workload, Complicated Situations Arising in Case Management, Tension in Relationships with Supervised Persons, Poor Cooperation with Community Institutions, Low Level of Pay, Low Level of Benefits, and Tension in Relationships with Colleagues or Hierarchical Superiors).

Next, the analysis identified the causal factors perceived as most important in generating high stress and emotional exhaustion levels. To model and explain the possible causal relationships, we used several types of regression models. However, after analyzing the conditions of application and the quality of the models obtained, we chose the multiple linear regression model [75].

A similar approach was applied to identify the determinants of emotional exhaustion. In this case, the correlation between the average exhaustion component variable resulting from the MBI-HSS and the variables corresponding to question 12 was analyzed. In this case, too, the final model was a multiple linear regression model.

Subject to the limitations of the assumptions underlying the multiple linear regression model, the model can explain the causal link between a set of cause variables and an effect variable using the following equation:

$$Y=X\beta +\epsilon$$

(1)

where

$Y$—matrix of effect variable values

$X$—matrix of values of the cause variables

$\beta$—matrix of regression coefficients of the model

$\epsilon$—matrix of model error values

The model should exhibit errors with an expected value equal with 0, their normal distribution, a constant dispersion (without heteroscedasticity) and not autocorrelated independence for the cause variables and between them and the errors. All these were validated using appropriate statistical tests:

• Jarque–Bera test, to check the normality of the errors [76]:

$$JB={\displaystyle \frac{n}{6}}\left(S^2+{\displaystyle \frac{1}{4}}{\left(k-3\right)}^2\right)$$

(2)

where:

$S$—skewness of error distribution

$K$—kurtosis of error distribution

• t-test to check the mathematical expectation of errors equal to 0 [77];
• Breusch-Pagan test, to check the constancy of the dispersion of errors [78]:

$$BP={\displaystyle \frac{n{\sum }_{i=1}^n{\left({\widehat{e}}^2-\overline{e}\right)}^2}{2{\left({\sum }_{i=1}^n{\widehat{e}}^2\right)}^2}}$$

(3)

where:

$n$—number of observations

$\overline{e}$—mean squared errors

${\widehat{e}}^2$—the estimated values from the auxiliary regression equation

$${\widehat{e}}^2={\alpha }_0+{\alpha }_1{X}_1+{\alpha }_2{X}_2+\cdots +{\alpha }_m{X}_m+u$$

(4)

• Durbin–Watson test, to check the autocorrelation of errors [79]:

$$DW={\displaystyle \frac{{\sum }_{i=2}^n{\left(e_i-e_{i-1}\right)}^2}{{\sum }_{i=1}^n{e_i}^2}}$$

(5)

• $VIF$ and $Tolerance$ calculation for multicollinearity assessment

To assess model quality, we used:

$$VI{F}_j={\displaystyle \frac{1}{1-R_j^2}}={\displaystyle \frac{1}{Toleranc{e}_j}}$$

(6)

where:

$R_j^2$ — $R^2$ obtained for the model regressing the variable $X_j$ on the other independent variables.

• Coefficient of determination ($R^2$):

  $$R^2={\displaystyle \frac{SSR}{SST}}=1-{\displaystyle \frac{SSE}{SST}}$$

  (7)
• Adjusted coefficient of determination (adjusted $R^2$):

  $$Adjusted R^2=1-\left(1-R^2\right){\displaystyle \frac{n-1}{n-k-1}}<R^2$$

  (8)
• Correlation ratio ($R$):

  $$R=\sqrt{R^2}$$

  (9)

2.3.3. Confirmatory Factor Analysis

In order to validate the theoretical structure that connects stress determinants, perceived stress, burnout, and intention to quit, we employed Confirmatory Factor Analysis (CFA) using structural equation modeling. CFA allowed us the simultaneous testing of measurement models and structural relationships while accounting for measurement error [80].

The CFA model used was based on the Job Demands–Resources framework on the empirical literature on burnout and tested the hypothesis of four latent constructs: Stress Determinants (organizational and interpersonal stressors), Perceived Stress (subjective experience of stress), Burnout (emotional exhaustion and depersonalization), and Intention to Quit (intention to quit the current job of the staff). The structural model assumed the following sequential relationships: Stress Determinants → Perceived Stress → Burnout → Intention to Quit, with demographic variables (age, gender, experience) as exogenous predictors.

The CFA model is defined by the following general structural equation framework:

Measurement Model:

$$y = \Lambda ᵧ\eta + \epsilon$$

(10)

$$x=\Lambda ₓ\xi +\delta$$

(11)

Structural Model:

$$\eta = B\eta + \Gamma \xi + \zeta$$

(12)

where:

−

$y$ and $x$ are vectors of observed endogenous and exogenous indicators;

−

$\eta$ and $\xi$ are vectors of latent endogenous and exogenous variables;

−

$\Lambda ᵧ$ and $\Lambda ₓ$ are factor loading matrices;

−

$B$ is the matrix of structural coefficients between latent endogenous variables;

−

$\Gamma$ is the matrix of structural coefficients from exogenous to endogenous latent variables;

−

$\epsilon$, $\delta$, and $\zeta$ are error term vectors.

The model identification was realized by the marker variable approach, fixing the first loading of each latent factor to 1.0. The model was estimated using Maximum Likelihood (ML) estimation, which provides optimal parameter estimates under normality assumptions [81]:

$$F_{ML} = log\left|\Sigma \right(\theta \left)\right| + tr\left[S{\Sigma }^{-1}\right(\theta \left)\right] - log\left|S\right| - p$$

(13)

where:

−

$S$ is the sample covariance matrix;

−

$\Sigma \left(\theta \right)$ is the model-implied covariance matrix;

−

$p$ is the number of observed variables.

The model fit was assessed using multiple indices to provide comprehensive evaluation [82]:

• Chi-square Test of Exact Fit:

  $${\chi }^2=(N-1) · F_{ML}$$

  (14)
• Comparative Fit Index (CFI):

$$CFI=1-{\displaystyle \frac{max\left({\chi }_M^2-d{f}_M, 0\right)}{max\left({\chi }_M^2-d{f}_M, {\chi }_B^2-d{f}_B, 0\right)}}$$

(15)

where subscripts M and B refer to the proposed model and baseline model, respectively.

