Psychosocial Risks Assessment in Cryopreservation Laboratories

Abstract

Background

Psychosocial risks are increasingly a type of risk analyzed in organizations beyond chemical, physical, and biological risks. To this type of risk, a greater attention has been given following the update of ISO 9001: 2015, more precisely the requirement 7.1.4 for the process operation environment. The update of this normative reference was intended to approximate OHSAS 18001: 2007 reference updated in 2018 with the publication of ISO 45001. Thus, the organizations are increasingly committed to achieving and demonstrating good occupational health and safety performance.

Methods

The aim of this study was to characterize the psychosocial risks in a cryopreservation laboratory and to develop a predictive model for psychosocial risk management. The methodology followed to collect the information was the inquiry by questionnaire that was applied to a sample comprising 200 employees.

Results

The results show that most of the respondents are aware of the psychosocial risks, identifying interpersonal relationships and emotional feelings as the main factors that lead to this type of risks. Furthermore, terms such as lack of resources, working hours, lab equipment, stress, and precariousness show strong correlation with psychosocial risks. The model presented in this study, based on artificial neural networks, exhibited good performance in the prediction of the psychosocial risks.

Conclusion

This work presents the development of an intelligent system that allows identifying the weaknesses of the organization and contributing to the enhancement of the psychosocial risks management.

1. Introduction

According to the Portuguese general Labor Law, the employers should identify the riskiness to which workers are exposed and perform the risk assessment. Thus, it is of utmost importance to understand this phenomenon in terms of occurrence, prevalence, and prevention in various professional activities [1]. Beyond the physical, chemical, and biological risks, due to the significant changes that have taken place in the working world, in recent years the psychosocial risks have emerged, with negative consequences for society, businesses, and workers [[1], [2], [3]]. Psychosocial risks derive from deficiencies in work design, organization, and management, as well as from a problematic social working context and may have psychological, physical, and social effects such as work-related stress, burnout, or depression [3].

The study of this theme began to develop in 1950, when the psychological aspects of the work became an object of research. Attention to this phenomenon became more pronounced in 2000, after an increase in serious accidents at work was noted. However, psychosocial risks are still identified today as a major challenge for occupational health and safety professionals [3,4].

2. State of art

2.1. Psychosocial risks

Psychosocial risks are a relatively recent concept and refer to conditions present in work situations, related to work organization, hierarchy, performance of the task, and the work environment, which may favor or impair work activity, as well as the quality of life, well-being, and health of workers [5]. These conditions, when favorable, foster the personal development of individuals. Conversely, when unfavorable, undermine the health and well-being, becoming a source of occupational stress with potential to cause psychological, physical, or social harm to individuals [4].

Psychosocial risks can be caused by a diversity of factors. Some may be intuitive, whereas others require detailed analysis to be identified as underlying causal factors. As a result, there are usually no quick fixes at hand, that is, a continuous and effective management process is required. To achieve this level, it is important to understand the most important underlying causal factors before selecting solutions [5]. The factors that may lead to psychosocial risks are related to content of the work, workload and work rate, working hours, control, environment and equipment, culture and organizational function, interpersonal relationships at work, organization role, development of career, and work–home interaction, as shown in Fig. 1A.

Fig. 1.

(A) Psychosocial risks factors. (B) Categories of psychosocial risks factors.

Content of the work refers to the lack of short work cycles, to the automatic or meaningless work, and to the under/over use of worker's skills (i.e., skills do not correspond to the tasks assigned). The lack of variety and complexity of tasks and the consequent monotony or repeatability can also be a source of suffering at work [4,6].

Workload and work rate is related to the inability to cope with the demands of the profession (e.g., when the worker feels that the demands of the job are excessive and cannot cope with them or the lack of sufficient requirements). This item is also associated to high levels of emotional pressure and mental burden, as well as the continued existence of difficult deadlines [6].

Working hours is related to working hours (e.g., shift work, night shifts, Sunday work, rigid and inflexible work hours, unpredictable hours, and long hours or those that do not allow socialization). These issues are seen as being inconsistent with the preservation of the well-being and influence temporal and emotional availability for personal and family relationships [6].

Control refers to the person's stress level that can be influenced by the person's level of control over workload and work rate, as well as other risk factors [6]. When a worker has control and influence over how his work is planned and carries out, it helps him cope with challenges ahead. Conversely, if the worker does not have the expected control, whether other people determine the pace or the way of work, the worker can feel stress. Lack of flexibility, as well as low participation in decision-making, contribute to stress and prevent a person from developing and using new skills [6].

