Implementation of Behavior-Based Safety in the Workplace: A Review of Conceptual and Empirical Literature

Abstract

Behavior-Based Safety (BBS) methodology is more and more implemented by companies in several industrial fields for preserving workers’ safety through a structured behavioral intervention with measurable effects as part of a socially sustainable approach for health and wellness care at the workplace. Although the traditional BBS components have been widely described in literature, its evolution over the years in terms of application strategies and complexity levels still requires some insights. Also, it is often difficult to identify clear boundaries between “standard” BBS and similar or complementary interventions. To clarify some of these aspects, in this work approximately 230 scientific papers about BBS and related topics, published from the 1970s until 2023, were retrieved from the Scopus database, classified in a structured way, and analyzed from different perspectives. Results highlight the main peculiarities, limits, and strengths of BBS in its evolution, from a conceptual as well as a practical point of view, with a particular focus on the trends observed in the last two decades. Outcomes suggest that future prospects of BBS will integrate the successful traditional principles of positive feedback and observation with new elements, including technological aspects, full involvement of the company’s levels, and attention to workers’ peculiarities, thus creating variants of the approach suitable for different contexts.

1. Introduction

Worker safety is an essential component of business for organizations, which aim to continuously improve their own safety records by reducing injuries. It has been acknowledged for many years that a significant reduction in injuries can be obtained through behavioral change programs, generally known as behavior-based safety (BBS) initiatives or behavioral safety programs (BSPs). These programs work by applying psychological research on human behavior to safety issues, starting from the knowledge of the scientific laws that govern how people act, and by adopting measurable behavioral indicators within a structured observation system, with the purpose of reducing accidents and injuries at the workplace through a structured system.

The introduction of behavior-based approaches in the workplace has mainly originated from the expectation of overcoming the results obtained by traditional methodologies as training and inspection-sanction systems, starting from the observation that knowledge does not necessarily lead to safe actions and that methods like blame and punishment cannot be used too extensively in practice because they worsen the working climate and endanger co-operation within the organization.

According to the cognitive theories of error, various types of human error can be identified [1], given the relevance of human behavior to the occurrence of accidents.

The BBS finds its foundation in the theoretical and scientific principles of behavioral psychology, a branch of psychology that studies human behavior based on what of it is directly observable. The so-called “ABC” three-contingency model was introduced to describe the basis of any behavioral safety intervention: by manipulating the contingencies A (antecedents) and C (consequences), it is possible to obtain, in fact, any change in terms of frequency, duration, and latency of the behavioral response B [2]. In standard training instead (when “courses” are treated, generally attended outside of working activity), significant antecedents can be observed (e.g., the content of a presentation or the text of a book) but the consequences (as feedback) are typically scarce. Skinner’s “theory of reinforcement” [3] had a significant influence on behavioral safety. A reinforcement occurs when the consequence of a behavior makes it more likely that the behavior will occur again in the future [4]. A positive reinforcement motivates an actor to continue expressing a given behavior and consists of a pleasant thing or condition as a consequence of his/her act. A negative reinforcement, instead, is the requirement to behave in such a way as to avoid possible consequent dangerous situations. Negative reinforcement strengthens behavior by stopping or removing an unpleasant experience. Punishment is different from reinforcement, since it is designed to weaken or eliminate behavior rather than to increase it [5].

The introduction of the term “Behavior-Based Safety” (BBS) is attributed to E.S. Geller, who coined it in 1979, referring to the use of applied behavior analysis methods to achieve continuous improvement in safety performance [6]. With regard to the structure of their implementation, BBS interventions typically consist of the following steps [7]:

(i)

Analysis of the baseline scenario, where the current safety level (i.e., the safety performance before the behavioral intervention) is evaluated for a targeted company. To estimate such performance, in-field data on accidents, near accidents, or unsafe behaviors can be collected by directly observing the workers for a defined timespan; alternatively, those data can be extracted from the accidents or injury records of the investigated company, when available. Typically, the analysis of the baseline scenario also includes the identification of the hazardous tasks to be monitored or improved by means of the behavioral intervention, together with the definition of a target safety performance level. In practice, the mission, values, and goals of the process are set, and correct behaviors are defined. It is important to note that, if the baseline period is too short, the workers can change behaviors in response to being part of a study rather than as the effect of the behavioral intervention: this is the so-called “Hawthorne effect” [8];

(ii)

Intervention, mainly consisting of providing feedback to the subjects (employees or other company professional profiles) about their safety performance on the targeted items and encouraging them to improve their compliance with safe practices. The presence of observers is still required, in order to carry out behavioral observations during the intervention phase and the baseline analysis. Awards, training, and the use of checklists are other distinctive points of the method. The number of participants can vary considerably, up to a few hundred, also as a function of the number of sites involved;

(iii)

Analysis of the post-intervention scenario. This step aims to quantify the safety performance of the workplace after the behavioral intervention, using some specific key performance indicators (KPIs), and to compare it with that of the baseline scenario, to evaluate the resulting improvements. Such a comparison can be supported by an analysis of the statistical significance of the observed outcomes, as long as the number of participants is sufficient, to test if any variation in the safety performance can be attributed to the behavioral intervention or if it is due to exogenous factors. Whenever such evaluation returns unsatisfactory results, a refinement or change of the intervention could be adopted, otherwise, if the desired targets are reached, the attention could be turned to another set of behaviors;

(iv)

Eventually, the analysis of a reversal scenario. During the reversal phase, the intervention is withdrawn, and the original working conditions are restored, for investigating whether the improved safety performance can be maintained even when the intervention phase ends during a follow-up period. The reversal phase is relevant, since many researchers are still doubtful about the persistence of results achieved by means of a behavioral intervention.

For real BBS applications, the general structure outlined above could, nevertheless, undergo substantial modifications and be applied in several variants with different degrees of complexity. Moreover, BBS is characterized by the measurability of its effects, so each BBS implementation can be deeply evaluated and compared with similar ones through a scientific approach.

Since their introduction, behavioral safety techniques have undergone several evolutionary changes, partially overlapping in time [9]. At the beginning, they were applied according to a “supervisory” process, where supervisors observed the workers’ behavior, gave feedback, and provided positive/negative reinforcements. An “employee-led” approach was thus proposed during the 1980s; in this case, the behavioral safety intervention is managed entirely by the employees, who carry out the observations and provide one-to-one feedback for both safe and risky behaviors observed. In the 1990s, a third cultural approach to behavioral safety, grounded on the partnership between employees and company’s top management, was introduced.

In case of BBS, the feedback mechanism [6,10] traditionally appears as the primary behavioral intervention. Velsor et al. in 1997 [11] defined it as ‘‘information about a person’s performance or behavior, or the impact of performance or behavior, that is intentionally delivered to that person in order to facilitate change or improvement’’. Feedback is often accompanied by goal setting (since a conscious planning of goals to be pursued is expected to produce safer behaviors), as well as by observations and checklists.

BSPs have been increasingly implemented within organizations, as a part of their accident prevention programs [12]. Nonetheless, their effectiveness has yet to be demonstrated in several industrial fields. Similarly, the optimal approach to identify unsafe behaviors [13], as well as the issue concerning the duration in time of the intervention effects [14], could be investigated further. In addition, while the basic and traditional components of BBS have been widely debated and reviewed in the literature [13], with a consequent first highlighting of its main strengths and limitations, the evolution of this approach during the years still requires some insights. Along the years, BBS has been applied in different ways and with different complexity levels, depending on the specific context, on the company’s objectives and also on the risk level, i.e., in high-hazard industries with safety-critical environments [15,16,17]. Moreover, in many application cases, it appears more and more difficult to identify clear boundaries between “standard” BBS and different or complementary interventions, like those involving a culture change approach. In some cases, even the terminology used for describing those interventions appears extremely diverse and variable. In summary, BBS and BSPs have gradually shown a plurality of new theoretical and practical development trends that deserve attention.

With the purpose of clarifying these issues and filling some of the existing research gaps on the subject, this paper proposes an updated literature analysis, considering a wide selection of conceptual studies and empirical applications of BBS in the workplace, with the main aim of showing how BBS has evolved over time. A wide timespan, ranging from the 1970s until 2023, has been considered for the analysis, for effectively capturing the key characteristics of BBS interventions, from their introduction up to recent times, as well as to highlight the strengths and limits of the approach. Following the same line of reasoning, different kinds of study are considered in the analysis, including both practical implementations and theoretical approaches. Those studies are categorized and analyzed under various perspectives, so as to identify those topics that reflect the main research trends on behavioral safety, especially in the last two decades, even when traditional BBS is not directly involved, but other similar, alternative, or complementary behavioral solutions are, so as to delineate future prospects of BBS interventions.

The remainder of the paper is organized as follows. Section 2 provides a description of: (i) the inclusion criteria used for progressively selecting papers from the Scopus database to obtain the final sample of interest, which was subdivided into a “List A” of documents [13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146] (most pertinent papers about BBS) and a “List B” [147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199,200,201,202,203,204,205,206,207,208,209,210,211,212,213,214,215,216,217,218,219,220,221,222,223,224,225,226,227,228,229,230,231,232,233,234,235,236,237,238,239,240,241,242,243,244,245,246,247,248] (papers that appear of interest but that are less expressively related to BBS or do not concern BBS in its complete and traditional form); (ii) the analysis grids to be applied to papers in list A, preventively subdivided in four categories; (iii) the methodology for graphically mapping the bibliographical coupling between papers in list B. In Section 3, a keyword analysis on papers of list A is firstly realized and commented, then, the results of the deep analysis of these papers, as obtained through the abovementioned grids, are shown through comparison graphs (original data from papers in List A [13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146] are available in the Supplementary Material in MS Excel format, release 2410 for Windows 365). Section 4 shows the bibliometric analysis of papers in List B (in particular, the graphical mapping of the bibliographical couplings) and describes the consequent identification of clusters of papers. Finally, a discussion about the obtained results and their implications is proposed in Section 5.

2. Materials and Methods

The following subsections illustrate the methods and criteria that have been applied for selecting the material to be reviewed and for subsequently cataloguing and analyzing the identified papers.

2.1. Inclusion Criteria for Papers Selection

As a first step of the review, several queries have been carried out on the Scopus database (in the “Article title, Abstract, Keywords” search field), focusing on words or expressions that are particularly significant and common in the BBS field (and in closely related topics). No specific boundaries were set for the publication timespan, so the overall search covers the period between the 1970s and 2023. The number of papers resulting from each query is summarized below:

• QUERY 1 = “behavioral safety”: 333 papers;
• QUERY 2 = “behavioral analysis” + “safety”: 183 papers;
• QUERY 3 = “behavior based safety”: 386 papers;
• QUERY 4 = “behavioral approach” + “safety”: 195 papers;
• QUERY 5 = “behavior approach” + “safety”: 10 papers.

In addition, it has been noticed that, before the 1990s, the behavioral interventions presenting a structure similar to BBS (whose official appearance was at the end of the 1970s only) were still mostly considered and related to the term “feedback”. Hence, the following additional query was made:

• QUERY 6 = “feedback” + ((“behavior”) OR (“safety performance”)) + ((“safety”) OR (“accident”)) + ((“work*”) OR (“staff”) OR (“personnel”) OR (“employ*”) OR (“occupation*”)): 1162 papers.

It should be noticed that searching for the single term “feedback”, even if combined with “safety”, would return a really huge number of papers, due to the fact that “feedback” is extensively used in various scientific fields. This is the reason why Query 6 includes additional words that contribute to better direct the Scopus search towards behavioral safety interventions at the workplace, with the beneficial effect of limiting the number of papers identified within a reasonable amount. In addition, the upper limit of the publication timespan for Query 6 was set at year 2019, given that it is mainly aimed at identifying older articles.

