Child and Adolescent Social Work Journal (2021) 38:175–180

https://doi.org/10.1007/s10560-020-00733-w

Using Machine Learning in Burnout Prediction: A Survey

Małgorzata Grządzielewska1

Accepted: 14 December 2020 / Published online: 15 January 2021

© The Author(s) 2021

Abstract
Accurate prediction provides a number of important benefits for research and decision-making. Occupational burnout is
intertwined with individual, cultural, and social factors, the resolution of which requires methods that can deal with large
amounts of data. The application of such methods capable of dealing with large datasets is a relatively novel research area
in social science. For this purpose, this article presents insights into machine learning methods, mainly related to prediction
tasks. A brief review of these techniques in burnout domain was applied. It is shown that the choice of a method depends on
the presence of certain dependent variables. This paper also presents a comparison between novel and traditional approaches,
which shows that the appropriateness of a technique depends on the aim of the research. The theoretical and practical impli-
cations of using machine learning methods in this context is also presented in the paper. It is found that a gap in the study of
burnout exists which requires the attention of social work researchers. Through machine learning techniques, new theoretical
models of burnout can be created. These algorithms can also provide new approaches to create data-driven interventions.
Burnout monitoring systems supported by machine-learning algorithms can also be used in recruitment processes and to
supervise employees. Applying machine learning methods in reducing burnout can also provide socio-economic benefits
such as help to reduce employee turnover and improve general working conditions.

Keywords Machine learning · Supervised machine learning · Prediction · Burnout · Social work

The meta-analysis of burnout shows that many factors influ- workload pressures can lead to poor decision-making and
ence the perceived level of this syndrome, from the indi- also the tendency to take short-sighted view of issues (Kel-
vidual level to the macroeconomic level. Each individual ler & Ho, 1988). McGee (1989) shows that social workers
may have a complex combination of factors, so there is a with higher levels of this syndrome made earlier decisions
need to deal with a large amount of data. Applying machine and were more likely not to change their minds even in the
learning to data from various source (e.g. EEG, sensors, face of new facts.
questionnaires) might help to identify and predict burnout Burned out personnel copes with stress by denying the
in workers. The results obtained with these methods can need to engage in particularly demanding cases. Therefore, it
be more accurate because they provide more techniques to is important to deal with this syndrome since the behaviour
examine non-linear relationships in the data than traditional of employees may have serious consequences especially for
statistical techniques. young clients.
Usually, burnout is assessed in white-collar (civil serv- Job burnout is a syndrome of emotional exhaustion,
ants), and the helping professions such as caregivers, teach- depersonalization, and reduced feelings of personal accom-
ers (Kim et al., 2011; Valente et al., 2011; Vercambre et al., plishment. A key dimension is emotional exhaustion where
2009) and social workers in child welfare practise (Lizano people feel overwhelmed by work and are physically and
& Mor Barak, 2012; Mcfadden, Cambpell & Taylor, 2014; emotionally fatigued. A second dimension—depersonaliza-
Travis & Mor Barak, 2010). Working under considerable tion—means that workers developing cynical feelings and
attitudes about their clients. The last dimension is reduced
personal accomplishment, where they feel dissatisfied and
* Małgorzata Grządzielewska negatively view their work (Maslach et al., 1996). A more
[email protected]
general definition from the International Classification of
1
Centre for Family Research, Nicolaus Copernicus University, Diseases, 11th revision (ICD-11), combines burnout with
ul. Lwowska 1, 87‑100 Toruń, Poland