• Tucker–Lewis Index (TLI):

  $$TLI={\displaystyle \frac{\left(\frac{{\chi }_B^2}{d{f}_B}\right)-\left(\frac{{\chi }_M^2}{d{f}_M}\right)}{\left(\frac{{\chi }_B^2}{d{f}_B}\right)-1}}$$

  (16)
• Root Mean Square Error of Approximation (RMSEA):

  $$RMSEA=\sqrt{{\displaystyle \frac{\mathit{max}\left(\frac{{\chi }_M^2-d{f}_M}{N-1}, 0\right)}{d{f}_M}}}$$

  (17)

We applied established fit criteria for model acceptability [83]:

−

CFI and TLI ≥ 0.90 (acceptable), ≥0.95 (good fit);

−

RMSEA ≤ 0.08 (acceptable), ≤0.06 (good fit);

−

Chi-square/df ratio < 3.0 (acceptable), <2.0 (good fit).

Modification indices (MI) were examined to identify potential model improvements:

$$M{I}_{\mathrm{i}\mathrm{j}} ={\displaystyle \frac{{\left[\frac{\partial F}{\partial \theta ᵢⱼ}\right]}^2}{\left[W^{-1}\right]ᵢⱼ}}$$

(18)

where:

−

${\theta }_{\mathrm{i}\mathrm{j}}$ represents a fixed parameter;

−

$W^{-1}$ is the inverse of the asymptotic covariance matrix.

Modifications were considered only if: (1) MI > 10.0, (2) expected parameter change (EPC) > 0.20, and (3) theoretically justifiable.

2.3.4. Multivariate and Cluster Analysis

Finally, the analysis continued with the identification of groups of employees with common characteristics in terms of perceived levels of stress and burnout, respectively, using visualization and clustering based on principal component analysis (PCA) and appropriate clustering methods. Determining the type of appropriate clustering method was performed in a preliminary analysis using several assessment indicators:

• Statistic GAP [84]:

$$GA{P}_n(k)={\displaystyle \frac{1}{B}}\sum _{b=1}^{B}log{(W}_{kb}^{*})-log{(W}_k)$$

(19)

where:

○

$W_k$—within-cluster sum of squared distances from the cluster means for the observed data.

○

$W_{kb}^{*}$—within-cluster sum of squared distances from the cluster means for the reference dataset used.

○

$B$—number of data sets sampled from the used reference distribution.

○

The standard deviation of the GAP statistic is:

$${SE}_k=\sqrt{{\displaystyle \frac{1}{B}}\sum _{b=1}^B{\left(log{W}_{kb}^{*}-{\displaystyle \frac{1}{B}}\sum _{b=1}^{B}log{W}_{kb}^{*}\right)}^2}$$

(20)

$$s_k={SE}_k\sqrt{\left(1+{\displaystyle \frac{1}{B}}\right)}$$

(21)

The optimal number of clusters is:

$$k^{*}=\{Gap\left(k\right)\ge GAP\left(k+1\right)-s_{k+1}\}$$

(22)

• Silhouette score [85]:

$$S={\displaystyle \frac{D_{nearest-clust}-{\mathrm{D}}_{intra-clust}}{\left({\mathrm{D}}_{nearest-clust},{\mathrm{D}}_{intra-clust}\right) }}$$

(23)

where:

○

$D_{intra-clust}$—mean intra-cluster distance.

○

$D_{nearest-clust}$—mean nearest-cluster distance.

• Davies–Bouldin Index [86]:

$$DB={\displaystyle \frac{1}{k}}\sum _{i=1}^k{\displaystyle \frac{d_i+d_j}{d_{ij}}}$$

(24)

where:

○

$d_i$—the average distance between each point from cluster $i$ and its centroid.

○

$d_j$—the average distance between each point from cluster $j$ and its centroid.

○

$d_{ij}$—the Euclidian distance between centroids of clusters $i$ and $j$.

• Calinski–Harabasz Index [87]:

$$CH={\displaystyle \frac{BSS/(k-1) }{WSS/(n-k)}}$$

(25)

where:

○

$k$—number of clusters.

○

$n$—total number of data points.

○

$BSS$—Between-Cluster Sum of Squares:

$$BSS=\sum _{i=1}^k{n}_i\cdot {\Vert m_i-m\Vert }^2$$

(26)

where:

■

$n_i$—the number of points in cluster $i$;

■

$m_i$ the mean of cluster $i$;

■

$m$—the overall mean of the dataset.

○

$WSS$—Within-Cluster Sum of Squares:

$$WSS=\sum _{i=1}^k\sum _{x\in C_i}{\Vert x_i-m_i\Vert }^2$$

(27)

where $x_i$ is a data point belonging to cluster $C_i$

For identifying a stronger structure of clusters and well-separated clusters, it is better to obtain higher values for statistic GAP, values for Silhouette score closer to 1, lower values for Davies–Bouldin Index and higher values for Calinski–Harabasz Index.

Several clustering methods were tested, including K-means, DBScan, Gaussian mixture model, agglomerative clustering and spectral clustering.

The results of these tests revealed that the conditions for using the K-means method [88] are met. This method can identify compact, hyperspherical and well-delineated clusters, as confirmed through preliminary analysis. However, the method requires the optimal number of clusters ($k$) to be specified in advance.

In order to ensure the robustness of the results obtained by the cluster analysis, a dataset was built with the values determined for the above indicators, whose median was used to establish the optimal number of clusters.

This method is preferred for the following reasons:

• The algorithm is simple and easy to implement.
• It can be easily adapted: distance function, initialization mode, stop criteria, etc., can be modified.
• Its temporal complexity is linear with respect to the number of observations ($N$), dimensions ($D$) and clusters ($K$).
• It is suitable for large datasets.
• It is invariant to data ordering.

Before using the K-means method, it is important to be aware of its limitations:

• The user must decide in advance how many clusters to create. There are cluster validation methods that can help in choosing the $K$ value.
• The method effectively detects only spherical clusters. This limitation can be alleviated by using other distance functions, such as the Mahalanobis distance, which allows ellipsoidal shapes.
• The use of Euclidean distance makes extreme values significantly influence the centroid position. This can be improved by eliminating outliers or using a more robust distance such as the $L1$ distance.
• The algorithm may get stuck in a suboptimal solution depending on the initial points chosen. Improper initializations can lead to empty clusters or slow convergence.