The six queries returned a list of 2099 total papers, after the necessary removal of duplicated results; these papers have been subsequently and progressively selected using the following criteria:

• Only papers published in peer-reviewed international journals and written in English have been kept. Other publication forms (e.g., books, conference proceedings, newspapers articles, unpublished works, doctoral dissertations, papers published on national journals, etc.) were excluded. This step reduced the number of papers to 1311;
• Articles of strict competence of medical/psychological/biological/pharmaceutical fields were not retained. This step reduced the number of papers to 956;
• Only papers dealing with human behavior have been kept. This means that, for instance, studies in which the word “behavior” referred to machines or software were excluded. This screening reduced the number of papers to 792;
• Only papers concerning safety have been kept. In fact, BBS could even be applied for guaranteeing optimal job performances [249], but working efficiently does not automatically entail working safely, and complementarily aiming for high productivity goals does not necessarily mean sacrificing safety. This step reduced the number of papers to 652;
• Papers that do not concern safety inside working places have been excluded (e.g., analyses of educational contexts). This step reduced the number of papers to 410.

The screening in step 1 was based on an analysis of paper title and publication outlet; in steps 2 and 3 it was based on title, abstract, and journal title; in steps 4 and 5, it was based on title and abstract.

At this stage of the analysis, it was finally necessary to focus on the specific case of BBS by distinguishing it from different behavioral approaches. In line with this, from an analysis of the abstract, the remaining 410 papers were divided into three groups:

• “List A”, including the most pertinent studies about BBS and feedback approaches (134 papers);
• “List B”, including studies that appear of interest but that are less focused on BBS or do not concern BBS in its full and traditional form (102 papers);
• Papers considered not pertinent enough to the BBS topic or with particular causes of exclusion from the analysis (174 papers). This third group was excluded from the subsequent analysis; because of the wide number of papers in this group, clarifying the reason for exclusion was appropriate. The main motivation for the exclusion was the lack of significant pertinence with the BBS subject, for example because, notwithstanding their overcoming of the previous selection phase, the focus was mainly on specific safety topics (e.g., risk analysis), or on particular contexts (e.g., the military one), or on psychological aspects. In some cases, moreover, the main proposed intervention appeared to be organizational or technical, rather than behavioral. Some other exclusion causes have also been identified. For example, papers about hand-hygiene problems in sanitary services have been excluded from the analysis because, in that case, the human error does not affect only workers’ safety, but rather patients’ safety, which would require a dedicated (and different) analysis. Similarly, papers involving stress and mobbing issues could be better analyzed in a separate study, since they are at the crossroad between employee safety and mental health/psychology. Finally, several papers have been excluded because, in their (generally brief) full-text, the commercial and/or informative aspects are dominant compared to the scientific data.

Papers belonging to List A and List B were considered consistent with the subject of the research and have been analyzed in the further phases of the present study (see Section 3 and Section 4), with criteria that will be exposed in Section 2.2.

As the very last selection step, the papers belonging to List A have undergone a further sub-classification, which led to identifying five types of study, namely:

(i)

Review papers (12): papers whose main focus is the analysis of previously published research about BBS;

(ii)

Case studies (78): papers which describe practical BBS implementations in a real working environment;

(iii)

Laboratory simulations (7): papers concerning BBS applications in a simulated working environment or in laboratory settings;

(iv)

Conceptual studies (26): theoretical analyses of the BBS methodology;

(v)

Surveys (11): papers based on questionnaire surveys and analysis of related answers or focus groups.

For articles presenting connections with more than one of such categories, the classification with the greatest relevance was adopted. Figure 1 summarizes the complete selection process as exposed.

The publication timespan of the most pertinent papers (List A) ranges from 1976 to 2023, with a good temporal continuity, especially since the mid-eighties, as shown in Figure 2, and with an increasing interest in the topic through the years, especially in the last two decades. In particular, if taking only the sub-categories of case studies and conceptual studies (i.e., the ones including the highest numbers of papers), the temporal distribution of the papers shows that practical applications of BBS have a dominant role across the whole timespan, but the consistent presence of many conceptual studies in the 2000–2010 period (and also in the subsequent decade, although to a lesser extent) deserves attention. In particular, such significant conceptual development of BBS could have had an influence on the subsequent development of empirical applications in their most modern declinations. This suggests that it is paramount to include in the present study a wide analysis of the BBS topic, including both applicative as well as theoretical aspects.

In Section 2.2, the criteria applied for carrying out the in-depth analysis of the above-mentioned five groups of papers are outlined, with a particular focus on the identification and depiction of the various paper attributes that can be used to describe the BBS implementation and its effectiveness. The set of possible values that have been selected as acceptable for each paper attribute (after a preliminary analysis of all manuscripts) is specified, too. Data not specified within the paper full-text were labelled as “N.A.” (Not Available). Such a classification reflects what was reported in the original Excel contents, available in the Supplementary Material. This latter, in fact, shows the complete classification of all “List A” papers through the above-mentioned attributes, excluding the case of strictly numerical parameters, a very limited set of qualitative values only is available for each attribute. Such a “closed-answer” classification through a “drop down” menu allows the better categorization of the concepts and makes the subsequent analysis more immediate. All additional details, where present, are reported in the single cell by comment. For each paper, bibliographic metadata (authors, title, publication year, publication journal) are also tabulated; these attributes are not recalled or detailed further in Section 2.2.1, Section 2.2.2, Section 2.2.3 and Section 2.2.4 as their meaning is intuitive.

2.2. Procedure for the Analysis of Papers in “List A”

2.2.1. Criteria for the Analysis of Review Papers

The attributes that have been chosen as the most effective criteria for deeply analyzing the 12 review papers are listed in

Table 1. Set of paper attributes used as analysis criteria for previous reviews.

Paper Attribute		Assignable Values
Geographical area of first author’s affiliation		Europe/North America/South America/Oceania/Asia/Africa (single countries are specified in Excel cell by comments)
Review type		Systematic review/Narrative review
Papers’ classification data (only for systematic reviews)	⮚ Number of reviewed papers	Any integer numeric value/N.A.
	⮚ From	Any calendar year/N.A.
	⮚ To	Any calendar year/N.A.
	⮚ Number of reviewed papers regarding feedback/total number of reviewed papers [%]	Any numeric value/N.A.
	⮚ Typology of reviewed papers	Case studies/Others/N.A.
Main investigated methodology of interest		Feedback/Feedback + Goal setting/Feedback + Coaching
Main investigated analysis context		Industry/Transport/Industry + transport/Mixed
Global evaluation of BBS/simple feedback effectiveness		Positive/Negative/Uncertain/N.A.
Global evaluation of long-term effectiveness of BBS/simple feedback		Positive/Negative/Uncertain/N.A.
Emerged main critical factors for successful BBS or simple feedback	⮚ Interpersonal trust/effective communication	YES/NO/N.A.
	⮚ Active role of managers and supervisors	YES/NO/N.A.
	⮚ Gradual application of the method/training	YES/NO/N.A.
	⮚ A priori presence of safe infrastructures (equipment, materials)	YES/NO/N.A.
	⮚ Attention to workers’ personality, attitudes, skills, cultural background, and the peculiarities of the workplace	YES/NO/N.A.
Emerged main critical factors for long-term effectiveness	⮚ Increment of safety culture, social norms and/or safety climate	YES/NO/N.A.
	⮚ Workers’ involvement	YES/NO/N.A.
	⮚ Training	YES/NO/N.A.
	⮚ Programming of long-term strategies	YES/NO/N.A.
	⮚ Attention to workers’ personality, attitudes, skills, cultural background and to the peculiarities of the workplace	YES/NO/N.A.

, together with the values attributable to each of them. The related analyses are reported in Section 3.2 and based on the Supplementary Material.

The distinction between “systematic review” and “narrative review” must be understood in light of the definitions given by Uman in 2011 [250] and Greenhalgh et al. in 2018 [251]. Narrative reviews have mainly a descriptive nature and the analyzed studies are chosen by the authors without explicitly indicating the selection criteria. Hence, narrative reviews can be affected by selection bias and uncertainty in deriving conclusions, but they can, nevertheless, provide interesting interpretations and critiques, since their main aim is to deepen a topic. Systematic reviews generally address narrowly focused questions and are based, instead, on a predetermined precise (“systematic”, in fact) search and analysis strategy, where all selection steps are explicitly outlined; this is why some papers’ classification data in Table 1 are available for systematic reviews only. Systematic and narrative reviews both provide significant contributions to the scientific literature, in a complementary way. Please note that some review authors [27,35] complain that many studies have a poor methodological quality of intervention and are not based on a structured theoretical framework. They also found difficulties in comparing studies with different baselines.

BBS can be applied in different ways [252] and, even if a feedback intervention supported by observation is the essential common element in all successful BBS applications [32], additional factors can play an important role. Accordingly, Table 1 also cites goal setting and coaching techniques.

Safety coaching has, instead, been defined as a process of observation and feedback able to support safe behaviors and provide constructive feedback on risky behaviors at the workplace [253], however, such an approach is centered on a dialogue between one individual (the safety coach) and another one (the worker). The coach, by discussing and making considerations starting from observation and evidence, tries to stimulate the self-awareness and personal responsibility of the second individual concerning safety [15], eventually leading to long-lasting culture change and resilience [70]. As far as the main analysis contexts reported in Table 1 are concerned, “Industry” is intended as an all-embracing category, also including constructions [48] and mining [85]. The “mixed” category groups more than one context, whose list is detailed in each related MS Excel cell by comment.

The global evaluation of behavioral interventions refers to the observations directly provided by the authors of the analyzed reviews inside their manuscripts. The distinction between BBS and simple feedback techniques, which is also recalled in other parts of this study, has to be clarified, too. Such a distinction is grounded in the fact that, especially before the 1990s, the term BBS was not yet widely used, even if feedback techniques associated with behavioral observations were already applied [37]. In more recent times, the application of feedback (without explicit reference to more structured BBS interventions) has been reported in the literature, mainly in relation to the operators of transport field, e.g., truck drivers [142] or food deliverers [81].

In the present study, the analysis of the selected papers highlighted some aspects that were found to be particularly significant for determining: (i) the effectiveness of the behavioral intervention; (ii) the permanence of the achieved results over time (as distinguished in Table 1). Long-term effectiveness has been defined in the literature as the occurrence of one or more target behaviors above baseline levels for a minimum time (Boyce and Geller in 2001 suggested at least two weeks [29]) after the withdrawal of the behavioral intervention. Such levels must be visually detectable on temporal data or reported in terms of statistical significance.

Interpersonal trust in management, as well as in co-workers’ abilities, is one of the elements that can contribute to providing good results from BBS interventions. It is usually accompanied by effective communication (between peers or between supervisors and supervisees) and good workers’ involvement. For example, the intervention effectiveness has often been observed to be higher when the subjects of the intervention also have the role of observers [76] or when workers are allowed to contribute to setting objectives, so as to prevent managers from proposing goals that are impossible to achieve [80].

The effectiveness of any BBS intervention can also be positively influenced by the active role of managers and supervisors in continuously supporting the process, through personal involvement, example setting, and provision of the necessary organizational resources to promote workplace safety, especially if workers can see that their supervisors are fully supportive of the implementation of BBS intervention plans [113]. In some cases, feedback procedures have even been applied to both front-line workers and their supervisors [254]. In general, the importance of personal involvement can be extended to all levels of the hierarchical scale in a generic working context where BBS is applied, including front-line workers, supervisors, managers, and even directors [105].