13
Vol.:(0123456789)

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

176 M. Grządzielewska

stress. The World Health Organization regards burnout as dimension of burnout and other variables (Tong, Wang &
a result of chronic stress in the workplace that has not been Peng, 2015).
successfully managed (https​://www.who.int/menta​l_healt​h/ Multivariate statistical analyses (MVAs) afford the
evide​nce/burn-out/en/). opportunity to analyse the relationship between many vari-
Most of the previous work depicts relationships between ables. Methods can be divided into determining a depend-
burnout and different factors such as: job satisfaction ent variable or not (Manly, 2005; Rosenblad, 2017). Several
(Stalker et al., 2007; Chen & Scannapieco, 2010); turno- techniques are presented below and broken down by either
ver and retention (Barak, Nissly & Levin, 2001; Kyonne, exploratory or predictive purposes.
2009; O’Donnell & Kirkner, 2009; Schwartz, 2008); and Explanatory modelling is concerned with testing causal
work environment (Perrone, 2007). Some articles focus on hypotheses, while the aim of predictive modelling is predict-
the prediction of burnout using statistical models, mainly ing new or future observations. Currently, statistical model-
logistic regression. The main aim of this paper is to present ling is used mostly for causal explanation, at the expense of
new methods of burnout prediction, based on machine learn- predictive modelling. Moreover, in many disciplines, there
ing techniques and compare them to a statistical, conven- is the assumption that predictive power (minimizing estima-
tional approach. To date, only a few studies show the use of tion error and bias) can be inferred from explanatory power
machine learning to predict burnout, with most of it focused (minimizing bias). It can lead to incorrect conclusions which
on mood and health disorders, e.g. depression, anxiety, or are confirmed by numerous studies (Shmueli, 2011).
suicide prevention. In this regard, this paper aims to sum- The most popular technique is structural equation model
marize using these methods to provide new knowledge to and its types. They represent, estimate and test the relation-
social work researchers. ship between latent and measured variables. These models
Subsequent sections of this paper are organized by start- are closely based on theory, making it possible to test theo-
ing with a review of the traditional approach to predict- retical constructs. To test theoretical models of burnout such
ing occupational burnout. Next, the methods of machine as JDCS model of job stress and JD-R model of burnout,
learning were presented and differentiated. Subsequently, a Kim & Stoner (2008) examined the effects of job auton-
machine learning method for burnout prediction was exhib- omy, role stress, and social support in predicting burnout
ited. A conclusion of the presented information and the asso- and turnover intention among social workers. Role stress
ciated implication is presented. had a positive effect on burnout but social support and job
autonomy had negative effect on turnover. A type of model
structural equation modelling is path analysis. For exam-
ple, path analysis was preformed to test the hypothesis of
Conventional, Statistical Methods conceptual model of burnout in a longitudinal study (Travis
et al., 2016). The main results showed that work–family
The use of data in the practice of social work has grown and role conflict were significant predictors of emotional
rapidly within the last decades. exhaustion and non-significant between role ambiguity and
Søbjerg et al. indicate that statistical approach should also burnout.
be used in making decisions about clients (Søbjerg et al., For prediction, a logistic regression model is used. It dif-
2020). Explanation, prediction, and description are the main fers from the linear regression model in that the dependent
purposes of statistical modelling (Hanna, 1969). According variable is binary. This model evaluates the likelihood of a
to Shmueli (2011), the same method can be used to solve certain event. For example, Stack (2004) showed that being
different problems; for example, linear regression model can a social worker increased the odds of suicide by 55.6%, com-
be descriptive if it is used for representing the association pared to the rest of the working age population. Sometimes
between the dependent and independent variables, explana- this model is limited to detection of variables affecting some
tory when used for causal inference, or prediction for the aim events—Strolin-Goltzman et al. (2008) used logistic regres-
of predicting new or future observations. sion to assess the effect of organizational factors on turnover
Descriptive analysis focus on summaries or representa- intentions in public child welfare systems. The result showed
tion about the sample but cannot explain why the phenom- that three of the five variables were statistically significant:
enon has occurred (Punch, 2005). A social work researcher salary and benefits, lack of other job options, and clarity of
may want to examine students’ addictions to their phones; performed task. For longitudinal studies, Cox regression is
descriptive research might aim to describe how many hours used. This model belongs to the class of survival models,
they use them (Decarlo, 2018). In articles on burnout for which determines the likelihood of occurrence of an event at
example, descriptive statistics is used to show sample char- a certain time by the the hazard ratios. In research on occu-
acteristics and burnout rates (Hamama, 2012). It can also be pational burnout, it is mainly used to study the relationships
used to present strength and direction of correlation between between the dependent variable and independent variables