Environment and equipment refers to inadequate availability, adequacy, or maintenance of equipment as well as precariousness and job insecurity. This item also stresses the poor environmental conditions, such as lack of space, poor lighting, excessive noise, or high temperatures, which may hinder workers' ability to concentrate [6].

Culture and organizational function relates to the levels of support and encouragement for problem-solving and personal development [6]. Positive support and feedback (from colleagues, from leaders, social support, or direct support for the profession) can help people overcome difficulties and lead to job satisfaction. This item is also related to communication, definition of organizational objectives, structural variables, hierarchical structure, leadership style, recognition at work, employees' freedom of expression, and changes within organization [4].

Interpersonal relationships at work are related to differences of opinion in a work environment. Work relationships can cause stress when people suffer discrimination, have poor relationships with superiors, colleagues, face interpersonal conflict, or lack social support [6]. In addition, inadequate, incomprehensible, or unbearable supervision should be included in this item. Another factor that cannot be overlooked and causes stress is bullying/mobbing [5]. Bullying may involve violence (physical, verbal, or psychological), intimidation, sexual harassment, and subtler acts such as physical or social isolation, excessive supervision, unfounded criticism, the impoverishment of tasks, evasion of information, or persecution at work [4].

Organization role refers to stress that arises from the lack of clarity about the roles and responsibilities that people have or when the roles and responsibilities give rise to conflict with colleagues, superiors, or customers. This factor also covers the stress that arises from the feeling that his role is incompatible with his skills and abilities [6].

Career development refers to the stress due to career stagnation, under-promotion, over-promotion, low compensation, or low commission. Career development also encompasses issues linked to job insecurity, career uncertainty, and low social value to work [4].

Work–home interaction is related with the conflict between work and family demands which can lead to conflicts of time, commitment, and support. In addition, it encompasses the stress that arises because of low home support and from dual-career problems [6].

Some of aforementioned factors can be grouped into generic categories, namely Work Conditions, Work Organization, and Leadership. The former one includes the factors Environment and Equipment and Content of Work, whereas the second comprises Workload and Work Rate and Working Hours. Finally, the third category groups the factors Culture and Organizational Function, Control and Organizational Role (Fig. 1B).

Recent studies show that psychosocial risk analysis in organizations is increasingly common in the industrial sector [[7], [8], [9]] and less frequent in the laboratory sector. Fergunson et al. [7] studied the psychosocial and biomechanical factors in furniture distribution workers, involving 454 participants at nine furniture distribution facilities during 6 months. The authors developed a multivariate logistic regression model that includes baseline functional performance probability, facility, perceived workload, intermediated reach distance number of exertions above threshold limit values, job tenure manual material handling, and age combined. The model sensitivity and specificity was of 68.5% and 71.9%, respectively.

Joensuu et al. [8] examined associations between job control, social support, and mental ill health in a multinational forest industry corporation. The 13,868 employees of a Finnish forest company with no previous hospital admissions for mental disorders responded to questionnaires on decision authority, skill discretion, coworker, and supervisor support [8]. The results show that high skill discretion was related with reduced risks of hospital admission for mental disorders, whereas high decision authority was connected with an elevated risk. Furthermore, the authors concluded that high decision authority was a risk factor for alcohol-related and depressive disorders, whereas good coworker support was linked with a reduced risk of nondepressive nonalcohol-related mental disorders. Supervisor support, in turns, was not associated with any mental disorders [8].

Metzler et al. [9] compared different methods for evaluating psychosocial hazards in the scope of risk assessment in a sample of 549 blast furnace workers of a German steel manufacturing. The authors highlighted that the risk management was strongly influenced by the choice of risk evaluation method because the measures to minimize the risks are directed only to the ones identified and in accordance with their level of priority.

In analysis laboratories, regardless of the sector of activity, the excessive pace of work, monotony, routine, as well as problems of interpersonal relationships, are examples of psychological risks for the workers. Bronkhorst [10] realized a hierarchical linear modeling of physical safety behavior to examine the relationship between job demands, job resources, safety climate, and safety behavior among employees working in health care. The author used a sample of 6230 health-care employees of 52 different Dutch organizations. Regardless of the focus (i.e., physical or psychological safety), this study shows that the consolidation of the safety climate increases employees' safety behavior. In addition, the authors point out that the organization's safety climate is an ideal target of intervention to avert and enhance negative physical and psychological health and safety outcomes, mainly in times of doubt and change.