The improvement in safe behaviors can also be aided by a gradual application of the method, that is a progressive increase in the complexity of BBS process (to allow workers to get used to it) and also a gradual transition from less intrusive interventions to more intrusive ones, as from group-level intervention to individualized feedback [33].

In many cases, safety cannot be obtained by the modification of behaviors alone, since the latter still has to be accompanied by the use and maintenance of proper equipment and materials, design engineering interventions, and proper organization in terms of work task schedule and application of administrative controls [13,145]. The literature reports that the effects of goal setting and feedback safety interventions are often disappointing because they are implemented in an environment that is globally unsupportive of safe performance [80].

The importance of considering workers’ personality, attitudes, skills, cultural background, as well as the peculiarities of the workplace (local laws, internal organization, complexity, temporary conditions, dimensions of the site) while organizing a BBS intervention, has also emerged. Individuals do not respond equally to interventions, and personalized approaches are therefore crucial for overcoming limitations of group-based interventions [83]. Personal work experience, physical health, and degree of proficiency in safety operation have to be taken into account [138]. Cross-cultural differences should also be considered when planning to apply interventions in different countries and cultures.

In any case, whenever productivity appears more financially rewarding to the worker than the concern for safety or, simply, anytime the worker does not internalize the relevance of safe behaviors, it is unlikely that behavioral interventions will be effective. That is why an important contribution to BBS effectiveness can originate from the acquisition of an adequate safety culture and the possibility of working in a context where the social norms support safety and there is an adequate safety climate.

Cooper in 2000 [255] defined safety culture as ‘that observable degree of effort by which all organizational members direct their attention and actions toward improving safety on a daily basis’. Safety culture can increase and extend the effect of behavioral interventions over time, also inducing the worker to act safely in his or her private life or towards society. It is related to greater awareness and, consequently, better safety performance. In 1998, Reason proposed a view of safety culture based on the importance of organizational learning from errors and near misses [256]. A proactive safety culture encourages employees to report incidents and near misses in a blame-free environment. The author promoted a systematic approach to identify and mitigate risks through a strong leadership commitment to safety initiatives, effective communication at all levels of the organization, and the employee’s engagement in and awareness of safety issues. Reason argued that safety culture is made up of various interacting elements, methodologies, mindsets, and management practices that impact the overall safety performance of the organization. Quite recently, the concept of safety culture and connected culture change approaches [257] increasingly emerged as complementary to classical BBS applications, although the importance of the cultural component appeared in the literature from the 1990s in various interpretations [164]. Management commitment, trust between peers, and communication can be regarded as safety culture elements [141]. Behavioral change is favored and maintained by a positive safety culture, but at the same time successful BBS programs can produce positive changes in people’s beliefs and culture [59].

Similarly, the effectiveness of behavioral interventions is strictly related to the concept of safety climate, which consists of the employees’ shared perceptions of the importance of safety and safety policies in their working environment, as well as workers’ perceived risk of injuries and accidents [206]. The safety climate can become the mediator that makes supervisor behaviors influence worker behaviors indirectly [113]. In general, once the social context proposes new safer behaviors and defines them as new social norms through the feedback process, the worker will more easily adopt correct behaviors [28].

2.2.2. Criteria for the Analysis of Case Studies and Laboratory Simulations

A set of 85 case studies and laboratory-simulated studies were identified through the initial selection process. The paper attributes that were identified as effective criteria for carrying out the subsequent in-depth analysis (see Section 3.3 and the Supplementary Material) are listed in

Table 2. Set of paper attributes used as analysis criteria for case studies and laboratory simulations.

Paper Attribute		Assignable Values
Geographical area of first author’s affiliation		Europe/North America/South America/Oceania/Asia/Africa (single countries are specified in Excel cell by comments)
Study typology		Case study/Laboratory simulation
Number of involved working sites		Any integer numeric value/N.A.
Geographical area of implementation		Europe/North America/South America/Oceania/Asia/Africa (single countries are specified in Excel cell by comments)
Implementation context		Industry/Office/Health services/Road and environment maintenance/Dining services/Transport/Mixed
Typology of participants (subjects of the intervention)		Employees or workers/Drivers/Students/Employees or workers + supervisors or managers/Supervisors or managers/Others
Number of actual participants		Any integer numeric value/N.A.
Presence of a BBS steering committee or other dedicated committees (the latter are specified in comments)		YES/NO
Duration of the baseline phase		Any duration/N.A.
Duration of the intervention phase		Any duration/N.A.
Year of implementation beginning		Any calendar year/N.A.
How feedback was provided		Verbal feedback/Graphic feedback/Verbal and graphic feedback/Automatic warning alarms/Written feedback/Material feedback/No feedback/N.A.
Feedback frequency		Daily/Twice a week/Weekly/Monthly/More rarely than monthly/In real-time/Just after observation/N.A.
Individual/group feedback		Individual/Group/Individual and group/N.A.
Goal setting		YES/NO
Additional tangible rewards		YES/NO
Training for studied subjects		YES/NO
Observation phase	⮚ Presence of observers	YES/NO
	⮚ Number of observers	Any integer numeric value/N.A.
	⮚ Evaluation of inter-observer reliability	YES/NO/N.A.
	⮚ Training for observers	YES/NO/N.A.
	⮚ Observers’ professional profile	Employees or workers/Employees or workers + supervisors or managers/Supervisors or managers/External observers/N.A.
	⮚ Observation duration (per observation session, unless otherwise specified in comments)	Any duration/N.A.
	⮚ Observations frequency	Several times a day/Daily/Several times a week/Weekly/Several times a month/Monthly/Continuously/N.A.
	⮚ Declared use of check-lists or observation forms	YES/NO/N.A.
KPIs		Accidents or injuries/Accidents or injuries + safe VS unsafe behaviors/Number of safe behaviors/Number of unsafe behaviors/Percentage of safe behaviors/Percentage of unsafe behaviors
Typology of analysis		Temporal trends + statistical analyses/Temporal trends/Statistical analyses + direct comparison between pre-intervention and post-intervention values
Intervention effectiveness		Not significant/Partially satisfactory/Absolutely satisfactory
Use of surveys (questionnaires/interviews/focus groups)		YES/NO
Presence of follow-up measurements in case of an intervention withdrawal		YES/NO
Duration of the withdrawal period		Any duration/N.A.
Follow-up results		Return to baseline/Partial persistency of obtained safety improvements/Total persistency of obtained safety improvements/Further improvement of safety performance/N.A.
Key factors emerged for a successful BBS (or simple feedback) intervention	⮚ Interpersonal trust/effective communication	YES/NO
	⮚ Workers’ involvement	YES/NO
	⮚ Active role and commitment of managers and supervisors	YES/NO
	⮚ Technology support	YES/NO
	⮚ Attention to workers’ personality, attitudes, skills, cultural background	YES/NO
	⮚ Attention to the peculiarities of the workplace	YES/NO
	⮚ A priori presence of pro-safety engineering and organizational solutions (equipment, materials, procedures)	YES/NO
	⮚ Attention to safety culture, social norms and/or safety climate	YES/NO
	⮚ Adequate feedback frequency	YES/NO
	⮚ Additional rewards	YES/NO
	⮚ Accuracy of observations	YES/NO
	⮚ Training	YES/NO
	⮚ Central role of safety coaching	YES/NO
Emerged main criticisms generally addressed to BBS or simple feedback	⮚ Privacy problems	YES/NO
	⮚ Risk of blaming workers or overburdening them with responsibility	YES/NO
	⮚ Underestimate of some risks	YES/NO
	⮚ Underreporting of accidents	YES/NO
Applied dedicated technologies		Brief description of the dedicated technology/NO

together with the values attributable to each of them.

When analyzing experimental/laboratory BBS applications, attention has been focused on solutions with positive feedback as the main type of intervention. On the other hand, studies using incentives, rewards, or training as exclusive interventions have been excluded. The elements have, nonetheless, been included in Table 2 as possible adjuvants of the effectiveness of the main intervention [74,82,111], together with goal setting [56,64,100,101]. The concept of incentive/reward programs could be theoretically viewed in an extended way, since even a classical behavior-based technique, as scheduled feedback between employers and their supervisors [229], can motivate workers, resulting in a “natural” incentive; positive final feedback could be thought as a reward as well [175]. In this paper, incentives, awards, and rewards are intended, in a more restricted way, as purely material incentives and tangible rewards, according to principles of the token economy [258]. They can include, for example, social celebrations, monetary rewards, lotteries, coupons, promise cards, raffles, team meals, and award ceremonies. Workers can be rewarded for their safe behavior or exceptional performance, for example if they work injury-free over a given period of time [144], even if sometimes this approach can increase the risk of unreported accidents.

The presence of training phases can be useful to get information about (i) the required work tasks, the correspondent safe behaviors (including ergonomic postures) to be adopted, or the consequences of unsafe behaviors [39,103] and about (ii) BBS methodology [68] and self-monitoring procedures.

In many cases, a BBS steering committee is established, with the role of overseeing and managing the process [43,44,45,51,61,98,99]. This committee is also responsible for driving the observation and feedback process.

Both the number and professional profile of the observers, as well as the frequency of observation and duration of each session, can vary significantly depending on the implementation. Observers can be trained before the intervention about how to correctly observe and judge the workers’ behaviors. Sometimes they can be external subjects, as university research personnel or the authors of the papers themselves [54,73,106]. The reliability of their evaluations is typically assessed through a preliminary comparative analysis of agreements/disagreements given by the various observers (inter-observer reliability). The observation of safe/unsafe behaviors is often carried out by means of proper checklists, which are used for assessing the employees’ behavior [71]. Although they are not the only way to collect observations [13], checklists are widely used in practice. Observers can sometimes be completely absent, although these situations typically reflect the very specific cases of the mere usage of accident statistics or of the workers’ self-management, in which solitary workers or workers with little supervision (e.g., drivers or miners) can self-monitor with auto-observation and auto-scoring [65,66,94].

Feedback can be applied in several ways, too. It can be addressed to individuals or to groups, and it can be realized through graphical tools such as graphs, public charts, posters, flyers, signs [41,47,55,69], or through direct comments given by observers or supervisors [38,46], or during meetings [53,78], or even by using other materials (e.g., a card system [93]). Whenever the intervention is strongly supported by more modern technologies, real-time feedback can be provided, for example through automatic warning alarms, along with a continuous observation process. The act of observing the safety performance of others can promote the observer’s own safety behavior, within the so-called “observer effect” [40,92,107], which also contributes to promoting the spread of safety culture.

For the purpose of this study, the intervention effectiveness (in the intervention and follow-up phases) has been evaluated in line with what was qualitatively expressed or shown by the authors of the papers analyzed, as the extreme heterogeneity of the KPIs used would make a quantitative comparison meaningless or unfeasible. The authors’ evaluations can be substantiated by the presence of control groups [60] or comparison sites [70,87,95], which are used to distinguish the improvements resulting from the BBS intervention from those due to exogenous factors (and therefore observed even in sites/groups not involved in the intervention).

In some cases, the analyzed studies had to consider the “response generalization” phenomenon, occurring when multiple behaviors, functionally related to a common outcome, improve as a result of intervening in one of them [124,259]. In this case, the effectiveness of the intervention should be assessed not only based on the targeted behaviors but also on the functionally related ones, not targeted by the intervention [81].