The K-means method is applied by following simple steps [89]:

• Initialization: Choose $K$ initial cluster centers for the set of $M$ points in $N$ dimensions.
• Assignment: Each point is assigned to the nearest-cluster center.
• Update: The cluster centers are recalculated as the average of the points in each cluster.
• Optimization: Check whether moving a point to another cluster would improve the result. If so, the transfer is made.
• Acceleration: Only clusters that have recently changed are monitored to streamline recalculations.
• Repeat: The assignment, update and optimization steps are repeated until no more changes occur.

2.3.5. Ethical Considerations

This study adhered to the ethical standards of sociological research. The conducted sociological research complies with national and international ethical standards governing research involving human subjects. All participants were informed about the aim of the research, the voluntary nature of their participation, and the confidentiality of their responses. Informed consent was obtained prior to completing the questionnaire. Participants were explicitly notified—both at the beginning and end of the online questionnaire—that by clicking the “Submit” button, they agree to provide anonymous and voluntary information that may be used for research purposes and scientific publication. Ethical approval was obtained from the Ethics Commission of the Faculty of Social Sciences, University of Craiova.

All participants provided informed consent through the online platform. Data anonymization procedures were implemented. No personal identifying information was collected beyond general demographic categories necessary for analysis.

3. Results

3.1. Sample Characteristics

The preliminary analysis revealed a sample that includes a majority of female staff (86.234%), with a normal distribution by age groups, with a uniform territorial distribution by probation services, with exceptions for towns with several penitentiaries (Bucharest, Constanța, Dolj, Iași, Suceava). The surveyed staff also show a normal distribution by work experience. The distribution of the interviewed staff by probation experience shows two majority groups: one of those with up to 10 years of experience and a second of those with more than 15 years of experience in the field. Almost half of the respondents had less than 10 years of probation experience. The distribution by educational background revealed that the majority of the staff had a law or social work degree.

3.2. Descriptive Analysis of Stress and Burnout

Descriptive statistical analysis of the results revealed that 45.344% of the respondents perceive a high or very high level of actual stress (level 4 or 5) and only 10.526% of the respondents would be more likely to choose to quit their current job. After aggregating the burnout data, the results showed that only 9.716% of the respondents felt personally exhausted, 2.834% of the respondents felt depersonalization effects, while 54.417% of the respondents felt that they had achieved personal accomplishments.

Mean current stress level was 3.500 (SD = 0.698), with 95% CI [3.420, 3.600]. The distribution showed the following: No/Low stress (1–2): 3.644%, Moderate stress (3): 51.012%, High/Very high stress (4–5): 45.344%.

3.3. Regression Analysis of Stress Determinants

Analysis of the determinants of current stress levels using multiple linear regression modeling revealed the following:

The final model obtained (

Table 2. Current stress model coefficients.

Model		Unstandardized Coefficients		t	Sig.
		B	Std. Error
1	(Constant)	2.206	0.236	9.366	0.000
	Current_Files	0.007	0.002	3.614	0.000
	Difficulty	0.219	0.038	5.814	0.000
	Low_salary	0.176	0.069	2.572	0.011
	Low_benefits	−0.216	0.068	−3.173	0.002

),

$$\begin{array}{cc}C\widehat{urrent\_stress}\hfill & =2.206+0.007\times Current\_Files+0.219\times Difficulty\hfill \\ & +0.176\times Low\_salary-0.216\times Low\_benefits\hfill \end{array}$$

(28)

although significant (Sig. for F-test = 0.000) has a low quality, being able to explain only 19.5% (R² = 0.195) of the variation in the current stress level by the variation in the remaining signified variables in the final model.

This finding indicates that there are other unmeasured factors with a potentially stronger influence on the current stress level, which were not captured by the questionnaire used. These may include personal variables such as individual coping strategies, resilience, mental health history, work–life balance, and organizational aspects such as management style, institutional culture, availability of psychological support, or recent organizational changes. Identifying and measuring these factors would allow further development and refinement of the current analysis.

A few points are worth noting:

■

The workload, as assessed by $Current\_Files$, has almost 0 influence on the current stress level.

■

The difficulty of cases in the pipeline has more influence on stress than the number of cases.

3.4. Regression Analysis of Emotional Exhaustion Determinants

The analysis further aimed to identify the factors that have significant influence in causing the emotional exhaustion effect. To this end, the variables that formed the basis for the stress assessment were used as explanatory variables for a multiple linear regression model modeling the relationship between them and the outcome variable emotional exhaustion.

The resulting model included only two variables with significant influence: faulty collab and current stress (

Table 3. Average exhaustion model coefficients.

Model		Unstandardized Coefficients		t	Sig.
		B	Std. Error
1	(Constant)	−1.766	0.332	−5.324	0.000
	Current_stress	0.891	0.089	9.966	0.000
	Faulty_collab	0.352	0.041	8.658	0.000

).

The resulting model

$$\widehat{Average\_Exhaustion }=-1.766+0.891\times Current\_stress+0.352\times Faulty\_collab$$

(29)

is significant (Sig. for F test =0.000) and has poor–moderate quality and can explain why the variation in the variables contained only 42.9% of the variation of the $Average\_Exhaustion$ variable (

Table 4. ANOVA for average exhaustion model.

	Sum of Squares	df	Mean Square	F	Sig.
Regression	172.031	2	86.015	90.569	0.000 a
Residual	228.884	241	0.950
Total	400.915	243

^a Predictors: (Constant), Faulty_collab, Current_stress.

).

Of the two explanatory variables, the variable $Current\_stress$ had the greatest influence on the perceived level of emotional exhaustion. Interpreting the value of its coefficient, it can be seen that any increase in the level of current stress is transmitted almost unattenuated directly into the level of the emotional exhaustion variable.