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Using Machine Learning in Burnout Prediction: A Survey 177

at different times. For example, Mohren et al. (2003) utilized involves using some variables to predict unknown values
this technique to study the relationships between burnout of other data and supervised methods are mainly used for
as a risk factor for common infections. Results show the this purpose. Different types of target variables require dif-
subscale Exhaustion in MBI-GSS is the strongest predictor ferent methods: classification is used for categorical vari-
for infections studied. ables, regression for continuous variables (Mitchell, 1997).
High-accuracy predictions can lead to discovering new, However, there are some exceptions: for example, k-Means-
potential variables. Moreover, Shmueli & Koppius (2011) Mode can deal with both types of data. Paul and Hoque
present how predictive models can contribute to theoreti- (2010) have applied these unsupervised models to predict
cal development. The aim of statistical prediction, in the the likelihood of diseases. The probability of a disease in a
broadest sense, is limited to understanding the conditional cluster was defined as the number of patients with the dis-
distribution of dependent variable y given other variables x ease divided by the total number of patients in the cluster.
(Varian, 2014). Thus, a need arose for methods that enable Results show that k-Means-Mode was better than the clas-
the prediction of future observations. sical k-Means algorithm.
At this point, it is worth emphasizing what was previ-
ously mentioned—prediction of y as a function of features
Machine Learning Methods x, is at the heart of machine learning, while statisticians and
data mining experts also focus on discovering characteristics
Machine learning is a vast field of knowledge and there is and patterns in data (Varian, 2014). While machine learning
no single definition. In general, the term is understood in methods can be used to refine explanatory models, they have
its broader and narrower sense. The broad understanding one major drawback—interpretability. Statistical methods
of machine learning refers to the study of algorithms and are simpler to understand, because they usually allow small
systems that improve their knowledge and results along with number of variables and present measure of association,
gaining experience (Flach, 2012). A more precise defini- such as odds ratio (OR) from logistic regression. Machine
tion says that a program learns through experience E, with learning methods are often difficult to interpret; this is par-
respect to the task class T and the efficiency measure P, if its ticularly observable in neural networks and less in LASSO
performance on tasks in T, as measured by P, appears with regression (Goodfellow et al., 2016).
experience E (Mitchell, 1997). The development of predictive machine learning models
Machine learning is a part of data mining process, that is primarily driven by data while conventional prediction
also includes methods used to discover connections and models are by theory (Shmueli, 2011). This is an advantage
patterns in large data sets. It should be remembered that for machine learning as it allows you to create new theories
machine learning and data mining tools are derived from and use large amounts of data. The integration of a huge
the methods of artificial intelligence and multidimensional number of factors is possible, because machine learning
statistics (Tan et al., 2005). copes with the problem of overfitting. This problem affects
Most models focus on two tasks: descriptive and predic- the predictor that performs well in-sample but fails out-of-
tive. It’s rarely used in explanatory modeling because sta- sample. Machine learning provides regularization techniques
tistical models do it better. To analyze the importance of the for this. This set of tools also provides more methods to
job-related factors in burnout, machine learning of gener- research nonlinear relationships between the data: random
alized linear algorithms was performed. These algorithms forest, support vector machines, classification and regression
were fitted to estimate the set of parameters by maximizing trees (CART), penalized regression (LASSO, LARS, and
the log-likelihood. elastic nets), and so on (Varian, 2014).
Results show that individuals with high neuroticism
seemed to have stronger levels of burnout when they faced
stress, while those with high general self-efficacy had the Using Machine Learning in Burnout
lowest burnout. Prediction
The main difference is that predictive models involve the
target variable while descriptive models do not. The basic Although machine learning has been advancing for several
distinction between machine learning methods is supervised years, it has only recently been used for behavioral sciences
and unsupervised learning. Supervised learning requires (DelPozo-Banos et al., 2018). For instance, these algorithms
labeled training data, in contrast to unsupervised methods. are used in computational psychiatry to improve diagnosis
Most descriptive models are part of unsupervised learning of mood disorders: stress (Silva et al., 2020), depression
(for example association rule discovery, clustering) but it (Webb et al., 2020) and suicidality (Kessler et al., 2015).
is also possible to use supervised learning to build these It allows possibilities to obtain detailed knowledge about
models, e.g. subgroup discovery (Flach, 2012). Prediction diseases (Huys et al., 2016). These methods are especially