As regards the KPIs used to evaluate safety performance, the safety level before and after the intervention is typically measured as the percentage of safe behaviors, which results in a Safety Performance Index. Other KPIs used to assess the safety performance are the number (or percentage) of injuries, accidents [75,84], and safe or unsafe behaviors. In some studies, productivity, job quality, and safety culture are evaluated as complementary indexes; they are marked in the Supplementary Material by a comment in the related cell. The evaluation of the intervention effectiveness could also include one or more health outcome measures, such as illnesses or fatalities. However, since many work-related health outcomes can originate long after the initial exposures to the hazard, this is not always feasible [13].

It also has to be observed that, in the present study, housekeeping is assumed to belong to safe behaviors that can be monitored and measured [49,97,102,260]. One way of preventing injuries consists, in fact, in keeping good order and tidiness (for example, a loss of balance due to stepping on objects can produce musculoskeletal problems).

In the analyzed papers, a common procedure is to provide a graphic representation of the considered KPI (for example, a safety index) as a temporal trend, from the baseline until the eventual withdrawal phase. In case of the unavailability of data all along the considered timespan, only a direct comparison between pre- and post-intervention values is presented, frequently including statistical analyses (e.g., mean, standard deviation, t-test, ANOVA).

The BBS practical intervention can be enriched by the further use of surveys (in the form of questionnaires, interviews, and focus groups) in various steps:

(i)

Before the intervention: to evaluate workers’ safety perception, to understand the motivations for performing unsafe behaviors [86], to define a safety performance measurement tool, to evaluate the safety psychology of workers, and to receive opinions about the proposed goals;

(ii)

After the intervention: to gather workers’ safety climate perceptions, to collect more information about their perception of the BBS program, and to verify if the intervention has increased personal safety awareness [58].

Table 2 lists the main possible critical factors for a successful BBS (or simple feedback) intervention. In addition to the considerations in Section 2.2.1, it has to be underlined that, in practical applications, technology has emerged as another important tool for improving BBS effectiveness. In particular, it often has an important role whenever a real-time feedback/observation system is required, including video-cameras [104], automatic alarms [42,79,112], tracking systems, observational databases, location systems [63], and on-board computer monitoring for drivers [57,96,136]. The importance of the active role and commitment of managers and supervisors still arises, for example, they can participate in periodic BBS meetings or provide additional feedback. In some cases, participants (drivers in particular) prefer to receive feedback from supervisors or managers than from technology alone [91].

Training can be added to the main feedback intervention for obtaining satisfactory results that, in some cases, would not have been achieved without such an association [90]. Not only can formal training be useful, but informal influences can also be so in terms of “tacit knowledge” [50,128], such as personal experience and knowledge acquired by mentoring or from co-workers.

Finally, Table 2 details some criticisms, generally addressed to BBS or simple feedback, that appeared or were mentioned in case studies/laboratory studies. First of all, workers can fear being spied on by managers through the observers and that their privacy could be violated, to such an extent that writing the observer’s name and the name of the observed person was optional or prohibited in certain cases [108,110]. The risk of blaming workers or overburdening them with responsibility is also a typical criticism directed at BBS. In 2016, Dekker and Breakey described the concept of “Just Culture” while emphasizing the importance of creating an environment where the outcomes of safety incidents are analyzed to identify underlying causes and implementing changes to prevent recurrence, rather than simply assigning blame [261]. According to Dekker and Breakey, a “Just Culture” fosters a blame-free environment where employees feel empowered to report issues and contribute to safety improvements, while encouraging dialogue between affected parties after incidents occur, and promoting understanding and collaborative solutions rather than punitive measures.

Other risks can be underestimated, such as minor injuries, e.g., metal splints and cuts which are often considered as “part of the job” by many workers, and rare (but very dangerous) events like environmental disasters in complex chemical and nuclear industries [89].

Finally, it can happen that accidents are underreported by workers, for example because they do not consider them serious enough, or they do not want to waste time on them [36], or because a workplace incentive program rewards low rates of injury reports [67], or simply for fear of censure [76].

2.2.3. Criteria for the Analysis of Conceptual Studies

A set of 26 conceptual studies was identified through the initial selection process. The paper attributes that have been chosen as criteria for carrying out the subsequent in-depth analysis (see Section 3.4 and the Supplementary Material) are listed in

Table 3. Set of paper attributes used as analysis criteria for conceptual studies.

Paper Attribute		Assignable Values
Geographical area of first author’s affiliation		Europe/North America/South America/Oceania/Asia/Africa (single countries are specified in Excel cell comments)
Study aim		Discussion and proposal of guidelines/Analysis of a specific phenomenon/Proposal of a model or methodology variant or tool/Proposal of research topics/Criticisms about the methodology
Proposed model (where applicable)		Brief description of the proposed model/N.A.
Analyzed specific phenomenon (where applicable)		Brief description of the analyzed specific phenomenon/N.A.
Main methodology of interest		Feedback and (or) reinforcement/BBS/Behavioral safety
Presence of supplementary empirical data		YES/NO
Presence of supplementary data from surveys		YES/NO
Contexts at very high risk are taken into consideration		YES/NO
Emerged main critical factors for successful BBS or simple feedback	⮚ Interpersonal trust/effective communication and attention to language	YES/NO
	⮚ Workers’ involvement	YES/NO
	⮚ Active role and commitment of managers and supervisors	YES/NO
	⮚ Technology support	YES/NO
	⮚ Attention to workers’ personality, attitudes, skills, cultural background	YES/NO
	⮚ Attention to the peculiarities of the workplace	YES/NO
	⮚ A priori presence of pro-safety engineering and organizational solutions (equipment, materials, procedures)	YES/NO
	⮚ Attention to safety culture, social norms and/or safety climate	YES/NO
	⮚ Training	YES/NO
	⮚ Central role of safety coaching	YES/NO
Emerged criticisms about BBS or simple feedback	⮚ Privacy problems	YES/NO
	⮚ Risk of blaming workers or overburdening them with responsibility	YES/NO
	⮚ Infrequent unsafe behaviors or low probability/high consequence risks are often not considered	YES/NO

, together with the values attributable to each of them.

Conceptual studies typically present theoretical analyses of the BBS methodology [122], although they sometimes focus on feedback and reinforcement phases only [117,133]. In a few cases, they target the general concept of behavioral safety. Papers here have been classified as “conceptual” even if they refer to methodologies different from BBS or if only a limited part of the paper is dedicated to BBS, regardless of the possible presence of empirical data from case studies [132]. Similarly, supplementary data, deriving from experimental applications or from surveys, can sometimes be included.

In the analyzed papers, four main theoretical subtopics were identified:

(i)

Discussion and proposal of guidelines for BBS application (for example, in relation to modern “person-based approaches”, which apply surveys, personal interviews, and other self-report measures to discover how individuals feel about certain situations and conditions, starting from the idea that observable behaviors do not completely represent a person’s actions [130]);

(ii)

Analysis of a specific phenomenon, in particular the normalization of deviance [115], response generalization [124], tacit knowledge [128], emotional intelligence [137], emotional safety culture [136], participatory safety management [131], and organizational discourse practices in BBS [129];

(iii)

Proposal of models or methodological variants or tools (e.g., models including safety culture [116]) or models based on a combination of decision support tools [138] or other variants [120,134,135];

(iv)

Proposal of research topics on BBS for further scientific insights [118,121];

(v)

Criticisms against the methodology (see, e.g., the study by Hopkins in 2006 [123]).

In relation to the phenomena mentioned in point (ii) of the list, some clarifications are needed. Emotional intelligence involves the ability to perceive, evaluate, and demonstrate emotions successfully [137]. This can be important in safety coaching as it allows the safety coach to recognize his/her own emotions for objective interpretations and feedback. The normalization of deviance is a long-term process in which individuals (or groups) admit a lower standard of performance which finally becomes acceptable. In relation to safety, it is characterized by an acceptance of risk related to deviant behavior that is never corrected, since it is ignored or undetected, until the deviant behaviors come to be considered normal [115].

Finally, the analysis of conceptual papers has identified previous studies that investigate contexts at very high risk as chemical and nuclear plants, and even constructions [114,125,126,127], since the application of BBS interventions in those contexts can require special measures and strategies [262].

2.2.4. Criteria for the Analysis of Survey Papers

A set of 11 surveys has been identified through the initial screening process. The paper attributes that have been chosen as effective criteria for carrying out the subsequent in-depth analysis (see Section 3.5 and the Supplementary Material) are listed in

Table 4. Set of paper attributes used as analysis criteria for surveys.

Paper Attribute		Assignable Values
Geographical area of first author’s affiliation		Europe/North America/South America/Oceania/Asia/Africa (single countries are specified in Excel cell by comment)
Geographical area of implementation		Europe/North America/South America/Oceania/Asia/Africa (single countries are specified in Excel cell by comment)
Application context		Industry/Transport/Health Services/Mixed
Number of working sites involved (where applicable)		Any integer numeric value/N.A.
Investigated behavioral intervention methodology		BBS/Simple feedback
Typology of participants to the survey		Employees or workers/Employees or workers + supervisors or managers/Supervisors or managers/Safety professionals/Drivers/Contractors
Number of respondents		Any integer numeric value
Survey methodology		Interviews/Questionnaires/Interviews + questionnaires
Additional use of focus groups		YES/NO
Emerged main critical factors for successful BBS or simple feedback	⮚ Interpersonal trust/effective communication	YES/NO
	⮚ Active role of managers and supervisors	YES/NO
	⮚ Mandatory participation	YES/NO
	⮚ Gradual application of the method/proper BBS training	YES/NO
	⮚ Technology support	YES/NO
	⮚ Attention to safety culture, social norms and/or safety climate	YES/NO
	⮚ Attention to workers’ personality, attitudes, skills, cultural background	YES/NO

together with the values attributable to each of them.

Surveys (realized through questionnaires, interviews, and/or focus groups) allow data to be found that are related to BBS as seen from the perspective of the subjects involved, who are mainly line workers, supervisors, and managers. Safety professionals turned out to be the subjects of surveys, too, since they can significantly contribute to providing a safe working environment, even through training, motivating, and monitoring the employees. The results of a survey can include: (i) participants’ opinions about BBS; (ii) participants’ expectations from BBS implementation; (iii) factors influencing participants’ involvement in BBS; and (iv) participants’ concerns about BBS. Sometimes, selected answers to surveys regarding the BBS approach can be used as possible KPIs in BBS applications [141].

At the same time, occupational accidents are generally negatively affected by personality dimensions as honesty, emotionality, and openness to experience, while they are positively influenced by conscientiousness [146]. Some of the surveys analyzed have also highlighted that even the mandatory nature of participation in the intervention can be included among possible crucial factors for a successful BBS. This means that mandatory processes do not necessarily reduce employee satisfaction, trust, or perceptions of personal control [140].

2.2.5. Keyword Analysis

A keyword analysis [263] was performed on papers in List A with the purpose of identifying and evaluating (especially in terms of presence and persistence over time) the various subtopics that BBS encompasses. At the same time, the most common nomenclature used in the field was derived. Keywords were retrieved automatically through the “Export” function of Scopus; 25 out of the 134 papers were excluded from the analysis because they were found to have no authors’ keywords. The resulting terms were merged into groups in the case (i) they were synonymous, or they referred to strictly related topics, (ii) they were different declinations of the same word (e.g., adjective and correspondent verb), or (iii) one was linguistically contained in the other. With such an approach, a list of 156 “grouped” keywords was obtained

2.3. Procedure for the Analysis of Papers in “List B”

During the initial selection process, 102 papers (List B) were classified as studies of potential interest but not specifically concerning BBS or focusing on BBS in its complete and traditional form. The publication timespan of papers in List B ranges from 1983 to 2023, but a good temporal continuity can be observed only after the 2000s (Figure 3), except for years 2020 and 2021, which were probably affected by the COVID-19 pandemic.