Aiming to estimate the association between psychosocial risk factors in the workplace and musculoskeletal disorders in nurses and aides Bernal et al. [11] examined 17 papers. Despite the low heterogeneity of cohorts, the authors identified associations between high psychosocial demands and low job control with prevalent and incident low back pain, prevalent shoulder pain, prevalent knee pain, and prevalent pain at any anatomical site.

All these studies show the main role of the psychosocial risk in organizations to minimize possible damage to the health of employees. In addition, the foregoing demonstrates the relevance of the development of models aiming to predict the level of psychosocial risks based on employees' experiences. In the present work, a predictive model based on artificial neural networks (ANNs) will be presented.

2.2. Artificial neural networks

ANNs are computational tools that aim to simulate the human brain and nervous system. The multilayer perceptron is one of the utmost common ANN architectures, in which neurons are assembled in layers and only forward connections exist [12]. ANNs are increasingly applied in data mining because of their good performance in prediction [13]. In last decades, several studies have been published showing the usefulness of ANNs to apprehend complex relationships between variables, in various areas of application (e.g., environment [14,15], health [16,17], and law [[18], [19], [20]] just to name a few.

3. Methods

3.1. Place of study

This study took place in a cryopreservation laboratory located in the north of Portugal. In Portugal, the quality and safety requirements for human tissues and cells are set by the Portuguese Ministry of Health [21]. However, the criteria implemented by the laboratory under study are even stricter. The laboratory participates in external quality assays, where national and international entities test and certify the reliability of the stored sample quality control tests. Furthermore, the laboratory has the accreditation of the American Association of Blood Banks, which sets the most stringent quality criteria in the industry worldwide [22]. The cryopreservation laboratories were chosen because the work performed in them requires immediate results, confidentiality, and secrecy and deals with large volumes of information. The pressure associated to these requirements can create conditions for the appearance of psychosocial risks in the workers, and it is important to study these risks.

3.2. Participants

This study included 200 participants aged between 17 and 80 years, with an average of 41 ± 23 years old. The gender distribution was 44.5% and 55.5% for male and female participants, respectively. The participants belong to different departmental areas in the laboratory (i.e., quality management, human resources, finance, administrative, commercial, and technical).

3.3. Data collection

Aiming to perform the purposes defined before, a versatile tool to data collection was used. After taking into consideration, the advantages and limitations intrinsic to possible techniques, the inquiry by the questionnaire was chosen because it has a well-defined structure and enables the conversion of the qualitative information into a quantitative [[23], [24], [25]].

A questionnaire aiming to evaluate the perception of psychosocial risks in the workplace was designed specifically for this study and applied to a cohort of 200 employees. The questionnaire was organized into three sections, where the former one includes the general questions related with workers' age, gender, academic qualifications, and departmental areas. The second one comprises statements related with the work conditions, organization, leadership, career development, interpersonal relationships, and emotional feelings. Finally, the third section comprises issues related with the workers' opinions about the psychosocial risks. In the first part of the questionnaire the answers are descriptive, whereas in the second one the Likert scale with four levels (strongly disagree, disagree, agree, and strongly agree) was used. In the third section, the respondents choose five terms (the ones that they consider more relevant) from a list of 12 terms, ranking them in accordance with their relevance, using a numeric scale that varies from 1 to 5 (Annex 1).

The validation of the questionnaire follows practices of Bell [26]. Thus, the questionnaire was evaluated by a group of experts (i.e., a group of auditors) that suggested some corrections. After expert analysis, the questionnaire was modified and applied to a restrict group of employees, not included in the sample, to assess it validity and to identify difficulties in the interpretation of the questionnaire. The updated version was applied individually to the entire sample, in person, by the researcher. The return rate was 90.9% (200 inquiries received in 220 delivered).

3.4. Qualitative data processing

Aiming the conversion of qualitative information into quantitative information followed the method proposed by Fernandes et al. [27]. As per this method, a set of n issues regarding a particular subject is itemized into a unitary area circle split into n slices, where the marks in the axis correspond to each one of the possible answers, as described in the section 4.3.

