What Is Risk Management? Transforming Workplace Safety with Data-Driven Insights

What can be done today to prevent worker injuries and deaths tomorrow? The answer lies in the notion of predictive analytics discovered through the use of risk management software that use historical data to identify the workplace danger waiting to strike.

The proactive shift: Anticipating hazards in workplace safety

Rather than reacting or responding to what’s already happened, EHS leaders should be looking to anticipate and forecast potential hazards and implement necessary change and prevention strategies before an employee is harmed.

Predicting negative events or incidents that might happen in a workplace is a matter of identifying “causal factors” through data analysis, but these clues are often not apparent and likely to be hidden in variables not captured in incident reports.

Comprehensive data sourcing for risk prevention

A causal factor can be defined as any major unplanned or unintended contributor to an incident that if eliminated would prevent or reduce the severity or frequency of the incident. Identifying causal factors requires digging deeper into information sources that might include equipment operation and process data, vehicle telemetry, weather, geospatial, socio-demographic, human resources (payroll, performance data) and training, industry and other data.

Through predictive modeling techniques and risk management software, it’s possible to pinpoint the driving factors behind workplace incidents and develop prevention strategies.

Overcoming data collection challenges in risk management

One of the greatest obstacles in analyzing and potentially exposing risks is finding and collecting data, which is usually scattered across various business systems and often difficult to access. Not all data is written in common syntax, so it must be interpreted and standardized, and data quality is always an issue.

Accurate data, and by extension, data analysis, will become increasingly more important for EHS. In its report, 4 Trends in Occupational Safety and Health to Expect in 2021, Columbia Southern University noted that: “It will be crucial for all businesses to use their safety data to perform predictive safety modeling. This modeling will aim to anticipate potential safety hazards and establish which conditions increase the occurrence of incidents.”

It goes on to say: “To remain competitive, companies will employ machine-learning-dependent safety software and attempt to stop accidents in their tracks before they occur.” Noted was the fact that “predictive analytics are only as reliable as the data they are based on.” 

The value of the risk assessment

Risk assessment seeks to identify and evaluate hazards found in a work environment before incidents happen, determine the level of danger that exists in those hazards and assess the likelihood of harm occurring. It is also the guiding principle behind job safety analysis (JSA) and job hazard analysis (JHA). Risk hazard needs to be evaluated with criteria that help to build a credible understanding of what is and is not acceptable. Most regulatory bodies require some form of risk assessment of hazards, and all follow a similar template that includes:

• Identifying risks to the worker associated with work activity
• Identifying hazards found in the work environment that pose a threat of loss
• Providing details of identified risks or hazards and context to build understanding
• Utilizing a measurement system to evaluate risk understanding and determining precautions
• Building controls that protect people and the work environment   

Building a Risk Matrix

A risk matrix is a great assessment tool for evaluating and estimating risk level. It helps to judge whether the hazard and possible risk are acceptable, scores it and then plots findings on a matrix chart to ultimately determine the level of control required to reduce the risks and mitigate incidents. If the activity rating rises above acceptable levels, then controls are warranted to lower scores. 

For every hazard identified during an inspection and for each associated activity, it’s important to ask the question “what if?”  What could be the worst-case outcome regarding a hazard or concerning activity?  Is it a fatality, significant injury, permanent disability or health effect?  Is it a minor injury, an environmental concern or something that could cause damage? A risk matrix is a tool to judge the likelihood and severity of harm, based on a criterion that includes:

• Severity – The degree or amount of expected loss
• Likelihood – How likely that the loss will occur
• Risk Rating – The probability and severity of the risk before and after control actions are taken

Below is a risk matrix example:

A five-point (5×5) matrix, such as the those presented below, estimate the likelihood (probability) and severity (consequence) within the five descriptive levels.

Likelihood scale

Rating Score	Qualitative Element	Definition
0	Impossible	No injury or illness, damage or other loss is possible.
1	Improbable	Loss, injury or illness could only occur under freak conditions. The situation is well managed, and all reasonable precautions have been taken.
2	Unlikely	This situation is generally well managed. However occasional lapses could occur. This also applies to situations where people are well trained and required to behave safely to protect themselves.
3	Likely	Insufficient or substandard controls in place. The loss is unlikely during normal operation however, it may occur in emergencies or non-routine conditions.
4	Very Likely	Serious failures in management controls exist. The effects of human behavior or other factors could cause an accident but are usually supported by this additional factor (e.g., ladder not appropriately secured, process upset, oil spilled on the floor, poorly trained personnel).
5	Almost Certain	Absence of management controls. If conditions remain unchanged, there is nearly 100 percent certainty that an accident will happen (e.g., broken rung on a ladder, live exposed electrical conductor).