3.5. Confirmatory Factor Analysis Results

A CFA was conducted in order to validate the theoretical structure which conceptualizes the relationships among organizational stress determinants ($STRS$), perceived stress ($PSTR$), burnout syndrome ($BURN$), and intention to quit ($QUIT$) within the Romanian probation system context. The final model obtained is presented in Figure 1.

The final model shown satisfactory fit with the empirical data, according to standardized evaluation indicators. The Comparative Fit Index (CFI = 0.951) exceeds the recommended threshold of 0.95, indicating excellent fit of the theoretical model with the observed data. The Tucker–Lewis Index (TLI = 0.934) falls within the acceptable range (0.90–0.95), confirming model stability. The Root Mean Square Error of Approximation (RMSEA = 0.075) falls within acceptable limits (≤0.08), suggesting reasonable approximation of the population covariance structure. The chi-square test (χ² = 176.21, df = 77, p < 0.001) indicates a statistically significant difference from the perfect model, but this value should be interpreted with caution given the sensitivity of this indicator to sample size.

The latent factors present satisfactory indicators of convergent and discriminant validity. For the $BURN$ construct, the standardized factor loadings are high: emotional exhaustion ($EE$, λ = 0.980) and depersonalization ($DP$, λ = 0.707), indicating adequate representation of burnout dimensions according to the Maslach model. The $PSTR$ construct is represented through current stress ($Current\_stress$, λ = 0.690) and maximum experienced stress ($Max\_stress$, λ = 0.655). The $STRS$ factor presents variable factor loadings, with the highest values for high caseload ($High\_files$, λ = 0.856), task difficulty ($Difficulty$, λ = 0.864), and complicated case management situations ($Complicated\_situations$, λ = 0.791).

The model confirms the existence of a complex causal sequence in the development of burnout and $QUIT$. Organizational $STRS$ exert a moderate effect on perceived stress ($\beta$ = 0.447, p < 0.001) and explain approximately 20% of its variation. This relationship suggests that objective organizational factors translate significantly into the subjective stress experience of probation counselors.

A remarkable finding is the identification of a dual pathway of influence toward $BURN$. On one hand, $PSTR$ directly influences burnout ($\beta$ = 0.393, p < 0.001), confirming the classic theoretical mechanism of cognitive mediation. On the other hand, $STRS$ also exert a direct effect on burnout ($\beta$= 0.451, p < 0.001), independent of mediation through $PSTR$. This dual pathway suggests that organizational factors can generate burnout both through cognitive stress evaluation processes and through direct resource depletion mechanisms.

$QUIT$ is primarily influenced by burnout ($\beta$ = 0.509, p < 0.001), representing a moderate relationship that confirms the specialized literature regarding organizational burnout consequences. The effect of gender on intention to quit ($\beta$ = 0.108, p < 0.05) is weak but statistically significant, suggesting minor differences between men and women regarding intentions to leave the profession.

Age ($Age\_C$) presents a weak but significant effect on $PSTR$ ($\beta$ = 0.176, p < 0.05), indicating that older employees tend to report slightly higher levels of $PSTR$. This finding may reflect either increased sensitivity to stressors with age or a greater capacity to recognize and report stress.

The analysis of modification indices (MI) reveals only one significant improvement suggestion: the correlation between high caseload and complicated case management situations ($High\_files$ ~~ $Complicated\_situations$, MI = 10.65, EPC = −0.175). This negative correlation suggests that, contrary to intuitive expectations, high caseload does not directly associate with increased complexity of individual cases, possibly reflecting differences in case type distribution among counselors. The absence of other significant modifications (MI > 10) confirms that the theoretical model adequately captures the structure of relationships among variables, without requiring major specification adjustments.

The results confirm the validity of a multidimensional model of occupational stress in the probation system, which integrates both objective organizational determinants and subjective evaluation processes. The identification of a dual pathway to burnout (direct and mediated) has important implications for organizational interventions, suggesting the need for differentiated strategies that address both objective work condition restructuring and stress management competency development. For more detailed information, see the Supplementary Materials.

3.6. Multivariate and Cluster Analysis Results

The next stage of the analysis aimed at identifying groups of staff with similar stress characteristics and in the last part of the analysis we aimed to identify the structure of groups of staff showing characteristics of emotional burnout.

3.6.1. Stress Cluster Analysis Results

All recorded demographic and socio-professional variables were used, together with the variables for the assessment of stress levels. On their basis, a principal component analysis (PCA) was performed. Of the principal components determined, the first two were retained in order to subsequently visualize the clusters using 2D graphical representation.

Preliminary analysis was performed to identify the characteristics of the clusters. Several clustering methods were tested: K-means, DBScan, Gaussian Mixture Model, Agglomerative Clustering and Spectral clustering. This indicated that of the clustering methods tested, the K-means method best identified the cluster structure, although the clustering quality assessment metrics (Silhouette: 0.25, Statistic GAP: 0.31, Davies-Bouldin: 1.56, Calinski-Harabasz: 86.56) indicated moderate quality.

To determine the optimal number of clusters, the method that ensures cluster robustness was used, by using the median of the dataset constructed from the number of optimal clusters indicated based on the calculated scores. The optimal number of clusters thus determined was $k$ = 3.

The cluster structure thus identified is presented in

Table 5. Clusters of staff with similar stress characteristics.

Cluster	No. Persons	Percentage
0	61	25.414
1	69	28.750
2	110	45.836
Total	240	100.00

:

The first two identified PCA equations are presented below:

$$\begin{array}{cc}PC1 = -0.086 & \times Age\_C - 0.085 \times Work\_experience \hfill \\ & - 0.113 \times Probation\_experience \hfill \\ & - 0.031 \times Current\_Files + 0.086 \times Current\_stress \hfill \\ & + 0.392\times Difficulty + 0.403\times Complicated\_situations \hfill \\ & + 0.401\times Stressful\_situations + 0.343\times Faulty\_collab \hfill \\ & + 0.375\times Low\_salary + 0.385\times Low\_benefits \hfill \\ & + 0.286 \times Tensions\_colleagues\hfill \end{array}$$

(30)

$$\begin{array}{cc}PC2=0.569& \times Age\_C +0.559\times Work\_experience\hfill \\ & +0.455\times Probation\_experience\hfill \\ & +0.007\times Current\_Files+0.186\times Current\_stress\hfill \\ & +0.140\times Difficulty\hfill \\ & +0.110\times Complicated\_situations\hfill \\ & +0.067\times Stressful\_situations\hfill \\ & +0.120\times Faulty\_collab-0.132\times Low\_salary\hfill \\ & -0.114\times Low\_benefits+0.202\times Tensions\_colleagues\hfill \end{array}$$

(31)

Analyzing the structure of the equation PC1, it can be seen that it actually measures a kind of professional and organizational pressure, representing an index of stress and dissatisfaction, in other words a dimension of evaluation of organizational stress perceived by employees. Its equation presents large positive values for the variables: $Difficulty$, $Complicated\_situations$, $Stressful\_situations$, $Low\_salary$, $Low\_benefits$, $Faulty\_collab$, $Tensions\_colleagues$.