3.5. Artificial neural networks

The software used to implement ANNs was the Waikato Environment for Knowledge Analysis, keeping the default software parameters [28,29]. Aiming to guarantee statistical significance of the results, 30 experiments were applied in all tests. In each simulation, the database was randomly split into two mutually exclusive partitions, i.e., the training set, with 2/3 of the data, used to build up the model, and the test set, with the remaining examples to evaluate its performance [30].

3.6. Ethical aspects of the study

The respondents took notice of the goals of the study participated voluntarily, without any pressure or coercion, and were informed that their grades would not be affected. The participants gave an informed consent to participate in the study. The study was conducted in compliance with the relevant laws and institutional guidelines and was approved by the relevant authorities.

4. Results and discussion

4.1. Sample characterization

Respondents were organized into age groups (i.e., 17–20, 21–30, 31–50, 51–70, and higher than 70 years old). 79.5% of participants are aged between 21 and 50 years. In all age groups, the percentage of female participants is higher, except in the group 31–50 where 22.5% of respondents are male and 20.5% are female (Fig. 2A). Regarding academic qualification, 14.5% of the cohort stated to have basic education, 51.5% declared to finish secondary education, 30.0% affirmed to have a degree, and 4.0% declared to have post graduate education (Fig. 2B). Fig. 2B also shows that the distribution of respondents by gender is very similar for the different types of academic qualifications. Concerning departmental areas to which respondents belong, 7.5% are allocated to the quality management department, 4.0% to the human resources department, 11.5% to the finance department, 10.5% to the administrative department, 21.0% to the commercial department, and 45.5% to the technical department (Fig. 2C). Fig. 2C also shows that in all the departments the percentage of female respondents is higher, except in the administrative department where 5.5% of respondents are male and 5.0%, female.

Fig. 2.

(A) Sample characterization in terms of age groups. (B) Sample characterization in terms of academic qualifications. (C) Sample characterization in terms of departmental areas.

To characterize the laboratory, the graph shown in Fig. 3 presents the distribution of the academic qualifications of the respondents by department. A perusal of Fig. 3 reveals that only in quality management and human resources departments the percentage of respondents that claim to have higher education are greater than the percentage of respondents that declare to have basic/secondary education. The remaining departments show an inverse trend, being the administrative, commercial, and technical departments those that present differences higher than 7.5% between respondents with basic/secondary and higher education.

Fig. 3.

Distribution of academic qualifications of the respondents by department.

4.2. Answer frequency analysis

Fig. 4 presents the results obtained in the second part of the questionnaire, where respondents expressed their opinion on the sets of statements regarding each of the psychosocial risk factors. The graphs show the frequency of answering to each factor statement (Table 1).

Fig. 4.

Respondents' agreement/disagreement with the statements regarding each factor.

Table 1.

Correspondence between the statements included in the questionnaire and the psychosocial risk factors

Factor	Working conditions	Work organization	Leadership	Career development	Interpersonal relationships	Emotional feelings
Statements	S1 – S4	S5 – S7	S8 – S10	S11 – S13	S14 – S16	S17 – S20

Regarding Working Conditions, the analysis of results shows that a percentage of respondents ranging between 66.5% and 70% consider they have good working conditions (S1), resources and equipment adequate to perform the work (S2). Furthermore, they consider that the work is not monotonous and routine (S3) and need learning and ongoing updates (S4). However, a percentage ranging from 15% to 19% has an opposite opinion, which may point out a need for improvement. With regard to Work Organization, the overwhelming majority of respondents consider that decision making and goal setting take into account the workers' opinions (S6). Taking a glance to the answers related to goal setting (S5) and the service distribution (S7) shows that the majority of the participants declare they are clear and evenly, respectively. However, a percentage of about 10% disagree, which may suggest that leaders should pay attention to these points. The results related to Leadership show that most respondents claim that their work is recognized (S10) and has a favorable opinion about leadership (S9). Nevertheless, 19% state that the guidelines and priorities are unclear (S8), being an issue that should be improved. Concerning Career Development, the overwhelming majority of participants have a very positive opinion. Only a percentage less than 10% is dissatisfied with the expectations of career development. The graphs related with Interpersonal Relationships (S14 to S16) show that more than 80% of respondents consider that there is a good relationship between colleagues. However, 14% and 18% of respondents have a negative opinion regarding communication/information sharing (S15) and mutual help/support (S16), respectively. These relatively high values of unfavorable opinions indicate that this is a point where improvements are needed. Finally, with regard to Emotional Feelings (S17 to S20), most participants have a favorable opinion, which ranges between 61.5% (S19, S20) and 83.5% (S18). In fact, the statements related with professional life/personal life overlapping (S19) and time available/tasks to perform (S20) collected the highest number of negative responses.

