Predictive-based safety provides a new approach as jobsites struggle with limiting workplace hazards

In the U.S. Bureau of Labor Statistics annual report on fatal work injuries for 2020, the news looks good at first glance. There were 4,764 fatal occupational injuries, which is the lowest number since 2013. 

However, the raw number of fatal injuries doesn’t account for the fact that there were a lot fewer people working during the first year of the pandemic. Many construction sites were completely shut down for months.

So, when we look at the fatal work injury rate per 100,000 FTE workers, we get 3.4, which is just slightly less than the figure of 3.5 in 2019. Both rates are still too high—this means someone died of a workplace injury in the U.S. roughly every two hours—but at least the figures are moving downward, if only marginally so.

Unfortunately, this moderately good news doesn’t extend to construction, where the fatality rate was 10.2 per 100,000, the worst for at least the last decade. And we can’t blame that on COVID. Those deaths were not counted for this report. It’s a disheartening step backwards for the industry. 

But it’s not a surprising one. Business Insider analyzed the BLS data to tally the top 34 deadliest jobs in America, and 14 of them are construction-related. That’s 41%.

Time for a new approach

The fact is that construction fatality and recordable incident rates largely have remained the same for more than a decade. And while traditional, behavior-based programs significantly improved safety in the industry, they’ve clearly run into a brick wall. 

Traditional behavior-based safety programs aim to change worker behavior by collecting observations on a form or checklist. The data from these checklists are then aggregated to identify the most common issues and inform what the organization emphasizes in site supervision and training. Interaction with the crews while making observations is typically optional.

The problem is that behavior-based safety programs are always looking backwards to understand what already happened. If we want to start seeing recordable incident and fatal accident rates decline again in a serious way, we need to start looking forward, using predictive analytics, AI, and modern data science to predict where the highest risks will be and then take measures to mitigate them. What we need is predictive-based safety (PBS).

Predictive-based safety

PBS differs from traditional safety programs in that it makes it easy for anyone on a jobsite to conduct data collection with a mobile device and have safety conversations with craft workers.

Once data is collected, AI analyzes it along with a wide range of other information that contractors already have such as payroll, site images, incidents, and even weather data that AI will crunch to predict the future. Specifically, PBS can identify the 20% of projects that carry 80% of safety risk over the next week.

And that’s not all. A good PBS program also will provide recommendations for reducing risk so safety teams can direct exactly the right measures to the highest risk sites.

The results have been nothing short of astonishing. Adopters of PBS typically have seen their recordable incident rates drop by 30% over a 12-month period, with some seeing reductions of up to 50%.

Insurance companies also are starting to recognize the power of PBS. The Hartford, for instance, is paying for pilot deployments of our construction-trained AI to enable PBS with select clients that it insures. With the data that PBS provides, construction companies can provide strong evidence that they are at a lower risk to negotiate a better insurance rate.

A predictive-based safety pioneer

We are now at the forefront of enabling and developing technology and methods so construction companies can effectively implement PBS. Our safety observations module makes it simple for everyone to collect on-site safety data and initiate valuable conversations. 

Safety monitoring transforms the site images that contractors already are collecting into safety observations. Our AI can identify more than 100 hazards, from slip, trip, and fall hazards to workers at height and improper ergonomics.

Finally, predictive analytics brings it all together to create an incident early warning system. Trained on more than 20 centuries worth of historical project data and incident reports, our construction-specific AI model analyzes data from the safety monitoring, safety observations, and a wide array of other data sources to predict safety incidents and provide recommendations for how to prevent them. 

With predictive-based safety, construction can break through this barrier to start making significant progress reducing accident and fatality rates, with the aim of everyone going home safely every night. That’s a goal worth working toward.

Learn how developing a predictive-based safety program with Oracle Construction and Engineering has the power to substantially reduce your organization’s risk.