Carrying out the same path for PC2, we observe that it actually evaluates the level of professional seniority (older, more experience). Its equation is dominated by the following variables: $Age\_C$, $Work\_experience$, $Probation\_experience$.

These principal components were used to visualize and interpret the results obtained after grouping (Figure 2):

Figure 2 the PC1-PC2 space clearly shows the three clusters obtained:

• Cluster 0 (purple)—includes 25.414% of the analyzed employees. By analyzing its characteristics, it can be summarized that it includes stressed young employees but who had no visible reaction to stress. The localization in the graph (PC1 negative, PC2 negative) leads to the conclusions that this group has a low degree of stress perception and consists of employees of young age and with little experience.

Due to the fact that in this first purple cluster, there was an overall low level of stress and an almost absent level of emotional/behavioral reaction, we are confronted with a very likely underreporting of stress, either due to a lack of ability to identify stress or due to a normative behavior denoting institutional conformity. How can we interpret this result? By looking at the psychosocial profile of these employees. It is characterized by “low involvement and minimal stress”, latent uncertainty or emotional detachment (transposed by fear or not caring to externalize about the difficulties of tasks or by low involvement in difficult cases), by defensive adaptability (understood as acceptance of status and because they have a possible profile characterized by a more detached or passive attitude towards the institutional context), but also by a lack of benchmarks (the system not providing an induction phase that provides for a period of coaching and not including professional recognition in the HR procedure, under the chapter on employee motivation). Naturally, from this perspective, the probation system is faced with organizational risks such as: silent, invisible, unnoticeable burnout(s), difficult to identify in the absence of proactive interventions, but also with possible spontaneous departures from the service (periods of suspension, sabbaticals or resignations) or even a decline in motivation.

From a sociological point of view, the employees whose choices have placed them in this cluster are in the early stages of professional socialization, where organizational norms are not yet internalized, their professional role is poorly defined, and they express an attitude inclined towards normativity, deriving from lack of experience and prestige and assume institutional customs and hierarchical directives (either because they are unaware of mechanisms to challenge legitimacy, or because they lack professional reactive maturity, or because they feel more protected by adopting attitudes of professional detachment/passivity/alienation), for fear of jeopardizing their position.

In order for the organization to eliminate the stress level of these employees, it would be advisable to introduce mentoring programs provided by those with seniority in the system (in order to achieve normative transfer), design structured organizational induction plans (to quickly familiarize them with the values, roles and expectations in probation), to provide comfortable contexts for feedback (by organizing case/case review groups to gauge positive perceptions or fears/insecurities expressed by young employees), to facilitate their access to Burnout risk awareness, to encourage them to build resilience and to strengthen their sense of belonging.

• Cluster 1 (teal)—includes 28.750% of the analyzed employees, namely stress resilient senior employees. Localization in the graph, with PC1 moderate (negative to zero) and PC2 positive, respectively, leads to the conclusions that this group of employees perceive low-moderate stress and have high age and experience.

Employees whose answers placed them in cluster 1 are characterized psychologically as ‘stress resilient’, which implies that they perceive low to moderate stress and have a high capacity for emotional regulation, showing intrinsic motivations such as job content, job security, etc. Psychologically, they appear to manage stress through experience and adaptive approaches (being stable and balanced, with healthy coping mechanisms), but they can also be vulnerable to prolonged stress or sudden change (but managing to mask their psychological consumption through emotional withdrawal). But they are also subject to organizational risks, such as ‘professional indifference’ (late-stage burnout), if they do not receive constant validation and recognition, or they can become isolated or even easily slip into ‘professional spleen’ in the absence of stimulation.

Sociologically, it is those employees who have gone through the full process of professional socialization, becoming the holders of symbolic and relational capital, who, within the social division of labor, are seen as anchors of balance, but who can also become professionally rigid, difficult to adapt in an unstable environment and resistant to change. Some manage to cope with the stress manifested as a form of tolerance learned through years of work experience, others are less resilient and reach a plateau (the so-called professional apathy characteristic of Burnout Phase 3, disengagement).

And in their case, in order for the organization to decrease their stress level, it should value their experience (i.e., involve them in training new employees, basically valuing them as facilitators of cohesion), diversify their responsibilities to keep them motivated and prevent them from becoming routinized, provide them with public validation (through symbolic awards, personal and professional development, etc.), allow them to be involved in strategic system decisions, and provide them with career coaching services (through periodic reassessment of internal values and career aspirations).

• Cluster 2 (yellow), includes almost half of the analyzed employees (45.836%), namely Overworked and Frustrated staff. Its location in the graph, with positive PC1 and medium-high PC2, respectively, indicates for this cluster, consisting of moderately experienced or senior employees, perceived high levels of stress and pressure.

Almost half of the respondents of this research form the group of overworked employees, who psychologically report “systematic stress”, complain of work overload with high caseload, i.e., huge volume and lack of recognition, poor collaboration with other institutions, report low salaries, low benefits and relational strain, showing frustration and emotional distancing. From a psychological point of view, staff in this cluster feel exhausted by tasks, perceive unfairness and inequity, and are at imminent risk of advanced burnout, with manifest symptoms, resulting in burnout and decreased motivation and job performance, absenteeism and conflict.

From a sociological standpoint, those in this cluster have an active critical conscience; they are employees who are aware of the system’s rigors but constantly challenge them, often expressing emotional detachment and displaying a kind of professional anomie.