The statements included in the questionnaire were elaborated so that a higher percentage of positive answers correspond to a lower psychosocial risk. Thus, the overall analysis of the results shown in Fig. 4 suggests that the factors Interpersonal Relationships and Emotional Feelings are the ones that more contribute to psychosocial risk in the organization under study. Conversely, Leadership and Career Development seem to have a minor contribution.

These results are in agreement with those obtained by Ando et al. [31] and MacDonald et al. [32]. The authors refer that a strong association between psychosocial risk factors in the workplace and musculoskeletal disorders exists, identifying the working conditions (e.g., monotony, routine and repetitive movements) as the main cause.

Fig. 5 presents the results obtained in the third part of the questionnaire, where respondents select and classify, based on their opinion, the five more relevant terms related with psychosocial risks. The frequency of term selection is presented in the bar chart and may be read on the left side scales. The frequency of term priority (scale on the right side) was computed considering only the respondents that chose that term. The analysis of the frequencies of term selection allows identifying three groups. The first one includes the terms Lack of Resources, Stress, and Concentration that were selected by more than 40% of respondents. The second group comprises the terms Working Hours, Control, Lab Equipment, and Precariousness that were chosen by 30% to 36% of respondents. Finally, the third group is formed by the terms Trials, Violence, Insecurity, Rhythm, and Routine, chosen by less than 20% of respondents.

Fig. 5.

Frequency of term selection versus term priority (considering only the respondents that choose the term).

Regarding the priority given to the selected terms, only a small percentage of participants who chose terms of group 1, classified them as the first priority (ranging from 0% to 8%) or as the second priority (varying between 5.4% and 34.1%). Lack of Resources, Stress, and Concentration were mainly chosen as the third, fourth, and fifth priority, respectively (Fig. 5). Despite the fact that they were selected by more than 40% of respondents, the terms of group 1 are not considered as the ones that best describe psychosocial risks. As regards to the terms of the second group, it should be emphasized that Precariousness, Control, and Lab Equipment were classified as the first priority by, respectively, 73.3%, 37.9%, and 35.5% of the participants who selected them. Furthermore, it should be highlight that Working Hours was classified as the second priority by 50% of the respondents who selected it. Finally, in the third group there was a similar trend. Insecurity and Trials were classified as the first priority, respectively, by 55.6% and 51.3% of the participants who selected them, whereas Routine was classified as the second priority by 95.8%. These results suggest that the terms of second and third groups, although chosen by a smaller number of respondents, were considered by them as the best to describe the psychosocial risks.

The graph presented in Fig. 6 shows the relative frequency of binary associations between terms selected by more than 30% of participants. Their analysis shows that any possible combination of terms of group 1 (Lack of Resources, Stress, and Concentration) was chosen by at least 33% of participants. Regarding associations between terms of group 1 and group 2, only the combination between Stress and Working Hours was selected by more than 33% of participants. The associations Stress and Precariousness, Concentration and Precariousness, Concentration and Working Hours and all possible combinations between Lack of Resources and terms of group 2 (Working Hours, Control, Lab Equipment and Precariousness) were chosen by a percentage of participants ranging between 20% and 33.3%. The remaining binary associations were selected by less than 20% of participants.

Fig. 6.

Binary associations between terms selected by more than 30% of respondents. (Bold stands for terms of first group, and no bold denotes terms of second group).

The frequency of association between sets of three terms was also studied (Fig. 7). It was found that the terms of the group 1 were simultaneously selected by a percentage of participants ranging between 20% and 33%, as well as all ternary associations between Precariousness and terms of first group. Sets of three terms containing Working Hours and terms of first group were selected by at least 15% of respondents. To finish the analysis of Fig. 7, it also noted that a percentage between 10% and 15% of respondents set associations between Rhythm, Lack of Resources and Stress; Routine, Lack of Resources and Stress; and Trials, Lack of Resources and Concentration.

Fig. 7.