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

178 M. Grządzielewska

important because they can analyze data from different Batata et al. (2018) proposes a benchmark of a classifica-
sources. For example, Kaczor and colleagues used machine tion method. They collected data from caregivers using a
learning techniques to detect stressful situations using digital self-estimation indicator and SRB burnout metric. Results
sensors worn by emergency medicine physicians and a self- showed that classification techniques from supervised learn-
assessment questionnaire (Kaczor et al. (2020). ing effectively predicted the level of burnout. The analysis of
Sometimes machine learning methods are compared with significant features from Decision Tree showed that burnout
the traditional approach. Mary & Jabasheela (2018) predicts was affected mostly by exhaustion and financial situation.
depression, stress, and anxiety using different machine learn- Zhernova et al. (2020) used Maslach Burnout Inventory
ing methods and compared results with a logistic regression to predict early prerequisites of burnout. Applying machine
model. It turned out that logistic regression had the highest learning approaches allowed to correctly predict burnout in
accuracy. Kessler et al. (2016) identified the level of depres- 70% of cases.
sion by survey and found that machine learning algorithms
achieved better results than conventional techniques.
Machine learning was also used to identify the level Discussion, Theoretical and Practical
of burnout. An example is the work of Bauernhofer et al. Implications
(2018), whose research sample included 103 patients clini-
cally diagnosed for occupational burnout. The patients were This article invites discussion on the use of machine learning
tested with questionnaires such as Maslach Burnout Inven- methods to predict burnout and their practical implications.
tory—General Survey (MBI-GS), the Recovery-Stress- This is one of the first reviews on this topic. Research from
Questionnaire for Work and the Beck Depression Inventory. other fields of science, such as computational psychometrics
Cluster analyses were performed to explore the burnout sub- and bioinformatics, show the importance of the development
types and MBI-GS subscales was used as clustering varia- of these methods and how much remains to be done. This
bles. Three burnout subtypes were identified: the burned-out work also shows a gap in the study of occupational burnout
subtype, the exhausted/cynical subtype, and the exhausted which requires researchers of social work to fill.
subtype. Main results showed that the burned-out subtype Division of prediction methods by the approach and aims
was more depressed than the others, but no difference was enabled to simply select a method for research purpose.
identified between burned-out and exhausted/cynical sub- Machine learning algorithms can lead to creating new theo-
types with stress and sociodemographic characteristics. retical models of burnout. The factors influencing burnout
The analysis of the literature showed that the use of may change because burnout is related to advancing glo-
machine learning methods to predict occupational burnout balization and technological progress (Chabot, 2019). This
deals with a few topics, of which some examples are pro- means that the new methods based on artificial intelligence
vided below with explanation. presented in this article, can lead to discovering newer,
Lee et al. (2020) used k-means to group about one thou- potential variables and combating the prevalence of burnout.
sand nurses working in a medical center in Taiwan into two This summary leads to an identification of potential direc-
classes (burnout and non-burnout states). Next, the convo- tions for future research.
lutional neural network (CNN) deep learning method was The main practical application of these models is to use
applied to predictive model to estimate 38 parameters for them through evidence-based approaches to create policies
burn-out sample. and new interventions based on data. Most interventions
Kurbatov et al. (2020) applied k-means unsupervised have focused on the individual worker (Schaufeli & Enz-
clustering (k-means analysis) and supervised clustering mann, 1998). However, each individual may have complex
(k-means cluster group) to identify and predict burnout in combinations of significant factors; therefore, it is impor-
surgical trainees. Results revealed three clusters with high- tant to provide methods that deal in large amounts of data.
risk, intermediate-risk, and low-risk of burnout. The high- Drawing conclusions from such database will enable deci-
risk cluster had a higher proportion of women and single or sion-makers to take actions more tailored to the individual
unmarried people. This may be attributable to the lack of a and investigate all workers in a certain unit. This method
work-life balance. will enable a team approach and provide the appropriate
A type of machine learning model—multitask learning resources to institutions with the largest number of employ-
technique—was used in the work of Taylor et al. (2020). ees in need. This work can indicate the need for managers
Based on data collected from surveys, sensors, and smart- and policymakers in the human services sector to design
phones, the future mood, stress, and health of the respond- new legislative actions aimed at managing burnout based on
ents were predicted. The empirical results showed that using prediction. Before any preventive actions can be taken, there
these methods to account for individual differences resulted is the need to know where to direct them so that the right
in significant performance improvements. people are reached. Supporting the decision- making process