In their case, in order for the probation system to adequately manage stress levels, it needs to take urgent organizational development measures, more specifically to conduct an internal analysis on the equitable distribution of tasks by reviewing job descriptions and institutional pressure, to conduct regular participatory organizational assessments, to reorganize internal communication circuits, to rethink the system of rewards and social recognition, and to offer employees affected by burnout the deconcentration of psychotherapy and counseling services, and time management and activity planning.

3.6.2. Emotional Burnout Cluster Analysis Results

All variables for assessing the level of personal burnout, depersonalization and burnout were used for this purpose, together with demographic and socio-professional variables.

The same steps were followed as for the detection of the group structure of stress.

Preliminary analysis of cluster characteristics was performed by testing several clustering methods, such as K-means, DBScan, Gaussian Mixture Model, Agglomerative Clustering and Spectral clustering. Again, among the clustering methods tested, the cluster structure was best identified by the K-means method. The clustering quality assessment metrics (Silhouette: 0.20, Statistic GAP: 0.37, Davies–Bouldin: 1.62, Calinski–Harabasz: 59.13) indicated moderate quality. The optimal number of clusters determined in the same manner as in the previous case was $k$= 3.

The cluster structure thus identified is presented in

Table 6. Clusters of staff showing characteristics of emotional burnout.

Cluster	No. Persons	Percentage
0	137	58.547
1	41	17.521
2	56	23.932
Total	234	100.00

:

The equations of the identified principal components are presented below:

$$\begin{array}{cc}PC1 = 0.016 \times & Age\_C + 0.006 \times Work\_experience \hfill \\ & + 0.048 \times Probation\_experience + 0.283 \times Drained \hfill \\ & + 0.286 \times Used\_up\_worday + 0.282 \times Tired \hfill \\ & + 0.058 \times Understand\_beneficiaries \hfill \\ & + 0.236 \times Treat\_benef\_impersonal \hfill \\ & + 0.262 \times Strain\_work\_people \hfill \\ & + 0.061 \times Effectively\_problems\_beneficiaries \hfill \\ & + 0.303 \times Burned\_out\_work \hfill \\ & - 0.041 \times Positively\_influencing + 0.208 \times Callous \hfill \\ & + 0.302 \times Hardening\_emotionally - 0.121 \times Energetic \hfill \\ & + 0.265 \times Frustrated + 0.265 \times Working\_too\_hard \hfill \\ & + 0.219 \times Don’t\_care + 0.243 \times Work\_people\_stress \hfill \\ & + 0.018 \times ConstAtmRelax - 0.055 \times Exhilarated \hfill \\ & - 0.078 \times Worthwhile\_things + 0.291 \times End\_of\_rope \hfill \\ & - 0.058 \times Emotional\_problems\_calmly \hfill \\ & + 0.194 \times Beneficiaries\_blame\_proble\hfill \end{array}$$

(32)

$$\begin{array}{cc}PC2=0.088\hfill & \times Age\_C+0.121\times Work\_experience\hfill \\ & +0.084\times Probation\_experience+0.098\times Drained\hfill \\ & +0.109\times Used\_up\_worday+0.018\times Tired\hfill \\ & +0.341\times Understand\_beneficiaries\hfill \\ & -0.022\times Treat\_benef\_impersonal\hfill \\ & -0.077\times Strain\_work\_people\hfill \\ & +0.327\times Effectively\_problems\_beneficiaries\hfill \\ & +0.097\times Burned\_out\_work\hfill \\ & +0.374\times Positively\_influencing-0.079\times Callous\hfill \\ & -0.005\times Hardening\_emotionally+0.273\times Energetic\hfill \\ & -0.001\times Frustrated+0.159\times Working\_too\_hard\hfill \\ & -0.133\times Don’t\_care-0.053\times Work\_people\_stress\hfill \\ & +0.337\times Relaxed\_atmosphere+0.379\times Exhilarated\hfill \\ & +0.336\times Worthwhile\_things+0.048\times End\_of\_rope\hfill \\ & +0.266\times Emotional\_problems\_calmly\hfill \\ & -0.029\times Beneficiaries\_blame\_problems\hfill \end{array}$$

(33)

Analyzing the structure of the PC1 equation, it can be observed that it actually assesses the cumulative level of burnout, physical and emotional exhaustion, and cynicism towards the beneficiaries. It presents high and positive values for the coefficients related to the variables: Used_up_workday, Drained, Burned_out_work, Hardening_emotionally, Frustrated, End_of_rope, Working_too_hard, Work_people_stress, Callous.

PC2, on the other hand, measures the level of Motivation, Well-being, Exhilarated and Perceived Effectiveness, representing, in other words, the component of positive motivation and positive perception of work. It shows coefficients with positive and high values for the variables: Exhilarated, Worthwhile_things, Relaxed_atmosphere, Emotional_problems_calmly, Positively_influencing, Effectively_problems_beneficiaries.

The graphical representation of the identified clusters using the two components is presented in Figure 3.

The three clusters identified have the following characteristics:

• Cluster 0 (purple), groups most of the employees (58.547%) and includes balanced employees with positive motivation and low-moderate stress. Its graphical localization determined by values for PC1 low or moderate values and positive values for PC2, leads to the identification of the following common characteristics for employees:

This group of people can be characterized psychologically by “adaptive functioning”, registering employees “who perform mechanistically”, through inertia and routine. These people have a low level of emotional exhaustion and depersonalization, but with high self-demand, with a positive relationship to professional achievement and managing to maintain a relatively balanced psychological state. They seem to internalize stress, not knowing how to ask for help and accumulating tensions, showing tendencies to minimize efforts and try to maintain control, not being able to recognize their limits, because they have set as a value the duty to the meaning of work. When they do not manage their time and emotions properly, they risk burnout, with various somatizations and ‘empathy BURN’. They have effective adaptive strategies, a good work–life balance and are potential stability factors in the institution.

From a sociological point of view, they are those employees who can be associated with the description of the ‘burnout by over-compliance’ typology, because they have internalized their professional roles and institutional values so well that they are no longer aware of their level of burnout. They exhibit increasing levels of obedience, their burnout stemming from lack of recognition.