Ternary associations between terms. (Bold stands for terms of first group, no bold denotes terms of second group, and italics indicate terms of third group).

The graph depicted in Fig. 8 shows the frequency of quaternary associations between terms selected by the participants. Its analysis shows that Lack of Resources, Concentration, and Stress were chosen simultaneously with Precariousness by 23% of participants and with Working Hours by 19%. Quaternary associations between terms of group 1 and the remaining terms of group 2, i.e., Control and Lab Equipment, exhibit frequencies less than 5%. Regarding terms of third group, only Violence and Insecurity were selected together with the terms of group 1.

Fig. 8.

Quaternary associations between terms. (Bold stands for terms of first group, no bold denotes terms of second group, and italics indicate terms of third group).

In addition, this analysis enables to identify, among the less selected factors, the ones that present a strong association with the most chosen by the respondents. A perusal to Fig. 5, Fig. 8 reveals that Precariousness (selected by only 30% of respondents) was associated in 23% of cases with the terms of group 1, being ticked as first priority by 73.3% of those who chose it. A similar result was obtained for Working Hours, selected by 36% of respondents, but linked in 19% of cases to the terms of group 1, being ticked as second priority by 50% of those who chose it.

4.3. Psychosocial risk assessment

Aiming to gather information about psychosocial risk factors, the second section of the questionnaire comprises statements related to work conditions, organization, leadership, career development, interpersonal relationships, and emotional feelings. Fig. 9 shows the answers of respondent #1 to the second part of the questionnaire.

Fig. 9.

The answers of respondent #1 to the second part of the questionnaire.

To quantify the qualitative information presented in Fig. 9, the method proposed by Fernandes et al. [27] was followed. For each dimension (i.e., work conditions, organization, leadership, career development, interpersonal relationships, and emotional feelings) the correspondent answers were itemized into a unitary area circle. The marks in the axis correspond to the possible answer, i.e., strongly agree, agree, disagree, and strongly disagree. Taking as an example the dimension work conditions, the answer of respondent #1 to statement 1 (S1) was strongly agree, and the correspondent area is computed as $\frac{1}{4}\times \pi \times {\left(\frac{1}{\sqrt{\pi }}\right)}^2=0.25$; the answer to the statement 2 (S2) was the alternative agree and the area is $\frac{1}{4}\times \pi \times {\left(\frac{3}{4}\times \frac{1}{\sqrt{\pi }}\right)}^2=0.14$. Finally, for the statements 3 and 4 (S3 and S4), the answers were disagree and the areas are $\frac{1}{4}\times \pi \times {\left(\frac{2}{4}\times \frac{1}{\sqrt{\pi }}\right)}^2=0.06$. The total area (i.e., 0.51) is the sum of the partial ones, being the quantitative value regarding the dimension work conditions for respondent #1 (Fig. 10). For the remaining dimensions, the procedure is similar, and the results are shown in Table 2.

Fig. 10.

A view of the qualitative data processing.

Table 2.

A fragment of the knowledge base for psychosocial risk assessment

Respondent	Working conditions	Work organization	Leadership	Career development	Interpersonal relationships	Emotional feelings
1	0.51	0.57	0.46	0.85	0.35	0.19
2	0.67	0.71	0.35	0.57	0.46	0.40
…	…	…	…	…	…	…
200	0.70	0.60	0.85	0.71	0.57	0.48

The performance of the ANN model can be assessed through the confusion matrix. Table 3 presents the confusion matrix for the ANN model shown in Fig. 11 (The values indicate the average of the 30 experiments). The values presented in Table 3 allow computing the model accuracy for training set (93.4%, i.e., 127 well classified in 136) and for test set (90.6%, i.e., 58 well classified in 64).

Table 3.

Confusion matrix regarding the ANN model for psychosocial risk assessment

Target	Predictive
	Training set			Test set
	Low	Medium	High	Low	Medium	High
Low	15	2	1	4	1	0
Medium	1	98	4	0	48	5
High	0	1	14	0	0	6

Fig. 11.

ANN model for psychosocial risk assessment.