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Using Machine Learning in Burnout Prediction: A Survey 179

in burnout monitoring systems with machine-learning tech- 2668–2679). Gothenburg, Sweden: IEEE. https:​ //doi.org/10.1109/
niques allows evaluation of government programs on this WSC.2018.86323​85
Bauernhofer, K., Bassa, D., Canazei, M., Jiménez, P., Paechter, M., &
topic. These methods can be used not only to intervene, but Papousek, I. (2018). Subtypes in clinical burnout patients enrolled
also to prevent burnout—these tools can also help with the in an employee rehabilitation program: Differences in burnout
recruitment process. profiles, depression, and recovery/resources-stress balance. BMC
Burnout meta-analysis emphasizes the role of individual Psychiatry, 18(1), 1.
Chabot, P. (2019). Global burnout. New York: Bloomsbury Publish-
factors in predicting burnout (Alarcon et al., 2009) so it may ing Inc.
be helpful to discover potentially burned-out new employ- Chen, S., & Scannapieco, M. (2010). The influence of job satisfac-
ees. It will avoid a situation where an employee starts work tion on child welfare worker’s desireto stay: An examination of
and then quits after a short term. These methods can also the interaction effect of self-efficacy and supportive supervision.
Children and Youth Services Review, 32(4), 482–486.
keep existing workers in the workplace by improving the Cox, T., Griffith, A. J., & Rial-Gonzalez, E. (2000). Research on
work situation. work-related stress. Report to the European Agency for Safety
Reducing workload is also important for socio-economic and Health at Work. Luxembourg: Office for Official Publica-
reasons—it is costly for individuals, employers and soci- tions of the European Communities.
DeCarlo, M. (2018). Scientific inquiry in social work. Roanoke, VA:
eties (Bährer-Kohler, 2013). Cox et al. (2000) found that Open Social Work Education.
about 60% of absence from work is related to stress in the DelPozo-Banos, M., John, A., Petkov, N., Berridge, D. M., Southern,
workplace. Machine learning can reduce the number of tasks K., Loyd, K., et al. (2018). Using neural networks with routine
by optimizing certain processes, which can significantly health records to identify suicide risk: Feasibility Study. JMIR
Mental Health, 5(2), e10144.
reduce the level of burnout. By predicting and identifying Flach, P. (2012). Machine learning: The art and science of algo-
the level of this syndrome among employees, the quality of rithms that make sense of data. Cambridge, MA: Cambridge
work performed and the atmosphere in the workplace can University Press. https​://doi.org/10.1017/CBO97​80511​97300​0.
improve, which can help in coping with excessive stress in Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning.
Cambridge, MA: MIT Press.
the workplace. Hamama, L. (2012). Burnout in social workers treating children as
related to demographic characteristics, work environment, and
Compliance with Ethical Standards social support. Social Work Research, 36(2), 113–125. https​://
doi.org/10.1093/swr/svs00​3.
Conflict of interest The author declares no conflict of interest. Hanna, J. F. (1969). Explanation, prediction, description, and infor-
mation theory. Synthese, 20, 308–334. https​://doi.org/10.1007/
BF004​13732​.
Open Access This article is licensed under a Creative Commons Attri- Huys, Q., Maia, T., & Frank, M. (2016). Computational psychiatry