They are in constant need of support and validation, which requires streamlining services in the probation system: digitization and even the introduction of AI, which could help eliminate repetitive tasks and make working hours more flexible. Revalorizing the social component of the profession could help to revitalize the sense of usefulness and meaning in providing help, which reinforces the perception of symbolic and public recognition of the probation officer’s work.

• Cluster 1 (teal), groups 17.521% of the analyzed staff, with the common determinant traits that they do not perceive themselves as overtired and lack motivation. The graphical localization of the cluster, with negative values for PC1 and PC2, leads to determine the following common characteristics: lower reported levels of emotional exhaustion, reduced involvement in professional activities, limited engagement in additional responsibilities, minimal perceived work-related stress, and a tendency to maintain a stable but low-intensity work rhythm.

This cluster is defined from a psychological perspective by “latent tension”. These individuals have a moderate level of burnout but high stress and have a high potential to drift into burnout if stress persists. They show exhaustion, low empathy, helplessness and lack of control, indifference, detachment and robotic in relationships, overt forms of heightened depersonalization and even anxiety. They are at a critical point, where an increase in stress or deterioration in the professional environment can lead to emotional decompensation.

From a sociological standpoint, the drift away from institutional values and loss of meaning at work leads employees who placed their responses within this cluster towards “organizational anomie”. We notice the erosion of social relations, the symbolic rupture from the system by questioning the credibility and relevance of norms and depersonalization as a defense mechanism.

For this category of employees, the organization should consider streamlining their work, either by reducing the number of files, limiting interactions with chronically difficult beneficiaries, or by temporary internal retraining, in order to get them out of routine and give a boost of creativity and/or diversity to their work. Measures such as streamlining internal communication, job rotation, organizational psychotherapy, alternating work and non-work periods, longer periods of leave, individual or mixed leave, can be solutions for them.

• Cluster 2 (yellow) includes 23.932% of the analyzed staff and more clearly includes staff who perceive a high level of stress and a low or moderate level of motivation. The graphical localization of the cluster, with high values for PC1 and low to medium values for PC2, leads to the identification of the following common characteristics for the contained staff: pronounced emotional exhaustion, frequent feelings of being overworked, reduced enthusiasm for daily tasks, limited participation in optional or extra-professional activities, a perception of insufficient institutional support, and a tendency to experience work-related tension on a regular basis.

From a psychological standpoint, this group of people can be described as exhibiting “burnout symptoms”. These individuals show high levels of emotional exhaustion, experience constant stress and poor collaboration and have a high risk of full-blown burnout, possibly also symptoms of depersonalization. They are the ‘functional disconnected’, emotionally and intellectually detached, who no longer attach personal meaning to their work, but perceive it as a bureaucratic obligation, without ennobling it with passion. In the work they do with the beneficiaries, relations become distant, indifference sets in, patience is lost and even intolerance develops. They are at an advanced stage of psychological burnout, experiencing systemic and emotional frustrations, with serious risks for their mental health and professional performance.

From a sociological standpoint, the lack of affective involvement in professional activity and the incongruence between values and norms label this cluster of employees as ‘cynical conformists’, for whom work becomes mechanical, bureaucratization becomes excessive, disengagement is progressively dynamic and emotional and intellectual disconnection becomes a psycho-professional protective mechanism.

For them, the probation system should reconfigure the organizational culture and open it to the employees by shaping it towards a participatory core, reenergize teams, redefine the decision-making boundaries, redesign activities to redefine the meaning of work and the social mission of the organization.

4. Discussion

The present study is the first comprehensive, national investigation of occupational stress and burnout among probation counselors in Romania, addressing a significant gap in the literature and at the same time bringing a new methodological approach to understanding burnout in probation systems. The results obtained in this study confirm international models of stress and burnout in probation work and additionally reveal unique characteristics of the Romanian context through advanced statistical modeling.

While previous studies have often relied on limited samples or have focused on a single location [16,70], our study used a stratified random sampling methodology across all 42 national probation services, with 247 participants, representing 35.901% of the active workforce. This approach contrasts with previous studies conducted in Romania, such as the qualitative study by Severin [16] which was conducted only on five probation counselors, and the study conducted by Bălă and Oancea [70], conducted on 101 counselors from 20 services, limited to the regional level. The integration of CFA represented a significant methodological improvement over previous research in the field of probation. While studies such as the one conducted by Wirkus et al. in 2021 [48] in Poland used the MBI without validating its factorial structure in their sample, our study included a comprehensive CFA validation. The results obtained (χ²(86) = 176.206, CFI = 0.951, RMSEA = 0.075) confirmed an acceptable factorial validity of the MBI-HSS among employees in the Romanian probation system. The approach allowed us to eliminate one of the systematic methodological limitations of existing studies, identified in the critical review conducted by Okroș and Vîrgă in 2022 [90]. They emphasized the importance of verifying measurement structures in specific occupational contexts.

Using structural equation modeling to test the complex relationships among $STRS$, $PSTR$, $BURN$, and $QUIT$ work allowed the identification of dual pathways to burnout:

−

Direct effects from organizational stressors ($\beta$ = 0.451, p < 0.001);

−

Mediated effects through perceived stress ($\beta$ = 0.393, p < 0.001).

In addition, it provides empirical evidence for theoretical models that have been untested in probation contexts.

This finding of dual pathways contradicts simpler models found in previous research. For example, while Finn and Kuck in 2005 [23] identified high workload and bureaucracy as the main stressors, our SEM analysis reveals that the relationship between organizational factors and burnout is more complex and is exerted through both cognitive appraisal mechanisms and direct resource depletion processes. A better and more nuanced understanding has important implications for the design of interventions and highlights that effective prevention of burnout must address both objective work conditions and subjective stress appraisal processes.

The use of principal component analysis and cluster analysis enabled a refined understanding of latent burnout profiles, which not only supports individualized interventions but also calls for a rethinking of uniform staff management strategies. This typological differentiation reinforces the need for both targeted prevention programs and institutional reforms aimed at professional well-being.

Our K-means cluster analysis identified three distinct psychosocial profiles, a novel contribution to probation research. The typology—stress-resistant seniors (28.750%), under-involved younger staff (25.414%), and frustrated overworked employees (45.836%)—provides a more nuanced understanding of burnout vulnerability than the uniform approaches typically used in the literature.