To compute the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the model, the confusion matrixes regarding each possible output were conceived (Table 4). Sensitivity evaluates the proportion of positives cases (i.e., Low, Medium, or High) that are correctly identified as such, whereas specificity translates the proportion of negative ones that are correctly identified (i.e., No-Low, No-Medium, or No-High). PPV stands for the proportion of Low, Medium, or High cases well classified, whereas NPV denotes the proportion of No-Low, No-Medium, or No-High cases well labeled [33,34]. Table 5 presents the values obtained for those metrics. Sensitivity and specificity exhibit high values, from 0.80 to 0.99, indicating that the model exhibits a good performance in the evaluation of psychosocial risks. Regarding PPV and NPV metrics, the values computed range between 0.94 and 0.99, except for PPV_output-High and NPV_{output-Medium} (0.74, 0.54 and 0.86, 0.67, respectively, for training and test). Those results show that the confidence that can be placed when the model classifies a case as High or No-Medium is lower. Despite those weaknesses, the overall performance of the model is not affected. In fact, the model should avoid classifying the High cases as Low or Medium, to identify all the problematic cases, i.e., workers at high psychosocial risk.

Table 4.

Confusion matrix regarding each output classes of the ANN model for psychosocial risk assessment

Target	Predictive				Target	Predictive				Target	Predictive
	Training set		Test set			Training set		Test set			Training set		Test set
	Low	No-low	Low	No-low		Medium	No-medium	Medium	No-medium		High	No-high	High	No-high
Low	15	3	4	1	Medium	98	5	48	5	High	14	1	6	0
No-Low	1	117	0	59	No-Medium	3	30	1	10	No-High	5	116	5	53

Table 5.

Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for each output classes of the ANN model, split by training and test

Output	Training set				Test set
	Sensitivity	Specificity	PPV	NPV	Sensitivity	Specificity	PPV	NPV
Low	0.83	0.99	0.94	0.97	0.80	1	1	0.98
Medium	0.95	0.90	0.97	0.86	0.91	0.91	0.98	0.67
High	0.93	0.96	0.74	0.99	1	0.91	0.54	1

5. Study limitations

The results obtained in this study were very interesting. However, it is important to mention some limitations that prevented a more detailed assessment of the psychosocial risks to which workers in this type of laboratory are exposed. The main limitation is related to the sample size. The reduced number of workers in some departmental areas did not allow a more detailed analysis (i.e., department by department) of the psychosocial risks associated with the performance of different tasks with workloads differentiated. With a larger sample, it would also be possible to study the influence of other variables such as age, gender, or academic qualifications.

The questionnaire used to collect the data was conceived to be general, i.e., to be applied to all employees, regardless of the sector of the laboratory to which they belong. However, with a larger sample, differentiated data collection tools could be designed specifically for each departmental area.

6. Conclusions

Nowadays, psychosocial risk assessment has been left to the discretion of each organization. However, the standards used (i.e., OHSAS 18001 and ISO 9001) are not clear as to how these risks should be measured. The certifications based on these standards do not guarantee that psychosocial risks are, in fact, controlled. This study showed that Interpersonal Relationships and Emotional Feelings are the factors that more contribute to psychosocial risks, particularly the issues related with the overlapping of working and personal lives, the lack of time to accomplish some tasks, the sharing of information, and peer support. This study also revealed that the concept of psychosocial risks is generally present among most respondents. Despite respondents having ticked the terms Lack of Resources, Stress, and Concentration as the ones that better describe the psychosocial risks, the terms Precariousness, Control, and Lab Equipment were the ones most often marked as first priority. Moreover, the terms Lack of Resources, Working Hours, and Lab Equipment were chosen simultaneously by a great number of respondents, as well as the terms Stress, Working Hours, and Precariousness. In addition, this work presents an intelligent decision support system that stands for a new approach to this problem using the ANN paradigm to assess psychosocial risks. The presented approach presents a worthy performance exhibiting sensitivities and specificities higher than 80%. This approach focuses on the processing of information, collected through inquiry by questionnaire, and aims to prevent recurrent events and to enhance the psychosocial risks management. Beyond the possibility of identifying the weakness of the organization, this system allows to concept and to design future improvement actions to promote the employees' quality of life. The results of this study cannot be generalized to all organizations because the employees are exposed to different risks depending on the sector and the type of activity in which they operate. However, these kinds of models can be used in any organization. For this, it is necessary to carry out an assessment of the risks to which employees are exposed and adjust the data collection tools (i.e., questionnaires).

Conflicts of interest

All authors have no conflicts of interest to declare.

Appendix A. Supplementary data

The following is/are the supplementary data to this article: