Introduction

Firefighters perform duties critical in preventing, controlling, suppressing, or extinguishing fires in various environments1. Unfortunately, firefighting is perceived as one of the most perilous professions due to its imminent nature with risks and hazards with flames, noise, toxic gases, smoke, and carbon monoxide, stemming from rigorous physical training, exercises, and emergency responses2.

These apparently threaten the health and safety of associated personnel3. For firefighters, auditory disturbances, such as noise, among many other types stand out as a significant risk. Noise is characterized by its intrusive and disruptive nature that poses a threat to human health4. In the context of firefighting, mitigating noise is often challenging because high decibel levels are used to serve as vital warning signals and an effective means of communication for imminent hazards5. The same study5 also noted that firefighters are consistently exposed to intense noise levels during their duties. For example, equipment at fire stations can emit sounds ranging from 81 dB(A) to 108 dB(A). Critical noise sources within fire stations and on fire trucks include sirens and emergency alarms, which can produce considerable noise exceeding 98 dB(A) or higher6, being detrimental to health upon continual exposure.

Exposure to noise is associated with a range of mental and psychological disturbances and physiological disorders in humans. While auditory impairment typically results from noise levels exceeding 85 decibels, exposure to lower intensities of sound can also induce adverse effects7, in relation to exposure duration. Noise levels greater than 40 to 54 decibels can produce cardiovascular illnesses, cognitive difficulties, and a decline in mental health and overall well-being8.

Furthermore, exposure to noise can lead to significant mental effects such as depression, stress, anxiety, and aggression. A study by Beutel et al., after studying the relationship between noise exposure, depression, and anxiety, observed that noise annoyance can predict the onset of depressive and anxiety symptoms9. Additionally, Asivandzadeh et al. found that noise exposure is positively correlated with job stress and psychological problems among workers10. Similarly, a study performed by Hasanbegovic et al. showed that noise exposure can induce stressful disorders and aggressive behavior in industrial workers11.

As another significant effect of noise exposure, consistent exposure to noise can also lead to cognitive failures. In the case of firefighters, these are often explained as errors or inaccuracies in the execution of a task that an individual is typically capable of accomplishing12. Fallah et al. observed a relationship between sound pressure level with cognitive failures and noise annoyance in the workplace13. These failures are critical as they can directly impact the performance of firefighters and lead to accidents14. Cognitive failures may weaken memory, attention, or action, as it has been found to be associated with distractibility, poor selective attention, and mental error15.

Despite the importance of noise and its relation to cognitive effects, past studies have not paid sufficient or adequate attention to the specific association between siren noise exposure and cognitive failures among firefighters. Additionally, the extent of the effect through various pathways remains unclear. As a valid analysis technique, Bayesian Networks (BNs), well-known probabilistic and graphical models in nature that were first introduced by Pearl16, offer a valuable method for analyzing this specific association. These probabilistic and graphical models allow for the representation of random variables and their conditional dependencies, and probabilistic rules for learning and inference17, enabling the evaluation of events and determining the probability that they were caused by multiple factors18. BNs have been widely utilized in various domains, including safety, health, and decision support systems. In comparison to conventional regression models, Bayesian networks provide several advantages. For example, regression models cannot include two correlated independent variables within the same model, which can lead to miss critical interactions between variables. In contrast, Bayesian networks effectively manage these interactions within their framework. Traditional regression models also have difficulty in accurately computing the effects between dependent variables as well as between independent variables. Additionally, the graphical model of relationships in Bayesian networks tends to be more intuitive. Most regression models operate under fixed assumptions on variable relationships, like linear correlations or independence, assumptions which may not always be valid19. On the other hand, Bayesian networks differ significantly in their approach to probabilistic reasoning20,21. They are designed to operate effectively under conditions of uncertainty, a capability that regression models lack. Moreover, while regression models depend on complete and quantitative data for parameter estimation, Bayesian networks can be developed with either incomplete or qualitative data22. Given these benefits, Bayesian networks are considered more effective and versatile than traditional regression models.

Given the potential impact of siren noise exposure on cognitive failures and, consequently, on the performance and safety of firefighters, this research aims to investigate the effect of siren noise exposure and its mental health disorders on cognitive failures among firefighters using a Bayesian network model for the first time. This approach and our findings will shed light on the complex relationships between siren noise exposure, mental health outcomes, and cognitive function in firefighting contexts.