Such a temporal distribution suggests that, in recent times, the traditional types of study of BBS have been enriched by the publication of many papers that are only partially correlated with the standard BBS approach or that present it in new versions or with different nomenclatures. The analysis of this category of papers thus offers the possibility to identify the most recent trends of research topics related to new variants of BBS or closely related subjects.

The full texts of all papers in List B were analyzed with the purpose of subdividing them into groups on the basis of common topics. The main motivation was to identify recent research trends in sectors that are closely adjacent to that of the standard BBS, especially in the last two decades. This preliminary analysis, inevitably influenced by the subjective analytical ability and critical thinking of the authors, has led to the identification of several subgroups of papers focused on the following topics:

(i)

Safety culture (e.g., reviews, studies on its interactions with behavioral interventions, proposals of application models);

(ii)

Safety climate (e.g., reviews, surveys about safety perception at the workplace, studies on the effect of safety climate on safety performance, studies on its interactions with leadership roles);

(iii)

Incentives and rewards (e.g., studies on incentives or external incentives in companies, analyses of their effectiveness);

(iv)

Leadership role (e.g., leadership–based behavioral interventions and approaches, analysis of supervisory safety practices, studies on leadership training);

(v)

Training (e.g., implementations of classical safety training, studies on technology-supported training, proposal of optimizations of training approaches depending on variables like workers’ personalities);

(vi)

Audit (e.g., description of global on-site systems for safety data collection, studies on electronic methods for more efficient mode data transfer in official audits);

(vii)

Alternative methods to classical BBS (e.g., proposal of methodological variations with different participation levels for peers and managers, studies on interventions based on cognitive processes, proposal of multi-level approaches as in the case of Social Ecological Frameworks);

(viii)

Preparatory phases to possible subsequent BBS interventions (e.g., preliminary behavioral analysis of workers with identification of safe/unsafe acts, general preliminary safety data collection, preliminary definition of performance targets in terms of health and safety, analyses of the relationships between behaviors and incidental rate in workplace);

(ix)

General safety analyses and complex system modeling (e.g., proposals of system dynamics models for behavior analysis or for improving the global safety system, descriptions of general pro-safety interventions where the behavioral aspect is one component among others).

For validating the subdivision of papers in List B shown above, it has appeared necessary to find a scientific method to confirm, deny, or partially change it through the adoption of an objective approach. Bibliometric analysis with graphical mapping, a well-known analysis instrument [264] that can be carried out using several commercial software packages, has been chosen as a suitable tool for that purpose. This analysis allows the most influential authors and studies, as well as prominent themes, to be identified in the desired timespan for a selected (and even very numerous) group of items that constitute a bibliometric network made of nodes (e.g., papers) and edges linking all pairs of nodes [265]. In particular, bibliographic coupling has been adopted as the type of relation between papers to be analyzed, since, during a brief preliminary testing phase, it proved to return more significant results compared to other solutions (e.g., direct citations, co-citations, keyword co-occurrence, or co-authorships) for the specific case. Two publications are bibliographically coupled if a third publication is cited by both publications [266], hence, the larger the number of references the two publications have in common, the stronger the bibliographic relation between them.

In this study, bibliographic coupling has been analyzed and displayed using the VOSviewer software package, release 1.6.12 for Windows [267] that was chosen mainly because of its free availability and its automatic compatibility with Scopus comma-separated values (“.csv”) export files. This tool provides distance-based visualizations of bibliometric networks, so the distance between two nodes approximately indicates how connected they are. In particular, the VOSviewer can support “overlay” visualizations, where the color of a node indicates a given property of the node itself (e.g., number of citations in the literature, year of publication). Section 4 presents the results of the bibliometric mapping carried out on 102 papers from List B. The resulting observations about the observable research trends are finally reported in Section 5.

3. In-Depth Analysis of Most Pertinent Papers About BBS (“List A”)

The full text of all papers in List A was analyzed by applying the criteria detailed in Section 2.2. As a result, several tables, available as Supplementary Material, were obtained. Related data were elaborated in aggregate form to identify common features and trends. The in-depth analysis of each paper category (according to the groups introduced in Section 2.1) is preceded by a general analysis of the author keywords for all papers of List A, whose publication outlet was also evaluated. The Journal of Organizational Behavior Management is the journal that published most of the selected studies (33 papers). Safety Science and the Journal of Safety Research also published a relevant number of studies (15 and 11 papers, respectively), while the Journal of Applied Behavior Analysis published seven papers. A good number of papers (four) were published by Accident Analysis and Prevention, the International Journal of Industrial Ergonomics, the International Journal of Occupational Safety and Ergonomics, and the Journal of Occupational Accidents. The remaining 52 articles are homogeneously distributed across a further 35 journals.

3.1. Analysis of the Authors’ Keywords

Table 5. Number of occurrences (n) of author keywords for papers in List A. (Note: only keywords with n > 2 are displayed).

Author Keywords	Total Number of Occurrences (n)
Behavior-based	48
Feedback	30
Behavioral safety	28
Observation/observer	18
Industry/industrial/factory/plant; culture/cultural	16
Occupational safety	15
Management	13
Driving/driver	12
Accident/incident; goal/goal-setting	11
Organizational behavior/psychology; construction	10
Organization/company/enterprise; risk; safety	9
Injury	8
Intervention; behavior	7
Behavior modification/change/covariance;	6
Coaching; training; participation; incentive/reward/promise card; performance; worker; unsafe/at-risk-behavior; safe behavior	5
Office; vehicle; human factor/error; response generalization; knowledge; review; hazard; employee; workplace	4
Long-term; protection/protective; supervision/supervisor; applied behavior analysis; research; effectiveness; safety program/system; social	3

shows the keywords that appeared more than twice in the literature, including their total number of occurrences (n). The solidus between different words indicates that the related terms have been grouped, as explained in the methodology. The same keywords are reported in

Table 6. First year of appearance of author keywords for papers in List A. (Note: only keywords with n > 2 are displayed).

Author Keywords	Year of First Appearance
Industry/industrial/factory/plant; occupational safety	1978
Management; hazard; accident/incident; employee; vehicle; training; organization/company/enterprise; feedback; supervision/supervisor	1980
Safety	1986
Driving/driver; response generalization	1991
Culture/cultural; behavior	1993
Goal/goal-setting	1994
Protection/protective; risk; performance	1995
Safety program/system; workplace; behavioural safety	1997
Behavior modification/change/covariance	1998
Behavior-based; injury; incentive/reward/promise card	1999
Social; review; research	2000
Long-term; observation/observer; worker; organizational behavior/psychology; applied behavior analysis	2001
Participation; office	2004
Human factor/error	2005
Safe behavior	2006
Coaching	2007
Unsafe/at-risk behavior; intervention	2008
Knowledge	2010
Effectiveness	2012
Construction	2013

, ordered by their first year of appearance in the literature.

By comparing data from Table 5 and Table 6, the following considerations emerge:

(i)

The keywords with the highest number of occurrences (n > 10 is taken here as the threshold for “high” values), e.g., “behavioral safety”, “industry/industrial/factory/plant”, “goal/goal-setting”, “feedback”, are mainly terms that appeared in the literature between the end of the 1970s and the mid-1990s. Due to their early introduction, they had more time to increment their presence in journals. Moreover, they refer to very classical features of BBS (e.g., goal setting and feedback, as discussed in the Introduction), or very important and evergreen application contexts (e.g., industries);

(ii)

The presence of “driving/driver” between the keywords with the highest n, although potentially surprising, could actually indicate that this kind of activity, in which the role of human behavior (and, consequently, of human errors) in safety is crucial, was, in fact, always considered worthy of investigation by the scientific community interested in BBS;

(iii)

The keyword “culture/cultural” exhibits a high number of occurrences starting from the 90s, proving that the subsequent proliferation of safety culture topics had solid foundations in previous years;

(iv)

“Feedback” and “behavior-based”, that are very representative and classical keywords in BBS field, have a very high n. It is important to note that “feedback” had a very early debut, at the beginning of 1980s, while the expression “behavior-based safety”, even if it was coined for the first time at the end of the 1970s, required time to be assimilated by the scientific community. In the sample of papers considered, this keyword actually appears for the first time at the end of the 1990s; the very high n value (the highest one in Table 5) demonstrates the explosion of interest that the theme had in the subsequent years. The different temporal introduction of the two keywords substantiates the choice of using a dedicated query to capture the oldest studies (cf. Section 2.1);

(v)

Some (few) keywords are, at the same time, characterized by a high n value and a quite recent debut in the scientific community. They are therefore related to topics that have been investigated a lot from the 2000s onwards. This is the case for “observation/observer”, that refers to the observation phase of BBS applications: this result underlines the importance of this specific BBS feature;

(vi)

Instead of focusing attention on keywords appearing for the first time after 2004 only (that is, relatively recently), it is evident that the dominant role is held by “construction”, with the highest number of occurrences. This appears, therefore, as the industrial context in which BBS has been mostly studied and applied in recent times. In any case, it can also be noted that, among the remaining terms, even “safe behavior” and “unsafe/at-risk-behavior” have a quite relevant number of appearances, showing how the classification into safe/unsafe behaviors is increasingly present as a safety evaluation criterion.

To check the persistence of some topics along the considered timespan, the temporal distribution on Scopus for six authors’ keywords that appear in publications in at least 10 different years has been derived and is shown in Figure 4.

According to Figure 2, the period between approximately 1975 and 1985 is poor in keyword occurrences and quite poor in publications in List A. Only “feedback” and “industry/industrial/factory/plant” are occasionally present in this timespan, while they reappear significantly from 2000 onwards in terms of “dominance”, i.e., the frequency of a concept being used as a keyword over time [263]. This confirms that: (i) the industrial context has represented since its origins (and still represents) the main application area of BBS; (ii) as anticipated in Section 2.2, the term “feedback” was used before the 1990s as a substitute for what would have later been called the “behavior-based approach” and has maintained its importance as a fundamental BBS component until today. In Figure 4, particularly high values of n are also observed in the period 2010–2015. That is why List A includes not only studies about the complete BBS approach, but also old as well as more modern studies about simpler feedback interventions.

In addition, all six considered keywords hardly appear at all in the decade 1985–1995, while their presence begins to be extremely intense starting just before the 2000s. The keyword “culture/cultural”, in particular, presents a generally lower number of appearances compared to the other considered keywords; it is mainly present in the last two decades, although a minimal presence in the 1990s can be observed. A generally uniform distribution of occurrences is observed for the keyword “behavioral safety”, which is, in fact, a very common expression, which could refer to a plurality of interventions (not limited to BBS). The highest values in Figure 4 are, nonetheless, referred to “behavior-based”, with a very significant presence in the last decade and graphically represented by high amplitude peaks, seeming to follow the trend already observed in Figure 2 for case studies, which, in fact, dominate the last years of scientific publications on BBS. Finally, there is a significant number of appearances for the keyword “observation/observer” starting from 2000, confirming what was already observed in relation to Table 5 and Table 6.

3.2. Analysis of Previous Reviews

Based on the criteria outlined in Section 2.2.1, the contents of the review papers were categorized in tables in the Supplementary Material. Related data were subsequently elaborated, so as to derive general considerations, as expressed below.

Seven out of a total of twelve review papers were carried out in North America (the US, in particular). Most studies are dedicated to experimental behavioral interventions applied in industrial and transport contexts, and just over half of them consist of systematic reviews. The topic investigated is not always limited to feedback (or BBS), since, in most cases, feedback interventions represent only a part (from 30% to 92%) of the papers that they analyze. For each study, the factors that emerged as critical for a successful BBS and also for the long-term persistence of its results have been identified.