This finding contrasts with previous research. Although Holgate and Clegg in 1991 [31] observed differences between experienced and new officers in burnout susceptibility, their binary classification fails to capture the complexity highlighted by our analysis. Similarly, studies by Gayman et al. in 2018 [33] and Andersen et al. in 2017 [32] found mixed results regarding demographic predictors of burnout. Their findings are clarified and nuanced by our cluster analysis, which revealed that demographic factors interact with organizational and psychological variables to create distinct risk profiles. The identification of the “under-involved young staff” cluster is particularly significant because it represents a previously unrecognized pattern of potential silent burnout, characterized by emotional detachment and defensive adaptability. This finding suggests that traditional assessments of burnout may fail to capture early-stage disengagement among newer employees who may appear well-adjusted while, in fact, are at risk for burnout.

The psychosocial profiling approach offers practical advantages over the universal interventions proposed in the literature. While Ersayan et al. in 2022 [35] recommended implementing targeted training programs to improve probation officers’ attitudes toward probationers—such as mental health awareness initiatives, burnout risk education, and cognitive–behavioral or acceptance-based workshops to reduce depersonalization and foster professional accomplishment—the results of our analysis suggest these interventions should be further tailored to specific psychological profiles. For example, the group of “stress-resistant seniors” would rather require opportunities for validation and experience exchange, while “frustrated overworked employees” would require immediate organizational restructuring and stress reduction measures.

The results of our analysis also align with broader European trends, while revealing specific vulnerabilities in Romania. The average of 93 supervision cases per counselor (data from 2023) significantly exceeds the European average, confirming Romania’s position among the most overburdened probation systems in Europe [67]. However, unlike previous research that treated high workload as uniformly problematic, our regression analysis revealed that current workload ($\beta$ = 0.213, p < 0.001) has a smaller impact on stress than task difficulty ($\beta$ = 0.382, p < 0.001), suggesting that case complexity may be more important than simple volume. This finding contradicts the assumptions underlying workload reduction policies and aligns with more nuanced perspectives emerging in the recent literature. Gladfelter and Haggis in 2024 [37], testing the Job Demands–Resources model in the United States, similarly found that qualitative aspects of job’s demands matter more than quantitative measures, supporting our conclusion that intervention strategies should focus on case management support and training rather than simply reducing workload.

Compared to other studies from Central and Eastern Europe, our study provides superior empirical support for policy recommendations. The Polish study by Wirkus et al. in 2021 [48] identified important relationships between coping styles and burnout, but did not systematically address the organizational factors that were included in our study. While their study’s focus on emotion-focused versus problem-focused coping provided valuable insights, our structural equation modeling highlights the organizational antecedents that necessitate certain coping styles and thus addresses the root causes rather than just the symptoms.

This study has several theoretical and practical implications. Theoretically, this study expands the burnout literature by applying multidimensional quantitative methods to an overlooked professional context. Practically, the results can inform the design of differentiated human resource policies, emphasizing the need for early identification of high-risk staff and the development of burnout prevention tools. Furthermore, findings advocate for institutional investments in emotional support programs, workload rebalancing, and professional development pathways—especially for younger and emotionally exhausted employees. This study provides a solid empirical basis for guiding national policies aimed at enhancing well-being, retention, and effectiveness within the Romanian probation system.

5. Research Limitations and Future Directions

Like any empirical approach, the present study has a number of limitations that need to be recognized in order to ensure an accurate interpretation of the results. First, part of the instrument used, the questionnaire based on the MBI-HSS and specific items on stress perception, is based solely on self-report, which may lead to institutional underreporting or conformity, especially among young or early-career employees. Second, the quantitative model did not include deep psychosocial variables such as individual coping style, level of resilience, support outside the workplace or perception of personal meaning of work, which may decisively influence the risk of burnout.

Another limitation is the temporal limitation, capturing a static picture of the phenomenon without being able to track the evolution of stress and burnout over time.

As for the regression models applied to identify predictors of stress and burnout, the results showed a modest quality of statistical explanatory power: only 19.5% of the variance of current stress is explained by the linear model identified, and in the case of burnout, only 42.9%. These data suggest that there are additional relevant factors that were not captured by the applied questionnaire, such as leadership style, organizational culture, perceived meaning of work, or work–life balance. Thus, regression models, although statistically significant, reflect reality only partially and should be interpreted with caution, in complementarity with qualitative and typological analysis.

Based on these limitations, future research directions may include longitudinal studies that follow the evolution of occupational stress and burnout over the medium and long term; the integration of qualitative methods (interviews, focus groups, participatory observation) to capture the subjective dimensions of occupational stress and the employee-institution relationship; and the extension of the analytical model by including psychological and contextual variables (e.g., resilience, coping, organizational climate, family support). Also, the development of institutional tools for continuous assessment and psychosocial intervention would allow active monitoring of burnout risk and early intervention.

6. Conclusions

This study represents the first nationwide analysis of occupational stress and burnout among probation counselors in Romania, offering both empirical data and theoretical insight into a professional category largely neglected by research. Using the Maslach Burnout Inventory—HSS and advanced statistical methods, the study identified high levels of emotional exhaustion and significant correlations between stress factors and burnout dimensions.

A key contribution of this research lies in the identification of three psychological profiles—stress-resistant senior staff, under-involved younger professionals, and overworked, frustrated employees—through a combination of regression and cluster analysis. This typology allows for a more nuanced understanding of burnout risk, revealing that vulnerability is not uniformly distributed among all counselors.

The integration of CFA confirmed the factorial validity of the MBI-HSS in this professional context, strengthening the reliability of our findings. Results revealed high prevalence rates of stress and burnout, with emotional exhaustion strongly predicted by $PSTR$.

This study also confirmed the relevance of contextual factors such as dissatisfaction with salary and workload, insufficient support, and perceived injustice in shaping burnout risks. These findings support the need for differentiated institutional policies, psychological interventions tailored to staff profiles and differentiated human resource strategies that address not only the visible symptoms of burnout, but also the subtle psychosocial dynamics that contribute to its progressive onset.