bution 4.0 International License, which permits use, sharing, adapta- as a bridge from neuroscience to clinical applications. Nature
tion, distribution and reproduction in any medium or format, as long Neuroscience, 19, 404–413. https​://doi.org/10.1038/nn.4238.
as you give appropriate credit to the original author(s) and the source, Kaczor, E. E., Carreiro, S., Stapp, J., Chapman, B., & Indic, P.
provide a link to the Creative Commons licence, and indicate if changes (2020). Objective measurement of physician stress in the Emer-
were made. The images or other third party material in this article are gency Department using a wearable sensor. In Proceedings of
included in the article’s Creative Commons licence, unless indicated the 53rd Hawaii International Conference on System Sciences,
otherwise in a credit line to the material. If material is not included in 2020 (pp. 3729–3738). Hawaii: University of Hawai’i at Mānoa.
the article’s Creative Commons licence and your intended use is not Keller, L., & Ho, J. (1988). Decision problem structuring: Generat-
permitted by statutory regulation or exceeds the permitted use, you will ing options. Systems, Man and Cybernetics, 18, 715–728.
need to obtain permission directly from the copyright holder. To view a Kessler, R. C., van Loo, H. M., Wardenaar, K. J., Bossarte, R. M.,
copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Brenner, L. A., Cai, T., et al. (2016). Testing a machine-learn-
ing algorithm to predict the persistence and severity of major
depressive disorder from baseline self-reports. Molecular Psy-
chiatry, 21(10), 1366.
References Kessler, R. C., Warner, C. H., Ivany, C., Petukhova, M. V., Rose, S.,
Bromet, E. J., et al. (2015). Predicting suicides after psychiatric
Alarcon, G., Eschleman, K. J., & Bowling, N. A. (2009). Relationships hospitalization in US Army soldiers: The Army Study to assess
between personality variables and burnout: A meta-analysis. Work risk and resilience in servicemembers (Army STARRS). JAMA
& Stress: An International Journal of Work, Health & Organisa- Psychiatry, 72(1), 49–57. https​://doi.org/10.1001/jamap​sychi​
tions, 23(3), 244–263. atry.2014.1754.
Bährer-Kohler, S. (2013). Burnout for experts: Prevention in the con- Kim, H., & Stoner, M. (2008). Burnout and turnover intention among
text of living and working. Boston, MA: Springer. https​://doi. social workers: Effects of role stress, job autonomy and social
org/10.1007/978-1-4614-4391-9. support. Administration in Social Work, 32, 5–25. https​://doi.
Barak, M. E. M., Nissly, J. A., & Levin, A. (2001). Antecedents to org/10.1080/03643​10080​19223​57.
retention and turnover among child welfare, social work, and other Kim, H., Chang, M., Rose, K., & Kim, S. (2011). Predictors of car-
human service employees: What can we learn from past research? egiver burden in caregivers of individuals with dementia. Jour-
A review and metanalysis. Social Service Review, 75(4), 625–661. nal of Advanced Nursing, 68(4), 846–855.
Batata, O., Augusto, V., & Xie, X. (2018). Mixed Machine Learn- Kurbatov, V., Shaughnessy, M., Baratta, V., Heller, D. R., Freedman-
ing and Agent-Based Simulation for Respite Care Evalu- Weiss, M., Resio, B. J., et al. (2020). Application of advanced
ation. In 2018 Winter Simulation Conference, WSC (pp. bioinformatics to understand and predict burnout among