Figure 5 shows that interpersonal trust, the use of effective communication, and the active role of managers and supervisors appear as the key elements for obtaining satisfactory behavioral results. The attention given to the prerogatives of each participant, as well as of the specific workplace, also seem to have a quite important role to this end. In addition, the gradual application of BBS and the a priori presence of safe infrastructures (equipment, materials) have shown to affect, albeit to a lesser extent, the final behavioral results. In relation to the persistence of the BBS effect during an eventual withdrawal phase (Figure 6), it is evident how safety culture, safety climate, and adequate social norms, which typically presuppose a lasting awareness of the importance of correct behaviors, have a really dominant role.

Even the involvement of workers in BBS interventions is almost at the same level, since it is, itself, usually a necessary condition for the development of a general pro-safety mentality. In six out of twelve reviews, a goal setting technique was applied together with feedback, but no specific effects of this choice on the final intervention effectiveness were observed.

Figure 7 shows that 90% of the reviews gave a positive evaluation of BBS applications, based on the opinions that emerged from the comments given by the authors of the papers that constitute their bibliographical references list. The remaining ones instead expressed an uncertain judgment, highlighting, for example, that the expected results of the BBS applications were achieved only partially (e.g., only some behavioral targets out of that set were reached, or effectiveness appeared differently, depending on the group of participants). In terms of long-term effectiveness, the uncertain evaluation was, instead, dominant (75% of papers). In any case, no fully negative BBS results emerged. In a few cases, the authors did not provide a clear evaluation of the implementation.

3.3. Analysis of Case Studies and Laboratory Simulations

Following the procedure detailed in Section 2.2.2, the main contents of the case studies and laboratory studies were mapped and are presented in the Supplementary Material. Related data were then elaborated with the purpose of deriving general considerations, such as those outlined below.

Figure 8a shows the distribution of case studies and laboratory simulations in different working contexts (excluding contexts with no more than two implementations): industry reaches the highest value (69%), followed by office (14%) and transport (13%). A small quota of papers target health services (e.g., hospitals). The number of sites involved is variable, however, in most cases (66%), one application site was considered, while in 8% of cases the existence of comparison sites is reported. It is important to specify that laboratory simulations mainly reproduce office activities (five out of seven studies) due to the simplicity of replicating, e.g., computer activity and related ergonomic issues. Looking at industrial applications, 31% of the implementations were carried out in the manufacturing industry. The second most frequent implementation context is construction (21%), while five studies propose BBS in the mining industry. The analyzed case studies and simulated BBS interventions were implemented in North America (USA, in particular) in 57% of cases, followed by Asia (mainly China and South Korea) and Europe (mainly UK and Finland), as shown in Figure 8b. In almost all cases (95.1%), the geographical area of experimental application and the geographical area of origin of the first author coincide. The interventions have mainly targeted front-line workers and employees (76% of cases), with 8% of the studies also including managers and/or supervisors as additional subjects. BBS interventions were also often applied to drivers (14%), confirming the strong involvement of the transport sector. BBS steering committees, or other dedicated committees, are present in 28% of cases.

Feedback is mainly provided through graphic methodologies (31%), verbal communication (24%), or both (22%), but even automatic warning alarms have been used in a significant percentage of studies (9%) adopting real-time systems. Feedback was applied to groups or individuals at an almost equal percentage.

The duration of baseline and intervention phases can vary a lot (from a few weeks to some years). The choice of the KPIs to be adopted fell on the percentage of safe behaviors (and related indexes) in 51% of cases, but even an evaluation of the final effectiveness based on accidents or injuries was found to be quite widespread (32% of cases). The analysis of such indicators has been made mainly through a combination of both temporal analysis and statistical elaborations (56% of cases).

BBS effectiveness was evaluated as fully satisfactory after the intervention in 88.2% of cases, partially satisfactory in 9.4% of cases, and not significant in 2.4% of cases only (again, based on opinions that emerged from the comments given by the authors of the papers). In those cases in which a follow-up after withdrawal had been implemented, a total preservation of the obtained safety improvements was observed in 40.6% of cases, a partial persistency in 32.4% of the cases, a further improvement of safety in 5.4% of cases, and a return to baseline in 21.6% of cases.

For each study, the critical factors for a successful BBS initiative have also been identified. Figure 9 shows the number of case studies and laboratory studies in which each critical factor for a successful BBS was identified (blue bars) out of all studies (i.e., 85 papers). Each factor can theoretically reach the total number of case studies, since they are not mutually alternative, rather, they can often coexist. In the same graph, partial values, relating to the subset of studies with a partially or totally satisfactory follow-up analysis after intervention withdrawal, are highlighted using orange bars to delineate possible influences that every critical factor might have had on the long-term persistence of the results.

Interpersonal trust and the use of effective communication, workers’ involvement, the commitment of managers and supervisors, and the attention to safety culture and safety climate appear as the crucial elements for obtaining satisfactory behavioral results in the short-term (with more than 25 occurrences). The presence of additional training and of an adequate feedback frequency, the attention towards the prerogatives of each individual participant, the use of dedicated support technologies, and the a priori presence of pro-safety engineering and organizational solutions follow immediately, with 14 to 21 occurrences in papers. Very similar proportions between the presence of the different critical factors can be identified if looking only at the orange bars; in any case, a few factors (the use of dedicated technologies, the a priori presence of pro-safety engineering, and organizational solutions and the peculiarities of the workplace) seem to have a slightly lower influence if examining only papers with a reversal phase.

BBS implementation supported by applied dedicated technologies has been concentrated in recent years, starting from around 2003 (Figure 10). Four main criticisms against BBS (or simple feedback) have emerged from the analyzed studies (Figure 11), including privacy problems (18 papers) and the risk of blaming workers or overburdening them with responsibility (16 papers). The underreporting of accidents covers an important number of cases, too (13 papers). The underestimation of specific risks is finally cited in six papers only.

Interestingly, it can be finally noticed that, when housekeeping is specifically used as KPI, results after withdrawal appear particularly satisfactory.

3.4. Analysis of Conceptual Papers

On the basis of the criteria detailed in Section 2.2.3, the contents of the conceptual papers were elaborated in tables in the Supplementary Material. Related data were reworked to derive general considerations, as expressed below.

The conceptual studies were carried out in North America (the US, in particular) in 57% of cases. The remaining studies are almost equally distributed between Europe, Asia, and Oceania. The main methodology applied is the full BBS (69% of cases), followed by simple feedback and reinforcement (23% of cases, concentrated in the period before 2006). Only two papers discuss other related topics of behavioral safety. In most cases (81%), high risk contexts are not taken into consideration. Sometimes, conceptual studies are enriched by the presence of supplementary empirical data (30.7% of cases) and of supplementary data from surveys (19.2% of cases). The three main criticisms of BBS (or simple feedback) are privacy problems, the risk of blaming workers or overburdening them with responsibility, and the underestimation of infrequent unsafe behaviors or low probability/high consequence risks. In any case, the percentages of papers that report these criticisms are not very high (3.8%, 7.7% and 11.5%, respectively).

Figure 12a shows the distribution of the study motivations for the conceptual papers: it is evident that the discussion of BBS with a proposal of guidelines is the most debated subject (30.8%), but also the analysis of specific phenomena, the proposal of BBS innovative models and methodological variants or tools exhibit a very high percentage of occurrences (both at 26.9%). However, Figure 12b, in which the same study motivations are presented in terms of the number of papers produced from 1975 to 2019, highlights that studies discussing BBS and proposing guidelines have a more uniform distribution on the entire timespan, while the analysis of specific phenomena and the proposal of innovative BBS models seem to be the most debated topics in recent years (i.e., starting from 2000). A limited number of papers are, instead, dedicated to criticisms against the BBS methodology and to the proposal of research topics. For each study analyzed in this paper, the critical factors for a successful BBS were identified. Figure 13 shows the number of conceptual papers in which each critical factor for a successful BBS initiative was identified (each factor can theoretically cover the total number of conceptual papers, since they are not mutually exclusive, rather, they often coexist). Interpersonal trust and the use of effective communication, the commitment of managers and supervisors, and the attention to safety culture and safety climate appear to be the most significant elements that allow satisfactory behavioral results to be obtained (12, 11, and 11 papers, respectively).

Even the workers’ involvement, the attention to the prerogatives of each individual participant and the specific workplace, and the a priori presence of pro-safety engineering and organizational solutions seem to have a quite important role to this end, being cited in 7, 7, 6, and 6 papers, respectively. In addition, the use of dedicated support technologies, additional training, or safety coaching (4, 5, and 3 papers, respectively) appear to influence, albeit to a lesser extent, the final behavioral results.

3.5. Analysis of Survey Papers

On the basis of the criteria detailed in Section 2.2.4, the contents of the surveys were elaborated in table form in the Supplementary Material. Related data were then reworked, so as to derive general considerations, as expressed below.

The surveys analyzed were mainly carried out in North America (the US, in particular) (four cases) and Asia (five cases), out of a total of 11 papers. The geographical area of BBS implementation is almost always the same as the land of origin of the first author, except in one case, in which a US author describes a survey that was realized in China. Industry is the main application context (seven out of a total of eleven studies), followed by transport (two cases). The subjects of the investigation are both full BBS (seven cases) and simple feedback (four cases). The number of participants in the survey varies a lot (from 56 to 2991), particularly as a function of the number of working sites involved (from 1 to more than 2000).

Surveys were submitted to different professional profiles, including primarily front-line workers or employees (for industry) and drivers (for transport). In some cases, even managers, supervisors, and safety professionals were involved. The main survey tools used were questionnaires, sometimes supported by interviews. For each study, the factors that emerged as critical for successful BBS implementation were identified. Figure 14 shows that the active role of supervisors/managers, the attention to safety culture and safety climate, and the interpersonal use of effective communication are among the most significant elements for obtaining satisfactory behavioral results. Safety culture, in particular, refers here to self-esteem, belonging, personal control, self-efficacy, and optimism, which are all measurable by means of surveys [7]. Even the attention to the prerogatives of each participant seems to have an important role to this end. Finally, forcing participation in the survey, graduating the BBS application, and using dedicated technologies have shown to positively affect, albeit to a lesser extent, the final behavioral results.

4. Bibliometric Analysis on Other Papers of Interest (“List B”)

The VOSviewer software tool has been used to analyze and visualize the bibliographic coupling between papers in List B. The papers list was exported as a “.csv” file from Scopus and opened directly in VOSviewer as bibliographical data for creating the citation map.

No threshold on the minimum number or citations for a document was set, however, papers presenting fewer than two connections were excluded from the analysis, since their contribution to the identification of common topics is expected to be marginal and potentially harmful, as it could lead to the identification of phantom and misleading relationships. Using this criterion, 20 papers were excluded since they were not connected to each other and 6 additional papers were excluded because they showed one connection only. Hence, a final set of 76 connected items was elaborated by the software; such an amount was considered adequate for representing the most important emerging topics for all the 102 papers. After elaboration, the VOSviewer finally produced the bibliographical map of papers shown in Figure 15. In this figure, only some author names are clearly visible since the VOSviewer applies an automatic graphical representation in which labels of nodes do not overlap.