Methods

This case-control study was performed on 92 firefighters in Kashan, Iran, in 2023. The project accounted for the Research Ethics Committee of Kashan University of Medical Sciences (Ethics code: IR.KAUMS.MEDNT.REC.1402.065) during the case-control study.

Sample size

To determine the sample size, the Cochran formula (Eq. 1) was utilized.

$$\:\varvec{n}=\frac{\frac{{\varvec{z}}^{2}\varvec{p}\varvec{q}}{{\varvec{d}}^{2}}}{1+\frac{1}{\varvec{N}}\left[\frac{{\varvec{z}}^{2}\varvec{p}\varvec{q}}{{\varvec{d}}^{2}}-\left.1\right]\right.}$$
(1)

In this equation, N is the population size, p is the estimated proportion of the population that has the attribute in question, q is 1 – p, and d is the margin of error. In this study, the values of N, p, q, and d were equal to 112, 1.96, 0.5, 0.5, and 0.05, respectively. the minimum sample size was computed by 87 individuals. 92 firefighters participated in this study into two groups including persons non-exposed and exposed to siren noise.

Participants

A total of 92 firefighters were chosen through a random selection process from among six fire stations located in Kashan, Iran. Participants were divided into two groups, including individuals exposed to siren noise and persons non-exposed to siren noise. Eligibility for participation in both groups was restricted to individuals aged 18 to 60, with at least one year of professional experience, no psychiatric conditions, and non-consumption of medication affecting cognitive performance. On the other hand, the criteria for exclusion encompassed those who were not prepared to engage with the research process and participants who provided partial responses to the survey.

During the selection process, a list of all firefighters (a total of 120 individuals) was provided. The medical records of all firefighters were reviewed along with the inclusion and exclusion criteria for the research to invite 103 individuals. Among them, 92 firefighters (89%) completed the questionnaire.

Data collection

Prior to data collection, the study provided the participants with general information about the study, including the objectives and procedures. To collect data related to study objectives, noise exposure for both groups was initially assessed based on the ISO 9612 standard. Workstations of the people were identified, and the sound equivalent level was measured at each position for a duration of 15 min by using a sound meter in A-weighting23. Subsequently, based on the duration of presence at each station and the measured sound equivalent level, the equivalent sound pressure level in A-weighting was calculated.

A sound meter in A-weighting was located at the firefighters’ ears for one minute to measure the sound level of the siren sound.

Subsequently, the individuals involved in the study were instructed to fill out the provided paper-based survey during their breaks. Although every participant possessed the ability to read and write, the researchers were available to address any questions and provide assistance in completing the questionnaires when necessary. The tools included demographic information, depression anxiety stress scales, the Buss-Perry aggression questionnaire, and the cognitive failure questionnaire.

Tools

Sound measurement To measure the sound level, CEL-440 sound level meters manufactured by Casella-CEL company in England were used.

Demographic information The demographic questionnaire was designed to collect essential information including age, experience, gender, and educational attainment level.

Depression Anxiety Stress Scales (DASS) The Depression Anxiety Stress Scales (DASS) is a concise evaluative instrument comprising 21 items that assess three distinct psychological aspects: depression, anxiety, and stress, with each aspect being addressed by seven specific questions. Respondents rate these questions on a four-point Likert scale, ranging from zero to three. To determine the severity within each category, the individual scores for the questions are aggregated. Higher totals indicate more severe symptoms of the respective disorder24. Median values were used for categorizing the levels of this variable. A total score below and above 11 points for each dimension was considered as low and high status, respectively. Sahebi et al. reported that Cronbach’s alpha values for the subscales measuring depression, anxiety, and stress were obtained by 0.77, 0.79, and 0.78, respectively25.

Buss-Perry aggression questionnaire

Buss and Perry constructed a survey to measure aggressive tendencies across different groups. This instrument contains 29 items that assess four facets of aggression: physical, verbal, anger, and hostility26. Respondents rate each item using a 5-point Likert scale ranging from one to five27. The aggregate of these scores yields an overall score. Elevated scores indicate greater aggression levels. Median values were used for categorizing the levels of this variable. A total score below and above 73 points was considered as low and high status, respectively. Karimi et al. assessed the reliability and validity of the Persian adaptation of this survey, revealing a Cronbach’s alpha coefficient of 0.89 for instrument28.

Cognitive failure questionnaire

The cognitive failure questionnaire (CFQ) was designed to evaluate cognitive failures in workplaces. The CFQ tool comprised 30 items to which the participants responded using a five-point Likert scale from zero to four. This questionnaire has been comprised of three subscales of memory, attention, and performance29. The sum of scores is used as the total value. A higher score indicated a higher cognitive failure likelihood. Median values were used for categorizing the levels of this variable. A total score below and above 60 were classified as low and high status, respectively. Hassanzadeh et al. obtained a content validity of 0.70 and Cronbach’s alpha coefficient of 0.96 for this questionnaire and confirmed its validity and reliability30.

Statistical analysis

Statistical analyses were conducted utilizing version 24 of the SPSS software. Initially, descriptive statistics were computed, followed by the utilization of the expectation-maximization (EM) method to impute missing data. In addition, GeNIe academic software version 2.3 was employed for Bayesian network analysis. In the constructed model within this software, in addition to noise exposure, the variables of depression, anxiety, stress, aggression, and cognitive failures were used in the model because the DASS questionnaire consists of three separate dimensions and there is no total score. So for its interpretation, the score of each of these three dimensions is separately reported, but the aggression and the cognitive failure questionnaires have a total score. In the theoretical model, we considered the siren noise exposure as an independent variable, the mental health disorders as mediator variables, and cognitive failure as a dependent variable regarding the assumptions of the study.

The EM algorithm, serving as a deterministic estimation technique, was utilized for Bayesian network parameter estimation. This method is particularly suitable for situations involving missing or underreported data, operating asymptotically to estimate unknown parameters31. Following the establishment of the theoretical structure of the Bayesian network, the model used the EM algorithm to generate a Conditional Probability Table (CPT)32.

Delta p sensitivity analysis was performed to examine the effects of the variables33. This involved setting the probability of one category in selected variables to 100% while observing variations in other variables. The categories included noise exposure (low and high), depression (low and high), stress (low and high), anxiety (low and high), aggression (low and high), and cognitive failures (low and high). Given that in the sensitivity analysis of Bayesian studies, all the states between the variables should be checked, the use of the Likert scale and a higher number of categories for each variable according to the law of probabilities creates many states that make it difficult to interpret the results. Also, the number of participants in the present study is low, and the use of a Likert scale and a higher number of groups causes inappropriate distribution of qualitative data.

Sensitivity analysis was conducted on all possible states, individually and in various combinations. In the end, the study implemented a 10-fold cross-validation procedure to assess the validity of the model. The dataset underwent a random segmentation into ten equal parts, with nine of these parts (9 subsamples) serving as the training set for the Bayesian network model. The tenth part (1 subsample) was reserved for validating the model’s performance34.

Results

The statistical summary of the variables of age and work experience in the year are presented. The mean ± standard deviation (SD) of the age and experience were 35.40 ± 3.83 years and 10.87 ± 2.66 years in the case group and 38.55 ± 5.73 years and 13.23 ± 5.87 years in the control group, respectively. The mean ± SD of the equivalent continuous sound pressure level was 71.47 ± 8.59 dB(A) in the exposed group and 46.70 ± 3.22 dB(A) in the non-exposed group. The peak sound pressure level due to exposure to the siren sound at the ear location of the participants was 82.71 ± 4.61 dB(A).

Table 1 presents the demographic characteristics of the firefighters in terms of five parameters, while Table 2 presents the six studied variables with their frequencies and relative percentages per group. Among the participants, 55.4% were identified with low cognitive failure and 44.6% with high cognitive failure.

Taking further on cognitive failures, Table 3 summarizes their CPT, which describes the distribution coefficient among the variables. These results were based on to distribution of the participants in proper groups for further analysis.

Table 1 Demographic characteristics of the participants.
Full size table
Table 2 Frequency and percent of the studied variables per each group.
Full size table
Table 3 The conditional probability table (CPT) for cognitive failures.
Full size table

Figure 1 illustrates the relationships among the studied variables based on the theoretical model. It accounts for the marginal probabilities of the variables according to the Bayesian network model. Subsequently, Table 4 represents the results of univariate sensitivity analysis, indicating at the high noise exposure with a 100-percent probability, the probability of the variables of high depression, high stress, high anxiety, and high aggression increased by 10, 14, 13, and 16%, respectively.

In the case of high states with a 100-percent probability for these four individual variables – noise exposure, depression, stress, and anxiety – the probability of high aggression increased by 16, 14, 14, and 14%, respectively. For the same, the probability of high cognitive failure increased by 8, 28, 14, 8, and 13, respectively. High cognitive failure with a 100-percent probability also increased the probability of high depression, high stress, high anxiety, and high aggression by 28, 15, 10, and 16, respectively.

Fig. 1
figure 1

The theoretical model for the marginal probabilities of the variables according to the Bayesian network model.

Full size image
Table 4 The results of univariate sensitivity analysis.
Full size table

Table 5 reports the results of the multivariate sensitivity analysis. The study identified a few interesting findings.

  • When considering two variables with a probability of 100%, the highest increase in the probability of cognitive failure (25%) was associated with high noise exposure and high stress.

  • Among the three variables with a probability of 100%, the highest increase in the probability of cognitive failure occurred with high noise exposure, high depression, and high stress (39%), as well as high noise exposure, high stress, and high anxiety (39%).

  • For the four variables with a probability of 100%, the combination of high noise exposure, high depression, high stress, and high anxiety resulted in the greatest increase in the probability of cognitive failure (51%).

  • All variables with a probability of 100% collectively contributed to a 57% increase in the probability of cognitive failure.

Table 5 The results of multivariate sensitivity analysis.
Full size table

Table 6 illustrates the influence values related to the relationships among the model factors. Regarding noise exposure, the two most influential values were linked with stress at 0.315 and anxiety at 0.292. On the other hand, cognitive failure was linked with depression at 0.564 and noise exposure at 0.312.

Table 6 The influence value related to the relationship between the factors in the model.
Full size table

Figure 2 displays a Receiver Operating Characteristic (ROC) curve that illustrates the validity of the Bayesian model applied, whereas Table 7 enumerates the confusion matrix associated with categorizing cognitive failure status. The area under the curve was equal to 0.842. The calculated sensitivity, specificity, and accuracy rates of the model were also 0.822, 0.766, and 0.739.

Fig. 2
figure 2

The ROC curve.

Full size image
Table 7 The confusion matrix associated with categorizing cognitive failure status.
Full size table

Discussion

Considering the pivotal role firefighters play in saving lives and contributing to the well-being of our society, maintaining their concentration and performance is of utmost importance. Cognitive failures are a significant factor that can impair firefighters’ performance. Siren noise is a probable source that can create mental health effects and lead to cognitive failures. Recognizing the importance of this issue, the present study investigated the impact of siren noise exposure and mental health disorders in relation to cognitive failures among firefighters.

The average noise exposure experienced by firefighters was measured at 71.47 dB(A), with the peak sound pressure level due to exposure to the siren sound at the ear location recorded at 82.71 dB(A). While these values fall below the 85 dB(A) threshold recommended by ACGIH, this type of noise possesses alarming properties that can induce annoyance in firefighters and potentially contribute to mental disorders. While these values fall below the threshold recommended by ACGIH35, this type of noise possesses alarming properties as it can induce annoyance in firefighters, leading to the initiation of mental disorders36. This corroborates findings by Balastegui et al. about the annoyance of emergency alerts37.

Further, the study’s results about high siren noise exposure indicate positive associations with depression, stress, anxiety, and aggression. Among these, the highest increase was observed with stress and aggression (Table 4). The model factor analysis indicated the most influence values of noise exposure were linked with stress (Table 6). A possible explanation of this is that stress and aggression are short-term effects due to exposure to noise while depression and anxiety are long-term effects.

The results of other studies also revealed that noise can lead to annoyance and stress and thereby, it can cause biological effects, depression, and anxiety38. One of the reasons that can explain these relationships is the activation of the autonomous nervous system by exposure to noise. This mechanism can provoke emotional reactions such as annoyance and stress and increase the excretion of cortisol hormone39. This situation can result in depression and anxiety over time40. The results of a study performed by Ekhlas et al. on non-industrial employees showed that exposure to noise with a level higher than 70 dB(A) had meaningful relationships with depression (OR = 5.22), anxiety (OR = 1.16), and stress (OR = 1.17)40. Gong et al. in a systematic review and meta-analysis observed that high noise annoyance could be associated with depression symptoms with an odds ratio of 1.23, anxiety disorders with an odds ratio of 1.55, and general mental health with an odds ratio of 2.1941. Yamin et al. concluded that there are highly positive correlations between noise pollution and depression, aggression, anxiety, and stress42. The results of these studies indicate that noise exposure can strongly create the stated mental disorders. However, the importance of each of these disorders may be different in various studies. It can be because of different noise sources with various sound levels and different studied populations. On the other hand, these mental disorders due to noise exposure can cause cognitive failure through three subscales of memory, attention, and performance. The results of a study performed by Notkin showed that depression, anxiety, and stress can significantly affect working memory in humans43. Ajilchi and Nejati also concluded that depression, anxiety, and stress can significantly disrupt executive functions and affect Selective and shifting attention and cognitive abilities. It seems that stress impair decision making and anxiety impair sustained attention44. Vuoksimaa et al. observed that there is a strong relationship between aggression and performance45.

The present study reveals that variables such as noise exposure, depression, stress, anxiety, and aggression are linked to an increased likelihood of high cognitive failures, with respective probabilities of 8, 28, 14, 8, and 13% (Table 4). Between these variables, depression and stress exhibit the most significant influence on cognitive failures (Table 4). Depression, in particular, has been associated with heightened human errors across various cognitive functions, identified by several past studies in relation to human functions, such as verbal fluency46, attention47, and working memory48. Beyond this, the study and its findings incorporate past findings about the relationship between depression and cognitive failures. Payne et al. examined the link between negative emotional states and cognitive lapses, finding a strong connection with instances of memory and attention deficits. Notably, the associations with fear, sadness, and guilt yielded correlation coefficients of 0.41, 0.28, and 0.43, respectively49. Further research indicated a significant relationship between cognitive dysfunction scores and the geriatric depression scale (GDS) in older adults50. Additionally, Sullivan and Payne’s work revealed that college students diagnosed with either seasonal or nonseasonal depression reported considerably more cognitive lapses than their non-depressed peers, with a correlation coefficient of 0.4751. Fisher and colleagues’ study also determined that depression had a more pronounced association with self-reported cognitive failures than either complicated grief or anxiety52.

Another significant finding of the study pertains to the impact of job stress on cognitive failure. Job stress arises when job requirements exceed the capabilities, resources, or needs of individuals. This negative psychological factor exerts a strong influence on employees53, potentially resulting in persistent psychological stress that hampers their ability to perform duties effectively and may lead to overwork54. Several studies have been conducted which show that there are strong associations between these variables. For example, Azizian and Fathi anticipated occurrences of cognitive failure by job stress and workload in the workplace and found a substantial positive correlation between cognitive failure with job stress (0.60) and workload (0.72)54. Similarly, Alyahya et al. explored the link between cognitive failures and the organizational and personal factors that influence them and revealed that stress had the highest factor loading (0.88) in the model among factors55. The findings of the current study align with the outcomes presented in these previous studies. The results of the present study are consistent with those reported by previous studies. Furthermore, an intriguing result of the study is the observation that the effect of siren noise on cognitive failures, mediated through its mental health disorders, surpasses the direct impact of noise exposure on these failures (Table 5). This highlights the importance of considering not just the physical aspects of noise exposure but also its psychological repercussions when assessing its impact on cognitive function.

However, the findings of the present study suggest that the combination of variables including noise exposure, depression, anxiety, and other mental health disorders had a greater impact on cognitive failures (Fig. 1; Table 5). This combination could increase cognitive failures by 57%. Fisher et al. examined the effect of anxiety, depression, and complicated grief on perceived cognitive failures and concluded that persons with three combined conditions had the greatest frequency of cognitive failures52. It is clear that combinations of variables can more strongly impress on a consequence.

As a limitation, this study was only performed among male firefighters. Also, given the limited samples of these studies, other variables affecting cognitive failures such as workload were not entered into the model in the present study. It is recommended that these relationships are investigated in the next studies. Moreover, we tried to select the nearest groups in terms of mental workload and working activities as exposed and non-exposed groups. However, these groups are not completely similar.

Conclusion

The results of this study highlight the significant impact of siren noise exposure on cognitive failures among firefighters, with varying degrees of effect. The experimental results and analyses indicate that noise exposure increased the likelihood of cognitive failures by 8%, while in combination with its mental health disorders, this probability rose to 57%. In addition, stress and aggression exhibited the highest increases in mental health disorders due to noise exposure, while depression and stress had the greatest effects on cognitive failures. These results underscore the importance of addressing both siren noise exposure and its associated mental health disorders in mitigating cognitive failures. It is recommended that measures be taken to reduce these effects, such as employing sirens with appropriate sound characteristics and providing intervention training to manage psychological consequences. Further studies should investigate the effectiveness of such interventions in minimizing cognitive failures among firefighters. As noted earlier, the scope of the study was limited to male firefighters. In accomplishing the study’s objective, it focused on the relationship between siren noise exposure and its mental health disorders in relation to cognitive function among firefighters, without considering other variables, such as workload. It is recommended that future studies address the intricacy of these additional relationships.