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

180 M. Grządzielewska

surgical trainees. Journal of Surgical Education, 77(3), 499– Silva, E., Aguiar, J., Reis, L. P., Sá, J. O. E., Gonçalves, J., & Carvalho,
507. https​://doi.org/10.1016/j.jsurg​.2019.11.008. V. (2020). Stress among Portuguese Medical Students: The EuS-
Kyonne, J. (2009). The role of teamwork in public child welfare tress Solution. Journal of Medical Systems, 44(2), 1–6. https:​ //doi.
caseworkers’ intentions to leave. Dissertation Abstracts Inter- org/10.1007/s1091​6-019-1520-1.
national A: The Humanities and Social Sciences, 70(04), Søbjerg, L. M., Taylor, B., Przeperski, J., Horvat, S., Nouman, H., &
1429–1729. Harvey, D. (2020). Using risk factor statistics in decision making:
Lee, Y. L., Chou, W., Chien, T. W., Chou, P. H., Yeh, Y. T., & Lee, H. Prospects and challenges. European Journal of Social Work. https​
F. (2020). An app developed for detecting nurse burnouts using ://doi.org/10.1080/13691​457.2020.17727​28.
the convolutional neural networks in microsoft excel: Population- Stalker, C. A., Mandell, D., Frensch, K. A., Harvey, C., & Wright, M.
Based Questionnaire Study. JMIR Medical Informatics, 8(5), (2007). Children welfare workers who are exhausted yet satisfied
e16528. https​://doi.org/10.2196/16528​. with their jobs: How do they do it? Child & Family Social Work,
Lizano, E. L., & Mor Barak, M. E. (2012). Workplace demands and 12(2), 182–191.
resources as antecedents of job burnout among public child wel- Stack, S. (2004). Suicide among social workers: A research note.
fare workers: A longitudinal study. Children and Youth Services Archives of Suicide Research, 8(4), 379–388. https ​ : //doi.
Review, 34(9), 1769–1776. org/10.1080/13811​11049​04767​89.
Manly, B. F. J. (2005). Multivariate statistical analyses. A Primer (3rd Strolin-Goltzman, J., McCarthy, M., Smith, B., Caringi, J., Bronstein,
ed.). London: Chapman and Hall. L., & Lawson, H. (2008). Should I stay or should I go? A compari-
Mary S.A., & Jabasheela L. (2018). An evaluation of classification son study of intention to leave among public child welfare systems
techniques for depression, anxiety and stress assessment. In Inter- with high and low turnover rates. Child Welfare, 87(4), 125–143.
national Conference for phoenixes on emerging current trends Tan, P.-N., Steinbach, M., & Kumar, V. (2005). Introduction to data
in engineering and management (PECTEAM 2018). Amsterdam: mining. Boston: Pearson Addison Wesley.
Atlantis Press. Taylor, S., Jaques, N., Nosakhare, E., Sano, A., & Picard, R. (2020).
Maslach, C., Jackson, S. E., & Leiter, M. P. (1996). Maslach Burn- Personalized multitask learning for predicting tomorrow’s mood,
out Inventory. Manual. Palo Alto, CA: Consulting Psychologists stress, and health. IEEE Transactions on Affective Computing,
Press. 11(2), 200–213. https​://doi.org/10.1109/taffc​.2017.27848​32.
Mcfadden, P., Campbell, A., & Taylor, B. (2014). Resilience and burn- Tong, J., Wang, L., & Peng, K. (2015). From person-environment
out in child protection social work: Individual and organisational misfit to job burnout: Theoretical extensions. Journal of Man-
themes from a systematic literature review. British Journal of agerial Psychology, 30(2), 169–182. https​://doi.org/10.1108/
Social Work. https​://doi.org/10.1093/bjsw/bct21​0. JMP-12-2012-0404.
McGee, R. A. (1989). Burnout and professional decision making: An Travis, D. J., & Mor Barak, M. E. (2010). Fight or flight? Factors
analogue study. Journal of Counseling Psychology, 36(3), 345– impacting child welfare workers’ propensity to seek positive
351. https​://doi.org/10.1037/0022-0167.36.3.345. change or disengage at work. Journal of Social Service Research,