The number and thickness of the connecting lines denote the quantity and intensity of links between papers. No specific weight criteria (e.g., number of citations or number of links) were applied when determining the size of circles representing nodes, since these criteria were judged as not fundamental for the scope of the present work. The temporal trend is, instead, fundamental and, accordingly, the VOSviewer “overlay visualization” was applied, in which the colors of lines and nodes reflect the publication year. As highlighted in the chromatic scale in Figure 15, yellow elements indicate the most recent years, while blue violet elements refer to 2000 and previous years.

As the second step of the analysis, graphic areas including strongly connected articles were manually identified, coupling the concept of distance-based clustering (on which the VOSviewer representation is based) with the paper categories previously hypothesized by the authors (cf. Section 2.3). The resulting areas, contoured by red lines, are shown in Figure 16.

The clusters of papers were numbered from 1 to 6, depending on the topic addressed, i.e.,:

• Subgroup 1—REAL-TIME APPROACHES (14 papers): real-time methodologies and technologies, intended to improve specific phases of BBS implementations (mainly the initial baseline assessment and the subsequent observations), are presented in this cluster of papers. Particular attention is paid to those technologies that are able to enhance quality and quickness in the identification of unsafe behaviors (e.g., intelligent video surveillance, implementation of Big Data-based platforms for image collection and automatic analysis, and image-based behavior recognition techniques). Real-time locating and tracking technologies and devices for automated monitoring of location and movements of workers, as well as equipment in order to prevent risk exposure and access to hazardous areas, eventually in connection with a general Building Information Model, are analyzed in some papers. Alert and warning technologies constitute a further topic of interest, together with their key advantages compared to more traditional ones (as direct site observation).
• Subgroup 2—PRE-INTERVENTION ACTIVITIES (six papers): these papers analyze those activities that can be seen as propaedeutic to real behavioral interventions, such as evaluations of safety baseline (safety audit for data collection, postural analysis on workers), identification of optimal performance targets for subsequent interventions, and evaluation of peculiar needs of the specific working context. Simple applications of natural or material incentives are also included.
• Subgroup 3—SAFETY CLIMATE AND LEADERSHIP ROLE (24 papers): these are collateral topics compared to the traditional BBS intervention. Safety climate is addressed by several papers in this cluster, in the form of a review, a presentation of practical cases in the working environment, or conceptual studies. The employees’ safety perception, the role of social norms, and the direct effects of safety climate on workers’ safety behaviors are examples of the themes analyzed and, somewhere, even modeled. Safety climate represents a crucial topic even in papers where the role of leaders and managers in BBS implementations are touched on, although to a lesser extent. In other studies, instead, the leadership role in behavioral interventions is the main subject (e.g., leadership-based interventions, specific training for leaders and managers, role of supervisors), while safety climate is just briefly addressed.
• Subgroup 4—SAFETY CULTURE (six papers): this cluster of papers focuses on safety culture in its connection to or difference from behavioral interventions. Indeed, the literature [257] has shown that BBS and safety culture can be considered complementary and can be merged into a more balanced and comprehensive approach to safety. Similarly, connections between this topic and safety climate are highlighted, even with reference to practical cases. Other specific topics with a potential cultural impact (e.g., “nudges” techniques pushing workers towards correct behaviors) are also addressed in this cluster. General reviews about safety culture are included, too.
• Subgroup 5—TRAINING (seven papers): these papers cover safety training in its more traditional form as well as in more technological evolutions (e.g., the use of simulator-based exercises) or through the analysis of some conceptual debates about training and learning techniques (e.g., the evaluation of possible intermediate steps between direct learning from experience and the passive application of legal regulations).
• Subgroup 6—BBS VARIANTS AND ALTERNATIVE METHODOLOGIES (19 papers): the last cluster brings together papers touching on different subjects, but linked by the presentation of solutions in the form of “alternative to”/“variant of” BBS or by the development of case studies in complex contexts, where behavioral aspects have to be included in a broader analysis of global safety. For example, the “stage of change” (SOC) approach, where behavior change is considered a dynamic process with different stages where the worker can be located, appears in this subgroup; the same holds true for the Social Ecological Framework, also investigated by Uchenna Okoye in 2016 [268] and Hoque et al. in 2019 [269], which is grounded in the fact that the safety performance of workers and the eventual applied behavior change techniques are significantly influenced in a multi-level system by individual, interpersonal, organizational, community environment, and even policy factors. In several papers, instead, dynamical models of complex systems (as critical and high-hazard industrial plants or big companies), are described; in some places, only the assessment of risk and safety issues is touched on, but in many papers a complete analysis of the retroactive effect of behavioral interventions and of their interactions with organizational and technical pro-safety solutions, in a multi-criteria perspective, is included.

Table 7. List of the papers included in each subgroup and of the papers excluded from the analysis.

Subgroup Name	Included Papers
Subgroup 1: Real-time approaches	Arslan et al., 2019 [150]; Chen et al., 2022 [244]; Cheng et al., 2022 [243]; Duan et al., 2023 [248]; Fang et al., 2020 [238]; Guo et al., 2016 [183]; Guo et al., 2018 [181]; Guo et al., 2018 [182]; Guo et al., 2021 [241]; Han and Lee, 2013 [185]; Han et al., 2014 [186]; Soltanmohammadlou et al., 2019 [218]; Teizer, 2016 [223]; Yu et al., 2017 [232].
Subgroup 2: Pre-intervention activities	Bigelow et al., 1998 [154]; Fante et al., 2007 [174]; Reber and Wallin, 1983 [214]; Sulzer-Azaroff and Fellner, 1984 [219]; Wilder et al., 2018 [228]; Winn et al., 2004 [229].
Subgroup 3: Safety climate and leadership role	Christian et al., 2009 [165]; Cooper and Phillips, 2004 [166]; Eliseo et al., 2012 [172]; Fang et al., 2015 [173]; Gravina et al., 2017 [179]; Gravina et al., 2019 [180]; Grill et al., 2023 [240]; Haas et al., 2016 [184]; Howard Quartey, 2017 [188]; Jiang and Probst, 2015 [189]; Jiang and Probst, 2016 [190]; Luria and Morag, 2012 [199]; Martínez-Córcoles et al., 2011 [200]; Newnam et al., 2012 [204]; Oah et al., 2018 [206]; Olsen, 2010 [209]; Patel and Jha, 2016 [211]; Pessemier and England, 2012 [213]; Tafvelin et al., 2019 [220]; Warmerdam et al., 2018 [227]; Zohar and Erev, 2007 [235]; Zohar and Luria, 2003 [236]; Weaver et al., 2023 [245]; Zohar and Polachek, 2014 [237].
Subgroup 4: Safety culture	Al-Refaie, 2013 [148]; Chen and Jin, 2013 [159]; Choudhry et al., 2007 [163]; Choudhry et al., 2007 [164]; Håvold and Nesset, 2009 [187]; Lindhout and Reniers, 2017 [196].
Subgroup 5: Training	Brandhorst and Kluge, 2021 [246]; Brandhorst and Kluge, 2022 [247]; Crichton, 2017 [169]; Demirkesen and Arditi, 2015 [171]; O’Connor and Flin, 2003 [208]; Papaleo et al., 2013 [210]; Vidal-Gomel, 2017 [226].
Subgroup 6: BBS variants and alternative methodologies	Arandia et al., 2020 [149]; Awolusi and Marks, 2017 [151]; Azadeh and Mohammad Fam, 2009 [152]; Bergsten et al., 2018 [153]; Bouloiz et al., 2013 [156]; Choi and Lee, 2018 [161]; Choi et al., 2017 [160]; Daǧdeviren and Yüksel, 2008 [170]; Fonseca et al., 2019 [176]; Goh et al., 2018 [177]; Jiang et al., 2015 [191]; Lai et al., 2011 [194]; Lunt et al., 2019 [198]; Oakman et al., 2016 [207]; Paul and Maiti, 2007 [212]; Shin et al., 2014 [216]; Sing et al., 2014 [217]; Zhang et al., 2022 [239]; Zhou et al., 2014 [233].
Papers excluded from the analysis	Alovosius et al., 2009 [147]; Bolton, 2001 [155]; Candefjord et al., 2015 [157]; Cape-li-Schellpfeffer et al., 2000 [158]; Choi and Loh, 2017 [162]; Cournoyer et al., 2011 [167]; Crawford, 1991 [168]; Fell-Carlson, 2004 [175]; Grabowski and Jankowski, 2015 [178]; Ismail and Ramli, 2023 [242]; Kilcup et al., 2007 [192]; Kuipers et al., 2016 [193]; Lindell, 1994 [195]; Lindsay, 1992 [197]; Mattila et al., 1994 [201]; Mazaheri et al., 2009 [202]; Mostia, 2009 [203]; Nouri et al., 2008 [205]; Rose and Harshbarger, 1991 [215]; Tait and Walker, 2000 [221]; Tam and Fung, 1998 [222]; Teo, 2007 [224]; Tong et al., 2019 [225]; Wong et al., 2009 [230]; Yogeswara et al., 2013 [231]; Zohar, 2002 [234].

shows the list of papers included in each subgroup, with the addition of the above-mentioned 26 papers that were excluded from the analysis since they showed no reciprocal connections or one connection only.

By comparing the topics addressed in the six subgroups of papers with those hypothesized previously by the authors (cf. Section 2.3), a good correspondence can be observed (e.g., in the identification of papers related to safety culture and training), although the groups in Figure 16 merge some themes. Safety climate and leadership (points ii and iv of the previous list, respectively) are mixed into subgroup 3; this appears reasonable since the behavior of managers and supervisors in a company can have a big influence on global safety perception. General safety analyses and variants of classical BBS (points xi and vii of the previous list, respectively) are merged, too, in subgroup 6; in some cases, in fact, BBS variants also consist of complex structured systems and vice versa. Complexity can also be related to Industry 4.0, since technological advancements and interconnected systems present new risks necessitating innovative management strategies, as observed in 2019 by Brocal et al. [270] The integration of cyber–physical systems, IoT, and data analytics engenders a more intricate operational landscape, requiring a thorough comprehension of potential risks, a proactive and cohesive strategy for risk management and dynamic risk management, and assessment strategies that adjust to evolving conditions. Proficiently managing emerging risks in Industry 4.0 can improve overall organizational performance and resilience, resulting in safer and more efficient operations. A final group includes “isolated” papers that were excluded from the bibliographical mapping because of the low number of links.

It has to be observed, instead, that category viii was split into subgroups 1 and 2, since the graphical mapping highlighted the importance of the group of papers expressly concerning real-time approaches and technologies (subgroup 1). Moreover, the VOSviewer overlay visualization, focusing on the temporal distribution of publications, shows that subgroup 1 generally referred to very recent years (from 2013 onwards), while subgroup 2 (pre-intervention activities) additionally contains papers published in the 1980s and 1990s.

Similarly, subgroup 6 mainly consists of very recent papers. Concerning subgroup 3, it can be observed that papers discussing leadership topics are generally more recent than papers covering the broader subject of safety climate. This latter theme is also closely linked to the papers included in subgroup 4 (safety culture), as shown by the high density of connections displayed in Figure 16; both topics have had, in fact, a parallel development in the literature, from the first years after 2000 onwards (see Table 7).

5. Results Interpretation and Discussion

5.1. Evolution of the BBS Methodology

The analyses proposed in Section 3 offer various ideas inspiring general observations and reflections on how BBS has evolved over time and how it currently stands in the general framework of safety management at the workplace. Bibliometric analysis, shown in Section 4, has confirmed and enriched these results by highlighting the evolutionary and adaptive trends of BBS-related topics over time.

Over the years, BBS approaches have been applied and studied, referring to several contexts, of which industry (including construction) and transport have been the dominant ones from the beginning until today. Feedback and observations have been the foundation of the BBS method since its origin, when the nomenclature used was not yet as diversified as today, although increasing attention paid to the observation phase (i.e., timing, methods, and reliability) emerged, especially since the 2000s.