Severity scale

Rating Score	Qualitative Element	Definition
0	None	No injury or illness, damage, sickness or other loss is possible.
1	Minor	Minor injury, illness or loss is possible (e.g., light cuts, scratches, insignificant damage to property).
2	Low	Significant injuries or illnesses are possible (e.g., sprains, bruises, lacerations and events needing medical care). Damage to property or process.
3	Medium	Temporary disability is possible (e.g., fractures, finger amputation). Lost workdays due to injury or illness; substantial damage or loss of property or process.
4	High	Permanent disability is possible (e.g., significant loss of movement, loss of limb, sight or hearing)
5	Major	Causing death to one or more people. Loss or damage is such that it could cause serious business disruption (e.g., major fire, explosion or structural damage).

FAQs about risk management

How do you effectively integrate and manage data from disparate sources for predictive analytics?

Effectively integrating and managing data from disparate sources for predictive analytics is a crucial step for EHS managers looking to predict and prevent workplace incidents. This can be achieved by implementing a robust data management system that allows for the aggregation of data from various platforms. Such a system should include protocols for validating the accuracy and consistency of the data, as well as tools for data cleaning and normalization to ensure that all data is in a usable format. Additionally, using middleware or specialized software that can interface with different databases and business systems can help in harmonizing disparate data types. EHS managers may need to work closely with IT professionals to create a seamless process that facilitates regular data collection and updates.

What specific machine learning algorithms or models are most effective for predicting workplace incidents, and how can they be implemented?

It depends on the nature of the data and the specific use case. Commonly used algorithms include decision trees, logistic regression, and neural networks, among others. These models can uncover complex patterns and relationships within the data that may not be immediately apparent. To implement these effectively, EHS managers should consider working with data scientists or analysts who can select the appropriate model, train it on historical data, and validate its predictions. Continuous monitoring and refinement of the model are necessary to improve its predictive capabilities as new data becomes available.

How can the effectiveness of implemented control measures be evaluated and continuously improved?

Evaluating the effectiveness of implemented control measures is vital for ensuring workplace safety. This can be done by setting up key performance indicators (KPIs) that are aligned with the safety goals and by monitoring these KPIs regularly to track progress. Incident rates, near-misses, and other safety-related incidents can be analyzed before and after the implementation of control measures to assess their impact. Additionally, soliciting feedback from employees about the perceived effectiveness of these controls can provide insights into areas that may need adjustment. Continuous improvement can be facilitated by a process of regular review and updating of risk assessments, control measures, and predictive models to reflect the changing work environment and emerging data patterns. This iterative process ensures that safety measures remain effective and responsive to the workplace’s current conditions.

Deliver the right program control for job safety

Coupled with the risk assessment and risk matrix described above, Intelex Vice President of Health and Safety Scott Gaddis recommends the application of a widely accepted approach called a hierarchy of controls. He describes it as a simple-to-understand process that’s useful in gauging the control appetite of an organization. It should serve as an overarching methodology for how to deliver the right level of program control for job safety and job hazard management. 

An example from Gaddis’s report, Walking-Working Surfaces and Pedestrian Safety, offers a methodology for creating a control hierarchy – in this case for walking-working surfaces risks – and highlights the most effective controls, emphasizes engineering solutions, reveals administrative controls and the necessary reliance on personal protective equipment. (See the illustration, below.)

Ranking	Control	Control Feature	Example of Control
Most effective	Elimination	Completely eliminate the hazard	– Removing a hazardous chemical entirely from a manufacturing process and replacing it with a non-hazardous one.
	Substitution	Reduce a hazardous situation, component, material, and/or piece of equipment that does not have the same level of hazard. Substitution often requires significant changes and can be difficult to implement.	– Install slip-resistant flooring – Improve material handling practices to reduce personnel movement outside localized workspace.
	Engineering	Isolate people from an identified hazard or risk. While this control is viewed not as protection as elimination or substitution, this focus still controls exposure at the source of the hazard.	-Install guard barriers to separate pedestrians and material handling equipment. – Stop contamination from getting to the walking-working surface. – Improve illumination in walking areas.
	Administrative	Take steps to follow and support safe work practices by altering the way workers perform their work and reduce risks they have on the job.	– Update safety policies, rules, supervision, schedules and training.
Least effective	PPE	Protection worn by a worker for protection or to reduce loss threshold from a hazard.	– Slip-resistant footwear – Cut-resistant gloves – High-viz clothing

Learn more tips and best practices for job safety and hazard analysis by downloading the Intelex Insight Report: The More You Know (Part 1): The Essentials of Job Safety Analysis.