Mitchell, T. (1997). Machine learning. New York: McGraw Hill. 36(3), 188–205.
Mohren, D., Swaen, G., Kant, I., Amelsvoort, L., Borm, P., & Galama, Travis, D. J., Lizano, E. L., & Mor Barak, M. E. (2016). ‘I’m So
J. (2003). Common infections and the role of burnout in a Dutch Stressed!’’: A Longitudinal Model of Stress, Burnout and
working population. Journal of Psychosomatic Research, 55, Engagement among Social Workers in Child Welfare Settings.’
201–208. https​://doi.org/10.1016/S0022​-3999(02)00517​-2. British Journal of Social Work, 46(4), 1076–1095. https​://doi.
O’Donnell, J., & Kirkner, S. L. (2009). A longitudinal study of factors org/10.1093/bjsw/bct20​5.
influencing the retention of Title IV-E master’s of social work Valente, L. E., Truzzi, A., Souza, W. F., Alves, G. S., Alves, C. E.,
graduates in public child welfare. Journal of Public Child Welfare, Sudo, F. K., et al. (2011). Health self-perception by dementia fam-
3(1), 64–86. ily caregivers: Sociodemographic and clinical factors. Arquivos
Paul, R., & Hoque, A. S. M. L. (2010). Clustering medical data to de Neuro-Psiquiatria, 69(5), 739–744.
predict the likelihood of diseases. In IEEE. 2010 Fifth Interna- Varian, H. R. (2014). Big data: New tricks for econometrics. Journal
tional Conference on Digital Information Management (ICDIM of Economic Perspectives, 28(2), 3–28.
2010) (pp. 44–49). Thunder Bay, ON: Institute of Electrical and Vercambre, M. N., Brosselin, P., Gilbert, F., Nerrière, E., & Kovess-
Electronics Engineers (IEEE). ISBN: 9781424475728. Masféty, V. (2009). Individual and contextual covariates of
Perrone, R. V. (2007). The impact of burnout, job satisfaction and work burnout: A cross-sectional nationwide study of French teachers.
environment on child welfare family case managers. Dissertation BioMed Central Public Health, 10, 333–344.
Abstracts International Section A: Humanities and Social Sci- Webb, C. A., Cohen, Z. D., Beard, C., Forgeard, M., Peckham, A.
ences, 67(12-A), 4702. D., & Björgvinsson, T. (2020). Personalized prognostic predic-
Punch, K. F. (2005). Introduction to social research: Quantitative and tion of treatment outcome for depressed patients in a naturalistic
qualitative approaches. London: SAGE. psychiatric hospital setting: A comparison of machine learning
Rosenblad, A. (2017). Multivariate statistical methods: A Primer (4th approaches. Journal of Consulting and Clinical Psychology,
Edition). Journal of Statistical Software. https​://doi.org/10.18637​ 88(1), 25–38. https​://doi.org/10.1037/ccp00​00451​.
/jss.v078.b03. Zhernova, P., Bodyanskiy, Y., Yatsenko, B., & Zavgorodnii, I. (2020).
Schaufeli, W. B., & Enzmann, D. (1998). The burnout companion to Detection and prevention of professional burnout using machine
study and practice: A critical analysis. Washington, DC: Taylor learning methods. In 2020 IEEE 15th International Conference
& Francis. on Advanced Trends in Radioelectronics, Telecommunications
Schwartz, S. L. (2008). Engaging our workforce: How job demands and Computer Engineering, TCSE (pp. 218–221). Lviv-Slavske,
and resources contribute to social worker burnout, engagement Ukraine: IEEE. https:​ //doi.org/10.1109/TCSET4​ 9122.​ 2020.23542​
and intent to leave. Dissertation Abstracts International A: The 6
Humanities and Social Sciences, 68(07), 3151–3151.
Shmueli, G. (2011). To explain or to predict? Statistical Science, 25(3), Publisher’s Note Springer Nature remains neutral with regard to
289–310. https​://doi.org/10.1214/10-STS33​0. jurisdictional claims in published maps and institutional affiliations.
Shmueli, G., & Koppius, O. (2011). Predictive analytics in information
systems research. MIS Quarterly, 35(3), 553–572.

13

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

[email protected]