5.2. Strengths and Limitations of the Approach

From the analyzed literature it is evident that positive feedback driven by observation appeared, in the analyzed papers, as the BBS component demonstrating the greatest persistence of effectiveness, while goal setting and material rewards did not always appear as key elements for the success of BBS initiatives. Frequently, when the methodology was considered successful by its implementors, performance monitoring was not entrusted only to the evaluation of outcomes (e.g., number of accidents), but mainly to the direct observation of behaviors. The latter also makes it possible to grasp “near misses” by acting in advance and providing workers with intermediate feedback, rather than an unpleasant definitive judgment on the final performance. In line with this point, criticisms against BBS refer exactly to the risk of blaming workers and of violating their privacy, so these eventualities should be evaluated and avoided in companies if the involvement of internal workers and approval from external stakeholders are desired. Interestingly, criticisms of BBS are particularly discussed in case study/laboratory papers and in conceptual papers, with a slightly different view. Taking into account the different nature of case studies vs. conceptual papers, this could indicate that practical implementations of the BBS approach, at least at a laboratory level, allow for the better capturing of the critical aspects of the methodology when used in industrial settings.

In any case, both the counting of accidents or injuries and the more complex assessment of behavior still appear as relevant in most recent applications.

The use of additional training (in traditional form or in more technological ones) has proved to be useful in combination with feedback, which, however, remains the preferred tool for obtaining satisfactory results. The a priori presence of technical solutions and appropriate organizational procedures is typically a necessary condition for the behavioral intervention to achieve complete efficacy in the considered context.

Another finding from the review is that many behavioral interventions involved improvements in the safety level of the workplace, partially or totally. Although this indicates that BBS has the potential to be effective in various contexts, care must be taken when interpreting this outcome, since few studies only assessed the statistical significance of the observed outcomes. A statistically significant reduction in the targeted outcome should actually be reached to demonstrate the effectiveness of the intervention, without being satisfied with anecdotal evidence alone [35]. Moreover, regardless of the thresholds that each investigator can set for certifying the achievement of a statistically significant outcome, achieving a minimum level of improvement in safe practices (or reduction of accidents), as claimed by company leaders, is a prerequisite for authorizing the continuation of the behavioral intervention [62]. This implicitly means that additional research would be appropriate for providing statistical evidence of BBS effectiveness.

Similar considerations hold true for the long-term preservation of BBS results; although there is evidence that long-term effects have somehow been achieved and safeguarded, partially or totally, it is not crucial to identify the exact factors that have affected that lasting effectiveness. In general, the results obtained allow us to surmise that interpersonal trust, the usage of effective communication, workers’ involvement, intense commitment of managers and supervisors, and attention to safety culture and safety climate have a very significant, positive influence on the results of practical BBS applications, both in the short and long terms. Again, this is particularly evident from the analysis of the case study/laboratory papers, thus confirming that the effectiveness of the results can be fully captured if implementing the BBS approach (at least at the laboratory level). Nonetheless, these aspects would benefit from additional investigation in the future. Another piece of evidence from the review outcomes is that results after withdrawal are generally satisfactory when housekeeping is used as KPI, as the employees can appreciate the change and learn to read the feedback directly from the environment [26]. Exploring this point in detail could represent an additional future research direction.

Evidence from the literature review also suggests that, when planning a behavioral intervention, the implementation context needs to be carefully evaluated. In fact, the literature reports numerous behavioral interventions in industrial fields, such as, primarily, the manufacturing industry and construction, where the BBS implementation is well established, but many other contexts are also touched upon. Again, regarding the implementation of the BBS initiative, regardless of the number of steps carried out, a common challenge emerging from the studies reviewed is that the implementation requires a significant amount of time to be completed. The analysis of both the baseline scenario and the post-intervention one is particularly time-consuming, because of the need to collect enough accident/injury data or behavioral observations. The intervention step may also have a considerable duration. Time obviously entails cost; in this regard, it should be noted that the present research does not expressly analyze the economic aspects of BBS implementation, even if authors are aware that cost represents a relevant point for evaluating whether the intervention could be sustained in time by companies [79]. The literature reports just some fragmented cost data provided in case studies; it is suggested that additional scientific work is dedicated to this topic.

5.3. Interrelated Aspects

Safety culture and safety climate represent a catalyst for correct behaviors in the long run, too; these aspects turned out to be emerging BBS topics in the bibliometric analysis, but their importance began to be evident already in the 1990s. The enduring effects of BBS programs are reinforced by continued practice in safe working procedures and a redefinition of individual awareness and group social norms. Safety culture plays a crucial role, and it is often represented by the lasting presence of new correct behavioral routines. However, a collective comprehension of risk is necessary for cultivating a safety culture. For this reason, in 2015 Montibeller and Von Winterfeldt [271] examined the influence of cognitive biases on decision-making and proposed strategies for their recognition to enhance decision quality. Cognitive biases are systematic deviations from norms or rationality in judgment, which can influence decision-making in complex scenarios and shape individuals’ perceptions of reality. Motivational bias could lead individuals to preferentially select information or interpret experiences in accordance with their desires, objectives, or beliefs. This may result in overestimating the likelihood of positive outcomes or in a minimization of adverse information. Strategies exist to assist individuals in recognizing and comprehending biases, such as educating them about prevalent biases to enhance recognition or fostering collaboration within teams to introduce diverse perspectives that mitigate individual biases. The implementation of these strategies can improve decision-making quality within organizational contexts, resulting in superior outcomes.

The analyses have also shown that safety climate is strictly related to the role of leadership in the considered working environment, although studies about leadership are generally more recent than papers about safety climate. That is why the active role and involvement of managers, supervisors, and other leadership figures appear nowadays as a significant topic toward which future research should be addressed.

5.4. Recent Trends

Overall, based on the above considerations, the analysis carried out has brought to light the following practical implications in the main research themes emerging from the clustering of the selected studies.

(i)

Real-time approaches: the use of digital technologies in workplace safety is more and more diffused, and key technologies such as intelligent video surveillance, Big Data platforms for automatic image collection and analysis, and image-based behavior recognition techniques, as well as real-time locating and tracking technologies, potentially integrated with BIM tools, can certainly augment the practical implementation of BBS. This is in line with recent studies fostering the use of Safety 4.0 solutions [272].

(ii)

Pre-intervention activities: this research trend also brought to light the relevance of the provision of safety training and safety instructions to enable workers to properly use work equipment and carry out maintenance operations. In line with extant studies, these activities are among those that are most characterized by workers’ risk-taking behaviours [273,274]. The trend of carrying out an ever-greater diversification of approaches, to provide more contextualized and effective answers, is confirmed too.

(iii)

Safety climate and safety culture: the analysis of the literature related to the implementation of safety culture and safety climate approaches to enhance workers’ safety behaviour demonstrates the attention that should be paid to workers’ personality, attitudes, skills, cultural background, as well as to the peculiarities of the workplace and working task, which might largely differ from one context to another [275]. Actually, both safety climate and culture impact workers’ perceptions of safety policies, procedures, and practices within a specific work environment [276].

The present study has also highlighted the development of empirical BBS applications in various declinations and variants, particularly in the first and second decades of the third millennium. These variants seem to be suitable for being customized and declined for being effective in specific working environments, mainly thanks to the availability of such modern technologies with real-time approaches and ever-increasing attention paid to the prerogatives of each individual, as well as of the specific workplace involved. The support given by these technologies has particularly emerged in recent times, and because of the newness of the topic, some aspects still have to be evaluated. For example, drivers sometimes fear that depending too much on technology will reduce their original safe driving skills [143].

In line with this set of considerations, the analysis of conceptual studies (as well as the bibliometric analysis) has highlighted the importance of discussing and proposing guidelines, but also, in the last two decades, the emerging fundamental role of specific phenomena influencing behaviors (i.e., response generalization), and especially the appearance of new specific methodology variants or tools. In this respect, the importance that recent studies have given to insights into the pre-intervention activities (e.g., evaluations of the safety baseline, identification of optimal performance targets, evaluation of peculiar needs of the specific working context) also confirms the trend of carrying out an ever-greater diversification of approaches to provide more contextualized and effective answers.

The role of language in attracting attention to BBS (and organizational behavior management in general) from all potential practitioners (educational context and/or training for organizational and community application), as discussed by Geller in 2002 [119], appears here as particularly critical due to the wide differentiation assumed over time by the nomenclature in the field.

6. Conclusions

The present study was designed with the goal of reviewing conceptual as well as empirical research in the international literature regarding the implementation of Behavior-Based Safety in the workplace over time, from the 1970s until 2023, so as to highlight its limits, strengths, and potential. A more general aim was to provide useful data and observations for helping companies guarantee safety in the workplace, thus increasing their social sustainability.

The high number of relevant studies identified, together with the various facets that emerged from BBS applications, pushed us to divide such an analysis into two parts, i.e., an in-depth review of the most pertinent content and an additional bibliometric graphical mapping for less related contents. This double process made it possible to enrich the quantitative analysis of studies with evaluations of a more qualitative nature, thus allowing for a comprehensive description of the complex web of application procedures and theoretical variants that BBS has assumed over time, even from the point of view of the nomenclature used. A great percentage of the classified original data have been made available to the reader as Supplementary Material, while analyses have been presented in aggregate form through graphs and tables.

As with any other review paper, it is important to point out that this study presents an inherent limitation related to the initial phase of paper selection, which required six queries on Scopus. It was verified by the authors during the long analysis phase that such a selection, although very wide, cannot claim to be 100% all-inclusive; a few useful papers about BBS or related topics have been excluded, even if they appear in literature, in most cases because of very short abstracts, an absence of keywords, or the use of non-standard nomenclature. In addition, a study by Leivo (2001) [260] was not identified, where the concept of company “housekeeping” is implicitly related to the “safety” concept, even if is not directly cited in the abstract but was several times in the main text. Similarly, in the case study presented by Menckel et al. in 1997 [254] and not recognized by the Scopus search, the concept of “behavior” was not clearly mentioned in the title/abstract/keywords. In other cases, some pertinent papers have not been directly identified by the queries because, although they have been published in referenced scientific journals, Scopus has indexed them only for a limited number of years (that is the case for the “Professional safety” journal), and therefore it does not show some of them. Despite these aspects, however, the research field investigated appears extremely wide, and the unintentionally excluded papers are few. Overall, the level of detail of the present study aligns with the initial purpose of realizing an effective review of the BBS literature.

From a theoretical point of view, the obtained results suggest that future prospects of BBS can be considered closely related, and its most successful traditional principle, i.e., positive feedback driven by observation, is able to be combined with the awareness of new needs and research trends that have emerged as fundamental in the last two decades, mainly: (i) real-time functionality through technological support, (ii) involvement of all hierarchical levels in companies, (iii) attention to workers’ (and workplace) peculiarities and, consequently, (iv) the ability to adapt the method to different contexts through more and more specific application variants. From a practical perspective, the results of the review highlight that the concept of behavioral safety is no longer confined to the principles of positive feedback and observation but is becoming a broader concept that also incorporates technological aspects and the involvement of all levels of the company. This allows safety to be viewed from a more holistic perspective, to be adaptable to different contexts, and be able to consider the specific needs of various types of worker. At the same time, the usage of more modern technologies would allow companies to implement more sophisticated indicators for measuring not only accident or injury reduction, but also worker involvement and awareness. Future research could investigate how to develop these indicators and assess their impact on overall safety.